Training in progress, step 300

adapter_config.json

@@ -14,13 +14,13 @@
   "r": 64,
   "revision": null,
   "target_modules": [
-    "o_proj",
-    "q_proj",
-    "v_proj",
+    "gate_proj",
     "k_proj",
+    "q_proj",
     "up_proj",
-    "gate_proj",
-    "down_proj"
+    "down_proj",
+    "v_proj",
+    "o_proj"
   ],
   "task_type": "CAUSAL_LM"
 }

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7eb64e89142d2edd045ef5062485db930543cdfb71826620065d3de56bb9bc95
+oid sha256:ee2ec909333d3905200788230f22925c9ffad5b1eaa153c525a871e6167df588
 size 6264

training_config.json

@@ -2,7 +2,7 @@
   "experiment_key": "base",
   "save_safetensors": true,
   "max_shard_size": "10GB",
-  "local_rank": 0,
+  "local_rank": 1,
   "use_gradient_checkpointing": true,
   "trainer_key": "lm",
   "force_fp32": false,
@@ -27,7 +27,7 @@
   "quantized_hub_private_repo": null,
   "dataset_key": "desc2json",
   "train_local_path_to_data": "./train.jsonl",
-  "eval_local_path_to_data": null,
+  "eval_local_path_to_data": "./eval.jsonl",
   "shuffle": true,
   "max_eval_samples": 1000,
   "add_eval_to_train_if_no_path": false,