Training in progress, step 100

adapter_config.json

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

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dda825c13ac7b2f848b970b524f518b6456014cb3e473ece417bae2eb6739174
+oid sha256:25e89b84367382ff8f0d486818a7e9a1c72e5409a56e029bc42bf1057d91069d
 size 335605144

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1a00d251b9e8778867a0be2ae762151214e219b14ed22e794563e9fffee1931b
+oid sha256:702ddecfce26dfe6b8a1289d399de5c5452d05721418eac40963e864cdcc977f
 size 6264

training_config.json

@@ -58,16 +58,16 @@
   "lora_dropout": 0.1,
   "raw_lora_target_modules": "all",
   "output_dir": "./outputs/",
-  "per_device_train_batch_size": 2,
+  "per_device_train_batch_size": 8,
   "do_eval": false,
   "per_device_eval_batch_size": null,
-  "gradient_accumulation_steps": 4,
+  "gradient_accumulation_steps": 2,
   "eval_accumulation_steps": null,
   "eval_delay": 0,
   "eval_steps": 1000,
   "warmup_steps": 100,
-  "max_steps": 5000,
-  "num_train_epochs": 3,
+  "max_steps": null,
+  "num_train_epochs": 5,
   "learning_rate": 0.0002,
   "max_grad_norm": 1.0,
   "weight_decay": 0.001,