diff --git a/layer_0/width_131k/average_l0_11/hparams.json b/layer_0/width_131k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_0/width_131k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_0/width_131k/average_l0_15/hparams.json b/layer_0/width_131k/average_l0_15/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_0/width_131k/average_l0_15/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_0/width_131k/average_l0_21/hparams.json b/layer_0/width_131k/average_l0_21/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_0/width_131k/average_l0_21/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_0/width_131k/average_l0_30/hparams.json b/layer_0/width_131k/average_l0_30/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_0/width_131k/average_l0_30/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_0/width_131k/average_l0_41/hparams.json b/layer_0/width_131k/average_l0_41/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_0/width_131k/average_l0_41/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_0/width_131k/average_l0_8/hparams.json b/layer_0/width_131k/average_l0_8/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_0/width_131k/average_l0_8/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_0/width_16k/average_l0_11/hparams.json b/layer_0/width_16k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_0/width_16k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_0/width_16k/average_l0_129/hparams.json b/layer_0/width_16k/average_l0_129/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_0/width_16k/average_l0_129/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_0/width_16k/average_l0_17/hparams.json b/layer_0/width_16k/average_l0_17/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_0/width_16k/average_l0_17/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_0/width_16k/average_l0_35/hparams.json b/layer_0/width_16k/average_l0_35/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_0/width_16k/average_l0_35/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_0/width_16k/average_l0_68/hparams.json b/layer_0/width_16k/average_l0_68/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_0/width_16k/average_l0_68/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_1/width_131k/average_l0_13/hparams.json b/layer_1/width_131k/average_l0_13/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_1/width_131k/average_l0_13/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_1/width_131k/average_l0_20/hparams.json b/layer_1/width_131k/average_l0_20/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_1/width_131k/average_l0_20/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_1/width_131k/average_l0_33/hparams.json b/layer_1/width_131k/average_l0_33/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_1/width_131k/average_l0_33/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_1/width_131k/average_l0_56/hparams.json b/layer_1/width_131k/average_l0_56/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_1/width_131k/average_l0_56/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_1/width_131k/average_l0_6/hparams.json b/layer_1/width_131k/average_l0_6/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_1/width_131k/average_l0_6/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_1/width_131k/average_l0_9/hparams.json b/layer_1/width_131k/average_l0_9/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_1/width_131k/average_l0_9/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_1/width_16k/average_l0_15/hparams.json b/layer_1/width_16k/average_l0_15/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_1/width_16k/average_l0_15/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_1/width_16k/average_l0_175/hparams.json b/layer_1/width_16k/average_l0_175/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_1/width_16k/average_l0_175/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_1/width_16k/average_l0_31/hparams.json b/layer_1/width_16k/average_l0_31/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_1/width_16k/average_l0_31/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_1/width_16k/average_l0_69/hparams.json b/layer_1/width_16k/average_l0_69/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_1/width_16k/average_l0_69/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_1/width_16k/average_l0_9/hparams.json b/layer_1/width_16k/average_l0_9/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_1/width_16k/average_l0_9/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_10/width_131k/average_l0_15/hparams.json b/layer_10/width_131k/average_l0_15/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_10/width_131k/average_l0_15/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_10/width_131k/average_l0_151/hparams.json b/layer_10/width_131k/average_l0_151/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_10/width_131k/average_l0_151/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_10/width_131k/average_l0_27/hparams.json b/layer_10/width_131k/average_l0_27/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_10/width_131k/average_l0_27/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_10/width_131k/average_l0_47/hparams.json b/layer_10/width_131k/average_l0_47/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_10/width_131k/average_l0_47/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_10/width_131k/average_l0_84/hparams.json b/layer_10/width_131k/average_l0_84/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_10/width_131k/average_l0_84/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_10/width_131k/average_l0_9/hparams.json b/layer_10/width_131k/average_l0_9/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_10/width_131k/average_l0_9/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_10/width_16k/average_l0_10/hparams.json b/layer_10/width_16k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_10/width_16k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_10/width_16k/average_l0_113/hparams.json b/layer_10/width_16k/average_l0_113/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_10/width_16k/average_l0_113/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_10/width_16k/average_l0_17/hparams.json b/layer_10/width_16k/average_l0_17/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_10/width_16k/average_l0_17/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_10/width_16k/average_l0_243/hparams.json b/layer_10/width_16k/average_l0_243/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_10/width_16k/average_l0_243/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_10/width_16k/average_l0_31/hparams.json b/layer_10/width_16k/average_l0_31/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_10/width_16k/average_l0_31/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_10/width_16k/average_l0_57/hparams.json b/layer_10/width_16k/average_l0_57/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_10/width_16k/average_l0_57/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_11/width_131k/average_l0_16/hparams.json b/layer_11/width_131k/average_l0_16/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_11/width_131k/average_l0_16/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_11/width_131k/average_l0_162/hparams.json b/layer_11/width_131k/average_l0_162/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_11/width_131k/average_l0_162/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_11/width_131k/average_l0_27/hparams.json b/layer_11/width_131k/average_l0_27/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_11/width_131k/average_l0_27/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_11/width_131k/average_l0_49/hparams.json b/layer_11/width_131k/average_l0_49/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_11/width_131k/average_l0_49/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_11/width_131k/average_l0_88/hparams.json b/layer_11/width_131k/average_l0_88/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_11/width_131k/average_l0_88/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_11/width_131k/average_l0_9/hparams.json b/layer_11/width_131k/average_l0_9/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_11/width_131k/average_l0_9/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_11/width_16k/average_l0_10/hparams.json b/layer_11/width_16k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_11/width_16k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_11/width_16k/average_l0_118/hparams.json b/layer_11/width_16k/average_l0_118/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_11/width_16k/average_l0_118/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_11/width_16k/average_l0_18/hparams.json b/layer_11/width_16k/average_l0_18/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_11/width_16k/average_l0_18/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_11/width_16k/average_l0_255/hparams.json b/layer_11/width_16k/average_l0_255/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_11/width_16k/average_l0_255/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_11/width_16k/average_l0_32/hparams.json b/layer_11/width_16k/average_l0_32/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_11/width_16k/average_l0_32/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_11/width_16k/average_l0_60/hparams.json b/layer_11/width_16k/average_l0_60/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_11/width_16k/average_l0_60/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_12/width_131k/average_l0_10/hparams.json b/layer_12/width_131k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_12/width_131k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_12/width_131k/average_l0_17/hparams.json b/layer_12/width_131k/average_l0_17/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_12/width_131k/average_l0_17/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_12/width_131k/average_l0_183/hparams.json b/layer_12/width_131k/average_l0_183/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_12/width_131k/average_l0_183/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_12/width_131k/average_l0_29/hparams.json b/layer_12/width_131k/average_l0_29/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_12/width_131k/average_l0_29/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_12/width_131k/average_l0_52/hparams.json b/layer_12/width_131k/average_l0_52/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_12/width_131k/average_l0_52/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_12/width_131k/average_l0_96/hparams.json b/layer_12/width_131k/average_l0_96/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_12/width_131k/average_l0_96/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_12/width_16k/average_l0_10/hparams.json b/layer_12/width_16k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_12/width_16k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_12/width_16k/average_l0_130/hparams.json b/layer_12/width_16k/average_l0_130/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_12/width_16k/average_l0_130/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_12/width_16k/average_l0_19/hparams.json b/layer_12/width_16k/average_l0_19/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_12/width_16k/average_l0_19/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_12/width_16k/average_l0_287/hparams.json b/layer_12/width_16k/average_l0_287/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_12/width_16k/average_l0_287/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_12/width_16k/average_l0_33/hparams.json b/layer_12/width_16k/average_l0_33/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_12/width_16k/average_l0_33/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_12/width_16k/average_l0_64/hparams.json b/layer_12/width_16k/average_l0_64/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_12/width_16k/average_l0_64/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_13/width_131k/average_l0_10/hparams.json b/layer_13/width_131k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_13/width_131k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_13/width_131k/average_l0_17/hparams.json b/layer_13/width_131k/average_l0_17/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_13/width_131k/average_l0_17/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_13/width_131k/average_l0_189/hparams.json b/layer_13/width_131k/average_l0_189/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_13/width_131k/average_l0_189/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_13/width_131k/average_l0_30/hparams.json b/layer_13/width_131k/average_l0_30/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_13/width_131k/average_l0_30/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_13/width_131k/average_l0_54/hparams.json b/layer_13/width_131k/average_l0_54/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_13/width_131k/average_l0_54/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_13/width_131k/average_l0_99/hparams.json b/layer_13/width_131k/average_l0_99/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_13/width_131k/average_l0_99/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_13/width_16k/average_l0_11/hparams.json b/layer_13/width_16k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_13/width_16k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_13/width_16k/average_l0_132/hparams.json b/layer_13/width_16k/average_l0_132/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_13/width_16k/average_l0_132/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_13/width_16k/average_l0_19/hparams.json b/layer_13/width_16k/average_l0_19/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_13/width_16k/average_l0_19/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_13/width_16k/average_l0_285/hparams.json b/layer_13/width_16k/average_l0_285/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_13/width_16k/average_l0_285/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_13/width_16k/average_l0_34/hparams.json b/layer_13/width_16k/average_l0_34/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_13/width_16k/average_l0_34/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_13/width_16k/average_l0_65/hparams.json b/layer_13/width_16k/average_l0_65/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_13/width_16k/average_l0_65/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_14/width_131k/average_l0_10/hparams.json b/layer_14/width_131k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_14/width_131k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_14/width_131k/average_l0_105/hparams.json b/layer_14/width_131k/average_l0_105/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_14/width_131k/average_l0_105/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_14/width_131k/average_l0_18/hparams.json