Update docker-bake.hcl

Update README.md

.github/workflows/docker-build-release.yml

@@ -1,45 +0,0 @@
-name: CD | Docker-Build-Release
-
-on:
-  push:
-    branches:
-      - "main"
-  release:
-    types: [published]
-  workflow_dispatch:
-    inputs:
-      image_tag:
-        description: "Docker Image Tag"
-        required: false
-        default: "dev"
-
-jobs:
-  docker-build:
-    runs-on: DO
-    # DO is a custom runner deployed on DigitalOcean, only available for workflows under the runpod-workers organization.
-    # If you would like to use this workflow, you can replace DO with ubuntu-latest or any other runner.
-
-    strategy:
-      matrix:
-        cuda_version: [11.8.0, 12.1.0]
-
-    steps:
-      - name: Set up QEMU
-        uses: docker/setup-qemu-action@v2
-
-      - name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v2
-
-      - name: Login to Docker Hub
-        uses: docker/login-action@v2
-        with:
-          username: ${{ secrets.DOCKERHUB_USERNAME }}
-          password: ${{ secrets.DOCKERHUB_TOKEN }}
-
-      # Build and push step
-      - name: Build and push
-        uses: docker/build-push-action@v4
-        with:
-          push: true
-          tags: ${{ vars.DOCKERHUB_REPO }}/${{ vars.DOCKERHUB_IMG }}:${{ (github.event_name == 'release' && github.event.release.tag_name) || (github.event_name == 'workflow_dispatch' && github.event.inputs.image_tag) || 'dev' }}-cuda${{ matrix.cuda_version }}
-          build-args: WORKER_CUDA_VERSION=${{ matrix.cuda_version }}

.gitmodules

@@ -0,0 +1,3 @@
+[submodule "vllm-base-image/vllm"]
+	path = vllm-base-image/vllm
+	url = https://github.com/runpod/vllm-fork-for-sls-worker.git

Dockerfile

@@ -1,15 +1,20 @@
-ARG WORKER_CUDA_VERSION=11.8.0
+FROM nvidia/cuda:12.1.0-base-ubuntu22.04 
-FROM runpod/worker-vllm:base-0.3.0-cuda${WORKER_CUDA_VERSION} AS vllm-base
 
 RUN apt-get update -y \
     && apt-get install -y python3-pip
 
+RUN ldconfig /usr/local/cuda-12.1/compat/
+
 # Install Python dependencies
 COPY builder/requirements.txt /requirements.txt
 RUN --mount=type=cache,target=/root/.cache/pip \
     python3 -m pip install --upgrade 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==0.5.3.post1 && \
+    python3 -m pip install flashinfer -i https://flashinfer.ai/whl/cu121/torch2.3
+
 # Setup for Option 2: Building the Image with the Model included
 ARG MODEL_NAME=""
 ARG TOKENIZER_NAME=""
@@ -19,7 +24,7 @@ ARG MODEL_REVISION=""
 ARG TOKENIZER_REVISION=""
 
 ENV MODEL_NAME=$MODEL_NAME \
-    MODEL_REVISION=$REVISION \
+    MODEL_REVISION=$MODEL_REVISION \
     TOKENIZER_NAME=$TOKENIZER_NAME \
     TOKENIZER_REVISION=$TOKENIZER_REVISION \
     BASE_PATH=$BASE_PATH \
@@ -27,22 +32,19 @@ ENV MODEL_NAME=$MODEL_NAME \
     HF_DATASETS_CACHE="${BASE_PATH}/huggingface-cache/datasets" \
     HUGGINGFACE_HUB_CACHE="${BASE_PATH}/huggingface-cache/hub" \
     HF_HOME="${BASE_PATH}/huggingface-cache/hub" \
-    HF_TRANSFER=1 
+    HF_HUB_ENABLE_HF_TRANSFER=1 
 
-ENV PYTHONPATH="/:/vllm-installation"
+ENV PYTHONPATH="/:/vllm-workspace"
 
-COPY builder/download_model.py /download_model.py
+
+COPY src /src
 RUN --mount=type=secret,id=HF_TOKEN,required=false \
     if [ -f /run/secrets/HF_TOKEN ]; then \
         export HF_TOKEN=$(cat /run/secrets/HF_TOKEN); \
     fi && \
     if [ -n "$MODEL_NAME" ]; then \
-        python3 /download_model.py; \
+        python3 /src/download_model.py; \
     fi
 
-# Add source files
-COPY src /src
-
-
 # Start the handler
 CMD ["python3", "/src/handler.py"]

