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_512_v5
+- daichira/structured-hard-sft-4k
 language:
 - en
 license: apache-2.0
@@ -33,10 +34,10 @@ 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: 768
 - Epochs: 1
-- Learning rate: 5e-05
-- LoRA: r=64, alpha=128
+- Learning rate: 2e-05
+- LoRA: r=64, alpha=32
 
 ## Usage
 
@@ -59,7 +60,7 @@ model = PeftModel.from_pretrained(model, adapter)
 
 ## Sources & Terms (IMPORTANT)
 
-Training data: u-10bei/structured_data_with_cot_dataset_512_v5
+Training data: u-10bei/structured_data_with_cot_dataset_512_v5, daichira/structured-hard-sft-4k
 
 Dataset License: MIT License. This dataset is used and distributed under the terms of the MIT License.
 Compliance: Users must comply with the MIT license (including copyright notice) and the base model's original terms of use.

adapter_config.json

@@ -20,7 +20,7 @@
   "layers_pattern": null,
   "layers_to_transform": null,
   "loftq_config": {},
-  "lora_alpha": 128,
+  "lora_alpha": 32,
   "lora_bias": false,
   "lora_dropout": 0.0,
   "megatron_config": null,
@@ -33,13 +33,13 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "v_proj",
-    "gate_proj",
+    "q_proj",
     "down_proj",
     "k_proj",
-    "q_proj",
-    "up_proj",
-    "o_proj"
+    "v_proj",
+    "o_proj",
+    "gate_proj",
+    "up_proj"
   ],
   "target_parameters": null,
   "task_type": "CAUSAL_LM",

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:12fe8314e503c3e4902aa1d6d4e620af394d34cd5ee87f64e46f9edd1ea0fdf1
+oid sha256:7d443f7a8f746b513b74d6349786c51a7f8eae744bba1b3788c295958c9e4f98
 size 528550256