b/layer_14/width_131k/average_l0_18/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_14/width_131k/average_l0_18/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_14/width_131k/average_l0_197/hparams.json b/layer_14/width_131k/average_l0_197/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_14/width_131k/average_l0_197/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_14/width_131k/average_l0_31/hparams.json b/layer_14/width_131k/average_l0_31/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_14/width_131k/average_l0_31/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_14/width_131k/average_l0_56/hparams.json b/layer_14/width_131k/average_l0_56/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_14/width_131k/average_l0_56/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_14/width_16k/average_l0_11/hparams.json b/layer_14/width_16k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_14/width_16k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_14/width_16k/average_l0_137/hparams.json b/layer_14/width_16k/average_l0_137/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_14/width_16k/average_l0_137/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_14/width_16k/average_l0_19/hparams.json b/layer_14/width_16k/average_l0_19/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_14/width_16k/average_l0_19/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_14/width_16k/average_l0_294/hparams.json b/layer_14/width_16k/average_l0_294/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_14/width_16k/average_l0_294/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_14/width_16k/average_l0_35/hparams.json b/layer_14/width_16k/average_l0_35/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_14/width_16k/average_l0_35/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_14/width_16k/average_l0_67/hparams.json b/layer_14/width_16k/average_l0_67/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_14/width_16k/average_l0_67/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_15/width_131k/average_l0_10/hparams.json b/layer_15/width_131k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_15/width_131k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_15/width_131k/average_l0_103/hparams.json b/layer_15/width_131k/average_l0_103/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_15/width_131k/average_l0_103/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_15/width_131k/average_l0_17/hparams.json b/layer_15/width_131k/average_l0_17/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_15/width_131k/average_l0_17/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_15/width_131k/average_l0_198/hparams.json b/layer_15/width_131k/average_l0_198/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_15/width_131k/average_l0_198/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_15/width_131k/average_l0_30/hparams.json b/layer_15/width_131k/average_l0_30/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_15/width_131k/average_l0_30/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_15/width_131k/average_l0_55/hparams.json b/layer_15/width_131k/average_l0_55/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_15/width_131k/average_l0_55/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_15/width_16k/average_l0_10/hparams.json b/layer_15/width_16k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_15/width_16k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_15/width_16k/average_l0_131/hparams.json b/layer_15/width_16k/average_l0_131/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_15/width_16k/average_l0_131/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_15/width_16k/average_l0_18/hparams.json b/layer_15/width_16k/average_l0_18/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_15/width_16k/average_l0_18/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_15/width_16k/average_l0_290/hparams.json b/layer_15/width_16k/average_l0_290/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_15/width_16k/average_l0_290/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_15/width_16k/average_l0_34/hparams.json b/layer_15/width_16k/average_l0_34/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_15/width_16k/average_l0_34/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_15/width_16k/average_l0_65/hparams.json b/layer_15/width_16k/average_l0_65/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_15/width_16k/average_l0_65/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_16/width_131k/average_l0_11/hparams.json b/layer_16/width_131k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_16/width_131k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_16/width_131k/average_l0_121/hparams.json b/layer_16/width_131k/average_l0_121/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_16/width_131k/average_l0_121/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_16/width_131k/average_l0_20/hparams.json b/layer_16/width_131k/average_l0_20/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_16/width_131k/average_l0_20/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_16/width_131k/average_l0_232/hparams.json b/layer_16/width_131k/average_l0_232/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_16/width_131k/average_l0_232/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_16/width_131k/average_l0_35/hparams.json b/layer_16/width_131k/average_l0_35/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_16/width_131k/average_l0_35/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_16/width_131k/average_l0_65/hparams.json b/layer_16/width_131k/average_l0_65/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_16/width_131k/average_l0_65/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_16/width_16k/average_l0_11/hparams.json b/layer_16/width_16k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_16/width_16k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_16/width_16k/average_l0_152/hparams.json b/layer_16/width_16k/average_l0_152/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_16/width_16k/average_l0_152/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_16/width_16k/average_l0_21/hparams.json b/layer_16/width_16k/average_l0_21/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_16/width_16k/average_l0_21/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_16/width_16k/average_l0_346/hparams.json b/layer_16/width_16k/average_l0_346/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_16/width_16k/average_l0_346/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_16/width_16k/average_l0_39/hparams.json b/layer_16/width_16k/average_l0_39/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_16/width_16k/average_l0_39/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_16/width_16k/average_l0_75/hparams.json b/layer_16/width_16k/average_l0_75/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_16/width_16k/average_l0_75/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_17/width_131k/average_l0_11/hparams.json b/layer_17/width_131k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_17/width_131k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_17/width_131k/average_l0_117/hparams.json b/layer_17/width_131k/average_l0_117/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_17/width_131k/average_l0_117/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_17/width_131k/average_l0_19/hparams.json b/layer_17/width_131k/average_l0_19/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_17/width_131k/average_l0_19/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_17/width_131k/average_l0_221/hparams.json b/layer_17/width_131k/average_l0_221/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_17/width_131k/average_l0_221/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_17/width_131k/average_l0_35/hparams.json b/layer_17/width_131k/average_l0_35/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_17/width_131k/average_l0_35/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_17/width_131k/average_l0_64/hparams.json b/layer_17/width_131k/average_l0_64/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_17/width_131k/average_l0_64/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_17/width_16k/average_l0_12/hparams.json b/layer_17/width_16k/average_l0_12/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_17/width_16k/average_l0_12/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_17/width_16k/average_l0_144/hparams.json b/layer_17/width_16k/average_l0_144/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_17/width_16k/average_l0_144/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_17/width_16k/average_l0_21/hparams.json b/layer_17/width_16k/average_l0_21/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_17/width_16k/average_l0_21/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_17/width_16k/average_l0_317/hparams.json b/layer_17/width_16k/average_l0_317/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_17/width_16k/average_l0_317/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_17/width_16k/average_l0_38/hparams.json b/layer_17/width_16k/average_l0_38/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_17/width_16k/average_l0_38/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_17/width_16k/average_l0_73/hparams.json b/layer_17/width_16k/average_l0_73/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_17/width_16k/average_l0_73/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_18/width_131k/average_l0_11/hparams.json b/layer_18/width_131k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_18/width_131k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_18/width_131k/average_l0_113/hparams.json b/layer_18/width_131k/average_l0_113/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_18/width_131k/average_l0_113/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_18/width_131k/average_l0_19/hparams.json b/layer_18/width_131k/average_l0_19/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_18/width_131k/average_l0_19/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_18/width_131k/average_l0_221/hparams.json b/layer_18/width_131k/average_l0_221/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_18/width_131k/average_l0_221/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_18/width_131k/average_l0_34/hparams.json b/layer_18/width_131k/average_l0_34/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_18/width_131k/average_l0_34/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_18/width_131k/average_l0_62/hparams.json b/layer_18/width_131k/average_l0_62/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_18/width_131k/average_l0_62/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_18/width_16k/average_l0_11/hparams.json b/layer_18/width_16k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_18/width_16k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_18/width_16k/average_l0_140/hparams.json b/layer_18/width_16k/average_l0_140/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_18/width_16k/average_l0_140/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_18/width_16k/average_l0_20/hparams.json b/layer_18/width_16k/average_l0_20/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_18/width_16k/average_l0_20/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_18/width_16k/average_l0_309/hparams.json b/layer_18/width_16k/average_l0_309/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_18/width_16k/average_l0_309/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_18/width_16k/average_l0_37/hparams.json b/layer_18/width_16k/average_l0_37/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_18/width_16k/average_l0_37/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_18/width_16k/average_l0_71/hparams.json b/layer_18/width_16k/average_l0_71/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_18/width_16k/average_l0_71/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_19/width_131k/average_l0_10/hparams.json b/layer_19/width_131k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_19/width_131k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_19/width_131k/average_l0_110/hparams.json b/layer_19/width_131k/average_l0_110/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_19/width_131k/average_l0_110/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_19/width_131k/average_l0_18/hparams.json b/layer_19/width_131k/average_l0_18/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_19/width_131k/average_l0_18/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_19/width_131k/average_l0_210/hparams.json b/layer_19/width_131k/average_l0_210/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_19/width_131k/average_l0_210/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_19/width_131k/average_l0_32/hparams.json b/layer_19/width_131k/average_l0_32/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_19/width_131k/average_l0_32/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_19/width_131k/average_l0_60/hparams.json b/layer_19/width_131k/average_l0_60/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_19/width_131k/average_l0_60/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_19/width_16k/average_l0_11/hparams.json b/layer_19/width_16k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_19/width_16k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_19/width_16k/average_l0_132/hparams.json b/layer_19/width_16k/average_l0_132/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_19/width_16k/average_l0_132/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_19/width_16k/average_l0_19/hparams.json b/layer_19/width_16k/average_l0_19/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_19/width_16k/average_l0_19/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_19/width_16k/average_l0_293/hparams.json b/layer_19/width_16k/average_l0_293/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_19/width_16k/average_l0_293/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_19/width_16k/average_l0_35/hparams.json b/layer_19/width_16k/average_l0_35/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_19/width_16k/average_l0_35/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_19/width_16k/average_l0_67/hparams.json b/layer_19/width_16k/average_l0_67/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_19/width_16k/average_l0_67/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_2/width_131k/average_l0_12/hparams.json b/layer_2/width_131k/average_l0_12/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_2/width_131k/average_l0_12/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_2/width_131k/average_l0_19/hparams.json b/layer_2/width_131k/average_l0_19/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_2/width_131k/average_l0_19/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_2/width_131k/average_l0_36/hparams.json b/layer_2/width_131k/average_l0_36/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_2/width_131k/average_l0_36/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_2/width_131k/average_l0_5/hparams.json b/layer_2/width_131k/average_l0_5/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_2/width_131k/average_l0_5/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_2/width_131k/average_l0_70/hparams.json b/layer_2/width_131k/average_l0_70/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_2/width_131k/average_l0_70/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_2/width_131k/average_l0_8/hparams.json b/layer_2/width_131k/average_l0_8/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_2/width_131k/average_l0_8/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_2/width_16k/average_l0_14/hparams.json b/layer_2/width_16k/average_l0_14/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_2/width_16k/average_l0_14/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_2/width_16k/average_l0_189/hparams.json b/layer_2/width_16k/average_l0_189/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_2/width_16k/average_l0_189/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_2/width_16k/average_l0_29/hparams.json b/layer_2/width_16k/average_l0_29/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_2/width_16k/average_l0_29/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_2/width_16k/average_l0_67/hparams.json b/layer_2/width_16k/average_l0_67/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_2/width_16k/average_l0_67/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_2/width_16k/average_l0_8/hparams.json b/layer_2/width_16k/average_l0_8/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_2/width_16k/average_l0_8/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_20/width_131k/average_l0_10/params.npz b/layer_20/width_131k/average_l0_10/params.npz
deleted file mode 100644
index fa03974850948a7752b710be4d72c7d54c00cd52..0000000000000000000000000000000000000000
--- a/layer_20/width_131k/average_l0_10/params.npz
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:1f4adda33cc9c1993a448424463e2af08cf3abe4fb802b0665c1e42894b3ff06
-size 3759160648
diff --git a/layer_20/width_131k/average_l0_11/hparams.json b/layer_20/width_131k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_20/width_131k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_20/width_131k/average_l0_114/hparams.json b/layer_20/width_131k/average_l0_114/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_20/width_131k/average_l0_114/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_20/width_131k/average_l0_12/params.npz b/layer_20/width_131k/average_l0_12/params.npz
deleted file mode 100644
index 7712816569532e2eee696c06dc45e3dd22c93fef..0000000000000000000000000000000000000000
--- a/layer_20/width_131k/average_l0_12/params.npz
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:5e2ef96b99851159510af95883afbadc70d69b2d076249163d97b5ce7e4d1fc9
-size 3759160648
diff --git a/layer_20/width_131k/average_l0_19/hparams.json b/layer_20/width_131k/average_l0_19/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_20/width_131k/average_l0_19/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_20/width_131k/average_l0_221/hparams.json b/layer_20/width_131k/average_l0_221/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_20/width_131k/average_l0_221/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_20/width_131k/average_l0_269/params.npz b/layer_20/width_131k/average_l0_269/params.npz
deleted file mode 100644
index 016789a6ce746c0640ddae6bd0064cccd839baa6..0000000000000000000000000000000000000000
--- a/layer_20/width_131k/average_l0_269/params.npz
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:7f686e081c14800489b25b4a1e0347ebf72ac043030cf67b094c0ece2f12085a
-size 3759160648
diff --git a/layer_20/width_131k/average_l0_276/hparams.json b/layer_20/width_131k/average_l0_276/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..1f57879b375201b7f3021c4c1b1b2535d51a7a9d
--- /dev/null
+++ b/layer_20/width_131k/average_l0_276/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0001}
\ No newline at end of file
diff --git a/layer_20/width_131k/average_l0_288/params.npz b/layer_20/width_131k/average_l0_288/params.npz
deleted file mode 100644
index 5cdf5438848be75752f2d560a9bd6fc9fe5d3eeb..0000000000000000000000000000000000000000
--- a/layer_20/width_131k/average_l0_288/params.npz
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:5a213724874d37b68f8738ec83a22f84222bb7b6aaf8e43bd5346dcd0f640d97
-size 3759160648
diff --git a/layer_20/width_131k/average_l0_34/hparams.json b/layer_20/width_131k/average_l0_34/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_20/width_131k/average_l0_34/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_20/width_131k/average_l0_51/params.npz b/layer_20/width_131k/average_l0_51/params.npz
deleted file mode 100644
index 6348555b044194e30468cfe626ec6a3d4ba18fac..0000000000000000000000000000000000000000
--- a/layer_20/width_131k/average_l0_51/params.npz
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:598d6e8fc0f9a691562a0878f11f5f720eda666dbf5244203066c283a303aac2
-size 3759160648
diff --git a/layer_20/width_131k/average_l0_53/hparams.json b/layer_20/width_131k/average_l0_53/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..abd6bc114e06a1e5032eb86de258ffa39487010f
--- /dev/null
+++ b/layer_20/width_131k/average_l0_53/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0006}
\ No newline at end of file
diff --git a/layer_20/width_131k/average_l0_54/params.npz b/layer_20/width_131k/average_l0_54/params.npz
deleted file mode 100644
index 13d9fa3feb993d4c5603d835cdaefdbc1bbcd584..0000000000000000000000000000000000000000
--- a/layer_20/width_131k/average_l0_54/params.npz
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:4906d92deaedc6b46f700bccf326022cf182b199713e809e27689000ae16d69d
-size 3759160648
diff --git a/layer_20/width_131k/average_l0_62/hparams.json b/layer_20/width_131k/average_l0_62/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_20/width_131k/average_l0_62/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_20/width_16k/average_l0_11/hparams.json b/layer_20/width_16k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_20/width_16k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_20/width_16k/average_l0_138/hparams.json b/layer_20/width_16k/average_l0_138/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_20/width_16k/average_l0_138/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_20/width_16k/average_l0_20/hparams.json b/layer_20/width_16k/average_l0_20/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_20/width_16k/average_l0_20/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_20/width_16k/average_l0_310/hparams.json b/layer_20/width_16k/average_l0_310/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_20/width_16k/average_l0_310/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_20/width_16k/average_l0_36/hparams.json b/layer_20/width_16k/average_l0_36/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_20/width_16k/average_l0_36/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_20/width_16k/average_l0_393/params.npz b/layer_20/width_16k/average_l0_393/params.npz
deleted file mode 100644
index 9083f38b7bd374b34e334f860051f34e50d6e5df..0000000000000000000000000000000000000000
--- a/layer_20/width_16k/average_l0_393/params.npz
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:41043883864ed119b8a88a99bf306600fcecb326d7698fcd77a7c61cf22cd764
-size 469908696
diff --git a/layer_20/width_16k/average_l0_408/hparams.json b/layer_20/width_16k/average_l0_408/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..1f57879b375201b7f3021c4c1b1b2535d51a7a9d
--- /dev/null
+++ b/layer_20/width_16k/average_l0_408/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0001}
\ No newline at end of file
diff --git a/layer_20/width_16k/average_l0_427/params.npz b/layer_20/width_16k/average_l0_427/params.npz
deleted file mode 100644
index d562e099df0a65866ce0a81d0c1e52633f880bf5..0000000000000000000000000000000000000000
--- a/layer_20/width_16k/average_l0_427/params.npz
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:148ac602d253b9c53f00fdba80e69ea84a77b3f43756b09fcb3b758359a41720
-size 469908696
diff --git a/layer_20/width_16k/average_l0_57/params.npz b/layer_20/width_16k/average_l0_57/params.npz
deleted file mode 100644
index e9eddc1aed366aa530387790c8961fd540d7e02f..0000000000000000000000000000000000000000
--- a/layer_20/width_16k/average_l0_57/params.npz
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:03c853a20e80fb15f6a7896fdb019e5f4e30935df714855a8d7edccc2693962f
-size 469908696
diff --git a/layer_20/width_16k/average_l0_58/hparams.json b/layer_20/width_16k/average_l0_58/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..abd6bc114e06a1e5032eb86de258ffa39487010f
--- /dev/null
+++ b/layer_20/width_16k/average_l0_58/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0006}
\ No newline at end of file
diff --git a/layer_20/width_16k/average_l0_68/hparams.json b/layer_20/width_16k/average_l0_68/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_20/width_16k/average_l0_68/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_20/width_262k/average_l0_10/params.npz b/layer_20/width_262k/average_l0_10/params.npz
deleted file mode 100644
index fbb99fd40285f525c6924819b766b0ec8cfa1991..0000000000000000000000000000000000000000
--- a/layer_20/width_262k/average_l0_10/params.npz
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:f408a471f467f648c125df1404270f633b92fc2c089ded933f7d43bfd8fc3d2c
-size 7518305656
diff --git a/layer_20/width_262k/average_l0_11/hparams.json b/layer_20/width_262k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_20/width_262k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_20/width_262k/average_l0_13/params.npz b/layer_20/width_262k/average_l0_13/params.npz
deleted file mode 100644
index 9a2342cdf56dd7f416f980354f890d67a8fa2e7e..0000000000000000000000000000000000000000
--- a/layer_20/width_262k/average_l0_13/params.npz
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:bdc08bc7e83f4dc4687c518054032d718fecbc8ab9a4a6174dfe69b4c3c1b906
-size 7518305656
diff --git a/layer_20/width_262k/average_l0_249/params.npz b/layer_20/width_262k/average_l0_249/params.npz
deleted file mode 100644
index 1e4e494c67370382a75b5d7b0fb7884d422a3b16..0000000000000000000000000000000000000000
--- a/layer_20/width_262k/average_l0_249/params.npz
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:98717d82ec8c14aec85c35e4d66f0533aa0f14541ee1dda724df2ea2afc095da
-size 7518305656
diff --git a/layer_20/width_262k/average_l0_259/hparams.json b/layer_20/width_262k/average_l0_259/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..1f57879b375201b7f3021c4c1b1b2535d51a7a9d
--- /dev/null
+++ b/layer_20/width_262k/average_l0_259/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0001}
\ No newline at end of file
diff --git a/layer_20/width_262k/average_l0_276/params.npz b/layer_20/width_262k/average_l0_276/params.npz
deleted file mode 100644
index 51455980cf37c658b34e0e049ba8989d211aed2b..0000000000000000000000000000000000000000
--- a/layer_20/width_262k/average_l0_276/params.npz
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:dddc2e2aeac811737cbcf389b1124270e7d2bac08346845aadfa25593b36d47f
-size 7518305656
diff --git a/layer_20/width_262k/average_l0_49/params.npz b/layer_20/width_262k/average_l0_49/params.npz
deleted file mode 100644
index 272e458826513756978d3a677679c44d481ba334..0000000000000000000000000000000000000000
--- a/layer_20/width_262k/average_l0_49/params.npz
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:f1914a981b43185b87a84f7badcbc1bfb7ca24c4a99eb4e30ffe63cb10f0afbf
-size 7518305656
diff --git a/layer_20/width_262k/average_l0_50/hparams.json b/layer_20/width_262k/average_l0_50/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..abd6bc114e06a1e5032eb86de258ffa39487010f
--- /dev/null
+++ b/layer_20/width_262k/average_l0_50/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0006}
\ No newline at end of file
diff --git a/layer_20/width_262k/average_l0_64/params.npz b/layer_20/width_262k/average_l0_64/params.npz
deleted file mode 100644
index 95b86dc38628e73f6e0aabdfbc8eb44de95ecaec..0000000000000000000000000000000000000000
--- a/layer_20/width_262k/average_l0_64/params.npz
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:ec7799e0965e5d3ebc24842f3f22bb9f8b0ecd5827ef0ab32607ebbc5e183e39
-size 7518305656
diff --git a/layer_20/width_32k/average_l0_11/hparams.json b/layer_20/width_32k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_20/width_32k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_20/width_32k/average_l0_334/params.npz b/layer_20/width_32k/average_l0_334/params.npz
deleted file mode 100644
index 972b66d3137e274fc031f60cf3a7488f4295562d..0000000000000000000000000000000000000000
--- a/layer_20/width_32k/average_l0_334/params.npz
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:a6aa3811d110a818b961164634b3f587486320fd2a29697ae5a73a83a9b33c24
-size 939801816
diff --git a/layer_20/width_32k/average_l0_344/hparams.json b/layer_20/width_32k/average_l0_344/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..1f57879b375201b7f3021c4c1b1b2535d51a7a9d
--- /dev/null
+++ b/layer_20/width_32k/average_l0_344/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0001}
\ No newline at end of file
diff --git a/layer_20/width_32k/average_l0_357/params.npz b/layer_20/width_32k/average_l0_357/params.npz
deleted file mode 100644
index 4f5f049162bf269e29e5b77e3bed92c8993feff4..0000000000000000000000000000000000000000
--- a/layer_20/width_32k/average_l0_357/params.npz
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:21af42d61568ee6327c40c9e991e2ecce73161e585c1fc296ee276e136f65d88
-size 939801816
diff --git a/layer_20/width_32k/average_l0_55/params.npz b/layer_20/width_32k/average_l0_55/params.npz
deleted file mode 100644
index 99508df050cd5c72c897d0708d73429c8f73ea28..0000000000000000000000000000000000000000
--- a/layer_20/width_32k/average_l0_55/params.npz
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:d6410ee73404ab78072a88c0c6dacf071f87e09be771ee2d122a91ba4c3e1fbd
-size 939801816
diff --git a/layer_20/width_32k/average_l0_57/hparams.json b/layer_20/width_32k/average_l0_57/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..abd6bc114e06a1e5032eb86de258ffa39487010f
--- /dev/null
+++ b/layer_20/width_32k/average_l0_57/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0006}
\ No newline at end of file
diff --git a/layer_20/width_524k/average_l0_10/hparams.json b/layer_20/width_524k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_20/width_524k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_20/width_524k/average_l0_229/params.npz b/layer_20/width_524k/average_l0_229/params.npz
deleted file mode 100644
index 5b3abe6cdbb54a4ba9ccf3c5236870c5a0233a01..0000000000000000000000000000000000000000
--- a/layer_20/width_524k/average_l0_229/params.npz
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:2b05e58e30fa9461250bcb3b47e6fbbb713c30aef5d5a223374896c7d069cca1
-size 15036595576
diff --git a/layer_20/width_524k/average_l0_24/params.npz b/layer_20/width_524k/average_l0_24/params.npz
deleted file mode 100644
index 9468e1be8c97aca72d530dc1ae7724e29ef4d035..0000000000000000000000000000000000000000
--- a/layer_20/width_524k/average_l0_24/params.npz
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:642051803e99230263bde5794d66ac7a5d978275cf08fd5f3ad488d0375835b4
-size 15036595576
diff --git a/layer_20/width_524k/average_l0_241/hparams.json b/layer_20/width_524k/average_l0_241/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..1f57879b375201b7f3021c4c1b1b2535d51a7a9d
--- /dev/null
+++ b/layer_20/width_524k/average_l0_241/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0001}
\ No newline at end of file
diff --git a/layer_20/width_524k/average_l0_298/params.npz b/layer_20/width_524k/average_l0_298/params.npz
deleted file mode 100644
index 7309c87881e5aaccb33eba193df2ed5388084714..0000000000000000000000000000000000000000
--- a/layer_20/width_524k/average_l0_298/params.npz
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:ac2ce6d73b136664e9440935db069e493ee27a772e8a1084adc227f6ea6bfd53
-size 15036595576
diff --git a/layer_20/width_524k/average_l0_46/params.npz b/layer_20/width_524k/average_l0_46/params.npz
deleted file mode 100644
index 8f95385605ce493d2a4716c2f19ae36cd4f76fe3..0000000000000000000000000000000000000000
--- a/layer_20/width_524k/average_l0_46/params.npz
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:f998a6445ccc2d0d6a0d1f609d8bcd877c545f934c50a9123be3b9d1846674a9
-size 15036595576
diff --git a/layer_20/width_524k/average_l0_48/hparams.json b/layer_20/width_524k/average_l0_48/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..abd6bc114e06a1e5032eb86de258ffa39487010f
--- /dev/null
+++ b/layer_20/width_524k/average_l0_48/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0006}
\ No newline at end of file
diff --git a/layer_20/width_524k/average_l0_78/params.npz b/layer_20/width_524k/average_l0_78/params.npz
deleted file mode 100644
index b67416029c7528f94e0b34654f4547ca4cd48951..0000000000000000000000000000000000000000
--- a/layer_20/width_524k/average_l0_78/params.npz
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:7a3a1f5d5f294c13df36bc742c4b4fa04eedc0a026a575b39d51c6ba58a17431
-size 15036595576
diff --git a/layer_20/width_65k/average_l0_11/hparams.json b/layer_20/width_65k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_20/width_65k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_20/width_65k/average_l0_292/params.npz b/layer_20/width_65k/average_l0_292/params.npz
deleted file mode 100644
index a1f2fb9034e62eec1fb1976e543f34ce9ff82361..0000000000000000000000000000000000000000
--- a/layer_20/width_65k/average_l0_292/params.npz
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:93ec2dd26cb0a2e0be091956173706a8f3a6a717ffc958785f4db32202f9caab
-size 1879588056
diff --git a/layer_20/width_65k/average_l0_298/hparams.json b/layer_20/width_65k/average_l0_298/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..1f57879b375201b7f3021c4c1b1b2535d51a7a9d
--- /dev/null
+++ b/layer_20/width_65k/average_l0_298/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0001}
\ No newline at end of file
diff --git a/layer_20/width_65k/average_l0_309/params.npz b/layer_20/width_65k/average_l0_309/params.npz
deleted file mode 100644
index bef7e2c4ad2e7de0a96a83ce93750f84a5931b1e..0000000000000000000000000000000000000000
--- a/layer_20/width_65k/average_l0_309/params.npz
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:1f1a54df814defe15b53066e089336eddac93b18531cd1d9cf11420cf7fc17b2
-size 1879588056
diff --git a/layer_20/width_65k/average_l0_54/params.npz b/layer_20/width_65k/average_l0_54/params.npz
deleted file mode 100644
index e30b1c3d00bd6741dcadc74dee4050bc91708f6a..0000000000000000000000000000000000000000
--- a/layer_20/width_65k/average_l0_54/params.npz
+++ /dev/null
@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:3c8461e7da9e256285f7b57bfd4899255e81bf4c2bd20153272cfa3fe971f487
-size 1879588056
diff --git a/layer_20/width_65k/average_l0_55/hparams.json b/layer_20/width_65k/average_l0_55/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..abd6bc114e06a1e5032eb86de258ffa39487010f
--- /dev/null
+++ b/layer_20/width_65k/average_l0_55/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0006}
\ No newline at end of file
diff --git a/layer_21/width_131k/average_l0_109/hparams.json b/layer_21/width_131k/average_l0_109/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_21/width_131k/average_l0_109/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_21/width_131k/average_l0_11/hparams.json b/layer_21/width_131k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_21/width_131k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_21/width_131k/average_l0_19/hparams.json b/layer_21/width_131k/average_l0_19/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_21/width_131k/average_l0_19/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_21/width_131k/average_l0_204/hparams.json b/layer_21/width_131k/average_l0_204/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_21/width_131k/average_l0_204/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_21/width_131k/average_l0_33/hparams.json b/layer_21/width_131k/average_l0_33/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_21/width_131k/average_l0_33/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_21/width_131k/average_l0_58/hparams.json b/layer_21/width_131k/average_l0_58/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_21/width_131k/average_l0_58/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_21/width_16k/average_l0_11/hparams.json b/layer_21/width_16k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_21/width_16k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_21/width_16k/average_l0_129/hparams.json b/layer_21/width_16k/average_l0_129/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_21/width_16k/average_l0_129/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_21/width_16k/average_l0_19/hparams.json b/layer_21/width_16k/average_l0_19/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_21/width_16k/average_l0_19/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_21/width_16k/average_l0_278/hparams.json b/layer_21/width_16k/average_l0_278/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_21/width_16k/average_l0_278/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_21/width_16k/average_l0_36/hparams.json b/layer_21/width_16k/average_l0_36/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_21/width_16k/average_l0_36/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_21/width_16k/average_l0_66/hparams.json b/layer_21/width_16k/average_l0_66/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_21/width_16k/average_l0_66/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_22/width_131k/average_l0_10/hparams.json b/layer_22/width_131k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_22/width_131k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_22/width_131k/average_l0_105/hparams.json b/layer_22/width_131k/average_l0_105/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_22/width_131k/average_l0_105/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_22/width_131k/average_l0_18/hparams.json b/layer_22/width_131k/average_l0_18/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_22/width_131k/average_l0_18/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_22/width_131k/average_l0_197/hparams.json b/layer_22/width_131k/average_l0_197/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_22/width_131k/average_l0_197/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_22/width_131k/average_l0_32/hparams.json b/layer_22/width_131k/average_l0_32/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_22/width_131k/average_l0_32/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_22/width_131k/average_l0_58/hparams.json b/layer_22/width_131k/average_l0_58/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_22/width_131k/average_l0_58/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_22/width_16k/average_l0_11/hparams.json b/layer_22/width_16k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_22/width_16k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_22/width_16k/average_l0_123/hparams.json b/layer_22/width_16k/average_l0_123/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_22/width_16k/average_l0_123/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_22/width_16k/average_l0_19/hparams.json b/layer_22/width_16k/average_l0_19/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_22/width_16k/average_l0_19/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_22/width_16k/average_l0_268/hparams.json b/layer_22/width_16k/average_l0_268/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_22/width_16k/average_l0_268/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_22/width_16k/average_l0_35/hparams.json b/layer_22/width_16k/average_l0_35/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_22/width_16k/average_l0_35/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_22/width_16k/average_l0_65/hparams.json b/layer_22/width_16k/average_l0_65/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_22/width_16k/average_l0_65/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_23/width_131k/average_l0_10/hparams.json b/layer_23/width_131k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_23/width_131k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_23/width_131k/average_l0_101/hparams.json b/layer_23/width_131k/average_l0_101/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_23/width_131k/average_l0_101/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_23/width_131k/average_l0_18/hparams.json b/layer_23/width_131k/average_l0_18/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_23/width_131k/average_l0_18/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_23/width_131k/average_l0_187/hparams.json b/layer_23/width_131k/average_l0_187/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_23/width_131k/average_l0_187/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_23/width_131k/average_l0_32/hparams.json b/layer_23/width_131k/average_l0_32/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_23/width_131k/average_l0_32/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_23/width_131k/average_l0_56/hparams.json b/layer_23/width_131k/average_l0_56/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_23/width_131k/average_l0_56/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_23/width_16k/average_l0_11/hparams.json b/layer_23/width_16k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_23/width_16k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_23/width_16k/average_l0_120/hparams.json b/layer_23/width_16k/average_l0_120/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_23/width_16k/average_l0_120/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_23/width_16k/average_l0_19/hparams.json b/layer_23/width_16k/average_l0_19/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_23/width_16k/average_l0_19/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_23/width_16k/average_l0_248/hparams.json b/layer_23/width_16k/average_l0_248/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_23/width_16k/average_l0_248/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_23/width_16k/average_l0_35/hparams.json b/layer_23/width_16k/average_l0_35/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_23/width_16k/average_l0_35/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_23/width_16k/average_l0_63/hparams.json b/layer_23/width_16k/average_l0_63/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_23/width_16k/average_l0_63/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_24/width_131k/average_l0_10/hparams.json b/layer_24/width_131k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_24/width_131k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_24/width_131k/average_l0_17/hparams.json b/layer_24/width_131k/average_l0_17/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_24/width_131k/average_l0_17/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_24/width_131k/average_l0_180/hparams.json b/layer_24/width_131k/average_l0_180/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_24/width_131k/average_l0_180/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_24/width_131k/average_l0_30/hparams.json b/layer_24/width_131k/average_l0_30/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_24/width_131k/average_l0_30/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_24/width_131k/average_l0_55/hparams.json b/layer_24/width_131k/average_l0_55/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_24/width_131k/average_l0_55/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_24/width_131k/average_l0_97/hparams.json b/layer_24/width_131k/average_l0_97/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_24/width_131k/average_l0_97/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_24/width_16k/average_l0_10/hparams.json b/layer_24/width_16k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_24/width_16k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_24/width_16k/average_l0_114/hparams.json b/layer_24/width_16k/average_l0_114/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_24/width_16k/average_l0_114/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_24/width_16k/average_l0_19/hparams.json b/layer_24/width_16k/average_l0_19/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_24/width_16k/average_l0_19/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_24/width_16k/average_l0_234/hparams.json b/layer_24/width_16k/average_l0_234/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_24/width_16k/average_l0_234/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_24/width_16k/average_l0_34/hparams.json b/layer_24/width_16k/average_l0_34/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_24/width_16k/average_l0_34/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_24/width_16k/average_l0_61/hparams.json b/layer_24/width_16k/average_l0_61/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_24/width_16k/average_l0_61/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_25/width_131k/average_l0_10/hparams.json b/layer_25/width_131k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_25/width_131k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_25/width_131k/average_l0_177/hparams.json b/layer_25/width_131k/average_l0_177/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_25/width_131k/average_l0_177/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_25/width_131k/average_l0_18/hparams.json b/layer_25/width_131k/average_l0_18/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_25/width_131k/average_l0_18/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_25/width_131k/average_l0_31/hparams.json b/layer_25/width_131k/average_l0_31/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_25/width_131k/average_l0_31/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_25/width_131k/average_l0_54/hparams.json b/layer_25/width_131k/average_l0_54/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_25/width_131k/average_l0_54/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_25/width_131k/average_l0_96/hparams.json b/layer_25/width_131k/average_l0_96/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_25/width_131k/average_l0_96/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_25/width_16k/average_l0_11/hparams.json b/layer_25/width_16k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_25/width_16k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_25/width_16k/average_l0_114/hparams.json b/layer_25/width_16k/average_l0_114/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_25/width_16k/average_l0_114/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_25/width_16k/average_l0_19/hparams.json b/layer_25/width_16k/average_l0_19/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_25/width_16k/average_l0_19/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_25/width_16k/average_l0_231/hparams.json b/layer_25/width_16k/average_l0_231/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_25/width_16k/average_l0_231/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_25/width_16k/average_l0_34/hparams.json b/layer_25/width_16k/average_l0_34/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_25/width_16k/average_l0_34/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_25/width_16k/average_l0_61/hparams.json b/layer_25/width_16k/average_l0_61/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_25/width_16k/average_l0_61/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_26/width_131k/average_l0_10/hparams.json b/layer_26/width_131k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_26/width_131k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_26/width_131k/average_l0_176/hparams.json b/layer_26/width_131k/average_l0_176/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_26/width_131k/average_l0_176/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_26/width_131k/average_l0_18/hparams.json b/layer_26/width_131k/average_l0_18/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_26/width_131k/average_l0_18/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_26/width_131k/average_l0_32/hparams.json b/layer_26/width_131k/average_l0_32/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_26/width_131k/average_l0_32/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_26/width_131k/average_l0_55/hparams.json b/layer_26/width_131k/average_l0_55/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_26/width_131k/average_l0_55/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_26/width_131k/average_l0_97/hparams.json b/layer_26/width_131k/average_l0_97/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_26/width_131k/average_l0_97/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_26/width_16k/average_l0_11/hparams.json b/layer_26/width_16k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_26/width_16k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_26/width_16k/average_l0_116/hparams.json b/layer_26/width_16k/average_l0_116/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_26/width_16k/average_l0_116/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_26/width_16k/average_l0_20/hparams.json b/layer_26/width_16k/average_l0_20/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_26/width_16k/average_l0_20/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_26/width_16k/average_l0_233/hparams.json b/layer_26/width_16k/average_l0_233/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_26/width_16k/average_l0_233/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_26/width_16k/average_l0_35/hparams.json b/layer_26/width_16k/average_l0_35/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_26/width_16k/average_l0_35/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_26/width_16k/average_l0_63/hparams.json b/layer_26/width_16k/average_l0_63/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_26/width_16k/average_l0_63/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_27/width_131k/average_l0_11/hparams.json b/layer_27/width_131k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_27/width_131k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_27/width_131k/average_l0_171/hparams.json b/layer_27/width_131k/average_l0_171/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_27/width_131k/average_l0_171/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_27/width_131k/average_l0_19/hparams.json b/layer_27/width_131k/average_l0_19/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_27/width_131k/average_l0_19/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_27/width_131k/average_l0_33/hparams.json b/layer_27/width_131k/average_l0_33/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_27/width_131k/average_l0_33/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_27/width_131k/average_l0_56/hparams.json b/layer_27/width_131k/average_l0_56/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_27/width_131k/average_l0_56/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_27/width_131k/average_l0_96/hparams.json b/layer_27/width_131k/average_l0_96/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_27/width_131k/average_l0_96/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_27/width_16k/average_l0_118/hparams.json b/layer_27/width_16k/average_l0_118/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_27/width_16k/average_l0_118/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_27/width_16k/average_l0_12/hparams.json b/layer_27/width_16k/average_l0_12/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_27/width_16k/average_l0_12/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_27/width_16k/average_l0_21/hparams.json b/layer_27/width_16k/average_l0_21/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_27/width_16k/average_l0_21/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_27/width_16k/average_l0_230/hparams.json b/layer_27/width_16k/average_l0_230/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_27/width_16k/average_l0_230/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_27/width_16k/average_l0_36/hparams.json b/layer_27/width_16k/average_l0_36/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_27/width_16k/average_l0_36/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_27/width_16k/average_l0_65/hparams.json b/layer_27/width_16k/average_l0_65/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_27/width_16k/average_l0_65/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_28/width_131k/average_l0_11/hparams.json b/layer_28/width_131k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_28/width_131k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_28/width_131k/average_l0_171/hparams.json b/layer_28/width_131k/average_l0_171/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_28/width_131k/average_l0_171/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_28/width_131k/average_l0_19/hparams.json b/layer_28/width_131k/average_l0_19/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_28/width_131k/average_l0_19/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_28/width_131k/average_l0_32/hparams.json b/layer_28/width_131k/average_l0_32/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_28/width_131k/average_l0_32/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_28/width_131k/average_l0_57/hparams.json b/layer_28/width_131k/average_l0_57/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_28/width_131k/average_l0_57/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_28/width_131k/average_l0_98/hparams.json b/layer_28/width_131k/average_l0_98/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_28/width_131k/average_l0_98/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_28/width_16k/average_l0_119/hparams.json b/layer_28/width_16k/average_l0_119/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_28/width_16k/average_l0_119/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_28/width_16k/average_l0_12/hparams.json b/layer_28/width_16k/average_l0_12/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_28/width_16k/average_l0_12/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_28/width_16k/average_l0_21/hparams.json b/layer_28/width_16k/average_l0_21/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_28/width_16k/average_l0_21/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_28/width_16k/average_l0_229/hparams.json b/layer_28/width_16k/average_l0_229/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_28/width_16k/average_l0_229/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_28/width_16k/average_l0_37/hparams.json b/layer_28/width_16k/average_l0_37/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_28/width_16k/average_l0_37/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_28/width_16k/average_l0_65/hparams.json b/layer_28/width_16k/average_l0_65/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_28/width_16k/average_l0_65/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_29/width_131k/average_l0_11/hparams.json b/layer_29/width_131k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_29/width_131k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_29/width_131k/average_l0_171/hparams.json b/layer_29/width_131k/average_l0_171/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_29/width_131k/average_l0_171/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_29/width_131k/average_l0_19/hparams.json b/layer_29/width_131k/average_l0_19/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_29/width_131k/average_l0_19/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_29/width_131k/average_l0_33/hparams.json b/layer_29/width_131k/average_l0_33/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_29/width_131k/average_l0_33/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_29/width_131k/average_l0_56/hparams.json b/layer_29/width_131k/average_l0_56/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_29/width_131k/average_l0_56/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_29/width_131k/average_l0_97/hparams.json b/layer_29/width_131k/average_l0_97/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_29/width_131k/average_l0_97/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_29/width_16k/average_l0_119/hparams.json b/layer_29/width_16k/average_l0_119/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_29/width_16k/average_l0_119/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_29/width_16k/average_l0_12/hparams.json b/layer_29/width_16k/average_l0_12/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_29/width_16k/average_l0_12/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_29/width_16k/average_l0_21/hparams.json b/layer_29/width_16k/average_l0_21/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_29/width_16k/average_l0_21/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_29/width_16k/average_l0_224/hparams.json b/layer_29/width_16k/average_l0_224/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_29/width_16k/average_l0_224/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_29/width_16k/average_l0_38/hparams.json b/layer_29/width_16k/average_l0_38/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_29/width_16k/average_l0_38/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_29/width_16k/average_l0_66/hparams.json b/layer_29/width_16k/average_l0_66/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_29/width_16k/average_l0_66/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_3/width_131k/average_l0_103/hparams.json b/layer_3/width_131k/average_l0_103/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_3/width_131k/average_l0_103/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_3/width_131k/average_l0_14/hparams.json b/layer_3/width_131k/average_l0_14/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_3/width_131k/average_l0_14/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_3/width_131k/average_l0_25/hparams.json b/layer_3/width_131k/average_l0_25/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_3/width_131k/average_l0_25/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_3/width_131k/average_l0_46/hparams.json b/layer_3/width_131k/average_l0_46/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_3/width_131k/average_l0_46/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_3/width_131k/average_l0_6/hparams.json b/layer_3/width_131k/average_l0_6/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_3/width_131k/average_l0_6/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_3/width_131k/average_l0_9/hparams.json b/layer_3/width_131k/average_l0_9/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_3/width_131k/average_l0_9/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_3/width_16k/average_l0_10/hparams.json b/layer_3/width_16k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_3/width_16k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_3/width_16k/average_l0_17/hparams.json b/layer_3/width_16k/average_l0_17/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_3/width_16k/average_l0_17/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_3/width_16k/average_l0_229/hparams.json b/layer_3/width_16k/average_l0_229/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_3/width_16k/average_l0_229/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_3/width_16k/average_l0_37/hparams.json b/layer_3/width_16k/average_l0_37/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_3/width_16k/average_l0_37/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_3/width_16k/average_l0_90/hparams.json b/layer_3/width_16k/average_l0_90/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_3/width_16k/average_l0_90/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_30/width_131k/average_l0_11/hparams.json b/layer_30/width_131k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_30/width_131k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_30/width_131k/average_l0_170/hparams.json b/layer_30/width_131k/average_l0_170/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_30/width_131k/average_l0_170/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_30/width_131k/average_l0_18/hparams.json