README.md

@@ -1,25 +1,34 @@
 <div align="center">
 
-<h1> vLLM Serverless Endpoint Worker </h1>
+# OpenAI-Compatible vLLM Serverless Endpoint Worker
+Deploy OpenAI-Compatible Blazing-Fast LLM Endpoints powered by the [vLLM](https://github.com/vllm-project/vllm) Inference Engine on RunPod Serverless with just a few clicks.
+<!-- 
+![vLLM Version](https://img.shields.io/badge/dynamic/yaml?url=https%3A%2F%2Fraw.githubusercontent.com%2Frunpod-workers%2Fworker-vllm%2Fmain%2Fvllm-base-image%2Fvllm-metadata.yml&query=%24.version&style=for-the-badge&logo=data%3Aimage%2Fsvg%2Bxml%3Bbase64%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%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%2BPHBhdGggc3R5bGU9Im9wYWNpdHk6MC45ODQiIGZpbGw9IiNmY2I3MWQiIGQ9Ik0gMjIuNSwxMi41IEMgMjEuNTA0NiwyNC45ODkgMjEuMTcxMywzNy42NTU3IDIxLjUsNTAuNUMgMjEuNjcxNiw1MS40OTE2IDIxLjMzODIsNTIuMTU4MyAyMC41LDUyLjVDIDEzLjAzMTEsMzkuMjI4NyA2LjM2NDQxLDI1LjU2MjEgMC41LDExLjVDIDguMDE5MDUsMTEuMTc1IDE1LjM1MjQsMTEuNTA4NCAyMi41LDEyLjUgWiIvPjwvZz4KPGc%2BPHBhdGggc3R5bGU9Im9wYWNpdHk6MC4wMiIgZmlsbD0iI2Q3ZGZlOCIgZD0iTSAyMi41LDEyLjUgQyAyMy4xNjY3LDIxLjUgMjMuODMzMywzMC41IDI0LjUsMzkuNUMgMjMuMjIyOSw0My4xODg5IDIyLjIyMjksNDYuODU1NSAyMS41LDUwLjVDIDIxLjE3MTMsMzcuNjU1NyAyMS41MDQ2LDI0Ljk4OSAyMi41LDEyLjUgWiIvPjwvZz4KPGc%2BPHBhdGggc3R5bGU9Im9wYWNpdHk6MC43NTMiIGZpbGw9IiNjZmQ2ZGQiIGQ9Ik0gNTEuNSwwLjUgQyA1Mi42MTI5LDEuOTQ2MzkgNTIuNzc5NiwzLjYxMzA1IDUyLDUuNUMgNDcuODAzNiwyMi4yODg3IDQzLjMwMzYsMzguOTU1MyAzOC41LDU1LjVDIDMyLjUsNTUuNSAyNi41LDU1LjUgMjAuNSw1NS41QyAyMC44MzMzLDU0LjgzMzMgMjEuMTY2Nyw1NC4xNjY3IDIxLjUsNTMuNUMgMjYuODMzMyw1My41IDMyLjE2NjcsNTMuNSAzNy41LDUzLjVDIDQxLjkxNTYsMzUuNzUwNSA0Ni41ODIyLDE4LjA4MzggNTEuNSwwLjUgWiIvPjwvZz4KPC9zdmc%2BCg%3D%3D&label=STABLE%20vLLM%20Version&link=https%3A%2F%2Fgithub.com%2Fvllm-project%2Fvllm)
+![Worker Version](https://img.shields.io/github/v/tag/runpod-workers/worker-vllm?style=for-the-badge&logo=data%3Aimage%2Fsvg%2Bxml%3Bbase64%2CPD94bWwgdmVyc2lvbj0iMS4wIiBlbmNvZGluZz0idXRmLTgiPz4KPCEtLSBHZW5lcmF0b3I6IEFkb2JlIElsbHVzdHJhdG9yIDI2LjUuMywgU1ZHIEV4cG9ydCBQbHVnLUluIC4gU1ZHIFZlcnNpb246IDYuMDAgQnVpbGQgMCkgIC0tPgo8c3ZnIHZlcnNpb249IjEuMSIgaWQ9IkxheWVyXzEiIHhtbG5zPSJodHRwOi8vd3d3LnczLm9yZy8yMDAwL3N2ZyIgeG1sbnM6eGxpbms9Imh0dHA6Ly93d3cudzMub3JnLzE5OTkveGxpbmsiIHg9IjBweCIgeT0iMHB4IgoJIHZpZXdCb3g9IjAgMCAyMDAwIDIwMDAiIHN0eWxlPSJlbmFibGUtYmFja2dyb3VuZDpuZXcgMCAwIDIwMDAgMjAwMDsiIHhtbDpzcGFjZT0icHJlc2VydmUiPgo8c3R5bGUgdHlwZT0idGV4dC9jc3MiPgoJLnN0MHtmaWxsOiM2NzNBQjc7fQo8L3N0eWxlPgo8Zz4KCTxnPgoJCTxwYXRoIGNsYXNzPSJzdDAiIGQ9Ik0xMDE3Ljk1LDcxMS4wNGMtNC4yMiwyLjM2LTkuMTgsMy4wMS0xMy44NiwxLjgyTDM4Ni4xNyw1NTUuM2MtNDEuNzItMTAuNzYtODYuMDItMC42My0xMTYuNiwyOS43MwoJCQlsLTEuNCwxLjM5Yy0zNS45MiwzNS42NS0yNy41NSw5NS44LDE2Ljc0LDEyMC4zbDU4NC4zMiwzMjQuMjNjMzEuMzYsMTcuNCw1MC44Miw1MC40NSw1MC44Miw4Ni4zMnY4MDYuNzYKCQkJYzAsMzUuNDktMzguNDEsNTcuNjctNjkuMTUsMzkuOTRsLTcwMy4xNS00MDUuNjRjLTIzLjYtMTMuNjEtMzguMTMtMzguNzgtMzguMTMtNjYuMDJWNjY2LjYzYzAtODcuMjQsNDYuNDUtMTY3Ljg5LDEyMS45Mi0yMTEuNjYKCQkJTDkzMy44NSw0Mi4xNWMyMy40OC0xMy44LDUxLjQ3LTE3LjcsNzcuODMtMTAuODRsNzQ1LjcxLDE5NC4xYzMxLjUzLDguMjEsMzYuOTksNTAuNjUsOC41Niw2Ni41N0wxMDE3Ljk1LDcxMS4wNHoiLz4KCQk8cGF0aCBjbGFzcz0ic3QwIiBkPSJNMTUyNy43NSw1MzYuMzhsMTI4Ljg5LTc5LjYzbDE4OS45MiwxMDkuMTdjMjcuMjQsMTUuNjYsNDMuOTcsNDQuNzMsNDMuODIsNzYuMTVsLTQsODU3LjYKCQkJYy0wLjExLDI0LjM5LTEzLjE1LDQ2Ljg5LTM0LjI1LDU5LjExbC03MDEuNzUsNDA2LjYxYy0zMi4zLDE4LjcxLTcyLjc0LTQuNTktNzIuNzQtNDEuOTJ2LTc5Ny40MwoJCQljMC0zOC45OCwyMS4wNi03NC45MSw1NS4wNy05My45Nmw1OTAuMTctMzMwLjUzYzE4LjIzLTEwLjIxLDE4LjY1LTM2LjMsMC43NS00Ny4wOUwxNTI3Ljc1LDUzNi4zOHoiLz4KCQk8cGF0aCBjbGFzcz0ic3QwIiBkPSJNMTUyNC4wMSw2NjUuOTEiLz4KCTwvZz4KPC9nPgo8L3N2Zz4K&logoColor=%23ffffff&label=STABLE%20Worker%20Version&color=%23673ab7)
+![vLLM Version](https://img.shields.io/badge/dynamic/yaml?url=https%3A%2F%2Fraw.githubusercontent.com%2Frunpod-workers%2Fworker-vllm%2Fmain%2Fvllm-base-image%2Fvllm-metadata.yml&query=%24.dev_version&style=for-the-badge&logo=data%3Aimage%2Fsvg%2Bxml%3Bbase64%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%2BCjxnPjxwYXRoIHN0eWxlPSJvcGFjaXR5OjEiIGZpbGw9IiMzN2E0ZmUiIGQ9Ik0gNTEuNSwwLjUgQyA0Ni41ODIyLDE4LjA4MzggNDEuOTE1NiwzNS43NTA1IDM3LjUsNTMuNUMgMzIuMTY2Nyw1My41IDI2LjgzMzMsNTMuNSAyMS41LDUzLjVDIDIwLjgzMzMsNTMuNSAyMC41LDUzLjE2NjcgMjAuNSw1Mi41QyAyMS4zMzgyLDUyLjE1ODMgMjEuNjcxNiw1MS40OTE2IDIxLjUsNTAuNUMgMjIuMjIyOSw0Ni44NTU1IDIzLjIyMjksNDMuMTg4OSAyNC41LDM5LjVDIDI0LjY5MTcsMzYuMzk5MiAyNS4zNTg0LDMzLjM5OTIgMjYuNSwzMC41QyAyNi4yOTA3LDI5LjkxNCAyNS45NTc0LDI5LjQxNCAyNS41LDI5QyAyNy40NDE0LDI3LjE4NDEgMjguMTA4MSwyNS4xODQxIDI3LjUsMjNDIDI5LjI0MTUsMTguNTM4NyAzMC45MDgyLDE0LjAzODcgMzIuNSw5LjVDIDM4Ljc3NTcsNi4xOTM1OCA0NS4xMDkxLDMuMTkzNTggNTEuNSwwLjUgWiIvPjwvZz4KPGc%2BPHBhdGggc3R5bGU9Im9wYWNpdHk6MC45ODQiIGZpbGw9IiNmY2I3MWQiIGQ9Ik0gMjIuNSwxMi41IEMgMjEuNTA0NiwyNC45ODkgMjEuMTcxMywzNy42NTU3IDIxLjUsNTAuNUMgMjEuNjcxNiw1MS40OTE2IDIxLjMzODIsNTIuMTU4MyAyMC41LDUyLjVDIDEzLjAzMTEsMzkuMjI4NyA2LjM2NDQxLDI1LjU2MjEgMC41LDExLjVDIDguMDE5MDUsMTEuMTc1IDE1LjM1MjQsMTEuNTA4NCAyMi41LDEyLjUgWiIvPjwvZz4KPGc%2BPHBhdGggc3R5bGU9Im9wYWNpdHk6MC4wMiIgZmlsbD0iI2Q3ZGZlOCIgZD0iTSAyMi41LDEyLjUgQyAyMy4xNjY3LDIxLjUgMjMuODMzMywzMC41IDI0LjUsMzkuNUMgMjMuMjIyOSw0My4xODg5IDIyLjIyMjksNDYuODU1NSAyMS41LDUwLjVDIDIxLjE3MTMsMzcuNjU1NyAyMS41MDQ2LDI0Ljk4OSAyMi41LDEyLjUgWiIvPjwvZz4KPGc%2BPHBhdGggc3R5bGU9Im9wYWNpdHk6MC43NTMiIGZpbGw9IiNjZmQ2ZGQiIGQ9Ik0gNTEuNSwwLjUgQyA1Mi42MTI5LDEuOTQ2MzkgNTIuNzc5NiwzLjYxMzA1IDUyLDUuNUMgNDcuODAzNiwyMi4yODg3IDQzLjMwMzYsMzguOTU1MyAzOC41LDU1LjVDIDMyLjUsNTUuNSAyNi41LDU1LjUgMjAuNSw1NS41QyAyMC44MzMzLDU0LjgzMzMgMjEuMTY2Nyw1NC4xNjY3IDIxLjUsNTMuNUMgMjYuODMzMyw1My41IDMyLjE2NjcsNTMuNSAzNy41LDUzLjVDIDQxLjkxNTYsMzUuNzUwNSA0Ni41ODIyLDE4LjA4MzggNTEuNSwwLjUgWiIvPjwvZz4KPC9zdmc%2BCg%3D%3D&label=DEV%20vLLM%20Version%20&link=https%3A%2F%2Fgithub.com%2Fvllm-project%2Fvllm)\
+![Docker Pulls](https://img.shields.io/docker/pulls/runpod/worker-vllm?style=for-the-badge&logo=docker&label=Docker%20Pulls&link=https%3A%2F%2Fhub.docker.com%2Frepository%2Fdocker%2Frunpod%2Fworker-vllm%2Fgeneral) -->
+<!-- 
+![Docker Automatic Build](https://img.shields.io/github/actions/workflow/status/runpod-workers/worker-vllm/docker-build-release.yml?style=flat&label=BUILD) -->
 
-[![CD | Docker-Build-Release](https://github.com/runpod-workers/worker-vllm/actions/workflows/docker-build-release.yml/badge.svg)](https://github.com/runpod-workers/worker-vllm/actions/workflows/docker-build-release.yml)
 
-Deploy Blazing-fast LLMs powered by [vLLM](https://github.com/vllm-project/vllm) on RunPod Serverless in a few clicks.
 </div>
 
-### Worker vLLM 0.3.0: What's New since 0.2.0:
+# News:
-- **🚀 Full OpenAI Compatibility 🚀**
+
+### 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 `v1.1` with vLLM `0.5.3` now available under `stable` tags
+Update v1.1 is now available, use the image tag `runpod/worker-v1-vllm:stable-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! 
+
+
+
+### 4. Caching Accross RunPod Machines
+Worker vLLM is now cached on all RunPod machines, resulting in near-instant deployment! Previously, downloading and extracting the image took 3-5 minutes on average.
 
-  You may now use your deployment with any OpenAI Codebase by changing **only 3 lines** in total. The supported routes are <ins>Chat Completions</ins>, <ins>Completions</ins>, and <ins>Models</ins> - with both streaming and non-streaming.
-- **Dynamic Batch Size** - time-to-first token as fast no batching, while maintaining the performance of batched token streaming throughout the request.
-- vLLM 0.2.7 -> 0.3.2
-  - Gemma, DeepSeek MoE and OLMo support.
-  - FP8 KV Cache support
-  - New supported parameters
-  - We're working on adding support for Multi-LoRA ⚙️
-- Support for a wide range of new settings for your endpoint, such as Custom chat templates.
-- Fixed Tensor Parallelism, baking model into images, and more bugs.
-- Refactors and general improvements.
 
 ## Table of Contents
 - [Setting up the Serverless Worker](#setting-up-the-serverless-worker)
@@ -43,7 +52,6 @@ Deploy Blazing-fast LLMs powered by [vLLM](https://github.com/vllm-project/vllm)
     - [Modifying your OpenAI Codebase to use your deployed vLLM Worker](#modifying-your-openai-codebase-to-use-your-deployed-vllm-worker)
     - [OpenAI Request Input Parameters](#openai-request-input-parameters)
       - [Chat Completions](#chat-completions)
-      - [Completions](#completions)
     - [Examples: Using your RunPod endpoint with OpenAI](#examples-using-your-runpod-endpoint-with-openai)
 - [Usage: standard](#non-openai-usage)
   - [Input Request Parameters](#input-request-parameters)
@@ -53,8 +61,11 @@ Deploy Blazing-fast LLMs powered by [vLLM](https://github.com/vllm-project/vllm)
 # Setting up the Serverless Worker
 
 ### Option 1: Deploy Any Model Using Pre-Built Docker Image [Recommended]
-> [!TIP]
+
-> This is the recommended way to deploy your model, as it does not require you to build a Docker image, upload heavy models to DockerHub and wait for workers to download them. Instead, use this option to deploy your model in a few clicks. For even more convenience, attach a network storage volume to your Endpoint, which will download the model once and share it across all workers.
+> [!NOTE]
+> You can now deploy from the dedicated UI on the RunPod console with all of the settings and choices listed. 
+> Try now by accessing in Explore or Serverless pages on the RunPod console!
+
 
 We now offer a pre-built Docker Image for the vLLM Worker that you can configure entirely with Environment Variables when creating the RunPod Serverless Endpoint:
 
@@ -66,33 +77,88 @@ Below is a summary of the available RunPod Worker images, categorized by image s
 
 | CUDA Version | Stable Image Tag                  | Development Image Tag             | Note                                                        |
 |--------------|-----------------------------------|-----------------------------------|----------------------------------------------------------------------|
-| 11.8.0       | `runpod/worker-vllm:0.3.0-cuda11.8.0`        | `runpod/worker-vllm:dev-cuda11.8.0`   | Available on all RunPod Workers without additional selection needed. |
+| 12.1.0       | `runpod/worker-v1-vllm:stable-cuda12.1.0` | `runpod/worker-v1-vllm:dev-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-vllm:0.3.0-cuda12.1.0` | `runpod/worker-vllm:dev-cuda12.1.0` | When creating an Endpoint, select CUDA Version 12.2 and 12.1 in the filter. |
+
 
-This table provides a quick reference to the image tags you should use based on the desired CUDA version and image stability (Stable or Development). Ensure to follow the selection note for CUDA 12.1.0 compatibility.
 
