biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[V1][Usage] Refactor speculative decoding configuration and tests (#14434)

Browse Source 
Signed-off-by: Shangming Cai <caishangming@linux.alibaba.com>
This commit is contained in:
shangmingc
2025-03-23 13:28:10 +08:00
committed by GitHub
parent 0661cfef7a
commit 50c9636d87
20 changed files with 1055 additions and 802 deletions
Download Patch File Download Diff File Expand all files Collapse all files

39
docs/source/features/spec_decode.md
View File

@@ -30,8 +30,10 @@ sampling_params = SamplingParams(temperature=0.8, top_p=0.95)
llm = LLM(
model="facebook/opt-6.7b",
tensor_parallel_size=1,
speculative_model="facebook/opt-125m",
num_speculative_tokens=5,
speculative_config={
"model": "facebook/opt-125m",
"num_speculative_tokens": 5,
},
)
outputs = llm.generate(prompts, sampling_params)
@@ -45,10 +47,14 @@ To perform the same with an online mode launch the server:
```bash
python -m vllm.entrypoints.openai.api_server --host 0.0.0.0 --port 8000 --model facebook/opt-6.7b \
--seed 42 -tp 1 --speculative_model facebook/opt-125m \
--num_speculative_tokens 5 --gpu_memory_utilization 0.8
--seed 42 -tp 1 --gpu_memory_utilization 0.8 \
--speculative_config '{"model": "facebook/opt-125m", "num_speculative_tokens": 5}'
```
:::{warning}
Note: Please use `--speculative_config` to set all configurations related to speculative decoding. The previous method of specifying the model through `--speculative_model` and adding related parameters (e.g., `--num_speculative_tokens`) separately will be deprecated in the next release.
:::
Then use a client:
```python
@@ -101,9 +107,11 @@ sampling_params = SamplingParams(temperature=0.8, top_p=0.95)
llm = LLM(
model="facebook/opt-6.7b",
tensor_parallel_size=1,
speculative_model="[ngram]",
num_speculative_tokens=5,
ngram_prompt_lookup_max=4,
speculative_config={
"method": "ngram",
"num_speculative_tokens": 5,
"prompt_lookup_max": 4,
},
)
outputs = llm.generate(prompts, sampling_params)
@@ -131,8 +139,10 @@ sampling_params = SamplingParams(temperature=0.8, top_p=0.95)
llm = LLM(
model="meta-llama/Meta-Llama-3.1-70B-Instruct",
tensor_parallel_size=4,
speculative_model="ibm-ai-platform/llama3-70b-accelerator",
speculative_draft_tensor_parallel_size=1,
speculative_config={
"model": "ibm-ai-platform/llama3-70b-accelerator",
"draft_tensor_parallel_size": 1,
},
)
outputs = llm.generate(prompts, sampling_params)
@@ -175,8 +185,10 @@ sampling_params = SamplingParams(temperature=0.8, top_p=0.95)
llm = LLM(
model="meta-llama/Meta-Llama-3-8B-Instruct",
tensor_parallel_size=4,
speculative_model="yuhuili/EAGLE-LLaMA3-Instruct-8B",
speculative_draft_tensor_parallel_size=1,
speculative_config={
"model": "yuhuili/EAGLE-LLaMA3-Instruct-8B",
"draft_tensor_parallel_size": 1,
},
)
outputs = llm.generate(prompts, sampling_params)
@@ -194,11 +206,10 @@ A few important things to consider when using the EAGLE based draft models:
be able to be loaded and used directly by vLLM after [PR 12304](https://github.com/vllm-project/vllm/pull/12304).
If you are using vllm version before [PR 12304](https://github.com/vllm-project/vllm/pull/12304), please use the
[script](https://gist.github.com/abhigoyal1997/1e7a4109ccb7704fbc67f625e86b2d6d) to convert the speculative model,
and specify `speculative_model="path/to/modified/eagle/model"`. If weight-loading problems still occur when using
the latest version of vLLM, please leave a comment or raise an issue.
and specify `"model": "path/to/modified/eagle/model"` in `speculative_config`. If weight-loading problems still occur when using the latest version of vLLM, please leave a comment or raise an issue.
2. The EAGLE based draft models need to be run without tensor parallelism
(i.e. speculative_draft_tensor_parallel_size is set to 1), although
(i.e. draft_tensor_parallel_size is set to 1 in `speculative_config`), although
it is possible to run the main model using tensor parallelism (see example above).
3. When using EAGLE-based speculators with vLLM, the observed speedup is lower than what is