update README.md
Browse files
README.md
CHANGED
|
@@ -36,7 +36,9 @@ while intermediate reasoning (Chain-of-Thought) is masked.
|
|
| 36 |
- Max sequence length: 2048
|
| 37 |
- Epochs: 3
|
| 38 |
- Learning rate: 1e-06
|
| 39 |
-
-
|
|
|
|
|
|
|
| 40 |
|
| 41 |
## Usage
|
| 42 |
|
|
@@ -46,7 +48,7 @@ from peft import PeftModel
|
|
| 46 |
import torch
|
| 47 |
|
| 48 |
base = "Qwen/Qwen3-4B-Instruct-2507"
|
| 49 |
-
adapter = "
|
| 50 |
|
| 51 |
tokenizer = AutoTokenizer.from_pretrained(base)
|
| 52 |
model = AutoModelForCausalLM.from_pretrained(
|
|
|
|
| 36 |
- Max sequence length: 2048
|
| 37 |
- Epochs: 3
|
| 38 |
- Learning rate: 1e-06
|
| 39 |
+
- Weight decay: 0.05
|
| 40 |
+
- Warmup ratio: 0.1
|
| 41 |
+
- LoRA: r=64, alpha=128, dropout=0
|
| 42 |
|
| 43 |
## Usage
|
| 44 |
|
|
|
|
| 48 |
import torch
|
| 49 |
|
| 50 |
base = "Qwen/Qwen3-4B-Instruct-2507"
|
| 51 |
+
adapter = "MF0619/SFTv13"
|
| 52 |
|
| 53 |
tokenizer = AutoTokenizer.from_pretrained(base)
|
| 54 |
model = AutoModelForCausalLM.from_pretrained(
|