Update README.md
Browse files
README.md
CHANGED
|
@@ -8,8 +8,11 @@ base_model:
|
|
| 8 |
- Qwen/Qwen2-7B-Instruct
|
| 9 |
---
|
| 10 |
|
| 11 |
-
在2张V800-80G上,
|
| 12 |
|
|
|
|
|
|
|
|
|
|
| 13 |
|
| 14 |
epoch 1:
|
| 15 |
total FLOS 824852GF
|
|
@@ -17,14 +20,14 @@ train loss: 1.6586
|
|
| 17 |
train_samples_per_second: 3.246
|
| 18 |
train_steps_per_second: 0.014
|
| 19 |
train_runtime: 37:01:33.17
|
| 20 |
-
perplexity: 7.2646328377141005
|
| 21 |
|
| 22 |
epoch 2:
|
| 23 |
train loss: 1.3895
|
| 24 |
train_samples_per_second: 3.11
|
| 25 |
train_steps_per_second: 0.014
|
| 26 |
train_runtime: 38:38:44.20
|
| 27 |
-
perplexity:
|
| 28 |
|
| 29 |
```
|
| 30 |
llamafactory-cli train \
|
|
|
|
| 8 |
- Qwen/Qwen2-7B-Instruct
|
| 9 |
---
|
| 10 |
|
| 11 |
+
在2张V800-80G上,
|
| 12 |
|
| 13 |
+
基于Qwen/Qwen2-7B-Instruct,
|
| 14 |
+
|
| 15 |
+
使用SylvanL/Traditional-Chinese-Medicine-Dataset-Pretrain进行了2个epoch的继续预训练(Continue Pre-train).
|
| 16 |
|
| 17 |
epoch 1:
|
| 18 |
total FLOS 824852GF
|
|
|
|
| 20 |
train_samples_per_second: 3.246
|
| 21 |
train_steps_per_second: 0.014
|
| 22 |
train_runtime: 37:01:33.17
|
| 23 |
+
average perplexity: 7.2646328377141005
|
| 24 |
|
| 25 |
epoch 2:
|
| 26 |
train loss: 1.3895
|
| 27 |
train_samples_per_second: 3.11
|
| 28 |
train_steps_per_second: 0.014
|
| 29 |
train_runtime: 38:38:44.20
|
| 30 |
+
average perplexity: 5.800355962033688
|
| 31 |
|
| 32 |
```
|
| 33 |
llamafactory-cli train \
|