Skip to content

GitLab

Commit 9986bee1 authored by raojy's avatar raojy 💬
Browse files

Update README.md

parent 9a37f010
Hide whitespace changes
Inline Side-by-side
Showing with 24 additions and 116 deletions
README.md
View file @ 9986bee1
......@@ -7,7 +7,7 @@
Qwen3.6-35B-A3B 是一款采用混合专家 (MoE) 架构并包含视觉编码器的多模态因果语言模型。总参数量为 35B，推理时激活参数量为 3B。此版本重点优化了智能体编程 (Agentic Coding) 的准确率，并引入了思维保留 (Thinking Preservation) 机制，特别适合长上下文和复杂的代码仓库开发任务。
<div align=center>
<img src="./doc/qwen3.5_397b_a17b_infra.jpg"/>
<img src="./doc/1.png"/>
</div>
## 环境依赖
......@@ -15,12 +15,12 @@ Qwen3.6-35B-A3B 是一款采用混合专家 (MoE) 架构并包含视觉编码器
| :------: |:-----------------------------------------:|
| DTK | 26.04 |
| python | 3.10.12 |
| transformers | 5.5.0 |
| vllm | 0.18.1+das.3266200.dtk2604 |
| triton | 3.4.0+git1ef59765 |
| torch | 2.10.0+das.opt1.dtk2604.20260325.g6b060a |
| transformers | 5.2.0 |
| vllm | 0.15.1+das.opt1.alpha.dtk2604.torch290.2604081832.gbcb2ba |
| triton | 3.3.0+das.opt2.dtk2604.torch290.20260331.g31542e |
| torch | 2.9.0+das.opt1.dtk2604.20260331.g4e3c1e7 |
当前仅支持定制镜像: harbor.sourcefind.cn:5443/dcu/admin/base/custom:vllm018-ubuntu22.04-dtk26.04-gemma4-0413
当前推荐使用镜像: harbor.sourcefind.cn:5443/dcu/admin/base/custom:vllm015-ubuntu22.04-dtk26.04-0409-modelzoo
- 挂载地址`-v` 根据实际模型情况修改
```bash
......@@ -38,7 +38,7 @@ docker run -it \
-u root \
-v /opt/hyhal/:/opt/hyhal/:ro \
-v /path/your_code_data/:/path/your_code_data/ \
harbor.sourcefind.cn:5443/dcu/admin/base/custom:vllm018-ubuntu22.04-dtk26.04-gemma4-0413 bash
harbor.sourcefind.cn:5443/dcu/admin/base/custom:vllm015-ubuntu22.04-dtk26.04-0409-modelzoo bash
```
更多镜像可前往[光源](https://sourcefind.cn/#/service-list)下载使用。
......@@ -51,119 +51,33 @@ docker run -it \
暂无
## 推理
> 如果出现`ImportError: librocm_smi64.so.2: cannot open shaned object file: No such file or directory`报错，系机器hyhal版本较低所致，请进行升级。
### vllm
**mtp**功能需添加以下参数：
```
--speculative-config.method mtp \
--speculative-config.num_speculative_tokens 2
```
#### 单机推理
```bash
## serve启动
vllm serve Qwen/Qwen3.5-35B-A3B \
--port 8001 \
--trust-remote-code \
--dtype bfloat16 \
--tensor-parallel-size 2 \
--gpu-memory-utilization 0.925 \
--default-chat-template-kwargs '{"enable_thinking": false}' \
--reasoning-parser qwen3 \
--enable-auto-tool-choice \
--tool-call-parser qwen3_coder
## client访问
curl http://localhost:8001/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
"model": "Qwen/Qwen3.5-35B-A3B",
"messages": [
{"role": "user", "content": "Type \"I love Qwen3.5\" backwards"}
],
"temperature": 0.6
}'
```
#### 多机推理
1. 加入环境变量
> 请注意：
> 每个节点上的环境变量都写到.sh文件中，保存后各个计算节点分别source`.sh`文件
>
> VLLM_HOST_IP：节点本地通信口ip，尽量选择IB网卡的IP，**避免出现rccl超时问题**
>
> NCCL_SOCKET_IFNAME和 GLOO_SOCKET_IFNAME：节点本地通信网口ip对应的名称
>
> 通信口和ip查询方法：ifconfig
>
> IB口状态查询：ibstat !!!一定要active激活状态才可用，各个节点要保持统一
```bash
export ALLREDUCE_STREAM_WITH_COMPUTE=1
export VLLM_HOST_IP=x.x.x.x # 对应计算节点的IP，选择IB口SOCKET_IFNAME对应IP地址
export NCCL_SOCKET_IFNAME=ibxxxx
export GLOO_SOCKET_IFNAME=ibxxxx
export NCCL_IB_HCA=mlx5_x:1 # 环境中的IB网卡名字
unset NCCL_ALGO
export NCCL_MIN_NCHANNELS=16
export NCCL_MAX_NCHANNELS=16
export NCCL_NET_GDR_READ=1
export HIP_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