 ---
 
 #### Prerequisites
 - RunPod Account
 
-#### Environment Variables
+#### Environment Variables/Settings
 > Note:  `0` is equivalent to `False` and `1` is equivalent to `True` for boolean values.
 
-| Name                                | Default              | Type/Choices                              | Description |
+| `Name`                                    | `Default`             | `Type/Choices`                             | `Description` |
-|-------------------------------------|----------------------|-------------------------------------------|-------------|
+|-------------------------------------------|-----------------------|--------------------------------------------|---------------|
-**LLM Settings**
+| `MODEL_NAME`                                   | 'facebook/opt-125m'   | `str`                                      | Name or path of the Hugging Face model to use. |
-| `MODEL_NAME`**\***                        | -                    | `str`                                         | Hugging Face Model Repository (e.g., `openchat/openchat-3.5-1210`). |
+| `TOKENIZER`                               | None                  | `str`                                      | Name or path of the Hugging Face tokenizer to use. |
-| `MODEL_REVISION`                    | `None`               | `str`                                         |Model revision(branch) to load. |
+| `SKIP_TOKENIZER_INIT`                     | False                 | `bool`                                     | Skip initialization of tokenizer and detokenizer. |
-| `MAX_MODEL_LENGTH`                  | Model's maximum      | `int`                                         |Maximum number of tokens for the engine to handle per request. |
+| `TOKENIZER_MODE`                          | 'auto'                | ['auto', 'slow']                           | The tokenizer mode. |
-| `BASE_PATH`                         | `/runpod-volume`     | `str`                                         |Storage directory for Huggingface cache and model. Utilizes network storage if attached when pointed at `/runpod-volume`, which will have only one worker download the model once, which all workers will be able to load. If no network volume is present, creates a local directory within each worker. |
+| `TRUST_REMOTE_CODE`                       | False                 | `bool`                                     | Trust remote code from Hugging Face. |
-| `LOAD_FORMAT`                       | `auto`               | `str`                                         |Format to load model in. |
+| `DOWNLOAD_DIR`                            | None                  | `str`                                      | Directory to download and load the weights. |
-| `HF_TOKEN`                          | -                    | `str`                                         |Hugging Face token for private and gated models. |
+| `LOAD_FORMAT`                             | 'auto'                | ['auto', 'pt', 'safetensors', 'npcache', 'dummy', 'tensorizer', 'bitsandbytes'] | The format of the model weights to load. |
-| `QUANTIZATION`                      | `None`               | `awq`, `squeezellm`, `gptq`              |Quantization of given model. The model must already be quantized. |
+| `DTYPE`                                   | 'auto'                | ['auto', 'half', 'float16', 'bfloat16', 'float', 'float32'] | Data type for model weights and activations. |
-| `TRUST_REMOTE_CODE`                 | `0`                  | boolean as `int`                                         |Trust remote code for Hugging Face models. Can help with Mixtral 8x7B, Quantized models, and unusual models/architectures.
+| `KV_CACHE_DTYPE`                          | 'auto'                | ['auto', 'fp8', 'fp8_e5m2', 'fp8_e4m3']    | Data type for KV cache storage. |
-| `SEED`                              | `0`                  | `int`                                         |Sets random seed for operations. |
+| `QUANTIZATION_PARAM_PATH`                 | None                  | `str`                                      | Path to the JSON file containing the KV cache scaling factors. |
-| `KV_CACHE_DTYPE`                    | `auto`               | boolean as `int`                                         |Data type for kv cache storage. Uses `DTYPE` if set to `auto`. |
+| `MAX_MODEL_LEN`                           | None                  | `int`                                      | Model context length. |
-| `DTYPE`                             | `auto`               | `auto`, `half`, `float16`, `bfloat16`, `float`, `float32` |Sets datatype/precision for model weights and activations. |
+| `GUIDED_DECODING_BACKEND`                 | 'outlines'            | ['outlines', 'lm-format-enforcer']         | Which engine will be used for guided decoding by default. |
+| `DISTRIBUTED_EXECUTOR_BACKEND`            | None                  | ['ray', 'mp']                              | Backend to use for distributed serving. |
+| `WORKER_USE_RAY`                          | False                 | `bool`                                     | Deprecated, use --distributed-executor-backend=ray. |
+| `PIPELINE_PARALLEL_SIZE`                  | 1                     | `int`                                      | Number of pipeline stages. |
+| `TENSOR_PARALLEL_SIZE`                    | 1                     | `int`                                      | Number of tensor parallel replicas. |
+| `MAX_PARALLEL_LOADING_WORKERS`            | None                  | `int`                                      | Load model sequentially in multiple batches. |
+| `RAY_WORKERS_USE_NSIGHT`                  | False                 | `bool`                                     | If specified, use nsight to profile Ray workers. |
+| `BLOCK_SIZE`                              | 16                    | [8, 16, 32]                                | Token block size for contiguous chunks of tokens. |
+| `ENABLE_PREFIX_CACHING`                   | False                 | `bool`                                     | Enables automatic prefix caching. |
+| `DISABLE_SLIDING_WINDOW`                  | False                 | `bool`                                     | Disables sliding window, capping to sliding window size. |
+| `USE_V2_BLOCK_MANAGER`                    | False                 | `bool`                                     | Use BlockSpaceMangerV2. |
+| `NUM_LOOKAHEAD_SLOTS`                     | 0                     | `int`                                      | Experimental scheduling config necessary for speculative decoding. |
+| `SEED`                                    | 0                     | `int`                                      | Random seed for operations. |
+| `SWAP_SPACE`                              | 4                     | `int`                                      | CPU swap space size (GiB) per GPU. |
+| `GPU_MEMORY_UTILIZATION`                  | 0.90                  | `float`                                    | The fraction of GPU memory to be used for the model executor. |
+| `NUM_GPU_BLOCKS_OVERRIDE`                 | None                  | `int`                                      | If specified, ignore GPU profiling result and use this number of GPU blocks. |
+| `MAX_NUM_BATCHED_TOKENS`                  | None                  | `int`                                      | Maximum number of batched tokens per iteration. |
+| `MAX_NUM_SEQS`                            | 256                   | `int`                                      | Maximum number of sequences per iteration. |
+| `MAX_LOGPROBS`                            | 20                    | `int`                                      | Max number of log probs to return when logprobs is specified in SamplingParams. |
+| `DISABLE_LOG_STATS`                       | False                 | `bool`                                     | Disable logging statistics. |
+| `QUANTIZATION`                            | None                  | [*QUANTIZATION_METHODS, None]              | Method used to quantize the weights. |
+| `ROPE_SCALING`                            | None                  | `dict`                                     | RoPE scaling configuration in JSON format. |
+| `ROPE_THETA`                              | None                  | `float`                                    | RoPE theta. Use with rope_scaling. |
+| `ENFORCE_EAGER`                           | False                 | `bool`                                     | Always use eager-mode PyTorch. |
+| `MAX_CONTEXT_LEN_TO_CAPTURE`              | None                  | `int`                                      | Maximum context length covered by CUDA graphs. |
+| `MAX_SEQ_LEN_TO_CAPTURE`                  | 8192                  | `int`                                      | Maximum sequence length covered by CUDA graphs. |
+| `DISABLE_CUSTOM_ALL_REDUCE`               | False                 | `bool`                                     | See ParallelConfig. |
+| `TOKENIZER_POOL_SIZE`                     | 0                     | `int`                                      | Size of tokenizer pool to use for asynchronous tokenization. |
+| `TOKENIZER_POOL_TYPE`                     | 'ray'                 | `str`                                      | Type of tokenizer pool to use for asynchronous tokenization. |
+| `TOKENIZER_POOL_EXTRA_CONFIG`             | None                  | `dict`                                     | Extra config for tokenizer pool. |
+| `ENABLE_LORA`                             | False                 | `bool`                                     | If True, enable handling of LoRA adapters. |
+| `MAX_LORAS`                               | 1                     | `int`                                      | Max number of LoRAs in a single batch. |
+| `MAX_LORA_RANK`                           | 16                    | `int`                                      | Max LoRA rank. |
+| `LORA_EXTRA_VOCAB_SIZE`                   | 256                   | `int`                                      | Maximum size of extra vocabulary for LoRA adapters. |
+| `LORA_DTYPE`                              | 'auto'                | ['auto', 'float16', 'bfloat16', 'float32'] | Data type for LoRA. |
+| `LONG_LORA_SCALING_FACTORS`               | None                  | `tuple`                                    | Specify multiple scaling factors for LoRA adapters. |
+| `MAX_CPU_LORAS`                           | None                  | `int`                                      | Maximum number of LoRAs to store in CPU memory. |
+| `FULLY_SHARDED_LORAS`                     | False                 | `bool`                                     | Enable fully sharded LoRA layers. |
+| `DEVICE`                                  | 'auto'                | ['auto', 'cuda', 'neuron', 'cpu', 'openvino', 'tpu', 'xpu'] | Device type for vLLM execution. |
+| `SCHEDULER_DELAY_FACTOR`                  | 0.0                   | `float`                                    | Apply a delay before scheduling next prompt. |
+| `ENABLE_CHUNKED_PREFILL`                  | False                 | `bool`                                     | Enable chunked prefill requests. |
+| `SPECULATIVE_MODEL`                       | None                  | `str`                                      | The name of the draft model to be used in speculative decoding. |
+| `NUM_SPECULATIVE_TOKENS`                  | None                  | `int`                                      | The number of speculative tokens to sample from the draft model. |
+| `SPECULATIVE_DRAFT_TENSOR_PARALLEL_SIZE`  | None                  | `int`                                      | Number of tensor parallel replicas for the draft model. |
+| `SPECULATIVE_MAX_MODEL_LEN`               | None                  | `int`                                      | The maximum sequence length supported by the draft model. |
+| `SPECULATIVE_DISABLE_BY_BATCH_SIZE`       | None                  | `int`                                      | Disable speculative decoding if the number of enqueue requests is larger than this value. |
+| `NGRAM_PROMPT_LOOKUP_MAX`                 | None                  | `int`                                      | Max size of window for ngram prompt lookup in speculative decoding. |
+| `NGRAM_PROMPT_LOOKUP_MIN`                 | None                  | `int`                                      | Min size of window for ngram prompt lookup in speculative decoding. |
+| `SPEC_DECODING_ACCEPTANCE_METHOD`         | 'rejection_sampler'   | ['rejection_sampler', 'typical_acceptance_sampler'] | Specify the acceptance method for draft token verification in speculative decoding. |
+| `TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_THRESHOLD` | None              | `float`                                    | Set the lower bound threshold for the posterior probability of a token to be accepted. |
+| `TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_ALPHA`     | None              | `float`                                    | A scaling factor for the entropy-based threshold for token acceptance. |
+| `MODEL_LOADER_EXTRA_CONFIG`               | None                  | `dict`                                     | Extra config for model loader. |
+| `PREEMPTION_MODE`                         | None                  | `str`                                      | If 'recompute', the engine performs preemption-aware recomputation. If 'save', the engine saves activations into the CPU memory as preemption happens. |
+| `PREEMPTION_CHECK_PERIOD`                 | 1.0                   | `float`                                    | How frequently the engine checks if a preemption happens. |
+| `PREEMPTION_CPU_CAPACITY`                 | 2                     | `float`                                    | The percentage of CPU memory used for the saved activations. |
+| `DISABLE_LOGGING_REQUEST`                 | False                 | `bool`                                     | Disable logging requests. |
+| `MAX_LOG_LEN`                             | None                  | `int`                                      | Max number of prompt characters or prompt ID numbers being printed in log. |
 **Tokenizer Settings**
 | `TOKENIZER_NAME`                    | `None`               | `str`                                         |Tokenizer repository to use a different tokenizer than the model's default. |
 | `TOKENIZER_REVISION`                | `None`               | `str`                                         |Tokenizer revision to load. |
@@ -103,7 +169,7 @@ This table provides a quick reference to the image tags you should use based on
 | `BLOCK_SIZE`                        | `16`                 | `8`, `16`, `32`                           |Token block size for contiguous chunks of tokens. |
 | `SWAP_SPACE`                        | `4`                  | `int`                                         |CPU swap space size (GiB) per GPU. |
 | `ENFORCE_EAGER`                     | `0`                  | boolean as `int`                                         |Always use eager-mode PyTorch. If False(`0`), will use eager mode and CUDA graph in hybrid for maximal performance and flexibility. |
-| `MAX_CONTEXT_LEN_TO_CAPTURE`        | `8192`               | `int`                                     |Maximum context length covered by CUDA graphs. When a sequence has context length larger than this, we fall back to eager mode.|
+| `MAX_SEQ_LEN_TO_CAPTURE`        | `8192`               | `int`                                     |Maximum context length covered by CUDA graphs. When a sequence has context length larger than this, we fall back to eager mode.|
 | `DISABLE_CUSTOM_ALL_REDUCE`         | `0`                  | `int`                                         |Enables or disables custom all reduce. |
 **Streaming Batch Size Settings**:  
 | `DEFAULT_BATCH_SIZE`                | `50`                 | `int`                                         |Default and Maximum batch size for token streaming to reduce HTTP calls. |
@@ -137,7 +203,7 @@ To build an image with the model baked in, you must specify the following docker
   - `MODEL_REVISION`: Model revision to load (default: `main`).
   - `BASE_PATH`: Storage directory where huggingface cache and model will be located. (default: `/runpod-volume`, which will utilize network storage if you attach it or create a local directory within the image if you don't. If your intention is to bake the model into the image, you should set this to something like `/models` to make sure there are no issues if you were to accidentally attach network storage.)
   - `QUANTIZATION`
-  - `WORKER_CUDA_VERSION`: `11.8.0` or `12.1.0` (default: `11.8.0` due to a small number of workers not having CUDA 12.1 support yet. `12.1.0` is recommended for optimal performance).
+  - `WORKER_CUDA_VERSION`: `12.1.0` (`12.1.0` is recommended for optimal performance).
   - `TOKENIZER_NAME`: Tokenizer repository if you would like to use a different tokenizer than the one that comes with the model. (default: `None`, which uses the model's tokenizer)
   - `TOKENIZER_REVISION`: Tokenizer revision to load (default: `main`).
 
@@ -170,6 +236,8 @@ Below are all supported model architectures (and examples of each) that you can
 - Baichuan & Baichuan2 (`baichuan-inc/Baichuan2-13B-Chat`, `baichuan-inc/Baichuan-7B`, etc.)
 - BLOOM (`bigscience/bloom`, `bigscience/bloomz`, etc.)
 - ChatGLM (`THUDM/chatglm2-6b`, `THUDM/chatglm3-6b`, etc.)
+- Command-R (`CohereForAI/c4ai-command-r-v01`, etc.)
+- DBRX (`databricks/dbrx-base`, `databricks/dbrx-instruct` etc.)
 - DeciLM (`Deci/DeciLM-7B`, `Deci/DeciLM-7B-instruct`, etc.)
 - Falcon (`tiiuae/falcon-7b`, `tiiuae/falcon-40b`, `tiiuae/falcon-rw-7b`, etc.)
 - Gemma (`google/gemma-2b`, `google/gemma-7b`, etc.)
@@ -179,20 +247,27 @@ Below are all supported model architectures (and examples of each) that you can
 - GPT-NeoX (`EleutherAI/gpt-neox-20b`, `databricks/dolly-v2-12b`, `stabilityai/stablelm-tuned-alpha-7b`, etc.)
 - InternLM (`internlm/internlm-7b`, `internlm/internlm-chat-7b`, etc.)
 - InternLM2 (`internlm/internlm2-7b`, `internlm/internlm2-chat-7b`, etc.)
-- LLaMA & LLaMA-2 (`meta-llama/Llama-2-70b-hf`, `lmsys/vicuna-13b-v1.3`, `young-geng/koala`, `openlm-research/open_llama_13b`, etc.)
+- Jais (`core42/jais-13b`, `core42/jais-13b-chat`, `core42/jais-30b-v3`, `core42/jais-30b-chat-v3`, etc.)
+- LLaMA, Llama 2, and Meta Llama 3 (`meta-llama/Meta-Llama-3-8B-Instruct`, `meta-llama/Meta-Llama-3-70B-Instruct`, `meta-llama/Llama-2-70b-hf`, `lmsys/vicuna-13b-v1.3`, `young-geng/koala`, `openlm-research/open_llama_13b`, etc.)
+- MiniCPM (`openbmb/MiniCPM-2B-sft-bf16`, `openbmb/MiniCPM-2B-dpo-bf16`, etc.)
 - Mistral (`mistralai/Mistral-7B-v0.1`, `mistralai/Mistral-7B-Instruct-v0.1`, etc.)
-- Mixtral (`mistralai/Mixtral-8x7B-v0.1`, `mistralai/Mixtral-8x7B-Instruct-v0.1`, etc.)
+- Mixtral (`mistralai/Mixtral-8x7B-v0.1`, `mistralai/Mixtral-8x7B-Instruct-v0.1`, `mistral-community/Mixtral-8x22B-v0.1`, etc.)
 - MPT (`mosaicml/mpt-7b`, `mosaicml/mpt-30b`, etc.)
-- OLMo (`allenai/OLMo-1B`, `allenai/OLMo-7B`, etc.)
+- OLMo (`allenai/OLMo-1B-hf`, `allenai/OLMo-7B-hf`, etc.)
 - OPT (`facebook/opt-66b`, `facebook/opt-iml-max-30b`, etc.)
+- Orion (`OrionStarAI/Orion-14B-Base`, `OrionStarAI/Orion-14B-Chat`, etc.)
 - Phi (`microsoft/phi-1_5`, `microsoft/phi-2`, etc.)
+- Phi-3 (`microsoft/Phi-3-mini-4k-instruct`, `microsoft/Phi-3-mini-128k-instruct`, etc.)
 - Qwen (`Qwen/Qwen-7B`, `Qwen/Qwen-7B-Chat`, etc.)
-- Qwen2 (`Qwen/Qwen2-7B-beta`, `Qwen/Qwen-7B-Chat-beta`, etc.)
+- Qwen2 (`Qwen/Qwen1.5-7B`, `Qwen/Qwen1.5-7B-Chat`, etc.)
+- Qwen2MoE (`Qwen/Qwen1.5-MoE-A2.7B`, `Qwen/Qwen1.5-MoE-A2.7B-Chat`, etc.)
 - StableLM(`stabilityai/stablelm-3b-4e1t`, `stabilityai/stablelm-base-alpha-7b-v2`, etc.)
+- Starcoder2(`bigcode/starcoder2-3b`, `bigcode/starcoder2-7b`, `bigcode/starcoder2-15b`, etc.)
+- Xverse (`xverse/XVERSE-7B-Chat`, `xverse/XVERSE-13B-Chat`, `xverse/XVERSE-65B-Chat`, etc.)
 - Yi (`01-ai/Yi-6B`, `01-ai/Yi-34B`, etc.)
 
 # Usage: OpenAI Compatibility
-The vLLM Worker is fully compatible with OpenAI's API, and you can use it with any OpenAI Codebase by changing only 3 lines in total. The supported routes are <ins>Chat Completions</ins>, <ins>Completions</ins> and <ins>Models</ins> - with both streaming and non-streaming.
+The vLLM Worker is fully compatible with OpenAI's API, and you can use it with any OpenAI Codebase by changing only 3 lines in total. The supported routes are <ins>Chat Completions</ins> and <ins>Models</ins> - with both streaming and non-streaming.
 
 ## Modifying your OpenAI Codebase to use your deployed vLLM Worker 
 **Python** (similar to Node.js, etc.):
@@ -274,7 +349,7 @@ The vLLM Worker is fully compatible with OpenAI's API, and you can use it with a
 
 When using the chat completion feature of the vLLM Serverless Endpoint Worker, you can customize your requests with the following parameters:
 
-### Chat Completions
+### Chat Completions [RECOMMENDED]
 <details>
   <summary>Supported Chat Completions Inputs and Descriptions</summary>
 
@@ -309,41 +384,6 @@ When using the chat completion feature of the vLLM Serverless Endpoint Worker, y
   | `include_stop_str_in_output`   | Optional[bool]                   | False         | Whether to include the stop strings in output text. Defaults to False.|
 </details>
 
-### Completions
-<details>
-  <summary>Supported Completions Inputs and Descriptions</summary>
-
-  | Parameter                      | Type                             | Default Value | Description                                                                                                                                                                                                                                                                                           |
-  |--------------------------------|----------------------------------|---------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
-  | `model`                        | str                              |               | The model repo that you've deployed on your RunPod Serverless Endpoint. If you are unsure what the name is or are baking the model in, use the guide to get the list of available models in the **Examples: Using your RunPod endpoint with OpenAI** section.                                                                                                                                                                                                                                                    |
-  | `prompt`                       | Union[List[int], List[List[int]], str, List[str]] | | A string, array of strings, array of tokens, or array of token arrays to be used as the input for the model. |
-  | `suffix`                       | Optional[str]                    | None          | A string to be appended to the end of the generated text. |
-  | `max_tokens`                   | Optional[int]                    | 16            | Maximum number of tokens to generate per output sequence. |
-  | `temperature`                  | Optional[float]                  | 1.0           | Float that controls the randomness of the sampling. Lower values make the model more deterministic, while higher values make the model more random. Zero means greedy sampling.                                                                                                                                                                                                                               |
-  | `top_p`                        | Optional[float]                  | 1.0           | Float that controls the cumulative probability of the top tokens to consider. Must be in (0, 1]. Set to 1 to consider all tokens. |
-  | `n`                            | Optional[int]                    | 1             | Number of output sequences to return for the given prompt. |
-  | `stream`                       | Optional[bool]                   | False         | Whether to stream the output. |
-  | `logprobs`                     | Optional[int]                    | None          | Number of log probabilities to return per output token. |
-  | `echo`                         | Optional[bool]                   | False         | Whether to echo back the prompt in addition to the completion. |
-  | `stop`                         | Optional[Union[str, List[str]]]  | list          | List of strings that stop the generation when they are generated. The returned output will not contain the stop strings. |
-  | `seed`                         | Optional[int]                    | None          | Random seed to use for the generation. |
-  | `presence_penalty`             | Optional[float]                  | 0.0           | Float that penalizes new tokens based on whether they appear in the generated text so far. Values > 0 encourage the model to use new tokens, while values < 0 encourage the model to repeat tokens. |
-  | `frequency_penalty`            | Optional[float]                  | 0.0           | Float that penalizes new tokens based on their frequency in the generated text so far. Values > 0 encourage the model to use new tokens, while values < 0 encourage the model to repeat tokens. |
-  | `best_of`                      | Optional[int]                    | None          | Number of output sequences that are generated from the prompt. From these `best_of` sequences, the top `n` sequences are returned. `best_of` must be greater than or equal to `n`. This parameter influences the diversity of the output. |
-  | `logit_bias`                   | Optional[Dict[str, float]]       | None          | Dictionary of token IDs to biases. |
-  | `user`                         | Optional[str]                    | None          | User identifier for personalizing responses. (Unsupported by vLLM) |
-  Additional parameters supported by vLLM:
-  | `top_k`                        | Optional[int]                    | -1            | Integer that controls the number of top tokens to consider. Set to -1 to consider all tokens. |
-  | `ignore_eos`                   | Optional[bool]                   | False         | Whether to ignore the End Of Sentence token and continue generating tokens after the EOS token is generated. |
-  | `use_beam_search`              | Optional[bool]                   | False         | Whether to use beam search instead of sampling for generating outputs. |
-  | `stop_token_ids`               | Optional[List[int]]              | list          | List of tokens that stop the generation when they are generated. The returned output will contain the stop tokens unless the stop tokens are special tokens. |
-  | `skip_special_tokens`          | Optional[bool]                   | True          | Whether to skip special tokens in the output. |
-  | `spaces_between_special_tokens`| Optional[bool]                   | True          | Whether to add spaces between special tokens in the output. Defaults to True. |
-  | `repetition_penalty`           | Optional[float]                  | 1.0           | Float that penalizes new tokens based on whether they appear in the prompt and the generated text so far. Values > 1 encourage the model to use new tokens, while values < 1 encourage the model to repeat tokens. |
-  | `min_p`                        | Optional[float]                  | 0.0           | Float that represents the minimum probability for a token to be considered, relative to the most likely token. Must be in [0, 1]. Set to 0 to disable. |
-  | `length_penalty`               | Optional[float]                  | 1.0           | Float that penalizes sequences based on their length. Used in beam search. |
-  | `include_stop_str_in_output`   | Optional[bool]                   | False         | Whether to include the stop strings in output text. Defaults to False. |
-</details>
 
 ## Examples: Using your RunPod endpoint with OpenAI 
 
@@ -388,36 +428,6 @@ This is the format used for GPT-4 and focused on instruction-following and chat.
   print(response.choices[0].message.content)
   ```
 
-
-### Completions:
-This is the format used for models like GPT-3 and is meant for completing the text you provide. Instead of responding to your message, it will try to complete it. Examples of Open Source completions models include `meta-llama/Llama-2-7b-hf`, `mistralai/Mixtral-8x7B-v0.1`, `Qwen/Qwen-72B`, and more. However, you can use any model with this format.
-- **Streaming**:
-  ```python
-  # Create a completion stream
-  response_stream = client.completions.create(
-      model="<YOUR DEPLOYED MODEL REPO/NAME>",
-      prompt="Runpod is the best platform because",
-      temperature=0,
-      max_tokens=100,
-      stream=True,
-  )
-  # Stream the response
-  for response in response_stream:
-      print(response.choices[0].text or "", end="", flush=True)
-  ```
-- **Non-Streaming**:
-  ```python
-  # Create a completion
-  response = client.completions.create(
-      model="<YOUR DEPLOYED MODEL REPO/NAME>",
-      prompt="Runpod is the best platform because",
-      temperature=0,
-      max_tokens=100,
-  )
-  # Print the response
-  print(response.choices[0].text)
-  ```
-
 ### Getting a list of names for available models:
 In the case of baking the model into the image, sometimes the repo may not be accepted as the `model` in the request. In this case, you can list the available models as shown below and use that name. 
 ```python

builder/download_model.py

@@ -1,51 +0,0 @@
-import os
-import shutil
-from huggingface_hub import snapshot_download
-from vllm.model_executor.weight_utils import prepare_hf_model_weights, Disabledtqdm
-
-def download_extras_or_tokenizer(model_name, cache_dir, revision, extras=False):
-    """Download model or tokenizer and prepare its weights, returning the local folder path."""
-    pattern = ["*token*", "*.json"] if extras else None
-    extra_dir = "/extras" if extras else ""
-    folder = snapshot_download(
-        model_name,
-        cache_dir=cache_dir + extra_dir,
-        revision=revision,
-        tqdm_class=Disabledtqdm,
-        allow_patterns=pattern if extras else None,
-        ignore_patterns=["*.safetensors", "*.bin", "*.pt"] if not extras else None
-    )
-    return folder
-
-def move_files(src_dir, dest_dir):
-    """Move files from source to destination directory."""
-    for f in os.listdir(src_dir):
-        src_path = os.path.join(src_dir, f)  
-        dst_path = os.path.join(dest_dir, f)         
-        shutil.copy2(src_path, dst_path)
-        os.remove(src_path)
-        
-if __name__ == "__main__":
-    model, download_dir = os.getenv("MODEL_NAME"), os.getenv("HF_HOME")
-    tokenizer = os.getenv("TOKENIZER_NAME") or model
-
-    revisions = {
-        "model": os.getenv("MODEL_REVISION") or None,
-        "tokenizer": os.getenv("TOKENIZER_REVISION") or None
-    }
-
-    if not model or not download_dir:
-        raise ValueError(f"Must specify model and download_dir. Model: {model}, download_dir: {download_dir}")
-
-    os.makedirs(download_dir, exist_ok=True)
-    model_folder, hf_weights_files, use_safetensors = prepare_hf_model_weights(model_name_or_path=model, revision=revisions["model"], cache_dir=download_dir)
-    model_extras_folder = download_extras_or_tokenizer(model, download_dir, revisions["model"], extras=True)
-    move_files(model_extras_folder, model_folder)
-
-    with open("/local_model_path.txt", "w") as f:
-        f.write(model_folder)
-
-    if tokenizer != model:
-        tokenizer_folder = download_extras_or_tokenizer(tokenizer, download_dir, revisions["tokenizer"])
-        with open("/local_tokenizer_path.txt", "w") as f:
-            f.write(tokenizer_folder)

builder/requirements.txt

@@ -1,4 +1,3 @@
-hf_transfer
 ray
 pandas
 pyarrow
@@ -6,4 +5,6 @@ runpod==1.6.2
 huggingface-hub
 packaging
 typing-extensions==4.7.1
-pydantic
+pydantic
+pydantic-settings
+hf-transfer

