Add README
Browse files
README.md
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: Qwen/Qwen3-4B-Instruct-2507
|
| 3 |
+
datasets:
|
| 4 |
+
- u-10bei/structured_data_with_cot_dataset_512_v2
|
| 5 |
+
language: [en]
|
| 6 |
+
license: apache-2.0
|
| 7 |
+
pipeline_tag: text-generation
|
| 8 |
+
tags: [sft, structured-output, qwen3]
|
| 9 |
+
---
|
| 10 |
+
# qwen3-4b-structeval-v15
|
| 11 |
+
V14-base + seed=777 + weight_decay=0.03 + TOML特化prompt(Codex第9回)
|
| 12 |
+
- lr=1.5e-05, epochs=2, r=8, alpha=8, BS=32
|
| 13 |
+
- dropout=0, seed=777, warmup=0.05
|
| 14 |
+
- weight_decay=0.03
|
| 15 |
+
- system_prompt: TOML特化+inline table禁止+1行1キー（学習/推論統一）
|
| 16 |
+
- enable_thinking=False, strip_cot=True(strict)
|