Update LoRA adapter with dataset v5 (added random schema and sorting constraints)

README.md

@@ -2,7 +2,6 @@
 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
@@ -14,7 +13,7 @@ tags:
 - structured-output
 ---
 
-qwen3-4b-structured-output-v5-hard-lora
+qwen3-4b-structured-output-v5-lora
 
 This repository provides a **LoRA adapter** fine-tuned from
 **Qwen/Qwen3-4B-Instruct-2507** using **QLoRA (4-bit, Unsloth)**.
@@ -34,9 +33,9 @@ while intermediate reasoning (Chain-of-Thought) is masked.
 
 - Base model: Qwen/Qwen3-4B-Instruct-2507
 - Method: QLoRA (4-bit)
-- Max sequence length: 2048
+- Max sequence length: 1024
 - Epochs: 1
-- Learning rate: 5e-06
+- Learning rate: 1e-06
 - LoRA: r=64, alpha=128
 
 ## Usage
@@ -60,7 +59,7 @@ model = PeftModel.from_pretrained(model, adapter)
 
 ## Sources & Terms (IMPORTANT)
 
-Training data: u-10bei/structured_data_with_cot_dataset_512_v5, daichira/structured-hard-sft-4k
+Training data: u-10bei/structured_data_with_cot_dataset_512_v5
 
 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

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

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:41fcfdaa6ef12a8cc3b423405313f350a32ec94c887d5a843a0bce7c4821476f
+oid sha256:83d9458465d6b27e4113e4935c1040d686a2156ad6e0c5f3e80a94b7329b8348
 size 528550256