Skip to content

Commit 2d07f9f

Browse files
authored
fix sh ddp_backend (#1360)
1 parent 94a0658 commit 2d07f9f

File tree

12 files changed

+27
-29
lines changed

12 files changed

+27
-29
lines changed

docs/source/LLM/Agent微调最佳实践.md

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -165,7 +165,7 @@ Final Answer: 如果您想要一款拍照表现出色的手机,我为您推荐
165165
| ms-bench | 60000(抽样) |
166166
| self-recognition | 3000(重复抽样) |
167167

168-
我们也支持使用自己的Agent数据集。数据集格式需要符合[自定义数据集](https://github.com/modelscope/swift/blob/main/docs/source/LLM/%E8%87%AA%E5%AE%9A%E4%B9%89%E4%B8%8E%E6%8B%93%E5%B1%95.md#%E8%87%AA%E5%AE%9A%E4%B9%89%E6%95%B0%E6%8D%AE%E9%9B%86)的要求。更具体地,Agent的response/system应该符合上述的Action/Action Input/Observation格式。
168+
我们也支持使用自己的Agent数据集。数据集格式需要符合[自定义数据集](%E8%87%AA%E5%AE%9A%E4%B9%89%E4%B8%8E%E6%8B%93%E5%B1%95.md#%E8%87%AA%E5%AE%9A%E4%B9%89%E6%95%B0%E6%8D%AE%E9%9B%86)的要求。更具体地,Agent的response/system应该符合上述的Action/Action Input/Observation格式。
169169

170170
我们将**MLP****Embedder**加入了lora_target_modules. 你可以通过指定`--lora_target_modules ALL`在所有的linear层(包括qkvo以及mlp和embedder)加lora. 这**通常是效果最好的**.
171171

docs/source/LLM/LLM微调文档.md

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -37,7 +37,7 @@ pip install -r requirements/llm.txt -U
3737
```
3838

3939
## 微调
40-
如果你要使用界面的方式进行微调与推理, 可以查看[界面训练与推理文档](https://github.com/modelscope/swift/blob/main/docs/source/GetStarted/%E7%95%8C%E9%9D%A2%E8%AE%AD%E7%BB%83%E6%8E%A8%E7%90%86.md).
40+
如果你要使用界面的方式进行微调与推理, 可以查看[界面训练与推理文档](../GetStarted/%E7%95%8C%E9%9D%A2%E8%AE%AD%E7%BB%83%E6%8E%A8%E7%90%86.md).
4141

4242
### 使用python
4343
```python

docs/source/LLM/LLM量化文档.md

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -305,7 +305,7 @@ CUDA_VISIBLE_DEVICES=0 swift sft \
305305
```
306306

307307
**注意**
308-
- hqq支持更多自定义参数,比如为不同网络层指定不同量化配置,具体请见[命令行参数](https://github.com/modelscope/swift/blob/main/docs/source/LLM/命令行参数.md)
308+
- hqq支持更多自定义参数,比如为不同网络层指定不同量化配置,具体请见[命令行参数](命令行参数.md)
309309
- eetq量化为8bit量化,无需指定quantization_bit。目前不支持bf16,需要指定dtype为fp16
310310
- eetq目前qlora速度比较慢,推荐使用hqq。参考[issue](https://github.com/NetEase-FuXi/EETQ/issues/17)
311311

docs/source/LLM/index.md

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -5,7 +5,7 @@
55
1. [LLM推理文档](LLM推理文档.md)
66
2. [LLM微调文档](LLM微调文档.md)
77
3. [DPO训练文档](DPO训练文档.md)
8-
4. [界面训练与推理](https://github.com/modelscope/swift/blob/main/docs/source/GetStarted/%E7%95%8C%E9%9D%A2%E8%AE%AD%E7%BB%83%E6%8E%A8%E7%90%86.md)
8+
4. [界面训练与推理](../GetStarted/%E7%95%8C%E9%9D%A2%E8%AE%AD%E7%BB%83%E6%8E%A8%E7%90%86.md)
99
5. [LLM评测文档](LLM评测文档.md)
1010
6. [LLM量化文档](LLM量化文档.md)
1111
7. [VLLM推理加速与部署](VLLM推理加速与部署.md)

docs/source/LLM/自我认知微调最佳实践.md

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -89,7 +89,7 @@ My name is Qianwen, which is also known as Tongyi Qianwen. I am a large-scale la
8989
如果尝试了上述方法后仍然无法改善睡眠问题,建议咨询医生或睡眠专家,以排除潜在的健康问题。
9090
"""
9191
```
92-
如果你要进行单样本推理, 可以参考[LLM推理文档](https://github.com/modelscope/swift/blob/main/docs/source/LLM/LLM%E6%8E%A8%E7%90%86%E6%96%87%E6%A1%A3.md#qwen-7b-chat)
92+
如果你要进行单样本推理, 可以参考[LLM推理文档](LLM%E6%8E%A8%E7%90%86%E6%96%87%E6%A1%A3.md#qwen-7b-chat)
9393

9494
使用CLI:
9595
```bash

docs/source_en/LLM/LLM-quantization.md

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -218,7 +218,7 @@ CUDA_VISIBLE_DEVICES=0 swift sft \
218218
```
219219

220220
**Note**
221-
- hqq supports more customizable parameters, such as specifying different quantization configurations for different network layers. For details, please see [Command Line Arguments](https://github.com/modelscope/swift/blob/main/docs/source_en/LLM/Command-line-parameters.md).
221+
- hqq supports more customizable parameters, such as specifying different quantization configurations for different network layers. For details, please see [Command Line Arguments](Command-line-parameters.md).
222222
- eetq quantization uses 8-bit quantization, and there's no need to specify quantization_bit. Currently, bf16 is not supported; you need to specify dtype as fp16.
223223
- Currently, eetq's qlora speed is relatively slow; it is recommended to use hqq instead. For reference, see the [issue](https://github.com/NetEase-FuXi/EETQ/issues/17).
224224

examples/pytorch/llm/scripts/atom_7b_chat/lora/sft.sh

Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -8,7 +8,6 @@ swift sft \
88
--tuner_backend peft \
99
--dtype AUTO \
1010
--output_dir output \
11-
--ddp_backend nccl \
1211
--dataset ms-bench \
1312
--num_train_epochs 3 \
1413
--max_length 2048 \

examples/pytorch/llm/scripts/codeqwen1half_7b_chat/lora/sft.sh

Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -8,7 +8,6 @@ swift sft \
88
--tuner_backend peft \
99
--dtype AUTO \
1010
--output_dir output \
11-
--ddp_backend nccl \
1211
--dataset leetcode-python-en \
1312
--num_train_epochs 3 \
1413
--max_length 2048 \

examples/pytorch/llm/scripts/dbrx-instruct/lora_mp/sft.sh

Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -9,7 +9,6 @@ swift sft \
99
--template_type AUTO \
1010
--dtype bf16 \
1111
--output_dir output \
12-
--ddp_backend nccl \
1312
--dataset blossom-math-zh \
1413
--num_train_epochs 1 \
1514
--max_length 1024 \

examples/pytorch/llm/scripts/deepseek-v2-chat/lora_mp/sft.sh

Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -7,7 +7,6 @@ swift sft \
77
--tuner_backend peft \
88
--dtype bf16 \
99
--output_dir output \
10-
--ddp_backend nccl \
1110
--dataset alpaca-zh#5000 \
1211
--num_train_epochs 1 \
1312
--max_length 1024 \

0 commit comments

Comments
 (0)