test with small T5 model and 5e-4 learning rate

adapter_config.json

@@ -19,14 +19,14 @@
   "megatron_core": "megatron.core",
   "modules_to_save": null,
   "peft_type": "LORA",
-  "r": 4,
+  "r": 6,
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "v",
-    "k",
     "q",
-    "o"
+    "v",
+    "o",
+    "k"
   ],
   "task_type": "SEQ_2_SEQ_LM",
   "use_dora": false,

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:16ae681e220ed2479ed57254e3ba20b8a8981959c63dc8f503bb8e860fdaabf0
-size 1402608
+oid sha256:dd6db6bb5fa7b02144cd17f4ea4c68f7224161dcc8c681c096a5edde4d382273
+size 2090840