Merge pull request #125 from runpod-workers/up-0.6.3

update vllm

.github/workflows/CI-runpod_dep.yml

@@ -19,32 +19,49 @@ jobs:
 
       - name: Check for new package version and update
         run: |
-          # Get current version
-          current_version=$(grep -oP 'runpod==\K[^"]+' ./builder/requirements.txt)
+          echo "Fetching the current runpod version from requirements.txt..."
           
-          # Get new version
+          # Get current version, allowing both == and ~= in the search pattern
+          current_version=$(grep -oP 'runpod[~=]{1,2}\K[^"]+' ./builder/requirements.txt)
+          echo "Current version: $current_version"
+
+          # Extract major and minor from current version
+          current_major_minor=$(echo $current_version | cut -d. -f1,2)
+          echo "Current major.minor: $current_major_minor"
+
+          echo "Fetching the latest runpod version from PyPI..."
+          
+          # Get new version from PyPI
           new_version=$(curl -s https://pypi.org/pypi/runpod/json | jq -r .info.version)
           echo "NEW_VERSION_ENV=$new_version" >> $GITHUB_ENV
+          echo "New version: $new_version"
+
+          # Extract major and minor from new version
+          new_major_minor=$(echo $new_version | cut -d. -f1,2)
+          echo "New major.minor: $new_major_minor"
 
           if [ -z "$new_version" ]; then
-              echo "Failed to fetch the new version."
+              echo "ERROR: Failed to fetch the new version from PyPI."
               exit 1
           fi
 
-          # Check if the version is already up-to-date
-          if [ "$current_version" = "$new_version" ]; then
-              echo "The package version is already up-to-date."
+          # Check if the major or minor version is different
+          if [ "$current_major_minor" = "$new_major_minor" ]; then
+              echo "No update needed. The new version ($new_major_minor) is within the allowed range (~= $current_major_minor)."
               exit 0
           fi
 
-          # Update requirements.txt
-          sed -i "s/runpod==.*/runpod==$new_version/" ./builder/requirements.txt
+          echo "New major/minor detected ($new_major_minor). Updating requirements.txt..."
+
+          # Update requirements.txt, preserving the existing constraint type (~= or ==)
+          sed -i "s/runpod[~=][^ ]*/runpod~=$new_version/" ./builder/requirements.txt
+          echo "requirements.txt has been updated."
 
       - name: Create Pull Request
         uses: peter-evans/create-pull-request@v3
         with:
           token: ${{ secrets.GITHUB_TOKEN }}
-          commit-message: Update package version
+          commit-message: Update runpod package version
           title: Update runpod package version
           body: The package version has been updated to ${{ env.NEW_VERSION_ENV }}
           branch: runpod-package-update

Dockerfile

@@ -12,7 +12,7 @@ RUN --mount=type=cache,target=/root/.cache/pip \
     python3 -m pip install --upgrade -r /requirements.txt
 
 # Install vLLM (switching back to pip installs since issues that required building fork are fixed and space optimization is not as important since caching) and FlashInfer 
-RUN python3 -m pip install vllm==0.5.5 && \
+RUN python3 -m pip install vllm==0.6.3 && \
     python3 -m pip install flashinfer -i https://flashinfer.ai/whl/cu121/torch2.3
 
 # Setup for Option 2: Building the Image with the Model included

README.md

@@ -18,9 +18,9 @@ Deploy OpenAI-Compatible Blazing-Fast LLM Endpoints powered by the [vLLM](https:
 ### 1. UI for Deploying vLLM Worker on RunPod console:
 ![Demo of Deploying vLLM Worker on RunPod console with new UI](media/ui_demo.gif)
 
-### 2. Worker vLLM `v1.3.1` with vLLM `0.5.5` now available under `stable` tags 
+### 2. Worker vLLM `v1.5.0` with vLLM `0.6.2` now available under `stable` tags 
 
-Update v1.3.1 is now available, use the image tag `runpod/worker-v1-vllm:v1.3.1stable-cuda12.1.0`.
+Update v1.5.0 is now available, use the image tag `runpod/worker-v1-vllm:v1.5.0stable-cuda12.1.0`.
 
 ### 3. OpenAI-Compatible [Embedding Worker](https://github.com/runpod-workers/worker-infinity-embedding) Released
 Deploy your own OpenAI-compatible Serverless Endpoint on RunPod with multiple embedding models and fast inference for RAG and more! 
@@ -82,7 +82,7 @@ Below is a summary of the available RunPod Worker images, categorized by image s
 
 | CUDA Version | Stable Image Tag                  | Development Image Tag             | Note                                                        |
 |--------------|-----------------------------------|-----------------------------------|----------------------------------------------------------------------|
-| 12.1.0       | `runpod/worker-v1-vllm:v1.3.1stable-cuda12.1.0` | `runpod/worker-v1-vllm:v1.3.1dev-cuda12.1.0` | When creating an Endpoint, select CUDA Version 12.3, 12.2 and 12.1 in the filter. |
+| 12.1.0       | `runpod/worker-v1-vllm:v1.5.0stable-cuda12.1.0` | `runpod/worker-v1-vllm:v1.5.0dev-cuda12.1.0` | When creating an Endpoint, select CUDA Version 12.3, 12.2 and 12.1 in the filter. |
 
 
 

builder/requirements.txt

@@ -1,10 +1,11 @@
 ray
 pandas
 pyarrow
-runpod==1.7.0
+runpod~=1.7.0
 huggingface-hub
 packaging
 typing-extensions==4.7.1
 pydantic
 pydantic-settings
 hf-transfer
+transformers

src/engine.py

@@ -11,6 +11,7 @@ from vllm import AsyncLLMEngine
 from vllm.entrypoints.openai.serving_chat import OpenAIServingChat
 from vllm.entrypoints.openai.serving_completion import OpenAIServingCompletion
 from vllm.entrypoints.openai.protocol import ChatCompletionRequest, CompletionRequest, ErrorResponse
+from vllm.entrypoints.openai.serving_engine import BaseModelPath
 
 from utils import DummyRequest, JobInput, BatchSize, create_error_response
 from constants import DEFAULT_MAX_CONCURRENCY, DEFAULT_BATCH_SIZE, DEFAULT_BATCH_SIZE_GROWTH_FACTOR, DEFAULT_MIN_BATCH_SIZE
@@ -124,11 +125,13 @@ class OpenAIvLLMEngine(vLLMEngine):
         
     async def _initialize_engines(self):
         self.model_config = await self.llm.get_model_config()
-        
+        self.base_model_paths = [
+            BaseModelPath(name=self.engine_args.model, model_path=self.engine_args.model)
+        ]
         self.chat_engine = OpenAIServingChat(
-            async_engine_client=self.llm, 
+            engine_client=self.llm, 
             model_config=self.model_config,
-            served_model_names=[self.served_model_name], 
+            base_model_paths=self.base_model_paths,
             response_role=self.response_role,
             chat_template=self.tokenizer.tokenizer.chat_template,
             lora_modules=None,
@@ -136,9 +139,9 @@ class OpenAIvLLMEngine(vLLMEngine):
             request_logger=None
         )
         self.completion_engine = OpenAIServingCompletion(
-            async_engine_client=self.llm, 
+            engine_client=self.llm, 
             model_config=self.model_config,
-            served_model_names=[self.served_model_name],
+            base_model_paths=self.base_model_paths,
             lora_modules=[],
             prompt_adapters=None,
             request_logger=None

src/engine_args.py

@@ -88,7 +88,8 @@ DEFAULT_ARGS = {
     "typical_acceptance_sampler_posterior_alpha": float(os.getenv('TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_ALPHA', 0)) or None,
     "qlora_adapter_name_or_path": os.getenv('QLORA_ADAPTER_NAME_OR_PATH', None),
     "disable_logprobs_during_spec_decoding": os.getenv('DISABLE_LOGPROBS_DURING_SPEC_DECODING', None),
-    "otlp_traces_endpoint": os.getenv('OTLP_TRACES_ENDPOINT', None)
+    "otlp_traces_endpoint": os.getenv('OTLP_TRACES_ENDPOINT', None),
+    "use_v2_block_manager": os.getenv('USE_V2_BLOCK_MANAGER', 'true')
 }
 
 def match_vllm_args(args):

src/utils.py

@@ -3,6 +3,7 @@ import logging
 from http import HTTPStatus
 from functools import wraps
 from time import time
+from vllm.entrypoints.openai.protocol import RequestResponseMetadata
 
 try:
     from vllm.utils import random_uuid
@@ -47,10 +48,14 @@ class JobInput:
         self.min_batch_size = int(min_batch_size) if min_batch_size else None 
         self.openai_route = job.get("openai_route")
         self.openai_input = job.get("openai_input")
+class DummyState:
+    def __init__(self):
+        self.request_metadata = None
         
 class DummyRequest:
     def __init__(self):
         self.headers = {}
+        self.state = DummyState()
     async def is_disconnected(self):
         return False
 

worker-config.json

@@ -1,5 +1,125 @@
 {
     "versions": {
+      "0.6.2": {
+        "imageName": "runpod/worker-v1-vllm:v1.5.0stable-cuda12.1.0",
+        "minimumCudaVersion": "12.1",
+        "categories": [
+          {
+            "title": "LLM Settings",
+            "settings": [
+              "TOKENIZER", "TOKENIZER_MODE", "SKIP_TOKENIZER_INIT", "TRUST_REMOTE_CODE",
+              "DOWNLOAD_DIR", "LOAD_FORMAT", "DTYPE", "KV_CACHE_DTYPE", "QUANTIZATION_PARAM_PATH",
+              "MAX_MODEL_LEN", "GUIDED_DECODING_BACKEND", "DISTRIBUTED_EXECUTOR_BACKEND",
+              "WORKER_USE_RAY", "RAY_WORKERS_USE_NSIGHT", "PIPELINE_PARALLEL_SIZE",
+              "TENSOR_PARALLEL_SIZE", "MAX_PARALLEL_LOADING_WORKERS", "ENABLE_PREFIX_CACHING",
+              "DISABLE_SLIDING_WINDOW", "USE_V2_BLOCK_MANAGER", "NUM_LOOKAHEAD_SLOTS",
+              "SEED", "NUM_GPU_BLOCKS_OVERRIDE", "MAX_NUM_BATCHED_TOKENS", "MAX_NUM_SEQS",
+              "MAX_LOGPROBS", "DISABLE_LOG_STATS", "QUANTIZATION", "ROPE_SCALING", "ROPE_THETA",
+              "TOKENIZER_POOL_SIZE", "TOKENIZER_POOL_TYPE", "TOKENIZER_POOL_EXTRA_CONFIG",
+              "ENABLE_LORA", "MAX_LORAS", "MAX_LORA_RANK", "LORA_EXTRA_VOCAB_SIZE",
+              "LORA_DTYPE", "LONG_LORA_SCALING_FACTORS", "MAX_CPU_LORAS", "FULLY_SHARDED_LORAS",
+              "DEVICE", "SCHEDULER_DELAY_FACTOR", "ENABLE_CHUNKED_PREFILL", "SPECULATIVE_MODEL",
+              "NUM_SPECULATIVE_TOKENS", "SPECULATIVE_DRAFT_TENSOR_PARALLEL_SIZE",
+              "SPECULATIVE_MAX_MODEL_LEN", "SPECULATIVE_DISABLE_BY_BATCH_SIZE",
+              "NGRAM_PROMPT_LOOKUP_MAX", "NGRAM_PROMPT_LOOKUP_MIN", "SPEC_DECODING_ACCEPTANCE_METHOD",
+              "TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_THRESHOLD", "TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_ALPHA",
+              "MODEL_LOADER_EXTRA_CONFIG", "PREEMPTION_MODE", "PREEMPTION_CHECK_PERIOD",
+              "PREEMPTION_CPU_CAPACITY", "MAX_LOG_LEN", "DISABLE_LOGGING_REQUEST"
+            ]
+          },
+          {
+            "title": "Tokenizer Settings",
+            "settings": [
+              "TOKENIZER_NAME", "TOKENIZER_REVISION", "CUSTOM_CHAT_TEMPLATE"
+            ]
+          },
+          {
+            "title": "System Settings",
+            "settings": [
+              "GPU_MEMORY_UTILIZATION", "MAX_PARALLEL_LOADING_WORKERS", "BLOCK_SIZE",
+              "SWAP_SPACE", "ENFORCE_EAGER", "MAX_SEQ_LEN_TO_CAPTURE", "DISABLE_CUSTOM_ALL_REDUCE"
+            ]
+          },
+          {
+            "title": "Streaming Settings",
+            "settings": [
+              "DEFAULT_BATCH_SIZE", "DEFAULT_MIN_BATCH_SIZE", "DEFAULT_BATCH_SIZE_GROWTH_FACTOR"
+            ]
+          },
+          {
+            "title": "OpenAI Settings",
+            "settings": [
+              "RAW_OPENAI_OUTPUT", "OPENAI_RESPONSE_ROLE", "OPENAI_SERVED_MODEL_NAME_OVERRIDE"
+            ]
+          },
+          {
+            "title": "Serverless Settings",
+            "settings": [
+              "MAX_CONCURRENCY", "DISABLE_LOG_STATS", "DISABLE_LOG_REQUESTS"
+            ]
+          }
+        ]
+      },
+      "0.6.1": {
+        "imageName": "runpod/worker-v1-vllm:v1.4.0stable-cuda12.1.0",
+        "minimumCudaVersion": "12.1",
+        "categories": [
+          {
+            "title": "LLM Settings",
+            "settings": [
+              "TOKENIZER", "TOKENIZER_MODE", "SKIP_TOKENIZER_INIT", "TRUST_REMOTE_CODE",
+              "DOWNLOAD_DIR", "LOAD_FORMAT", "DTYPE", "KV_CACHE_DTYPE", "QUANTIZATION_PARAM_PATH",
+              "MAX_MODEL_LEN", "GUIDED_DECODING_BACKEND", "DISTRIBUTED_EXECUTOR_BACKEND",
+              "WORKER_USE_RAY", "RAY_WORKERS_USE_NSIGHT", "PIPELINE_PARALLEL_SIZE",
+              "TENSOR_PARALLEL_SIZE", "MAX_PARALLEL_LOADING_WORKERS", "ENABLE_PREFIX_CACHING",
+              "DISABLE_SLIDING_WINDOW", "USE_V2_BLOCK_MANAGER", "NUM_LOOKAHEAD_SLOTS",
+              "SEED", "NUM_GPU_BLOCKS_OVERRIDE", "MAX_NUM_BATCHED_TOKENS", "MAX_NUM_SEQS",
+              "MAX_LOGPROBS", "DISABLE_LOG_STATS", "QUANTIZATION", "ROPE_SCALING", "ROPE_THETA",
+              "TOKENIZER_POOL_SIZE", "TOKENIZER_POOL_TYPE", "TOKENIZER_POOL_EXTRA_CONFIG",
+              "ENABLE_LORA", "MAX_LORAS", "MAX_LORA_RANK", "LORA_EXTRA_VOCAB_SIZE",
+              "LORA_DTYPE", "LONG_LORA_SCALING_FACTORS", "MAX_CPU_LORAS", "FULLY_SHARDED_LORAS",
+              "DEVICE", "SCHEDULER_DELAY_FACTOR", "ENABLE_CHUNKED_PREFILL", "SPECULATIVE_MODEL",
+              "NUM_SPECULATIVE_TOKENS", "SPECULATIVE_DRAFT_TENSOR_PARALLEL_SIZE",
+              "SPECULATIVE_MAX_MODEL_LEN", "SPECULATIVE_DISABLE_BY_BATCH_SIZE",
+              "NGRAM_PROMPT_LOOKUP_MAX", "NGRAM_PROMPT_LOOKUP_MIN", "SPEC_DECODING_ACCEPTANCE_METHOD",
+              "TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_THRESHOLD", "TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_ALPHA",
+              "MODEL_LOADER_EXTRA_CONFIG", "PREEMPTION_MODE", "PREEMPTION_CHECK_PERIOD",
+              "PREEMPTION_CPU_CAPACITY", "MAX_LOG_LEN", "DISABLE_LOGGING_REQUEST"
+            ]
+          },
+          {
+            "title": "Tokenizer Settings",
+            "settings": [
+              "TOKENIZER_NAME", "TOKENIZER_REVISION", "CUSTOM_CHAT_TEMPLATE"
+            ]
+          },
+          {
+            "title": "System Settings",
+            "settings": [
+              "GPU_MEMORY_UTILIZATION", "MAX_PARALLEL_LOADING_WORKERS", "BLOCK_SIZE",
+              "SWAP_SPACE", "ENFORCE_EAGER", "MAX_SEQ_LEN_TO_CAPTURE", "DISABLE_CUSTOM_ALL_REDUCE"
+            ]
+          },
+          {
+            "title": "Streaming Settings",
+            "settings": [
+              "DEFAULT_BATCH_SIZE", "DEFAULT_MIN_BATCH_SIZE", "DEFAULT_BATCH_SIZE_GROWTH_FACTOR"
+            ]
+          },
+          {
+            "title": "OpenAI Settings",
+            "settings": [
+              "RAW_OPENAI_OUTPUT", "OPENAI_RESPONSE_ROLE", "OPENAI_SERVED_MODEL_NAME_OVERRIDE"
+            ]
+          },
+          {
+            "title": "Serverless Settings",
+            "settings": [
+              "MAX_CONCURRENCY", "DISABLE_LOG_STATS", "DISABLE_LOG_REQUESTS"
+            ]
+          }
+        ]
+      },
       "0.5.5": {
         "imageName": "runpod/worker-v1-vllm:v1.3.1stable-cuda12.1.0",
         "minimumCudaVersion": "12.1",
@@ -119,110 +239,6 @@
             ]
           }
         ]
-      },
-      "0.5.3": {
-        "imageName": "runpod/worker-v1-vllm:stable-cuda12.1.0",
-        "minimumCudaVersion": "12.1",
-        "categories": [
-          {
-            "title": "LLM Settings",
-            "settings": [
-              "TOKENIZER", "TOKENIZER_MODE", "SKIP_TOKENIZER_INIT", "TRUST_REMOTE_CODE",
-              "DOWNLOAD_DIR", "LOAD_FORMAT", "DTYPE", "KV_CACHE_DTYPE", "QUANTIZATION_PARAM_PATH",
-              "MAX_MODEL_LEN", "GUIDED_DECODING_BACKEND", "DISTRIBUTED_EXECUTOR_BACKEND",
-              "WORKER_USE_RAY", "RAY_WORKERS_USE_NSIGHT", "PIPELINE_PARALLEL_SIZE",
-              "TENSOR_PARALLEL_SIZE", "MAX_PARALLEL_LOADING_WORKERS", "ENABLE_PREFIX_CACHING",
-              "DISABLE_SLIDING_WINDOW", "USE_V2_BLOCK_MANAGER", "NUM_LOOKAHEAD_SLOTS",
-              "SEED", "NUM_GPU_BLOCKS_OVERRIDE", "MAX_NUM_BATCHED_TOKENS", "MAX_NUM_SEQS",
-              "MAX_LOGPROBS", "DISABLE_LOG_STATS", "QUANTIZATION", "ROPE_SCALING", "ROPE_THETA",
-              "TOKENIZER_POOL_SIZE", "TOKENIZER_POOL_TYPE", "TOKENIZER_POOL_EXTRA_CONFIG",
-              "ENABLE_LORA", "MAX_LORAS", "MAX_LORA_RANK", "LORA_EXTRA_VOCAB_SIZE",
-              "LORA_DTYPE", "LONG_LORA_SCALING_FACTORS", "MAX_CPU_LORAS", "FULLY_SHARDED_LORAS",
-              "DEVICE", "SCHEDULER_DELAY_FACTOR", "ENABLE_CHUNKED_PREFILL", "SPECULATIVE_MODEL",
-              "NUM_SPECULATIVE_TOKENS", "SPECULATIVE_DRAFT_TENSOR_PARALLEL_SIZE",
-              "SPECULATIVE_MAX_MODEL_LEN", "SPECULATIVE_DISABLE_BY_BATCH_SIZE",
-              "NGRAM_PROMPT_LOOKUP_MAX", "NGRAM_PROMPT_LOOKUP_MIN", "SPEC_DECODING_ACCEPTANCE_METHOD",
-              "TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_THRESHOLD", "TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_ALPHA",
-              "MODEL_LOADER_EXTRA_CONFIG", "PREEMPTION_MODE", "PREEMPTION_CHECK_PERIOD",
-              "PREEMPTION_CPU_CAPACITY", "MAX_LOG_LEN", "DISABLE_LOGGING_REQUEST"
-            ]
-          },
-          {
-            "title": "Tokenizer Settings",
-            "settings": [
-              "TOKENIZER_NAME", "TOKENIZER_REVISION", "CUSTOM_CHAT_TEMPLATE"
-            ]
-          },
-          {
-            "title": "System Settings",
-            "settings": [
-              "GPU_MEMORY_UTILIZATION", "MAX_PARALLEL_LOADING_WORKERS", "BLOCK_SIZE",
-              "SWAP_SPACE", "ENFORCE_EAGER", "MAX_SEQ_LEN_TO_CAPTURE", "DISABLE_CUSTOM_ALL_REDUCE"
-            ]
-          },
-          {
-            "title": "Streaming Settings",
-            "settings": [
-              "DEFAULT_BATCH_SIZE", "DEFAULT_MIN_BATCH_SIZE", "DEFAULT_BATCH_SIZE_GROWTH_FACTOR"
-            ]
-          },
-          {
-            "title": "OpenAI Settings",
-            "settings": [
-              "RAW_OPENAI_OUTPUT", "OPENAI_RESPONSE_ROLE", "OPENAI_SERVED_MODEL_NAME_OVERRIDE"
-            ]
-          },
-          {
-            "title": "Serverless Settings",
-            "settings": [
-              "MAX_CONCURRENCY", "DISABLE_LOG_STATS", "DISABLE_LOG_REQUESTS"
-            ]
-          }
-        ]
-      },
-      "0.4.2": {
-        "imageName": "runpod/worker-vllm:stable-cuda12.1.0",
-        "minimumCudaVersion": "12.1",
-        "categories": [
-          {
-            "title": "LLM Settings",
-            "settings": [
-              "MODEL_REVISION", "MAX_MODEL_LEN", "BASE_PATH", "LOAD_FORMAT", "QUANTIZATION",
-              "TRUST_REMOTE_CODE", "SEED", "KV_CACHE_DTYPE", "DTYPE"
-            ]
-          },
-          {
-            "title": "Tokenizer Settings",
-            "settings": [
-              "TOKENIZER_NAME", "TOKENIZER_REVISION", "CUSTOM_CHAT_TEMPLATE"
-            ]
-          },
-          {
-            "title": "System Settings",
-            "settings": [
-              "GPU_MEMORY_UTILIZATION", "MAX_PARALLEL_LOADING_WORKERS", "BLOCK_SIZE",
-              "SWAP_SPACE", "ENFORCE_EAGER", "MAX_SEQ_LEN_TO_CAPTURE", "DISABLE_CUSTOM_ALL_REDUCE"
-            ]
-          },
-          {
-            "title": "Streaming Settings",
-            "settings": [
-              "DEFAULT_BATCH_SIZE", "DEFAULT_MIN_BATCH_SIZE", "DEFAULT_BATCH_SIZE_GROWTH_FACTOR"
-            ]
-          },
-          {
-            "title": "OpenAI Settings",
-            "settings": [
-              "RAW_OPENAI_OUTPUT", "OPENAI_RESPONSE_ROLE", "OPENAI_SERVED_MODEL_NAME_OVERRIDE"
-            ]
-          },
-          {
-            "title": "Serverless Settings",
-            "settings": [
-              "MAX_CONCURRENCY", "DISABLE_LOG_STATS", "DISABLE_LOG_REQUESTS"
-            ]
-          }
-        ]
       }
     },
     "schema": {