vllm
vllm copied to clipboard
vllm-project
[Frontend] Support OpenAI batch file format
This PR adds basic support for the OpenAI batch file format (for chat endpoints).
The following help page summarizes the usage options:
# python -m vllm.entrypoints.openai.run_batch --help
usage: run_batch.py [-h] -i INPUT_FILE -o OUTPUT_FILE [--response-role RESPONSE_ROLE] [--model MODEL] [--tokenizer TOKENIZER] [--skip-tokenizer-init] [--revision REVISION]
[--code-revision CODE_REVISION] [--tokenizer-revision TOKENIZER_REVISION] [--tokenizer-mode {auto,slow}] [--trust-remote-code] [--download-dir DOWNLOAD_DIR]
[--load-format {auto,pt,safetensors,npcache,dummy,tensorizer}] [--dtype {auto,half,float16,bfloat16,float,float32}] [--kv-cache-dtype {auto,fp8}]
[--quantization-param-path QUANTIZATION_PARAM_PATH] [--max-model-len MAX_MODEL_LEN] [--guided-decoding-backend {outlines,lm-format-enforcer}] [--worker-use-ray]
[--pipeline-parallel-size PIPELINE_PARALLEL_SIZE] [--tensor-parallel-size TENSOR_PARALLEL_SIZE] [--max-parallel-loading-workers MAX_PARALLEL_LOADING_WORKERS]
[--ray-workers-use-nsight] [--block-size {8,16,32}] [--enable-prefix-caching] [--use-v2-block-manager] [--num-lookahead-slots NUM_LOOKAHEAD_SLOTS] [--seed SEED]
[--swap-space SWAP_SPACE] [--gpu-memory-utilization GPU_MEMORY_UTILIZATION] [--num-gpu-blocks-override NUM_GPU_BLOCKS_OVERRIDE]
[--max-num-batched-tokens MAX_NUM_BATCHED_TOKENS] [--max-num-seqs MAX_NUM_SEQS] [--max-logprobs MAX_LOGPROBS] [--disable-log-stats]
[--quantization {aqlm,awq,fp8,gptq,squeezellm,gptq_marlin,marlin,deepspeedfp,None}] [--enforce-eager] [--max-context-len-to-capture MAX_CONTEXT_LEN_TO_CAPTURE]
[--max-seq_len-to-capture MAX_SEQ_LEN_TO_CAPTURE] [--disable-custom-all-reduce] [--tokenizer-pool-size TOKENIZER_POOL_SIZE]
[--tokenizer-pool-type TOKENIZER_POOL_TYPE] [--tokenizer-pool-extra-config TOKENIZER_POOL_EXTRA_CONFIG] [--enable-lora] [--max-loras MAX_LORAS]
[--max-lora-rank MAX_LORA_RANK] [--lora-extra-vocab-size LORA_EXTRA_VOCAB_SIZE] [--lora-dtype {auto,float16,bfloat16,float32}] [--max-cpu-loras MAX_CPU_LORAS]
[--fully-sharded-loras] [--device {auto,cuda,neuron,cpu}] [--image-input-type {pixel_values,image_features}] [--image-token-id IMAGE_TOKEN_ID]
[--image-input-shape IMAGE_INPUT_SHAPE] [--image-feature-size IMAGE_FEATURE_SIZE] [--scheduler-delay-factor SCHEDULER_DELAY_FACTOR] [--enable-chunked-prefill]
[--speculative-model SPECULATIVE_MODEL] [--num-speculative-tokens NUM_SPECULATIVE_TOKENS] [--speculative-max-model-len SPECULATIVE_MAX_MODEL_LEN]
[--speculative-disable-by-batch-size SPECULATIVE_DISABLE_BY_BATCH_SIZE] [--ngram-prompt-lookup-max NGRAM_PROMPT_LOOKUP_MAX]
[--ngram-prompt-lookup-min NGRAM_PROMPT_LOOKUP_MIN] [--model-loader-extra-config MODEL_LOADER_EXTRA_CONFIG]
[--served-model-name SERVED_MODEL_NAME [SERVED_MODEL_NAME ...]] [--engine-use-ray] [--disable-log-requests] [--max-log-len MAX_LOG_LEN]
vLLM OpenAI-Compatible batch runner.
options:
-h, --help show this help message and exit
-i INPUT_FILE, --input-file INPUT_FILE
The path or url to a single input file. Currently supports local file paths, or the http protocol (http or https). If a URL is specified, the file should be
available via HTTP GET.
-o OUTPUT_FILE, --output-file OUTPUT_FILE
The path or url to a single output file. Currently supports local file paths, or web (http or https) urls. If a URL is specified, the file should be
available via HTTP PUT.
--response-role RESPONSE_ROLE
The role name to return if `request.add_generation_prompt=true`.
--model MODEL Name or path of the huggingface model to use.
...
Assuming the following file exists:
# cat ../batch.jsonl
{"custom_id": "request-1", "method": "POST", "url": "/v1/chat/completions", "body": {"model": "meta-llama/Meta-Llama-3-8B-Instruct", "messages": [{"role": "system", "content": "You are a helpful assistant."},{"role": "user", "content": "Hello world!"}],"max_tokens": 1000}}
{"custom_id": "request-2", "method": "POST", "url": "/v1/chat/completions", "body": {"model": "meta-llama/Meta-Llama-3-8B-Instruct", "messages": [{"role": "system", "content": "You are an unhelpful assistant."},{"role": "user", "content": "Hello world!"}],"max_tokens": 1000}}
The program is run as python -m vllm.entrypoints.openai.run_batch -i ../batch.jsonl -o ../results.jsonl --model meta-llama/Meta-Llama-3-8B-Instruct (this produces the typical logging from the async engine, which I've omitted).
Finally, the resulting file produced is:
# cat ../results.jsonl
{"id":"vllm-383d1c59835645aeb2e07d004d62a826","custom_id":"request-1","response":{"id":"cmpl-61c020e54b964d5a98fa7527bfcdd378","object":"chat.completion","created":1715633336,"model":"meta-llama/Meta-Llama-3-8B-Instruct","choices":[{"index":0,"message":{"role":"assistant","content":"Hello! It's great to meet you! I'm here to help with any questions or tasks you may have. What's on your mind today?"},"logprobs":null,"finish_reason":"stop","stop_reason":null}],"usage":{"prompt_tokens":25,"total_tokens":56,"completion_tokens":31}},"error":null}
{"id":"vllm-42e3d09b14b04568afa3f1797751a267","custom_id":"request-2","response":{"id":"cmpl-f44d049f6b3a42d4b2d7850bb1e31bcc","object":"chat.completion","created":1715633336,"model":"meta-llama/Meta-Llama-3-8B-Instruct","choices":[{"index":0,"message":{"role":"assistant","content":"*silence*"},"logprobs":null,"finish_reason":"stop","stop_reason":null}],"usage":{"prompt_tokens":27,"total_tokens":32,"completion_tokens":5}},"error":null}
FIX #4777
BEFORE SUBMITTING, PLEASE READ THE CHECKLIST BELOW AND FILL IN THE DESCRIPTION ABOVE
PR Checklist (Click to Expand)
Thank you for your contribution to vLLM! Before submitting the pull request, please ensure the PR meets the following criteria. This helps vLLM maintain the code quality and improve the efficiency of the review process.
PR Title and Classification
Only specific types of PRs will be reviewed. The PR title is prefixed appropriately to indicate the type of change. Please use one of the following:
[Bugfix]for bug fixes.[CI/Build]for build or continuous integration improvements.[Doc]for documentation fixes and improvements.[Model]for adding a new model or improving an existing model. Model name should appear in the title.[Frontend]For changes on the vLLM frontend (e.g., OpenAI API server,LLMclass, etc.)[Kernel]for changes affecting CUDA kernels or other compute kernels.[Core]for changes in the core vLLM logic (e.g.,LLMEngine,AsyncLLMEngine,Scheduler, etc.)[Hardware][Vendor]for hardware-specific changes. Vendor name should appear in the prefix (e.g.,[Hardware][AMD]).[Misc]for PRs that do not fit the above categories. Please use this sparingly.
Note: If the PR spans more than one category, please include all relevant prefixes.
Code Quality
The PR need to meet the following code quality standards:
- We adhere to Google Python style guide and Google C++ style guide.
- Pass all linter checks. Please use
format.shto format your code. - The code need to be well-documented to ensure future contributors can easily understand the code.
- Include sufficient tests to ensure the project to stay correct and robust. This includes both unit tests and integration tests.
- Please add documentation to
docs/source/if the PR modifies the user-facing behaviors of vLLM. It helps vLLM user understand and utilize the new features or changes.
Notes for Large Changes
Please keep the changes as concise as possible. For major architectural changes (>500 LOC excluding kernel/data/config/test), we would expect a GitHub issue (RFC) discussing the technical design and justification. Otherwise, we will tag it with rfc-required and might not go through the PR.
What to Expect for the Reviews
The goal of the vLLM team is to be a transparent reviewing machine. We would like to make the review process transparent and efficient and make sure no contributor feel confused or frustrated. However, the vLLM team is small, so we need to prioritize some PRs over others. Here is what you can expect from the review process:
- After the PR is submitted, the PR will be assigned to a reviewer. Every reviewer will pick up the PRs based on their expertise and availability.
- After the PR is assigned, the reviewer will provide status update every 2-3 days. If the PR is not reviewed within 7 days, please feel free to ping the reviewer or the vLLM team.
- After the review, the reviewer will put an
action-requiredlabel on the PR if there are changes required. The contributor should address the comments and ping the reviewer to re-review the PR. - Please respond to all comments within a reasonable time frame. If a comment isn't clear or you disagree with a suggestion, feel free to ask for clarification or discuss the suggestion.
Thank You
Finally, thank you for taking the time to read these guidelines and for your interest in contributing to vLLM. Your contributions make vLLM a great tool for everyone!
My only concern is that this is not really an OpenAI-compatible API - but rather an OpenAI-compatible file format. I think this is fine, but perhaps we should make this explicit in the documentation ...
Yes I agree, I tried to use the wording "batch file format" everywhere but when writing the documentation I can be even more explicit. I alluded to it in the ticket (and I think you noticed it too on slack), but the actual openai api is tightly coupled to its file concept and opens up a whole can of worms around life cycle management/long lived state.
See comments in code - my biggest feedback is around the is_aborted - I would rather this PR not touch that function signature.
Updated to maintain backwards compatibility in the signature, but let me know if you have something specific in mind.
We should add embedding models as a follow up
Yup, agreed
This looks great.
Can you just add a short example of run_batch.py in /examples
