Upload LoRA adapter (README written by author)

README.md

@@ -1,7 +1,7 @@
 ---
 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
@@ -13,7 +13,7 @@ tags:
 - structured-output
 ---
 
-qwen3-4b-structured-output-lora_ver002
+qwen3-4b-structured-output-lora_ver003
 
 This repository provides a **LoRA adapter** fine-tuned from
 **Qwen/Qwen3-4B-Instruct-2507** using **QLoRA (4-bit, Unsloth)**.
@@ -35,7 +35,7 @@ while intermediate reasoning (Chain-of-Thought) is masked.
 - Method: QLoRA (4-bit)
 - Max sequence length: 512
 - Epochs: 1
-- Learning rate: 1e-06
+- Learning rate: 5e-06
 - LoRA: r=64, alpha=128
 
 ## Usage
@@ -59,7 +59,7 @@ model = PeftModel.from_pretrained(model, adapter)
 
 ## Sources & Terms (IMPORTANT)
 
-Training data: u-10bei/structured_data_with_cot_dataset_v2
+Training data: daichira/structured-5k-mix-sft
 
 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

@@ -22,7 +22,7 @@
   "loftq_config": {},
   "lora_alpha": 128,
   "lora_bias": false,
-  "lora_dropout": 0.0,
+  "lora_dropout": 0.05,
   "megatron_config": null,
   "megatron_core": "megatron.core",
   "modules_to_save": null,
@@ -33,12 +33,12 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
+    "v_proj",
+    "k_proj",
     "down_proj",
-    "up_proj",
     "q_proj",
     "o_proj",
-    "v_proj",
-    "k_proj",
+    "up_proj",
     "gate_proj"
   ],
   "target_parameters": null,

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cec8d4982c73e26ec387086bc5e6c9d9b14bca047a0560a82e5b0f30711ef135
+oid sha256:a742c4599ce15888f90418f72e449e34af558b599e4bc065294e402b3435f54d
 size 528550256