Cycle 0: New best model with eval_loss 0.666352 (Trained with Unsloth)

adapter_config.json

@@ -1,7 +1,7 @@
 {
   "alpha_pattern": {},
   "auto_mapping": null,
-  "base_model_name_or_path": "unsloth/Llama-3.2-3B",
+  "base_model_name_or_path": "unsloth/Qwen2.5-0.5B",
   "bias": "none",
   "fan_in_fan_out": false,
   "inference_mode": true,
@@ -10,25 +10,23 @@
   "layers_pattern": null,
   "layers_to_transform": null,
   "loftq_config": {},
-  "lora_alpha": 256,
-  "lora_dropout": 0.01,
+  "lora_alpha": 128,
+  "lora_dropout": 0.1,
   "megatron_config": null,
   "megatron_core": "megatron.core",
-  "modules_to_save": [
-    "lm_head"
-  ],
+  "modules_to_save": [],
   "peft_type": "LORA",
-  "r": 128,
+  "r": 64,
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "k_proj",
-    "up_proj",
-    "v_proj",
     "down_proj",
     "o_proj",
+    "k_proj",
+    "gate_proj",
     "q_proj",
-    "gate_proj"
+    "v_proj",
+    "up_proj"
   ],
   "task_type": "CAUSAL_LM",
   "use_dora": false,

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3d85d3a171ea05239da5f54dbfaea0d53598fa7f5ed69e887551d76c83a3fa12
-size 1566101640
+oid sha256:8ba8c750c044543b111412a82e6e9525ff45f11fe4f6aaa4169d563514c4a111
+size 140815952