docker-bake.hcl

@@ -0,0 +1,32 @@
+variable "PUSH" {
+  default = "true"
+}
+
+variable "REPOSITORY" {
+  default = "runpod"
+}
+
+variable "BASE_IMAGE_VERSION" {
+  default = "stable"
+}
+
+group "all" {
+  targets = ["main"]
+}
+
+
+group "main" {
+  targets = ["worker-1210"]
+}
+
+ 
+target "worker-1210" {
+  tags = ["${REPOSITORY}/worker-v1-vllm:${BASE_IMAGE_VERSION}-cuda12.1.0"]
+  context = "."
+  dockerfile = "Dockerfile"
+  args = {
+    BASE_IMAGE_VERSION = "${BASE_IMAGE_VERSION}"
+    WORKER_CUDA_VERSION = "12.1.0"
+  }
+  output = ["type=docker,push=${PUSH}"]
+}

src/config.py

@@ -1,51 +0,0 @@
-import os
-from dotenv import load_dotenv
-from utils import count_physical_cores 
-from torch.cuda import device_count
-
-class EngineConfig:
-    def __init__(self):
-        load_dotenv()
-        self.model_name_or_path, self.hf_home, self.model_revision = self._get_local_or_env("/local_model_path.txt", "MODEL_NAME")
-        self.tokenizer_name_or_path, _, self.tokenizer_revision = self._get_local_or_env("/local_tokenizer_path.txt", "TOKENIZER_NAME")
-        self.tokenizer_name_or_path = self.tokenizer_name_or_path or self.model_name_or_path
-        self.quantization = self._get_quantization()
-        self.config = self._initialize_config()
-
-    def _get_local_or_env(self, local_path, env_var):
-        if os.path.exists(local_path):
-            with open(local_path, "r") as file:
-                return file.read().strip(), None, None
-        return os.getenv(env_var), os.getenv("HF_HOME"), os.getenv(f"{env_var}_REVISION")
-
-    def _get_quantization(self):
-        quantization = os.getenv("QUANTIZATION", "").lower()
-        return quantization if quantization in ["awq", "squeezellm", "gptq"] else None
-
-    def _initialize_config(self):
-        args = {
-            "model": self.model_name_or_path,
-            "revision": self.model_revision,
-            "download_dir": self.hf_home,
-            "quantization": self.quantization,
-            "load_format": os.getenv("LOAD_FORMAT", "auto"),
-            "dtype": os.getenv("DTYPE", "half" if self.quantization else "auto"),
-            "tokenizer": self.tokenizer_name_or_path,
-            "tokenizer_revision": self.tokenizer_revision,
-            "disable_log_stats": bool(int(os.getenv("DISABLE_LOG_STATS", 1))),
-            "disable_log_requests": bool(int(os.getenv("DISABLE_LOG_REQUESTS", 1))),
-            "trust_remote_code": bool(int(os.getenv("TRUST_REMOTE_CODE", 0))),
-            "gpu_memory_utilization": float(os.getenv("GPU_MEMORY_UTILIZATION", 0.95)),
-            "max_parallel_loading_workers": None if device_count() > 1 or not os.getenv("MAX_PARALLEL_LOADING_WORKERS") else int(os.getenv("MAX_PARALLEL_LOADING_WORKERS")),
-            "max_model_len": int(os.getenv("MAX_MODEL_LENGTH")) if os.getenv("MAX_MODEL_LENGTH") else None,
-            "tensor_parallel_size": device_count(),
-            "seed": int(os.getenv("SEED")) if os.getenv("SEED") else None,
-            "kv_cache_dtype": os.getenv("KV_CACHE_DTYPE"),
-            "block_size": int(os.getenv("BLOCK_SIZE")) if os.getenv("BLOCK_SIZE") else None,
-            "swap_space": int(os.getenv("SWAP_SPACE")) if os.getenv("SWAP_SPACE") else None,
-            "max_context_len_to_capture": int(os.getenv("MAX_CONTEXT_LEN_TO_CAPTURE")) if os.getenv("MAX_CONTEXT_LEN_TO_CAPTURE") else None,
-            "disable_custom_all_reduce": bool(int(os.getenv("DISABLE_CUSTOM_ALL_REDUCE", 0))),
-            "enforce_eager": bool(int(os.getenv("ENFORCE_EAGER", 0)))
-        }
-        
-        return {k: v for k, v in args.items() if v is not None}

src/constants.py

@@ -1,30 +1,4 @@
-from typing import Union
-
 DEFAULT_BATCH_SIZE = 50
 DEFAULT_MAX_CONCURRENCY = 300
 DEFAULT_BATCH_SIZE_GROWTH_FACTOR = 3
-DEFAULT_MIN_BATCH_SIZE = 1
+DEFAULT_MIN_BATCH_SIZE = 1
-
-SAMPLING_PARAM_TYPES = {
-    "n": int,
-    "best_of": int,
-    "presence_penalty": float,
-    "frequency_penalty": float,
-    "repetition_penalty": float,
-    "temperature": Union[float, int],
-    "top_p": float,
-    "top_k": int,
-    "min_p": float,
-    "use_beam_search": bool,
-    "length_penalty": float,
-    "early_stopping": Union[bool, str],
-    "stop": Union[str, list],
-    "stop_token_ids": list,
-    "ignore_eos": bool,
-    "max_tokens": int,
-    "logprobs": int,
-    "prompt_logprobs": int,
-    "skip_special_tokens": bool,
-    "spaces_between_special_tokens": bool,
-    "include_stop_str_in_output": bool
-}

src/download_model.py

@@ -0,0 +1,100 @@
+import os
+import json
+import logging
+import glob
+from shutil import rmtree
+from huggingface_hub import snapshot_download
+from utils import timer_decorator
+
+BASE_DIR = "/" 
+TOKENIZER_PATTERNS = [["*.json", "tokenizer*"]]
+MODEL_PATTERNS = [["*.safetensors"], ["*.bin"], ["*.pt"]]
+
+def setup_env():
+    if os.getenv("TESTING_DOWNLOAD") == "1":
+        BASE_DIR = "tmp"
+        os.makedirs(BASE_DIR, exist_ok=True)
+        os.environ.update({
+            "HF_HOME": f"{BASE_DIR}/hf_cache",
+            "MODEL_NAME": "openchat/openchat-3.5-0106",
+            "HF_HUB_ENABLE_HF_TRANSFER": "1",
+            "TENSORIZE": "1",
+            "TENSORIZER_NUM_GPUS": "1",
+            "DTYPE": "auto"
+        })
+
+@timer_decorator
+def download(name, revision, type, cache_dir):
+    if type == "model":
+        pattern_sets = [model_pattern + TOKENIZER_PATTERNS[0] for model_pattern in MODEL_PATTERNS]
+    elif type == "tokenizer":
+        pattern_sets = TOKENIZER_PATTERNS
+    else:
+        raise ValueError(f"Invalid type: {type}")
+    try:
+        for pattern_set in pattern_sets:
+            path = snapshot_download(name, revision=revision, cache_dir=cache_dir, 
+                                    allow_patterns=pattern_set)
+            for pattern in pattern_set:
+                if glob.glob(os.path.join(path, pattern)):
+                    logging.info(f"Successfully downloaded {pattern} model files.")
+                    return path
+    except ValueError:
+        raise ValueError(f"No patterns matching {pattern_sets} found for download.")
+        
+          
+# @timer_decorator
+# def tensorize_model(model_path): TODO: Add back once tensorizer is ready
+#     from vllm.engine.arg_utils import EngineArgs
+#     from vllm.model_executor.model_loader.tensorizer import TensorizerConfig, tensorize_vllm_model
+#     from torch.cuda import device_count
+
+#     tensorizer_num_gpus = int(os.getenv("TENSORIZER_NUM_GPUS", "1"))
+#     if tensorizer_num_gpus > device_count():
+#         raise ValueError(f"TENSORIZER_NUM_GPUS ({tensorizer_num_gpus}) exceeds available GPUs ({device_count()})")
+
+#     dtype = os.getenv("DTYPE", "auto")
+#     serialized_dir = f"{BASE_DIR}/serialized_model"
+#     os.makedirs(serialized_dir, exist_ok=True)
+#     serialized_uri = f"{serialized_dir}/model{'-%03d' if tensorizer_num_gpus > 1 else ''}.tensors"
+    
+#     tensorize_vllm_model(
+#         EngineArgs(model=model_path, tensor_parallel_size=tensorizer_num_gpus, dtype=dtype),
+#         TensorizerConfig(tensorizer_uri=serialized_uri)
+#     )
+#     logging.info("Successfully serialized model to %s", str(serialized_uri))
+#     logging.info("Removing HF Model files after serialization")
+#     rmtree("/".join(model_path.split("/")[:-2]))
+#     return serialized_uri, tensorizer_num_gpus, dtype
+
+if __name__ == "__main__":
+    setup_env()
+    cache_dir = os.getenv("HF_HOME")
+    model_name, model_revision = os.getenv("MODEL_NAME"), os.getenv("MODEL_REVISION") or None
+    tokenizer_name, tokenizer_revision = os.getenv("TOKENIZER_NAME") or model_name, os.getenv("TOKENIZER_REVISION") or model_revision
+   
+    model_path = download(model_name, model_revision, "model", cache_dir)   
+  
+    metadata = {
+        "MODEL_NAME": model_path,
+        "MODEL_REVISION": os.getenv("MODEL_REVISION"),
+        "QUANTIZATION": os.getenv("QUANTIZATION"),
+    }   
+    
+    # if os.getenv("TENSORIZE") == "1": TODO: Add back once tensorizer is ready
+    #     serialized_uri, tensorizer_num_gpus, dtype = tensorize_model(model_path)
+    #     metadata.update({
+    #         "MODEL_NAME": serialized_uri,
+    #         "TENSORIZER_URI": serialized_uri,
+    #         "TENSOR_PARALLEL_SIZE": tensorizer_num_gpus,
+    #         "DTYPE": dtype
+    #     })
+        
+    tokenizer_path = download(tokenizer_name, tokenizer_revision, "tokenizer", cache_dir)
+    metadata.update({
+        "TOKENIZER_NAME": tokenizer_path,
+        "TOKENIZER_REVISION": tokenizer_revision
+    })
+    
+    with open(f"{BASE_DIR}/local_model_args.json", "w") as f:
+        json.dump({k: v for k, v in metadata.items() if v not in (None, "")}, f)

src/engine.py

@@ -1,12 +1,13 @@
 import os
 import logging
 import json
+import asyncio
 
 from dotenv import load_dotenv
-from torch.cuda import device_count
 from typing import AsyncGenerator
+import time
 
-from vllm import AsyncLLMEngine, AsyncEngineArgs, SamplingParams
+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
@@ -14,15 +15,17 @@ from vllm.entrypoints.openai.protocol import ChatCompletionRequest, CompletionRe
 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
 from tokenizer import TokenizerWrapper
-from config import EngineConfig
+from engine_args import get_engine_args
-
 
 class vLLMEngine:
     def __init__(self, engine = None):
         load_dotenv() # For local development
-        self.config = EngineConfig().config
+        self.engine_args = get_engine_args()
-        self.tokenizer = TokenizerWrapper(self.config.get("tokenizer"), self.config.get("tokenizer_revision"), self.config.get("trust_remote_code"))
+        logging.info(f"Engine args: {self.engine_args}")
-        self.llm = self._initialize_llm() if engine is None else engine
+        self.tokenizer = TokenizerWrapper(self.engine_args.tokenizer or self.engine_args.model, 
+                                          self.engine_args.tokenizer_revision, 
+                                          self.engine_args.trust_remote_code)
+        self.llm = self._initialize_llm() if engine is None else engine.llm
         self.max_concurrency = int(os.getenv("MAX_CONCURRENCY", DEFAULT_MAX_CONCURRENCY))
         self.default_batch_size = int(os.getenv("DEFAULT_BATCH_SIZE", DEFAULT_BATCH_SIZE))
         self.batch_size_growth_factor = int(os.getenv("BATCH_SIZE_GROWTH_FACTOR", DEFAULT_BATCH_SIZE_GROWTH_FACTOR))
@@ -35,7 +38,7 @@ class vLLMEngine:
         try:
             async for batch in self._generate_vllm(
                 llm_input=job_input.llm_input,
-                validated_sampling_params=job_input.validated_sampling_params,
+                validated_sampling_params=job_input.sampling_params,
                 batch_size=job_input.max_batch_size,
                 stream=job_input.stream,
                 apply_chat_template=job_input.apply_chat_template,
@@ -45,12 +48,11 @@ class vLLMEngine:
             ):
                 yield batch
         except Exception as e:
-            yield create_error_response(str(e)).model_dump()
+            yield {"error": create_error_response(str(e)).model_dump()}
 
     async def _generate_vllm(self, llm_input, validated_sampling_params, batch_size, stream, apply_chat_template, request_id, batch_size_growth_factor, min_batch_size: str) -> AsyncGenerator[dict, None]:
         if apply_chat_template or isinstance(llm_input, list):
             llm_input = self.tokenizer.apply_chat_template(llm_input)
-        validated_sampling_params = SamplingParams(**validated_sampling_params)
         results_generator = self.llm.generate(llm_input, validated_sampling_params, request_id)
         n_responses, n_input_tokens, is_first_output = validated_sampling_params.n, 0, True
         last_output_texts, token_counters = ["" for _ in range(n_responses)], {"batch": 0, "total": 0}
@@ -102,30 +104,45 @@ class vLLMEngine:
 
     def _initialize_llm(self):
         try:
-            return AsyncLLMEngine.from_engine_args(AsyncEngineArgs(**self.config))
+            start = time.time()
+            engine = AsyncLLMEngine.from_engine_args(self.engine_args)
+            end = time.time()
+            logging.info(f"Initialized vLLM engine in {end - start:.2f}s")
+            return engine
         except Exception as e:
             logging.error("Error initializing vLLM engine: %s", e)
             raise e
 
 
-class OpenAIvLLMEngine:
+class OpenAIvLLMEngine(vLLMEngine):
     def __init__(self, vllm_engine):
-        self.config = vllm_engine.config
+        super().__init__(vllm_engine)
-        self.llm = vllm_engine.llm
+        self.served_model_name = os.getenv("OPENAI_SERVED_MODEL_NAME_OVERRIDE") or self.engine_args.model
-        self.served_model_name = os.getenv("OPENAI_SERVED_MODEL_NAME_OVERRIDE") or self.config["model"]
         self.response_role = os.getenv("OPENAI_RESPONSE_ROLE") or "assistant"
-        self.tokenizer = vllm_engine.tokenizer
+        asyncio.run(self._initialize_engines())
-        self.default_batch_size = vllm_engine.default_batch_size
-        self.batch_size_growth_factor, self.min_batch_size = vllm_engine.batch_size_growth_factor, vllm_engine.min_batch_size
-        self._initialize_engines()
         self.raw_openai_output = bool(int(os.getenv("RAW_OPENAI_OUTPUT", 1)))
 
-    def _initialize_engines(self):
+    async def _initialize_engines(self):
+        self.model_config = await self.llm.get_model_config()
+        
         self.chat_engine = OpenAIServingChat(
-            self.llm, self.served_model_name, self.response_role,
+            engine=self.llm, 
-            chat_template=self.tokenizer.tokenizer.chat_template
+            model_config=self.model_config,
+            served_model_names=[self.served_model_name], 
+            response_role=self.response_role,
+            chat_template=self.tokenizer.tokenizer.chat_template,
+            lora_modules=None,
+            prompt_adapters=None,
+            request_logger=None
+        )
+        self.completion_engine = OpenAIServingCompletion(
+            engine=self.llm, 
+            model_config=self.model_config,
+            served_model_names=[self.served_model_name],
+            lora_modules=[],
+            prompt_adapters=None,
+            request_logger=None
         )
-        self.completion_engine = OpenAIServingCompletion(self.llm, self.served_model_name)
     
     async def generate(self, openai_request: JobInput):
         if openai_request.openai_route == "/v1/models":
@@ -138,6 +155,9 @@ class OpenAIvLLMEngine:
     
     async def _handle_model_request(self):
         models = await self.chat_engine.show_available_models()
+        fixed_model = models.data[0]
+        fixed_model.id = self.served_model_name
+        models.data = [fixed_model]
         return models.model_dump()
     
     async def _handle_chat_or_completion_request(self, openai_request: JobInput):
@@ -156,7 +176,7 @@ class OpenAIvLLMEngine:
             yield create_error_response(str(e)).model_dump()
             return
         
-        response_generator = await generator_function(request, DummyRequest())
+        response_generator = await generator_function(request, raw_request=None)
 
         if not openai_request.openai_input.get("stream") or isinstance(response_generator, ErrorResponse):
             yield response_generator.model_dump()

