[0.9.1]support deepseek w4a8 quantization #1320
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Signed-off-by: pichangping <1337510399@qq.com>
Signed-off-by: pichangping <1337510399@qq.com>
Signed-off-by: pichangping <1337510399@qq.com>
Signed-off-by: pichangping <1337510399@qq.com>
wangxiyuan
changed the title
29 tasks
wangxiyuan
approved these changes
Jun 20, 2025
| def load_weights(self, weights: Iterable[tuple[str, | ||
| torch.Tensor]]) -> set[str]: | ||
| weights = filter(lambda x: ".module." not in x[0], weights) | ||
| # weights = ((name, data) for name, data in weights if ".module." not in name) |
Collaborator
There was a problem hiding this comment.
remove the comment code
Collaborator
|
Please refactor w8w8 and w4a8 code together in the future to make the code clean. Thanks |
NNUCJ
pushed a commit
to NNUCJ/vllm-ascend
that referenced
this pull request
Jun 23, 2025
### What this PR does / why we need it? Supports Deepseek-R1 w4a8 quantization. Since R1 w4a8 uses mixed quantization, only the MOE layer uses w4a8_dynamic quantization, so we added the w4a8_dynamic.py file, which includes the AscendW4A8DynamicFusedMoEMethod class. ### Does this PR introduce _any_ user-facing change? no, use `--quantization=ascend` is engouh. ### How was this patch tested? #### 1.How to get weights using Modelslim ##### Installation steps Use the branch master, the commit id is: 298e175d69b3b855111a1e09bbe2fcd12fdb4e24 git clone https://gitee.com/ascend/msit.git cd msit/msmodelslim bash install.sh ##### The required transformers environment pip install transformers==4.48.2 ##### Generate w4a8 weights cd /example/DeepSeek Command reference: msmodelslim/example/DeepSeek/README.md Execute the [pre-check](https://gitee.com/ascend/msit/blob/master/msmodelslim/example/DeepSeek/README.md#运行前必检) and [DeepSeek-R1 w4a8 mix quantization](https://gitee.com/ascend/msit/blob/master/msmodelslim/example/DeepSeek/README.md#deepseek-r1-w4a8-混合量化前三层-mlpw8a8-dynamic-量化mla共享专家w8a8量化路由专家w4a8-dynamic量化) chapter Reference command：python3 quant_deepseek_w4a8.py --model_path {Original weight path} --save_path {Generate weight path} --mindie_format ##### Adapt to vllm-ascend Since mindie_format generates mindie format, some adaptation modifications are needed for vllm-ascend to use it: `quant_model_description_w8a8_dynamic.json` rename to `quant_model_description.json`, and change `"group_size": 0` to `"group_size": 256` Modification in `config.json`：`"model_type":deepseekv2` is changed to `"model_type":deepseek_v3` ; `quantization_config` is removed; #### 2.How to run w4a8 TP + EP： python -m vllm.entrypoints.openai.api_server --model=$1 --trust-remote-code -tp $2 --enable_expert_parallel --quantization ascend --port $3 --max-model-len $4 --max-num-seqs $5 --enforce-eager eg: python -m vllm.entrypoints.openai.api_server --model=/weightpath/w4a8_4_layer --trust-remote-code -tp 4 --enable_expert_parallel --quantization ascend --port 8002 --max-model-len 2048 --max-num-seqs 128 --enforce-eager DP+TP+EP: python -m vllm.entrypoints.openai.api_server --model=$1 --trust-remote-code -tp $2 -dp $3 --enable_expert_parallel --quantization ascend --port $4 --max-model-len $5 --max-num-seqs $6 --enforce-eager eg: python -m vllm.entrypoints.openai.api_server --model=/weightpath/w4a8_4_layer --trust-remote-code -tp 2 -dp2 --enable_expert_parallel --quantization ascend --port 8002 --max-model-len 2048 --max-num-seqs 128 --enforce-eager #### 3.Use constraints export VLLM_USE_V1=1 # v1 --------- Signed-off-by: pichangping <1337510399@qq.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What this PR does / why we need it?
Supports Deepseek-R1 w4a8 quantization.
Since R1 w4a8 uses mixed quantization, only the MOE layer uses w4a8_dynamic quantization, so we added the w4a8_dynamic.py file, which includes the AscendW4A8DynamicFusedMoEMethod class.
Does this PR introduce any user-facing change?
no, use
--quantization=ascendis engouh.How was this patch tested?
1.How to get weights using Modelslim
Installation steps
Use the branch master, the commit id is: 298e175d69b3b855111a1e09bbe2fcd12fdb4e24
git clone https://gitee.com/ascend/msit.git
cd msit/msmodelslim
bash install.sh
The required transformers environment
pip install transformers==4.48.2
Generate w4a8 weights
cd /example/DeepSeek
Command reference: msmodelslim/example/DeepSeek/README.md Execute the pre-check and DeepSeek-R1 w4a8 mix quantization chapter
Reference command：python3 quant_deepseek_w4a8.py --model_path {Original weight path} --save_path {Generate weight path} --mindie_format
Adapt to vllm-ascend
Since mindie_format generates mindie format, some adaptation modifications are needed for vllm-ascend to use it:
quant_model_description_w8a8_dynamic.jsonrename toquant_model_description.json, and change"group_size": 0to"group_size": 256Modification in
config.json："model_type":deepseekv2is changed to"model_type":deepseek_v3;quantization_configis removed;2.How to run w4a8
TP + EP：
python -m vllm.entrypoints.openai.api_server --model=$1 --trust-remote-code -tp $2 --enable_expert_parallel --quantization ascend --port $3 --max-model-len $4 --max-num-seqs $5 --enforce-eager
eg: python -m vllm.entrypoints.openai.api_server --model=/weightpath/w4a8_4_layer --trust-remote-code -tp 4 --enable_expert_parallel --quantization ascend --port 8002 --max-model-len 2048 --max-num-seqs 128 --enforce-eager
DP+TP+EP:
python -m vllm.entrypoints.openai.api_server --model=$1 --trust-remote-code -tp $2 -dp $3 --enable_expert_parallel --quantization ascend --port $4 --max-model-len $5 --max-num-seqs $6 --enforce-eager
eg: python -m vllm.entrypoints.openai.api_server --model=/weightpath/w4a8_4_layer --trust-remote-code -tp 2 -dp2 --enable_expert_parallel --quantization ascend --port 8002 --max-model-len 2048 --max-num-seqs 128 --enforce-eager
3.Use constraints
export VLLM_USE_V1=1 # v1