diff --git a/101/edges.pkl b/101/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..e5ca91be0d15dfebb7961bc819259c28cd200595
--- /dev/null
+++ b/101/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a2cb02e0518a50c2b53be59ea3c3788ec80c0d3c6c1e70070bd7115abdfae77
+size 113
diff --git a/101/ll_model.pth b/101/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..6eb2fd27a89b4b1bcdbb356cc08893a45c151706
--- /dev/null
+++ b/101/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a8fd5916d51408c7bd9536f5f7384aee78c453e6b3ea6c16714b9b92d21c7386
+size 14890
diff --git a/101/ll_model_cfg.pkl b/101/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..c5b574e112d5993bda07934df4a4c1e43ea9d974
--- /dev/null
+++ b/101/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7cd0768a4f665f6e168d8f8dcd31ee2ef833847442d11558cb1fdb65b71fab66
+size 1093
diff --git a/101/meta.json b/101/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..de46c8e730f709f9f311b0e7ec7e367865c9d5ec
--- /dev/null
+++ b/101/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-101-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/103/edges.pkl b/103/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..dab032da89d9c83c4731d7e2178d960620963db1
--- /dev/null
+++ b/103/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4df9e5eb129553e810673c7ea9f138df364f9461861df969e00972518748db4b
+size 1256
diff --git a/103/ll_model.pth b/103/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..42f30d1dab678429f9b8f7ede5ecbe7cc6e9a46f
--- /dev/null
+++ b/103/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d2deec61e135c0e16e0f0c1884c91141e8486825d8c697045e32c422cc2862bb
+size 106310
diff --git a/103/ll_model_cfg.pkl b/103/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..526ba5cc763abe7f4ac96eae1ca017d811866b1f
--- /dev/null
+++ b/103/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a7628439ed419829404ac8619bc94b8424d1eb5f6e00ce2384a591ef1771ab98
+size 1100
diff --git a/103/meta.json b/103/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..a1d74f806a0cd62169013ff04cf0ea545a7b70e2
--- /dev/null
+++ b/103/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-103-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/110/edges.pkl b/110/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..82c8bc7bcaa20f5f8535fedfd0e1f08743da6359
--- /dev/null
+++ b/110/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ce0c25059854bda5408738429c5404c6ae0931901591143f73367b3f0ae29f00
+size 598
diff --git a/110/ll_model.pth b/110/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..002d172ad23e24fb0e88cd6c3b42611f0871460a
--- /dev/null
+++ b/110/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2aab75c41c36e2d937b1707d84aec6d2fadf74696ed059c6d3667b6e7e868eaf
+size 54890
diff --git a/110/ll_model_cfg.pkl b/110/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..77913101467e2b6073dc18b1d5c63e2fd7e36c99
--- /dev/null
+++ b/110/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:20f14b516d1a8a2e436ddabe3bf7a1b3d5ef10df0bcf62228affc607ec26c56e
+size 1100
diff --git a/110/meta.json b/110/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..1d9ca4bda93726325e6317341dc36378429f3776
--- /dev/null
+++ b/110/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-110-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/111/edges.pkl b/111/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..87ce57ef1615f409f3add1e7e6061bf9d3f54caf
--- /dev/null
+++ b/111/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f9324e9a3dab349582cbae63c998b4251b92b875e49caa79826b450eadc9dcb
+size 1251
diff --git a/111/ll_model.pth b/111/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..f688028beda46e5729a37b8e06a6cbdf82c68616
--- /dev/null
+++ b/111/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c512e7e3b1fca95ffd88f36b14824ddd5b8127c76ea6819ca21a4e16273ed36e
+size 106310
diff --git a/111/ll_model_cfg.pkl b/111/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..fce67ee3bf24809d7ff43648cd3dec0c90d9b08d
--- /dev/null
+++ b/111/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c689fcff94029e6296340b4c2a5909a422b8d3fc7b26005911b3d760538c56e
+size 1100
diff --git a/111/meta.json b/111/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..66df8d4f0c8a7271463e2dc00a86f9f1d5ec708f
--- /dev/null
+++ b/111/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-111-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/113/edges.pkl b/113/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..c62f93d6eb747c0d989b74e9933f46448019f806
--- /dev/null
+++ b/113/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1aaf72d92f0df42e61bc20fd40672c902ce2e227f0df638b4b38111a09c06c93
+size 2847
diff --git a/113/ll_model.pth b/113/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..2e83ff44449e63a394212afd638961b38e03bb1b
--- /dev/null
+++ b/113/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ea7ec0d5b5e4ba6da236fedcb698feeef7c6987a179c5f31f8fd78a367ccf4a7
+size 2688376
diff --git a/113/ll_model_cfg.pkl b/113/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..e885a959ef28537152a4d7f554a8b844b6989b36
--- /dev/null
+++ b/113/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b5d64e11240c419a545b94fea7aabfae5ed77782f4f4de712846bc7ba453c7c9
+size 1103
diff --git a/113/meta.json b/113/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..67662fb2c7bc64487f07555d68ee837124784c93
--- /dev/null
+++ b/113/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-113-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/114/edges.pkl b/114/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..e5ca91be0d15dfebb7961bc819259c28cd200595
--- /dev/null
+++ b/114/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a2cb02e0518a50c2b53be59ea3c3788ec80c0d3c6c1e70070bd7115abdfae77
+size 113
diff --git a/114/ll_model.pth b/114/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..653085acc4531cc019d8c214d8ded068cc640a43
--- /dev/null
+++ b/114/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:941f9c19ab892cf7546c8535cd3728477616d467c294886c82425a87d1b9bd93
+size 14698
diff --git a/114/ll_model_cfg.pkl b/114/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..bc54bd65411c646fb7534e0c2760481904059ff4
--- /dev/null
+++ b/114/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d5209f9d9e7a92d79c658406442cc48849fe4256a1186472614bf141344c7d5
+size 1093
diff --git a/114/meta.json b/114/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..2b07d5cfb8edbe5f7bb5c5f13f3d992803d78310
--- /dev/null
+++ b/114/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-114-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/122/edges.pkl b/122/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..e5ca91be0d15dfebb7961bc819259c28cd200595
--- /dev/null
+++ b/122/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a2cb02e0518a50c2b53be59ea3c3788ec80c0d3c6c1e70070bd7115abdfae77
+size 113
diff --git a/122/ll_model.pth b/122/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..8b4a3af6269370c046892aa98398203f362fe53d
--- /dev/null
+++ b/122/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4c428334ca09901df8c1a557fb8ff91f94faad58da12c8cec5cc54aba42ee5cd
+size 14890
diff --git a/122/ll_model_cfg.pkl b/122/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..c5b574e112d5993bda07934df4a4c1e43ea9d974
--- /dev/null
+++ b/122/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7cd0768a4f665f6e168d8f8dcd31ee2ef833847442d11558cb1fdb65b71fab66
+size 1093
diff --git a/122/meta.json b/122/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..dc0a7909e6f23ef7f141b498b0bf588b4922e478
--- /dev/null
+++ b/122/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-122-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/124/edges.pkl b/124/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..7aa6d087ebbf22e3e88a3e50ec42082bf41ad8f6
--- /dev/null
+++ b/124/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94f304942c0a1b31c525866b45e6cb08898c68c8babe262aee1720e0d7ae703f
+size 1110
diff --git a/124/ll_model.pth b/124/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..24ce61be0c152f0d8026ae20baaa5d4841efc74f
--- /dev/null
+++ b/124/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d3204efb7c2c92f7a8807743e44c1e6c9c65d6828a9aaa22606b0874ccbe3293
+size 105478
diff --git a/124/ll_model_cfg.pkl b/124/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..0255d8d1c9571ea5aef667940b76fc68dd034f39
--- /dev/null
+++ b/124/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5d9b99b585bf898a743c98e402a9c1ef1600b3d43d8e79eeadd11b86b7d2c626
+size 1100
diff --git a/124/meta.json b/124/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..8fe8a6e50d57c0aa463ce4888eb88a9d4ae8b08e
--- /dev/null
+++ b/124/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-124-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/129/edges.pkl b/129/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..bab701bbaeec723cab176304deb0d1d43b1a754d
--- /dev/null
+++ b/129/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b01d4b9b1e5740fd2934079065cb57785bb111918826a8e8ad7b5b4af00c7bcc
+size 677
diff --git a/129/ll_model.pth b/129/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..5186704fb1bbd93b9f60c7fb65e9fb824ad5bc6e
--- /dev/null
+++ b/129/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f497c70d4f781598a2d86469fc884a7aee35013d2ca2851e03a40077224fc6ab
+size 21190
diff --git a/129/ll_model_cfg.pkl b/129/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..c13128b094d4d58d7841d2f117880ad35598f65b
--- /dev/null
+++ b/129/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a4c611ecd0f8c57d44f555697540a82578e20f4dbe88bb1bd25132bd06244fa2
+size 1100
diff --git a/129/meta.json b/129/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..ec2484d026d6587aa551af103b38b05cecf36170
--- /dev/null
+++ b/129/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-129-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/14/edges.pkl b/14/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..e940027173ac0abdd876eb239aa0bc8873367a4c
--- /dev/null
+++ b/14/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3156b2e517b8c45e76ea28f8c2ccbcda226f636cd44a809c908e167871d2cdab
+size 546
diff --git a/14/ll_model.pth b/14/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..d32815ee9dd47ddd179450e2203218565f0bd8b0
--- /dev/null
+++ b/14/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:83ce6aabb3fdb734d8e602755d05c15332bd422572385cbcb33cd0e41191c637
+size 19754
diff --git a/14/ll_model_cfg.pkl b/14/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..2d1652a666a5aa26837759c033db9a427b08f5ee
--- /dev/null
+++ b/14/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:27866dc79bdac527552f8e0d39e80579033fc10ab193f1bdedef3660205608dd
+size 1100
diff --git a/14/meta.json b/14/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..b042f0bcdd1f081a715899978d8af8c387b664d8
--- /dev/null
+++ b/14/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-14-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/2/edges.pkl b/2/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..1cf4c63dfc41213f5fed4e26d5d0ab36350fe3f0
--- /dev/null
+++ b/2/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dd8ee919e31a1590d5e02e7d6aa4625e2d9e3002e3bbb74402f5a49e7d0be36b
+size 1251
diff --git a/2/ll_model.pth b/2/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..f5e3e961bc1897770611f62ef25655d98d8680e4
--- /dev/null
+++ b/2/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2b2ef568269ef820d7d6975fed90fabf517f044f26f9b10a8a3724a7da3c4f83
+size 647458
diff --git a/2/ll_model_cfg.pkl b/2/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..987a0e6f920c5c653bb29adc2bcc3207ebff1b4e
--- /dev/null
+++ b/2/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:274196ba87c5a5cc25f5e3ed93a4c7dace5a7924905804c2ee6d2112c460f534
+size 1102
diff --git a/2/meta.json b/2/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..6a08c8f7d0192704aeb0a85ed8289abfe232cab7
--- /dev/null
+++ b/2/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-2-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/24/edges.pkl b/24/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..49002b046ab1fb20ff1c513545fa6313bde45b60
--- /dev/null
+++ b/24/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9626f34a27e68c1fd617876205908e46e8e9ca7ebce03801219894dfb4994d42
+size 537
diff --git a/24/ll_model.pth b/24/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..e6056fca3253cc09fa96e85ca11c02169576e3c0
--- /dev/null
+++ b/24/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd7c40f800ce7c3fc207b7693033c2fa25ff89e591c74ef8e89ae07a22bbc9b0
+size 141930
diff --git a/24/ll_model_cfg.pkl b/24/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..69ba915674651936f4461ab2cdfcae5f4485419a
--- /dev/null
+++ b/24/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3887932bc979926a0304b1ca63f6b7e05f0dae97b836c1922ce2f12445a0c383
+size 1093
diff --git a/24/meta.json b/24/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..ae20a79719ca8fa19b39c7471e5c8eb8c3d0decd
--- /dev/null
+++ b/24/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-24-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/25/edges.pkl b/25/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..3626717ecf51aa9a4c5c10a0ed55562874971992
--- /dev/null
+++ b/25/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd358562dc56469dbf03912dac0c5d3d5489b0f2dadcb85a7cd7c24f9582a74a
+size 1260
diff --git a/25/ll_model.pth b/25/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..135f80526666e1eb61ebb0e280bd34f8a09ff2de
--- /dev/null
+++ b/25/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0e72ecd925276352d1240da0b40019b38ef254cbfdad61b234d96672fc4ae2e2
+size 398954
diff --git a/25/ll_model_cfg.pkl b/25/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..7304195e44c86bd47cfecd5d1a1c6f934e6648f6
--- /dev/null
+++ b/25/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c5df91fe7e78fbfc9fe45fe234027fb7c15ecd4c9a0b7a4776da0bf067e09db5
+size 1102
diff --git a/25/meta.json b/25/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..8019c7b4f417333df5c50ad888e246dfea854fb3
--- /dev/null
+++ b/25/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-25-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/30/edges.pkl b/30/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..e5ca91be0d15dfebb7961bc819259c28cd200595
--- /dev/null
+++ b/30/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a2cb02e0518a50c2b53be59ea3c3788ec80c0d3c6c1e70070bd7115abdfae77
+size 113
diff --git a/30/ll_model.pth b/30/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..db6de21febb420b8bcc10e261dbf5f4f374874f0
--- /dev/null
+++ b/30/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f102ec21037cdaa7b2e9d9c93f84a6b98769b4a569c4c76c385f79eb820ceecd
+size 14890
diff --git a/30/ll_model_cfg.pkl b/30/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..c5b574e112d5993bda07934df4a4c1e43ea9d974
--- /dev/null
+++ b/30/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7cd0768a4f665f6e168d8f8dcd31ee2ef833847442d11558cb1fdb65b71fab66
+size 1093
diff --git a/30/meta.json b/30/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..93507f15dbff18c00266dd6edbd6314ba52a0133
--- /dev/null
+++ b/30/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-30-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/31/edges.pkl b/31/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..e5ca91be0d15dfebb7961bc819259c28cd200595
--- /dev/null
+++ b/31/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a2cb02e0518a50c2b53be59ea3c3788ec80c0d3c6c1e70070bd7115abdfae77
+size 113
diff --git a/31/ll_model.pth b/31/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..c5ef9b59504dde38ebf17532e842435668c5a561
--- /dev/null
+++ b/31/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb2d624604a144aad5e649ba82eb780a981378af5531fd7cb1d622b9a25088e2
+size 14506
diff --git a/31/ll_model_cfg.pkl b/31/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..bec58f03b391b6444e30d732934b7a482dd9b295
--- /dev/null
+++ b/31/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:17ef9951acb5bdc623a4b7219f35bedc242d0baaca76446dfd5b20db12f6e5ba
+size 1093
diff --git a/31/meta.json b/31/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..93f1b61e433084762771057545155e5032556db7
--- /dev/null
+++ b/31/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-31-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/39/edges.pkl b/39/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..561f7bfca5220cf67f60fd0c328c61d7666fb995
--- /dev/null
+++ b/39/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:361f8218dc908a7666d7991c895eb12b63d62380fd3b4a1a0fc045aaa24a8438
+size 592
diff --git a/39/ll_model.pth b/39/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..602399e8dc10840cd41f7ffb0231d8afd84884b7
--- /dev/null
+++ b/39/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6ff423c2683ec5427ee528b33cb92dca0c5ebc5b88496fb94fa0ee4e47bd88de
+size 1453098
diff --git a/39/ll_model_cfg.pkl b/39/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..6e195e9dfe4735c944662a1f28f988cad4373b00
--- /dev/null
+++ b/39/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:67220fd480f5a350b213369f638baa23f204a87beed579bdbf736046a917e553
+size 1096
diff --git a/39/meta.json b/39/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..40a32c389d5f8bf386e5d23647e3834b3173ef4e
--- /dev/null
+++ b/39/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-39-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/40/edges.pkl b/40/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..bdbefa18d08b742c27ecd0f2dfe7bebb1bfa0751
--- /dev/null
+++ b/40/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b6bcafa323f1f17a36bae3ba387d1d9a386a70723eaa3784fb7ca2d5d4f7631
+size 168
diff --git a/40/ll_model.pth b/40/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..2b9f95a71d396cfb0e003a2334c5309aed5209c0
--- /dev/null
+++ b/40/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e571af0a38a4a83420a8c7650e4a42c9bbf798ebcd94c0d42c88bf21cca597cc
+size 15018
diff --git a/40/ll_model_cfg.pkl b/40/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..6dd50ccad0a40c1489ec943c9451e501eb820d54
--- /dev/null
+++ b/40/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d2dbb3ecaf912cc92d49d41c90cf076299565216c44d0ed1d11eabd9ba76a0f0
+size 1093
diff --git a/40/meta.json b/40/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..7fdaadfaa44df22f5514908c51e39cba067b0c9f
--- /dev/null
+++ b/40/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-40-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/44/edges.pkl b/44/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..e940027173ac0abdd876eb239aa0bc8873367a4c
--- /dev/null
+++ b/44/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3156b2e517b8c45e76ea28f8c2ccbcda226f636cd44a809c908e167871d2cdab
+size 546
diff --git a/44/ll_model.pth b/44/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..72d8ae38d25b23cb11443304e8f4b9f98d4015b8
--- /dev/null
+++ b/44/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd6dfc6ecc892d2a5335f19eea48e09588427b91dc276616c1c201854a990a4e
+size 72426
diff --git a/44/ll_model_cfg.pkl b/44/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..337dfe4da80b1513059cc5f1bfda337ad057462f
--- /dev/null
+++ b/44/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:04187613894b7f89b32d1355605a0a21fde32558454a58fbaa3ac1a3386cae65
+size 1100
diff --git a/44/meta.json b/44/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..1130dd5ab0b89c69d185eeefea91381ec8d57f01
--- /dev/null
+++ b/44/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-44-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/45/edges.pkl b/45/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..5640ad5e4c4d8fb835c50086292f1a84f527fd94
--- /dev/null
+++ b/45/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b704042e6ae583b9f2e39c3f8fdf4b2d7cefaf7542dad1862d724668b66d6431
+size 743
diff --git a/45/ll_model.pth b/45/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..a85ef39347275d54d402c39e1d11b4d7b8dca6aa
--- /dev/null
+++ b/45/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5a2bb7d289c690eaf5f032e23b7aed4ab12f4a9da45898ec9e4ce1c15319b66b
+size 106886
diff --git a/45/ll_model_cfg.pkl b/45/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..59c12b4a4ee780fab67c35fd3192abbe2c9566d7
--- /dev/null
+++ b/45/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f75e600820887795cdc0cd09b7820aae1688993c5048bc1999b59ff6040813b
+size 1100
diff --git a/45/meta.json b/45/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..132c1faf59358c4ef9f9d9312484cd42c8d14454
--- /dev/null
+++ b/45/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-45-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/51/edges.pkl b/51/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..e5ca91be0d15dfebb7961bc819259c28cd200595
--- /dev/null
+++ b/51/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a2cb02e0518a50c2b53be59ea3c3788ec80c0d3c6c1e70070bd7115abdfae77
+size 113
diff --git a/51/ll_model.pth b/51/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..6f1dcccd40e931053eb5cd84295589db9188be9d
--- /dev/null
+++ b/51/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec7d24c99b385746e947502d30ece3030bdcc574e078d3ad25df1496c67a63f8
+size 15978
diff --git a/51/ll_model_cfg.pkl b/51/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..3d2302a3363574dc594821938281a180c83033e9
--- /dev/null
+++ b/51/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc4802365d0777899396186daa97c34aadedda12c8390fa7dabf1fc09b3283fa
+size 1093
diff --git a/51/meta.json b/51/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..2ff9aa68ed1bb95b442727d27a4d00f660647e39
--- /dev/null
+++ b/51/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-51-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/56/edges.pkl b/56/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..145ad9019256687adb7de5136c70accdaedd74a3
--- /dev/null
+++ b/56/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:994f3bdfbe5b148e1da38018a24a1567c1d86e5de1c18e9b4d62af358812c709
+size 189
diff --git a/56/ll_model.pth b/56/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..9af3658609446090b0a798571700c79d5c54c97c
--- /dev/null
+++ b/56/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c84d8bb57d109858f6e731435abbcaba6ab1ac7b40b3d47ab9fb30f08ed1a92d
+size 14698
diff --git a/56/ll_model_cfg.pkl b/56/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..534db4cd2638721923161790690b188cc23685b9
--- /dev/null
+++ b/56/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:45f4ad485e76a96603c114b396ae4c3f08a4b56f04ae6914ec4f8cde506d521b
+size 1093
diff --git a/56/meta.json b/56/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..70bea4946bedaee3130f120f2a1851cec2d3fe7a
--- /dev/null
+++ b/56/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-56-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/58/edges.pkl b/58/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..b2d757789822b41d846a57559018a46f42ba2cc6
--- /dev/null
+++ b/58/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4edacbb2f328f7d3526a6658a58484634a73124ea79cace68231ae0bb1ab20ea
+size 1196
diff --git a/58/ll_model.pth b/58/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..c3e0d148b408b7e0623cefea1c73715a5d5aa52b
--- /dev/null
+++ b/58/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:95278b3bf4897a27606f35ae5ed251fb022d9e4158f7575d32ab42d6fa142e20
+size 173318
diff --git a/58/ll_model_cfg.pkl b/58/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..72d1ecf1a3f69742594de23b537f329ce56b853f
--- /dev/null
+++ b/58/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8cc71fd34570fd0bef7a3f511b8675d7cbaec6db017ce3cfbee7a74f8a23e501
+size 1100
diff --git a/58/meta.json b/58/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..b82866dad8ed0822b6e068f15d2d04fc0d5faa50
--- /dev/null
+++ b/58/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-58-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/63/edges.pkl b/63/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..e940027173ac0abdd876eb239aa0bc8873367a4c
--- /dev/null
+++ b/63/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3156b2e517b8c45e76ea28f8c2ccbcda226f636cd44a809c908e167871d2cdab
+size 546
diff --git a/63/ll_model.pth b/63/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..2c31656aa34614a107f9b67d8c9b976d47422aff
--- /dev/null
+++ b/63/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:73cd4fe1e237568f33000e0b014f3e265e540c029db551bcb9a6356b73eda684
+size 72426
diff --git a/63/ll_model_cfg.pkl b/63/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..337dfe4da80b1513059cc5f1bfda337ad057462f
--- /dev/null
+++ b/63/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:04187613894b7f89b32d1355605a0a21fde32558454a58fbaa3ac1a3386cae65
+size 1100
diff --git a/63/meta.json b/63/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..a405e58bad6fe2bc1dfd5632c22bf5ca8c9fd1e6
--- /dev/null
+++ b/63/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-63-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/69/edges.pkl b/69/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..e5ca91be0d15dfebb7961bc819259c28cd200595
--- /dev/null
+++ b/69/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a2cb02e0518a50c2b53be59ea3c3788ec80c0d3c6c1e70070bd7115abdfae77
+size 113
diff --git a/69/ll_model.pth b/69/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..b77b155b79af09601a390d47c8eea31480ad8467
--- /dev/null
+++ b/69/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0bc7ed1e23136922f0e8572224f0fefb931f4bb0fd5206638c01d13976b7c2f2
+size 14570
diff --git a/69/ll_model_cfg.pkl b/69/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..67dc1de03664d8f95f764c714db04b2596defced
--- /dev/null
+++ b/69/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76f0aa58b6db19df2b1dd007801bbfca7cea64393b109f2b2e66f64a4e603fd8
+size 1093
diff --git a/69/meta.json b/69/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..098b0641cd7be81b9308b09256440b7ad587a571
--- /dev/null
+++ b/69/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-69-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/79/edges.pkl b/79/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..e5ca91be0d15dfebb7961bc819259c28cd200595
--- /dev/null
+++ b/79/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a2cb02e0518a50c2b53be59ea3c3788ec80c0d3c6c1e70070bd7115abdfae77
+size 113
diff --git a/79/ll_model.pth b/79/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..b1013702099189f559c6a092bc8f5b5458173518
--- /dev/null
+++ b/79/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0be8b31dc4c57d669f5d918b5115ec91e4d21633dca171a4eb28f23d78187012
+size 14570
diff --git a/79/ll_model_cfg.pkl b/79/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..67dc1de03664d8f95f764c714db04b2596defced
--- /dev/null
+++ b/79/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76f0aa58b6db19df2b1dd007801bbfca7cea64393b109f2b2e66f64a4e603fd8
+size 1093
diff --git a/79/meta.json b/79/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..f99d0e704fb80e02276a929048d6fb2cf0940f3a
--- /dev/null
+++ b/79/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-79-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/82/edges.pkl b/82/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..286adb7882954d9ea1f3ff1cc6a98fcb11347e1f
--- /dev/null
+++ b/82/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a79289804c5b5c05a361921e0a87b6c2e043b858ff57a78334865d1aba23ccaa
+size 798
diff --git a/82/ll_model.pth b/82/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..0ed375e1081a43029b8024832ded266a8ccdcefd
--- /dev/null
+++ b/82/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5aa2c86e46ca4718f4043c735231bc6b3b3bd86ba56aed5861c523886e0ea874
+size 140706
diff --git a/82/ll_model_cfg.pkl b/82/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..b38e09efe6d3119f0c45895074a68a4393f8feda
--- /dev/null
+++ b/82/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:78a927568daff80c1fa61e40868a6c36b5d297604342ffecc99b57af12d6e00d
+size 1100
diff --git a/82/meta.json b/82/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..aeddd974465bc159f15ecb1469534d726ccc25fc
--- /dev/null
+++ b/82/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-82-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/86/edges.pkl b/86/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..e5ca91be0d15dfebb7961bc819259c28cd200595
--- /dev/null
+++ b/86/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a2cb02e0518a50c2b53be59ea3c3788ec80c0d3c6c1e70070bd7115abdfae77
+size 113
diff --git a/86/ll_model.pth b/86/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..caf83c1207d81fdf8d1767a6cabac81a3116a1c9
--- /dev/null
+++ b/86/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:32dedebc0feaadf6b14339c165d7a603376c57f13ae2280430cbb115060c2c1f
+size 14570
diff --git a/86/ll_model_cfg.pkl b/86/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..67dc1de03664d8f95f764c714db04b2596defced
--- /dev/null
+++ b/86/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76f0aa58b6db19df2b1dd007801bbfca7cea64393b109f2b2e66f64a4e603fd8
+size 1093
diff --git a/86/meta.json b/86/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..63cb0f995ae2140e9c8f4fe8fff84f5b82a4797c
--- /dev/null
+++ b/86/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-86-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/87/edges.pkl b/87/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..e5ca91be0d15dfebb7961bc819259c28cd200595
--- /dev/null
+++ b/87/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a2cb02e0518a50c2b53be59ea3c3788ec80c0d3c6c1e70070bd7115abdfae77
+size 113
diff --git a/87/ll_model.pth b/87/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..ca317f2eb053c9fce0b90600353b364538a6ccc0
--- /dev/null
+++ b/87/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8dcfaa6128cfb7efe557adc545f7d21218edbc84730e1f19d917ad55c32da235
+size 14570
diff --git a/87/ll_model_cfg.pkl b/87/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..67dc1de03664d8f95f764c714db04b2596defced
--- /dev/null
+++ b/87/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76f0aa58b6db19df2b1dd007801bbfca7cea64393b109f2b2e66f64a4e603fd8
+size 1093
diff --git a/87/meta.json b/87/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..e8f96abfa5e27cbce249ba20825e3957ba37c4ad
--- /dev/null
+++ b/87/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-87-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file
diff --git a/93/edges.pkl b/93/edges.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..dab032da89d9c83c4731d7e2178d960620963db1
--- /dev/null
+++ b/93/edges.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4df9e5eb129553e810673c7ea9f138df364f9461861df969e00972518748db4b
+size 1256
diff --git a/93/ll_model.pth b/93/ll_model.pth
new file mode 100644
index 0000000000000000000000000000000000000000..0648e2acb9af6f20ef484f7735e29982142e285b
--- /dev/null
+++ b/93/ll_model.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d53d0579b86e0fe99c920125b284f6a55a8ccb57093ce1ee0cf5e1eff4df655
+size 80198
diff --git a/93/ll_model_cfg.pkl b/93/ll_model_cfg.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..e446ee48729a067c302b402fe982c943890c000d
--- /dev/null
+++ b/93/ll_model_cfg.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:177166a23aa03f8ac1b27fae7ef37385e6ecbf16fb2eb78ddf081045afd8e05a
+size 1100
diff --git a/93/meta.json b/93/meta.json
new file mode 100644
index 0000000000000000000000000000000000000000..086002889daed9fb049fe18f98441d0ef9bbab35
--- /dev/null
+++ b/93/meta.json
@@ -0,0 +1 @@
+{"output_dir": "/circuits-benchmark/results", "atol": 0.05, "lr": 0.001, "use_single_loss": true, "iit_weight": 1.0, "behavior_weight": 0.4, "strict_weight": 0.4, "epochs": 1000, "early_stop_accuracy_threshold": 99.9, "act_fn": "gelu", "use_wandb": true, "wandb_project": "iit-train", "wandb_name": "case-93-seed-67-s-0.4-b-0.4-iit-1", "save_model_to_wandb": true, "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "linear", "model_pair": "strict", "same_size": false, "seed": 67, "batch_size": 256, "include_mlp": false, "detach_while_caching": true, "scheduler_val_metric": ["val/accuracy", "val/IIA", "val/strict_accuracy"], "siit_sampling": "sample_all", "val_iia_sampling": "all"}
\ No newline at end of file