export VLLM_RPC_TIMEOUT=1800000
# 海光CPU绑定核
export VLLM_NUMA_BIND=1
export VLLM_RANK0_NUMA=0
export VLLM_RANK1_NUMA=1
export VLLM_RANK2_NUMA=2
export VLLM_RANK3_NUMA=3
export VLLM_RANK4_NUMA=4
export VLLM_RANK5_NUMA=5
export VLLM_RANK6_NUMA=6
export VLLM_RANK7_NUMA=7
```
2. 启动RAY集群
> x.x.x.x 对应第一步 VLLM_HOST_IP
```bash
# head节点执行
ray start --head --node-ip-address=x.x.x.x --port=6379 --num-gpus=8 --num-cpus=32
# worker节点执行
ray start --address='x.x.x.x:6379' --num-gpus=8 --num-cpus=32
```
3. 启动vllm server
```bash
## serve启动
vllm serve Qwen/Qwen3.5-397B-A17B \
vllm serve Qwen/Qwen3.6-35B-A3B \
--port 8001 \
--tensor-parallel-size 16 \
--distributed-executor-backend ray \
--trust-remote-code \
--dtype bfloat16 \
--gpu-memory-utilization 0.925 \
--default-chat-template-kwargs '{"enable_thinking": false}' \
--reasoning-parser qwen3 \
--enable-auto-tool-choice \
--tool-call-parser qwen3_coder
--tensor-parallel-size 4 \
--gpu-memory-utilization 0.925
## client访问
curl http://localhost:8001/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
"model": "Qwen/Qwen3.5-397B-A17B",
"messages": [
{"role": "user", "content": "Type \"I love Qwen3.5\" backwards"}
],
"temperature": 0.6
}'
curl -X POST "http://localhost:8001/v1/chat/completions" -H "Content-Type: application/json" -d '{
"model": "Qwen/Qwen3.6-35B-A3B",
"messages": [
{"role": "system", "content": "你是一个有用的助手。"},
{"role": "user", "content": "你好，请做一下简单的自我介绍。"}
],
"stream": false
}'
```
## 效果展示
<div align=center>
<img src="./doc/result-dcu.jpg"/>
<img src="./doc/2.png"/>
</div>
### 精度
......@@ -172,18 +86,12 @@ DCU与GPU精度一致，推理框架：vllm。
## 预训练权重
| 模型名称 | 权重大小 | DCU型号 | 最低卡数需求 | 下载地址 |
|:------:|:----:|:----------:|:------:|:---------------------:|
| Qwen3.5-397B-A17B | 397B | K100AI,BW1000 | 16 | [Hugging Face](https://huggingface.co/Qwen/Qwen3.5-397B-A17B) |
| Qwen3.5-397B-A17B-INT8 | 397B | BW1000 | 8 | [Modelscope](https://www.modelscope.cn/models/metax-tech/Qwen3.5-397B-A17B-W8A8) |
| Qwen3.5-122B-A10B | 122B | K100AI,BW1000 | 8 | [Hugging Face](https://huggingface.co/Qwen/Qwen3.5-122B-A10B) |
| Qwen3.5-35B-A3B | 35B | K100AI,BW1000 | 2 | [Hugging Face](https://huggingface.co/Qwen/Qwen3.5-35B-A3B) |
| Qwen3.5-27B | 27B | K100AI,BW1000 | 2 | [Hugging Face](https://huggingface.co/Qwen/Qwen3.5-27B) |
| Qwen3.5-9B | 9B | K100AI,BW1000 | 1 | [Hugging Face](https://huggingface.co/Qwen/Qwen3.5-9B) |
| Qwen3.5-4B | 4B | K100AI,BW1000 | 1 | [Hugging Face](https://huggingface.co/Qwen/Qwen3.5-4B) |
| Qwen3.5-2B | 2B | K100AI,BW1000 | 1 | [Hugging Face](https://huggingface.co/Qwen/Qwen3.5-2B) |
| Qwen3.5-0.8B | 0.8B | K100AI,BW1000 | 1 | [Hugging Face](https://huggingface.co/Qwen/Qwen3.5-0.8B) |
| Qwen3.6-35B-A3B | 397B | K100AI,BW1000 | 4 | [Hugging Face](https://huggingface.co/Qwen/Qwen3.6-35B-A3B) |
| Qwen3.6-35B-A3B-FP8 | 397B | BW1000 | 2 | [Hugging Face](https://huggingface.co/Qwen/Qwen3.6-35B-A3B-FP8) |
## 源码仓库及问题反馈
- https://developer.sourcefind.cn/codes/modelzoo/qwen3.5_vllm
- https://developer.sourcefind.cn/codes/modelzoo/qwen3.6_vllm
## 参考资料
- https://github.com/QwenLM/Qwen3.5
- https://github.com/QwenLM/Qwen3.6
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment