Text Generation
PEFT
Safetensors
English
qlora
lora
structured-output
MF0619 commited on
Commit
c5df05e
·
verified ·
1 Parent(s): 1c45812

Upload 8 files

Browse files
Files changed (3) hide show
  1. README.md +5 -5
  2. adapter_config.json +3 -3
  3. adapter_model.safetensors +1 -1
README.md CHANGED
@@ -1,7 +1,7 @@
1
  ---
2
  base_model: Qwen/Qwen3-4B-Instruct-2507
3
  datasets:
4
- - u-10bei/structured_data_with_cot_dataset_512_v5
5
  language:
6
  - en
7
  license: apache-2.0
@@ -13,7 +13,7 @@ tags:
13
  - structured-output
14
  ---
15
 
16
- <【課題】ここは自分で記入して下さい>
17
 
18
  This repository provides a **LoRA adapter** fine-tuned from
19
  **Qwen/Qwen3-4B-Instruct-2507** using **QLoRA (4-bit, Unsloth)**.
@@ -33,9 +33,9 @@ while intermediate reasoning (Chain-of-Thought) is masked.
33
 
34
  - Base model: Qwen/Qwen3-4B-Instruct-2507
35
  - Method: QLoRA (4-bit)
36
- - Max sequence length: 512
37
  - Epochs: 1
38
- - Learning rate: 1e-06
39
  - LoRA: r=64, alpha=128
40
 
41
  ## Usage
@@ -59,7 +59,7 @@ model = PeftModel.from_pretrained(model, adapter)
59
 
60
  ## Sources & Terms (IMPORTANT)
61
 
62
- Training data: u-10bei/structured_data_with_cot_dataset_512_v5
63
 
64
  Dataset License: MIT License. This dataset is used and distributed under the terms of the MIT License.
65
  Compliance: Users must comply with the MIT license (including copyright notice) and the base model's original terms of use.
 
1
  ---
2
  base_model: Qwen/Qwen3-4B-Instruct-2507
3
  datasets:
4
+ - u-10bei/structured_data_with_cot_dataset_512_v2
5
  language:
6
  - en
7
  license: apache-2.0
 
13
  - structured-output
14
  ---
15
 
16
+ qwen3-4b-lora-for-LLM2025A
17
 
18
  This repository provides a **LoRA adapter** fine-tuned from
19
  **Qwen/Qwen3-4B-Instruct-2507** using **QLoRA (4-bit, Unsloth)**.
 
33
 
34
  - Base model: Qwen/Qwen3-4B-Instruct-2507
35
  - Method: QLoRA (4-bit)
36
+ - Max sequence length: 1024
37
  - Epochs: 1
38
+ - Learning rate: 2e-05
39
  - LoRA: r=64, alpha=128
40
 
41
  ## Usage
 
59
 
60
  ## Sources & Terms (IMPORTANT)
61
 
62
+ Training data: u-10bei/structured_data_with_cot_dataset_512_v2
63
 
64
  Dataset License: MIT License. This dataset is used and distributed under the terms of the MIT License.
65
  Compliance: Users must comply with the MIT license (including copyright notice) and the base model's original terms of use.
adapter_config.json CHANGED
@@ -33,12 +33,12 @@
33
  "rank_pattern": {},
34
  "revision": null,
35
  "target_modules": [
36
- "v_proj",
37
  "down_proj",
38
- "up_proj",
39
- "gate_proj",
40
  "k_proj",
 
41
  "o_proj",
 
 
42
  "q_proj"
43
  ],
44
  "target_parameters": null,
 
33
  "rank_pattern": {},
34
  "revision": null,
35
  "target_modules": [
 
36
  "down_proj",
 
 
37
  "k_proj",
38
+ "gate_proj",
39
  "o_proj",
40
+ "v_proj",
41
+ "up_proj",
42
  "q_proj"
43
  ],
44
  "target_parameters": null,
adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:6264b0406c21bc014bb8d2a4756b9fdfec1b65c089a0555cc1e445f18ecc6cec
3
  size 528550256
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9b75d61e23d84f4fb3b7c2c6d23545f8454f2a79836d514687a8de806cc458b6
3
  size 528550256