b/layer_30/width_131k/average_l0_18/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_30/width_131k/average_l0_18/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_30/width_131k/average_l0_32/hparams.json b/layer_30/width_131k/average_l0_32/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_30/width_131k/average_l0_32/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_30/width_131k/average_l0_56/hparams.json b/layer_30/width_131k/average_l0_56/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_30/width_131k/average_l0_56/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_30/width_131k/average_l0_95/hparams.json b/layer_30/width_131k/average_l0_95/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_30/width_131k/average_l0_95/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_30/width_16k/average_l0_12/hparams.json b/layer_30/width_16k/average_l0_12/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_30/width_16k/average_l0_12/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_30/width_16k/average_l0_120/hparams.json b/layer_30/width_16k/average_l0_120/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_30/width_16k/average_l0_120/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_30/width_16k/average_l0_21/hparams.json b/layer_30/width_16k/average_l0_21/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_30/width_16k/average_l0_21/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_30/width_16k/average_l0_226/hparams.json b/layer_30/width_16k/average_l0_226/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_30/width_16k/average_l0_226/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_30/width_16k/average_l0_37/hparams.json b/layer_30/width_16k/average_l0_37/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_30/width_16k/average_l0_37/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_30/width_16k/average_l0_66/hparams.json b/layer_30/width_16k/average_l0_66/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_30/width_16k/average_l0_66/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_31/width_131k/average_l0_10/hparams.json b/layer_31/width_131k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_31/width_131k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_31/width_131k/average_l0_160/hparams.json b/layer_31/width_131k/average_l0_160/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_31/width_131k/average_l0_160/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_31/width_131k/average_l0_18/hparams.json b/layer_31/width_131k/average_l0_18/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_31/width_131k/average_l0_18/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_31/width_131k/average_l0_31/hparams.json b/layer_31/width_131k/average_l0_31/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_31/width_131k/average_l0_31/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_31/width_131k/average_l0_52/hparams.json b/layer_31/width_131k/average_l0_52/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_31/width_131k/average_l0_52/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_31/width_131k/average_l0_92/hparams.json b/layer_31/width_131k/average_l0_92/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_31/width_131k/average_l0_92/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_31/width_16k/average_l0_11/hparams.json b/layer_31/width_16k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_31/width_16k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_31/width_16k/average_l0_114/hparams.json b/layer_31/width_16k/average_l0_114/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_31/width_16k/average_l0_114/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_31/width_16k/average_l0_20/hparams.json b/layer_31/width_16k/average_l0_20/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_31/width_16k/average_l0_20/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_31/width_16k/average_l0_218/hparams.json b/layer_31/width_16k/average_l0_218/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_31/width_16k/average_l0_218/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_31/width_16k/average_l0_35/hparams.json b/layer_31/width_16k/average_l0_35/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_31/width_16k/average_l0_35/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_31/width_16k/average_l0_63/hparams.json b/layer_31/width_16k/average_l0_63/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_31/width_16k/average_l0_63/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_32/width_131k/average_l0_10/hparams.json b/layer_32/width_131k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_32/width_131k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_32/width_131k/average_l0_158/hparams.json b/layer_32/width_131k/average_l0_158/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_32/width_131k/average_l0_158/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_32/width_131k/average_l0_18/hparams.json b/layer_32/width_131k/average_l0_18/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_32/width_131k/average_l0_18/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_32/width_131k/average_l0_30/hparams.json b/layer_32/width_131k/average_l0_30/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_32/width_131k/average_l0_30/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_32/width_131k/average_l0_51/hparams.json b/layer_32/width_131k/average_l0_51/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_32/width_131k/average_l0_51/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_32/width_131k/average_l0_88/hparams.json b/layer_32/width_131k/average_l0_88/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_32/width_131k/average_l0_88/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_32/width_16k/average_l0_11/hparams.json b/layer_32/width_16k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_32/width_16k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_32/width_16k/average_l0_111/hparams.json b/layer_32/width_16k/average_l0_111/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_32/width_16k/average_l0_111/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_32/width_16k/average_l0_20/hparams.json b/layer_32/width_16k/average_l0_20/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_32/width_16k/average_l0_20/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_32/width_16k/average_l0_219/hparams.json b/layer_32/width_16k/average_l0_219/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_32/width_16k/average_l0_219/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_32/width_16k/average_l0_34/hparams.json b/layer_32/width_16k/average_l0_34/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_32/width_16k/average_l0_34/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_32/width_16k/average_l0_61/hparams.json b/layer_32/width_16k/average_l0_61/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_32/width_16k/average_l0_61/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_33/width_131k/average_l0_10/hparams.json b/layer_33/width_131k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_33/width_131k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_33/width_131k/average_l0_165/hparams.json b/layer_33/width_131k/average_l0_165/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_33/width_131k/average_l0_165/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_33/width_131k/average_l0_18/hparams.json b/layer_33/width_131k/average_l0_18/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_33/width_131k/average_l0_18/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_33/width_131k/average_l0_30/hparams.json b/layer_33/width_131k/average_l0_30/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_33/width_131k/average_l0_30/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_33/width_131k/average_l0_51/hparams.json b/layer_33/width_131k/average_l0_51/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_33/width_131k/average_l0_51/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_33/width_131k/average_l0_91/hparams.json b/layer_33/width_131k/average_l0_91/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_33/width_131k/average_l0_91/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_33/width_16k/average_l0_11/hparams.json b/layer_33/width_16k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_33/width_16k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_33/width_16k/average_l0_114/hparams.json b/layer_33/width_16k/average_l0_114/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_33/width_16k/average_l0_114/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_33/width_16k/average_l0_20/hparams.json b/layer_33/width_16k/average_l0_20/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_33/width_16k/average_l0_20/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_33/width_16k/average_l0_228/hparams.json b/layer_33/width_16k/average_l0_228/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_33/width_16k/average_l0_228/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_33/width_16k/average_l0_34/hparams.json b/layer_33/width_16k/average_l0_34/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_33/width_16k/average_l0_34/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_33/width_16k/average_l0_63/hparams.json b/layer_33/width_16k/average_l0_63/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_33/width_16k/average_l0_63/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_34/width_131k/average_l0_10/hparams.json b/layer_34/width_131k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_34/width_131k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_34/width_131k/average_l0_163/hparams.json b/layer_34/width_131k/average_l0_163/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_34/width_131k/average_l0_163/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_34/width_131k/average_l0_17/hparams.json b/layer_34/width_131k/average_l0_17/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_34/width_131k/average_l0_17/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_34/width_131k/average_l0_30/hparams.json b/layer_34/width_131k/average_l0_30/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_34/width_131k/average_l0_30/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_34/width_131k/average_l0_51/hparams.json b/layer_34/width_131k/average_l0_51/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_34/width_131k/average_l0_51/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_34/width_131k/average_l0_89/hparams.json b/layer_34/width_131k/average_l0_89/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_34/width_131k/average_l0_89/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_34/width_16k/average_l0_11/hparams.json b/layer_34/width_16k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_34/width_16k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_34/width_16k/average_l0_114/hparams.json b/layer_34/width_16k/average_l0_114/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_34/width_16k/average_l0_114/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_34/width_16k/average_l0_19/hparams.json b/layer_34/width_16k/average_l0_19/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_34/width_16k/average_l0_19/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_34/width_16k/average_l0_229/hparams.json b/layer_34/width_16k/average_l0_229/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_34/width_16k/average_l0_229/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_34/width_16k/average_l0_34/hparams.json b/layer_34/width_16k/average_l0_34/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_34/width_16k/average_l0_34/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_34/width_16k/average_l0_60/hparams.json b/layer_34/width_16k/average_l0_60/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_34/width_16k/average_l0_60/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_35/width_131k/average_l0_10/hparams.json b/layer_35/width_131k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_35/width_131k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_35/width_131k/average_l0_17/hparams.json b/layer_35/width_131k/average_l0_17/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_35/width_131k/average_l0_17/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_35/width_131k/average_l0_171/hparams.json b/layer_35/width_131k/average_l0_171/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_35/width_131k/average_l0_171/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_35/width_131k/average_l0_30/hparams.json b/layer_35/width_131k/average_l0_30/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_35/width_131k/average_l0_30/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_35/width_131k/average_l0_51/hparams.json b/layer_35/width_131k/average_l0_51/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_35/width_131k/average_l0_51/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_35/width_131k/average_l0_94/hparams.json b/layer_35/width_131k/average_l0_94/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_35/width_131k/average_l0_94/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_35/width_16k/average_l0_11/hparams.json b/layer_35/width_16k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_35/width_16k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_35/width_16k/average_l0_120/hparams.json b/layer_35/width_16k/average_l0_120/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_35/width_16k/average_l0_120/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_35/width_16k/average_l0_19/hparams.json b/layer_35/width_16k/average_l0_19/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_35/width_16k/average_l0_19/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_35/width_16k/average_l0_246/hparams.json b/layer_35/width_16k/average_l0_246/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_35/width_16k/average_l0_246/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_35/width_16k/average_l0_34/hparams.json b/layer_35/width_16k/average_l0_34/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_35/width_16k/average_l0_34/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_35/width_16k/average_l0_61/hparams.json b/layer_35/width_16k/average_l0_61/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_35/width_16k/average_l0_61/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_36/width_131k/average_l0_10/hparams.json b/layer_36/width_131k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_36/width_131k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_36/width_131k/average_l0_17/hparams.json b/layer_36/width_131k/average_l0_17/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_36/width_131k/average_l0_17/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_36/width_131k/average_l0_180/hparams.json b/layer_36/width_131k/average_l0_180/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_36/width_131k/average_l0_180/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_36/width_131k/average_l0_30/hparams.json b/layer_36/width_131k/average_l0_30/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_36/width_131k/average_l0_30/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_36/width_131k/average_l0_51/hparams.json b/layer_36/width_131k/average_l0_51/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_36/width_131k/average_l0_51/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_36/width_131k/average_l0_93/hparams.json b/layer_36/width_131k/average_l0_93/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_36/width_131k/average_l0_93/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_36/width_16k/average_l0_11/hparams.json b/layer_36/width_16k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_36/width_16k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_36/width_16k/average_l0_120/hparams.json b/layer_36/width_16k/average_l0_120/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_36/width_16k/average_l0_120/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_36/width_16k/average_l0_19/hparams.json b/layer_36/width_16k/average_l0_19/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_36/width_16k/average_l0_19/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_36/width_16k/average_l0_252/hparams.json b/layer_36/width_16k/average_l0_252/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_36/width_16k/average_l0_252/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_36/width_16k/average_l0_34/hparams.json b/layer_36/width_16k/average_l0_34/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_36/width_16k/average_l0_34/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_36/width_16k/average_l0_61/hparams.json b/layer_36/width_16k/average_l0_61/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_36/width_16k/average_l0_61/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_37/width_131k/average_l0_10/hparams.json b/layer_37/width_131k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_37/width_131k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_37/width_131k/average_l0_18/hparams.json b/layer_37/width_131k/average_l0_18/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_37/width_131k/average_l0_18/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_37/width_131k/average_l0_184/hparams.json b/layer_37/width_131k/average_l0_184/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_37/width_131k/average_l0_184/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_37/width_131k/average_l0_30/hparams.json b/layer_37/width_131k/average_l0_30/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_37/width_131k/average_l0_30/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_37/width_131k/average_l0_53/hparams.json b/layer_37/width_131k/average_l0_53/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_37/width_131k/average_l0_53/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_37/width_131k/average_l0_96/hparams.json b/layer_37/width_131k/average_l0_96/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_37/width_131k/average_l0_96/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_37/width_16k/average_l0_11/hparams.json b/layer_37/width_16k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_37/width_16k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_37/width_16k/average_l0_124/hparams.json b/layer_37/width_16k/average_l0_124/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_37/width_16k/average_l0_124/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_37/width_16k/average_l0_20/hparams.json b/layer_37/width_16k/average_l0_20/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_37/width_16k/average_l0_20/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_37/width_16k/average_l0_266/hparams.json b/layer_37/width_16k/average_l0_266/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_37/width_16k/average_l0_266/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_37/width_16k/average_l0_34/hparams.json b/layer_37/width_16k/average_l0_34/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_37/width_16k/average_l0_34/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_37/width_16k/average_l0_63/hparams.json b/layer_37/width_16k/average_l0_63/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_37/width_16k/average_l0_63/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_38/width_131k/average_l0_10/hparams.json b/layer_38/width_131k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_38/width_131k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_38/width_131k/average_l0_101/hparams.json b/layer_38/width_131k/average_l0_101/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_38/width_131k/average_l0_101/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_38/width_131k/average_l0_18/hparams.json b/layer_38/width_131k/average_l0_18/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_38/width_131k/average_l0_18/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_38/width_131k/average_l0_194/hparams.json b/layer_38/width_131k/average_l0_194/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_38/width_131k/average_l0_194/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_38/width_131k/average_l0_30/hparams.json b/layer_38/width_131k/average_l0_30/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_38/width_131k/average_l0_30/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_38/width_131k/average_l0_53/hparams.json b/layer_38/width_131k/average_l0_53/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_38/width_131k/average_l0_53/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_38/width_16k/average_l0_11/hparams.json b/layer_38/width_16k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_38/width_16k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_38/width_16k/average_l0_128/hparams.json b/layer_38/width_16k/average_l0_128/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_38/width_16k/average_l0_128/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_38/width_16k/average_l0_20/hparams.json b/layer_38/width_16k/average_l0_20/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_38/width_16k/average_l0_20/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_38/width_16k/average_l0_286/hparams.json b/layer_38/width_16k/average_l0_286/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_38/width_16k/average_l0_286/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_38/width_16k/average_l0_34/hparams.json b/layer_38/width_16k/average_l0_34/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_38/width_16k/average_l0_34/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_38/width_16k/average_l0_64/hparams.json b/layer_38/width_16k/average_l0_64/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_38/width_16k/average_l0_64/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_39/width_131k/average_l0_10/hparams.json b/layer_39/width_131k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_39/width_131k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_39/width_131k/average_l0_18/hparams.json b/layer_39/width_131k/average_l0_18/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_39/width_131k/average_l0_18/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_39/width_131k/average_l0_199/hparams.json b/layer_39/width_131k/average_l0_199/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_39/width_131k/average_l0_199/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_39/width_131k/average_l0_30/hparams.json b/layer_39/width_131k/average_l0_30/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_39/width_131k/average_l0_30/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_39/width_131k/average_l0_54/hparams.json b/layer_39/width_131k/average_l0_54/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_39/width_131k/average_l0_54/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_39/width_131k/average_l0_99/hparams.json b/layer_39/width_131k/average_l0_99/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_39/width_131k/average_l0_99/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_39/width_16k/average_l0_11/hparams.json b/layer_39/width_16k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_39/width_16k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_39/width_16k/average_l0_131/hparams.json b/layer_39/width_16k/average_l0_131/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_39/width_16k/average_l0_131/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_39/width_16k/average_l0_19/hparams.json b/layer_39/width_16k/average_l0_19/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_39/width_16k/average_l0_19/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_39/width_16k/average_l0_298/hparams.json b/layer_39/width_16k/average_l0_298/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_39/width_16k/average_l0_298/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_39/width_16k/average_l0_34/hparams.json b/layer_39/width_16k/average_l0_34/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_39/width_16k/average_l0_34/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_39/width_16k/average_l0_64/hparams.json b/layer_39/width_16k/average_l0_64/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_39/width_16k/average_l0_64/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_4/width_131k/average_l0_101/hparams.json b/layer_4/width_131k/average_l0_101/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_4/width_131k/average_l0_101/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_4/width_131k/average_l0_16/hparams.json b/layer_4/width_131k/average_l0_16/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_4/width_131k/average_l0_16/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_4/width_131k/average_l0_28/hparams.json b/layer_4/width_131k/average_l0_28/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_4/width_131k/average_l0_28/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_4/width_131k/average_l0_51/hparams.json b/layer_4/width_131k/average_l0_51/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_4/width_131k/average_l0_51/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_4/width_131k/average_l0_6/hparams.json b/layer_4/width_131k/average_l0_6/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_4/width_131k/average_l0_6/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_4/width_131k/average_l0_9/hparams.json b/layer_4/width_131k/average_l0_9/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_4/width_131k/average_l0_9/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_4/width_16k/average_l0_10/hparams.json b/layer_4/width_16k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_4/width_16k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_4/width_16k/average_l0_18/hparams.json b/layer_4/width_16k/average_l0_18/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_4/width_16k/average_l0_18/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_4/width_16k/average_l0_234/hparams.json b/layer_4/width_16k/average_l0_234/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_4/width_16k/average_l0_234/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_4/width_16k/average_l0_37/hparams.json b/layer_4/width_16k/average_l0_37/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_4/width_16k/average_l0_37/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_4/width_16k/average_l0_91/hparams.json b/layer_4/width_16k/average_l0_91/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_4/width_16k/average_l0_91/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_40/width_131k/average_l0_10/hparams.json b/layer_40/width_131k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_40/width_131k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_40/width_131k/average_l0_17/hparams.json b/layer_40/width_131k/average_l0_17/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_40/width_131k/average_l0_17/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_40/width_131k/average_l0_193/hparams.json b/layer_40/width_131k/average_l0_193/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_40/width_131k/average_l0_193/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_40/width_131k/average_l0_29/hparams.json b/layer_40/width_131k/average_l0_29/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_40/width_131k/average_l0_29/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_40/width_131k/average_l0_49/hparams.json b/layer_40/width_131k/average_l0_49/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_40/width_131k/average_l0_49/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_40/width_131k/average_l0_94/hparams.json b/layer_40/width_131k/average_l0_94/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_40/width_131k/average_l0_94/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_40/width_16k/average_l0_10/hparams.json b/layer_40/width_16k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_40/width_16k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_40/width_16k/average_l0_125/hparams.json b/layer_40/width_16k/average_l0_125/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_40/width_16k/average_l0_125/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_40/width_16k/average_l0_18/hparams.json b/layer_40/width_16k/average_l0_18/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_40/width_16k/average_l0_18/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_40/width_16k/average_l0_292/hparams.json b/layer_40/width_16k/average_l0_292/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_40/width_16k/average_l0_292/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_40/width_16k/average_l0_32/hparams.json b/layer_40/width_16k/average_l0_32/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_40/width_16k/average_l0_32/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_40/width_16k/average_l0_61/hparams.json b/layer_40/width_16k/average_l0_61/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_40/width_16k/average_l0_61/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_41/width_131k/average_l0_10/hparams.json b/layer_41/width_131k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_41/width_131k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_41/width_131k/average_l0_15/hparams.json b/layer_41/width_131k/average_l0_15/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_41/width_131k/average_l0_15/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_41/width_131k/average_l0_175/hparams.json b/layer_41/width_131k/average_l0_175/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_41/width_131k/average_l0_175/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_41/width_131k/average_l0_26/hparams.json b/layer_41/width_131k/average_l0_26/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_41/width_131k/average_l0_26/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_41/width_131k/average_l0_45/hparams.json b/layer_41/width_131k/average_l0_45/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_41/width_131k/average_l0_45/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_41/width_131k/average_l0_84/hparams.json b/layer_41/width_131k/average_l0_84/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_41/width_131k/average_l0_84/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_41/width_16k/average_l0_10/hparams.json b/layer_41/width_16k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_41/width_16k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_41/width_16k/average_l0_113/hparams.json b/layer_41/width_16k/average_l0_113/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_41/width_16k/average_l0_113/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_41/width_16k/average_l0_16/hparams.json b/layer_41/width_16k/average_l0_16/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_41/width_16k/average_l0_16/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_41/width_16k/average_l0_270/hparams.json b/layer_41/width_16k/average_l0_270/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_41/width_16k/average_l0_270/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_41/width_16k/average_l0_28/hparams.json b/layer_41/width_16k/average_l0_28/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_41/width_16k/average_l0_28/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_41/width_16k/average_l0_52/hparams.json b/layer_41/width_16k/average_l0_52/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_41/width_16k/average_l0_52/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_5/width_131k/average_l0_10/hparams.json b/layer_5/width_131k/average_l0_10/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_5/width_131k/average_l0_10/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_5/width_131k/average_l0_16/hparams.json b/layer_5/width_131k/average_l0_16/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_5/width_131k/average_l0_16/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_5/width_131k/average_l0_29/hparams.json b/layer_5/width_131k/average_l0_29/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_5/width_131k/average_l0_29/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_5/width_131k/average_l0_51/hparams.json b/layer_5/width_131k/average_l0_51/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_5/width_131k/average_l0_51/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_5/width_131k/average_l0_6/hparams.json b/layer_5/width_131k/average_l0_6/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_5/width_131k/average_l0_6/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_5/width_131k/average_l0_94/hparams.json b/layer_5/width_131k/average_l0_94/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_5/width_131k/average_l0_94/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_5/width_16k/average_l0_11/hparams.json b/layer_5/width_16k/average_l0_11/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_5/width_16k/average_l0_11/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_5/width_16k/average_l0_193/hparams.json b/layer_5/width_16k/average_l0_193/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_5/width_16k/average_l0_193/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_5/width_16k/average_l0_20/hparams.json b/layer_5/width_16k/average_l0_20/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_5/width_16k/average_l0_20/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_5/width_16k/average_l0_37/hparams.json b/layer_5/width_16k/average_l0_37/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_5/width_16k/average_l0_37/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_5/width_16k/average_l0_77/hparams.json b/layer_5/width_16k/average_l0_77/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_5/width_16k/average_l0_77/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_6/width_131k/average_l0_12/hparams.json b/layer_6/width_131k/average_l0_12/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_6/width_131k/average_l0_12/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_6/width_131k/average_l0_120/hparams.json b/layer_6/width_131k/average_l0_120/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_6/width_131k/average_l0_120/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_6/width_131k/average_l0_21/hparams.json b/layer_6/width_131k/average_l0_21/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_6/width_131k/average_l0_21/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_6/width_131k/average_l0_36/hparams.json b/layer_6/width_131k/average_l0_36/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_6/width_131k/average_l0_36/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_6/width_131k/average_l0_66/hparams.json b/layer_6/width_131k/average_l0_66/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_6/width_131k/average_l0_66/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_6/width_131k/average_l0_7/hparams.json b/layer_6/width_131k/average_l0_7/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_6/width_131k/average_l0_7/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_6/width_16k/average_l0_14/hparams.json b/layer_6/width_16k/average_l0_14/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_6/width_16k/average_l0_14/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_6/width_16k/average_l0_224/hparams.json b/layer_6/width_16k/average_l0_224/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_6/width_16k/average_l0_224/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_6/width_16k/average_l0_25/hparams.json b/layer_6/width_16k/average_l0_25/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_6/width_16k/average_l0_25/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_6/width_16k/average_l0_47/hparams.json b/layer_6/width_16k/average_l0_47/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_6/width_16k/average_l0_47/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_6/width_16k/average_l0_8/hparams.json b/layer_6/width_16k/average_l0_8/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_6/width_16k/average_l0_8/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_6/width_16k/average_l0_93/hparams.json b/layer_6/width_16k/average_l0_93/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_6/width_16k/average_l0_93/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_7/width_131k/average_l0_119/hparams.json b/layer_7/width_131k/average_l0_119/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_7/width_131k/average_l0_119/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_7/width_131k/average_l0_13/hparams.json b/layer_7/width_131k/average_l0_13/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_7/width_131k/average_l0_13/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_7/width_131k/average_l0_21/hparams.json b/layer_7/width_131k/average_l0_21/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_7/width_131k/average_l0_21/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_7/width_131k/average_l0_38/hparams.json b/layer_7/width_131k/average_l0_38/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_7/width_131k/average_l0_38/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_7/width_131k/average_l0_65/hparams.json b/layer_7/width_131k/average_l0_65/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_7/width_131k/average_l0_65/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_7/width_131k/average_l0_8/hparams.json b/layer_7/width_131k/average_l0_8/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_7/width_131k/average_l0_8/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_7/width_16k/average_l0_14/hparams.json b/layer_7/width_16k/average_l0_14/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_7/width_16k/average_l0_14/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_7/width_16k/average_l0_198/hparams.json b/layer_7/width_16k/average_l0_198/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_7/width_16k/average_l0_198/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_7/width_16k/average_l0_25/hparams.json b/layer_7/width_16k/average_l0_25/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_7/width_16k/average_l0_25/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_7/width_16k/average_l0_46/hparams.json b/layer_7/width_16k/average_l0_46/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_7/width_16k/average_l0_46/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_7/width_16k/average_l0_8/hparams.json b/layer_7/width_16k/average_l0_8/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_7/width_16k/average_l0_8/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_7/width_16k/average_l0_92/hparams.json b/layer_7/width_16k/average_l0_92/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_7/width_16k/average_l0_92/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_8/width_131k/average_l0_129/hparams.json b/layer_8/width_131k/average_l0_129/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_8/width_131k/average_l0_129/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_8/width_131k/average_l0_14/hparams.json b/layer_8/width_131k/average_l0_14/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_8/width_131k/average_l0_14/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_8/width_131k/average_l0_24/hparams.json b/layer_8/width_131k/average_l0_24/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_8/width_131k/average_l0_24/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_8/width_131k/average_l0_41/hparams.json b/layer_8/width_131k/average_l0_41/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_8/width_131k/average_l0_41/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_8/width_131k/average_l0_72/hparams.json b/layer_8/width_131k/average_l0_72/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_8/width_131k/average_l0_72/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_8/width_131k/average_l0_8/hparams.json b/layer_8/width_131k/average_l0_8/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_8/width_131k/average_l0_8/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_8/width_16k/average_l0_16/hparams.json b/layer_8/width_16k/average_l0_16/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_8/width_16k/average_l0_16/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_8/width_16k/average_l0_207/hparams.json b/layer_8/width_16k/average_l0_207/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_8/width_16k/average_l0_207/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_8/width_16k/average_l0_28/hparams.json b/layer_8/width_16k/average_l0_28/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_8/width_16k/average_l0_28/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_8/width_16k/average_l0_51/hparams.json b/layer_8/width_16k/average_l0_51/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_8/width_16k/average_l0_51/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_8/width_16k/average_l0_9/hparams.json b/layer_8/width_16k/average_l0_9/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_8/width_16k/average_l0_9/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_8/width_16k/average_l0_99/hparams.json b/layer_8/width_16k/average_l0_99/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_8/width_16k/average_l0_99/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_9/width_131k/average_l0_134/hparams.json b/layer_9/width_131k/average_l0_134/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_9/width_131k/average_l0_134/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_9/width_131k/average_l0_14/hparams.json b/layer_9/width_131k/average_l0_14/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_9/width_131k/average_l0_14/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_9/width_131k/average_l0_25/hparams.json b/layer_9/width_131k/average_l0_25/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_9/width_131k/average_l0_25/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_9/width_131k/average_l0_42/hparams.json b/layer_9/width_131k/average_l0_42/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_9/width_131k/average_l0_42/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_9/width_131k/average_l0_75/hparams.json b/layer_9/width_131k/average_l0_75/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_9/width_131k/average_l0_75/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_9/width_131k/average_l0_8/hparams.json b/layer_9/width_131k/average_l0_8/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_9/width_131k/average_l0_8/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file
diff --git a/layer_9/width_16k/average_l0_100/hparams.json b/layer_9/width_16k/average_l0_100/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a70646e9efa0c8fa4208f223f9eba8e5a46fc4e
--- /dev/null
+++ b/layer_9/width_16k/average_l0_100/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.00025}
\ No newline at end of file
diff --git a/layer_9/width_16k/average_l0_16/hparams.json b/layer_9/width_16k/average_l0_16/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eeb5df8292b414f42cf76096bddfa0458cc71dc
--- /dev/null
+++ b/layer_9/width_16k/average_l0_16/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.002}
\ No newline at end of file
diff --git a/layer_9/width_16k/average_l0_209/hparams.json b/layer_9/width_16k/average_l0_209/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0045df170b345fdc2665507991c8074aaed19c
--- /dev/null
+++ b/layer_9/width_16k/average_l0_209/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.000125}
\ No newline at end of file
diff --git a/layer_9/width_16k/average_l0_28/hparams.json b/layer_9/width_16k/average_l0_28/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..156f5c12cf1e6683871e4b4828e48a0dbd881802
--- /dev/null
+++ b/layer_9/width_16k/average_l0_28/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.001}
\ No newline at end of file
diff --git a/layer_9/width_16k/average_l0_51/hparams.json b/layer_9/width_16k/average_l0_51/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d86843bf8e50d07a194064d36382ca9944f4663
--- /dev/null
+++ b/layer_9/width_16k/average_l0_51/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.0005}
\ No newline at end of file
diff --git a/layer_9/width_16k/average_l0_9/hparams.json b/layer_9/width_16k/average_l0_9/hparams.json
new file mode 100644
index 0000000000000000000000000000000000000000..719ea6c1324be2969f0bf077fd15046d9c36715d
--- /dev/null
+++ b/layer_9/width_16k/average_l0_9/hparams.json
@@ -0,0 +1 @@
+{"sparsity_lambda": 0.004}
\ No newline at end of file