Upload LoRA adapter (README written by author)

README.md

@@ -2,6 +2,7 @@
 base_model: Qwen/Qwen3-4B-Instruct-2507
 datasets:
 - u-10bei/structured_data_with_cot_dataset_v2
+- daichira/structured-5k-mix-sft
 language:
 - en
 license: apache-2.0
@@ -33,7 +34,7 @@ while intermediate reasoning (Chain-of-Thought) is masked.
 
 - Base model: Qwen/Qwen3-4B-Instruct-2507
 - Method: QLoRA (4-bit)
-- Max sequence length: 1024
+- Max sequence length: 2048
 - Epochs: 1
 - Learning rate: 2e-06
 - LoRA: r=128, alpha=256
@@ -61,6 +62,7 @@ model = PeftModel.from_pretrained(model, adapter)
 
 Training data:
 - u-10bei/structured_data_with_cot_dataset_v2
+- daichira/structured-5k-mix-sft
 
 **License Note**:
 The datasets used in this training may have different licenses (e.g., MIT, Apache-2.0, or others).

adapter_config.json

@@ -33,13 +33,13 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "gate_proj",
-    "q_proj",
+    "o_proj",
     "down_proj",
     "v_proj",
     "up_proj",
+    "gate_proj",
     "k_proj",
-    "o_proj"
+    "q_proj"
   ],
   "target_parameters": null,
   "task_type": "CAUSAL_LM",

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:aa3fe320559edebee1c1412e277c64a549dc7359deca26a57443f765ac77c5fa
+oid sha256:fc6ff46f22d6376c43514e83497c3d2ab951d3769d18f56383a6e7dcfec95b57
 size 1057033224