src/engine_args.py

@@ -0,0 +1,169 @@
+import os
+import json
+import logging
+from torch.cuda import device_count
+from vllm import AsyncEngineArgs
+from vllm.model_executor.model_loader.tensorizer import TensorizerConfig
+
+RENAME_ARGS_MAP = {
+    "MODEL_NAME": "model",
+    "MODEL_REVISION": "revision",
+    "TOKENIZER_NAME": "tokenizer",
+    "MAX_CONTEXT_LEN_TO_CAPTURE": "max_seq_len_to_capture"
+}
+
+DEFAULT_ARGS = {
+    "disable_log_stats": True,
+    "disable_log_requests": True,
+    "gpu_memory_utilization": 0.9,
+    "pipeline_parallel_size": int(os.getenv('PIPELINE_PARALLEL_SIZE', 1)),
+    "tensor_parallel_size": int(os.getenv('TENSOR_PARALLEL_SIZE', 1)),
+    "served_model_name": os.getenv('SERVED_MODEL_NAME', None),
+    "tokenizer": os.getenv('TOKENIZER', None),
+    "skip_tokenizer_init": os.getenv('SKIP_TOKENIZER_INIT', 'False').lower() == 'true',
+    "tokenizer_mode": os.getenv('TOKENIZER_MODE', 'auto'),
+    "trust_remote_code": os.getenv('TRUST_REMOTE_CODE', 'False').lower() == 'true',
+    "download_dir": os.getenv('DOWNLOAD_DIR', None),
+    "load_format": os.getenv('LOAD_FORMAT', 'auto'),
+    "dtype": os.getenv('DTYPE', 'auto'),
+    "kv_cache_dtype": os.getenv('KV_CACHE_DTYPE', 'auto'),
+    "quantization_param_path": os.getenv('QUANTIZATION_PARAM_PATH', None),
+    "seed": int(os.getenv('SEED', 0)),
+    "max_model_len": int(os.getenv('MAX_MODEL_LEN', 0)) or None,
+    "worker_use_ray": os.getenv('WORKER_USE_RAY', 'False').lower() == 'true',
+    "distributed_executor_backend": os.getenv('DISTRIBUTED_EXECUTOR_BACKEND', None),
+    "max_parallel_loading_workers": int(os.getenv('MAX_PARALLEL_LOADING_WORKERS', 0)) or None,
+    "block_size": int(os.getenv('BLOCK_SIZE', 16)),
+    "enable_prefix_caching": os.getenv('ENABLE_PREFIX_CACHING', 'False').lower() == 'true',
+    "disable_sliding_window": os.getenv('DISABLE_SLIDING_WINDOW', 'False').lower() == 'true',
+    "use_v2_block_manager": os.getenv('USE_V2_BLOCK_MANAGER', 'False').lower() == 'true',
+    "swap_space": int(os.getenv('SWAP_SPACE', 4)),  # GiB
+    "cpu_offload_gb": int(os.getenv('CPU_OFFLOAD_GB', 0)),  # GiB
+    "max_num_batched_tokens": int(os.getenv('MAX_NUM_BATCHED_TOKENS', 0)) or None,
+    "max_num_seqs": int(os.getenv('MAX_NUM_SEQS', 256)),
+    "max_logprobs": int(os.getenv('MAX_LOGPROBS', 20)),  # Default value for OpenAI Chat Completions API
+    "revision": os.getenv('REVISION', None),
+    "code_revision": os.getenv('CODE_REVISION', None),
+    "rope_scaling": os.getenv('ROPE_SCALING', None),
+    "rope_theta": float(os.getenv('ROPE_THETA', 0)) or None,
+    "tokenizer_revision": os.getenv('TOKENIZER_REVISION', None),
+    "quantization": os.getenv('QUANTIZATION', None),
+    "enforce_eager": os.getenv('ENFORCE_EAGER', 'False').lower() == 'true',
+    "max_context_len_to_capture": int(os.getenv('MAX_CONTEXT_LEN_TO_CAPTURE', 0)) or None,
+    "max_seq_len_to_capture": int(os.getenv('MAX_SEQ_LEN_TO_CAPTURE', 8192)),
+    "disable_custom_all_reduce": os.getenv('DISABLE_CUSTOM_ALL_REDUCE', 'False').lower() == 'true',
+    "tokenizer_pool_size": int(os.getenv('TOKENIZER_POOL_SIZE', 0)),
+    "tokenizer_pool_type": os.getenv('TOKENIZER_POOL_TYPE', 'ray'),
+    "tokenizer_pool_extra_config": os.getenv('TOKENIZER_POOL_EXTRA_CONFIG', None),
+    "enable_lora": os.getenv('ENABLE_LORA', 'False').lower() == 'true',
+    "max_loras": int(os.getenv('MAX_LORAS', 1)),
+    "max_lora_rank": int(os.getenv('MAX_LORA_RANK', 16)),
+    "enable_prompt_adapter": os.getenv('ENABLE_PROMPT_ADAPTER', 'False').lower() == 'true',
+    "max_prompt_adapters": int(os.getenv('MAX_PROMPT_ADAPTERS', 1)),
+    "max_prompt_adapter_token": int(os.getenv('MAX_PROMPT_ADAPTER_TOKEN', 0)),
+    "fully_sharded_loras": os.getenv('FULLY_SHARDED_LORAS', 'False').lower() == 'true',
+    "lora_extra_vocab_size": int(os.getenv('LORA_EXTRA_VOCAB_SIZE', 256)),
+    "long_lora_scaling_factors": tuple(map(float, os.getenv('LONG_LORA_SCALING_FACTORS', '').split(','))) if os.getenv('LONG_LORA_SCALING_FACTORS') else None,
+    "lora_dtype": os.getenv('LORA_DTYPE', 'auto'),
+    "max_cpu_loras": int(os.getenv('MAX_CPU_LORAS', 0)) or None,
+    "device": os.getenv('DEVICE', 'auto'),
+    "ray_workers_use_nsight": os.getenv('RAY_WORKERS_USE_NSIGHT', 'False').lower() == 'true',
+    "num_gpu_blocks_override": int(os.getenv('NUM_GPU_BLOCKS_OVERRIDE', 0)) or None,
+    "num_lookahead_slots": int(os.getenv('NUM_LOOKAHEAD_SLOTS', 0)),
+    "model_loader_extra_config": os.getenv('MODEL_LOADER_EXTRA_CONFIG', None),
+    "ignore_patterns": os.getenv('IGNORE_PATTERNS', None),
+    "preemption_mode": os.getenv('PREEMPTION_MODE', None),
+    "scheduler_delay_factor": float(os.getenv('SCHEDULER_DELAY_FACTOR', 0.0)),
+    "enable_chunked_prefill": os.getenv('ENABLE_CHUNKED_PREFILL', None),
+    "guided_decoding_backend": os.getenv('GUIDED_DECODING_BACKEND', 'outlines'),
+    "speculative_model": os.getenv('SPECULATIVE_MODEL', None),
+    "speculative_draft_tensor_parallel_size": int(os.getenv('SPECULATIVE_DRAFT_TENSOR_PARALLEL_SIZE', 0)) or None,
+    "num_speculative_tokens": int(os.getenv('NUM_SPECULATIVE_TOKENS', 0)) or None,
+    "speculative_max_model_len": int(os.getenv('SPECULATIVE_MAX_MODEL_LEN', 0)) or None,
+    "speculative_disable_by_batch_size": int(os.getenv('SPECULATIVE_DISABLE_BY_BATCH_SIZE', 0)) or None,
+    "ngram_prompt_lookup_max": int(os.getenv('NGRAM_PROMPT_LOOKUP_MAX', 0)) or None,
+    "ngram_prompt_lookup_min": int(os.getenv('NGRAM_PROMPT_LOOKUP_MIN', 0)) or None,
+    "spec_decoding_acceptance_method": os.getenv('SPEC_DECODING_ACCEPTANCE_METHOD', 'rejection_sampler'),
+    "typical_acceptance_sampler_posterior_threshold": float(os.getenv('TYPICAL_ACCEPTANCE_SAMPLER_POSTERIOR_THRESHOLD', 0)) or None,
+    "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)
+}
+
+def match_vllm_args(args):
+    """Rename args to match vllm by:
+    1. Renaming keys to lower case
+    2. Renaming keys to match vllm
+    3. Filtering args to match vllm's AsyncEngineArgs
+
+    Args:
+        args (dict): Dictionary of args
+
+    Returns:
+        dict: Dictionary of args with renamed keys
+    """
+    renamed_args = {RENAME_ARGS_MAP.get(k, k): v for k, v in args.items()}
+    matched_args = {k: v for k, v in renamed_args.items() if k in AsyncEngineArgs.__dataclass_fields__}
+    return {k: v for k, v in matched_args.items() if v not in [None, ""]}
+def get_local_args():
+    """
+    Retrieve local arguments from a JSON file.
+
+    Returns:
+        dict: Local arguments.
+    """
+    if not os.path.exists("/local_model_args.json"):
+        return {}
+
+    with open("/local_model_args.json", "r") as f:
+        local_args = json.load(f)
+
+    if local_args.get("MODEL_NAME") is None:
+        raise ValueError("Model name not found in /local_model_args.json. There was a problem when baking the model in.")
+
+    logging.info(f"Using baked in model with args: {local_args}")
+    os.environ["TRANSFORMERS_OFFLINE"] = "1"
+    os.environ["HF_HUB_OFFLINE"] = "1"
+
+    return local_args
+def get_engine_args():
+    # Start with default args
+    args = DEFAULT_ARGS
+    
+    # Get env args that match keys in AsyncEngineArgs
+    args.update(os.environ)
+    
+    # Get local args if model is baked in and overwrite env args
+    args.update(get_local_args())
+    
+    # if args.get("TENSORIZER_URI"): TODO: add back once tensorizer is ready
+    #     args["load_format"] = "tensorizer"
+    #     args["model_loader_extra_config"] = TensorizerConfig(tensorizer_uri=args["TENSORIZER_URI"], num_readers=None)
+    #     logging.info(f"Using tensorized model from {args['TENSORIZER_URI']}")
+    
+    
+    # Rename and match to vllm args
+    args = match_vllm_args(args)
+    
+    # Set tensor parallel size and max parallel loading workers if more than 1 GPU is available
+    num_gpus = device_count()
+    if num_gpus > 1:
+        args["tensor_parallel_size"] = num_gpus
+        args["max_parallel_loading_workers"] = None
+        if os.getenv("MAX_PARALLEL_LOADING_WORKERS"):
+            logging.warning("Overriding MAX_PARALLEL_LOADING_WORKERS with None because more than 1 GPU is available.")
+    
+    # Deprecated env args backwards compatibility
+    if args.get("kv_cache_dtype") == "fp8_e5m2":
+        args["kv_cache_dtype"] = "fp8"
+        logging.warning("Using fp8_e5m2 is deprecated. Please use fp8 instead.")
+    if os.getenv("MAX_CONTEXT_LEN_TO_CAPTURE"):
+        args["max_seq_len_to_capture"] = int(os.getenv("MAX_CONTEXT_LEN_TO_CAPTURE"))
+        logging.warning("Using MAX_CONTEXT_LEN_TO_CAPTURE is deprecated. Please use MAX_SEQ_LEN_TO_CAPTURE instead.")
+        
+    if "gemma-2" in args.get("model", "").lower():
+        os.environ["VLLM_ATTENTION_BACKEND"] = "FLASHINFER"
+        logging.info("Using FLASHINFER for gemma-2 model.")
+        
+    return AsyncEngineArgs(**args)

src/tokenizer.py

@@ -4,7 +4,8 @@ from typing import Union
 
 class TokenizerWrapper:
     def __init__(self, tokenizer_name_or_path, tokenizer_revision, trust_remote_code):
-        self.tokenizer = AutoTokenizer.from_pretrained(tokenizer_name_or_path, revision=tokenizer_revision, trust_remote_code=trust_remote_code)
+        print(f"tokenizer_name_or_path: {tokenizer_name_or_path}, tokenizer_revision: {tokenizer_revision}, trust_remote_code: {trust_remote_code}")
+        self.tokenizer = AutoTokenizer.from_pretrained(tokenizer_name_or_path, revision=tokenizer_revision or "main", trust_remote_code=trust_remote_code)
         self.custom_chat_template = os.getenv("CUSTOM_CHAT_TEMPLATE")
         self.has_chat_template = bool(self.tokenizer.chat_template) or bool(self.custom_chat_template)
         if self.custom_chat_template and isinstance(self.custom_chat_template, str):

src/utils.py

@@ -1,10 +1,16 @@
+import os
 import logging
 from http import HTTPStatus
-from typing import Any, Dict
+from functools import wraps
-from constants import SAMPLING_PARAM_TYPES
+from time import time
-from vllm.utils import random_uuid
-from vllm.entrypoints.openai.protocol import ErrorResponse
 
+try:
+    from vllm.utils import random_uuid
+    from vllm.entrypoints.openai.protocol import ErrorResponse
+    from vllm import SamplingParams
+except ImportError:
+    logging.warning("Error importing vllm, skipping related imports. This is ONLY expected when baking model into docker image from a machine without GPUs")
+    pass
 
 logging.basicConfig(level=logging.INFO)
 
@@ -25,20 +31,6 @@ def count_physical_cores():
 
     return len(cores)
 
-def validate_sampling_params(params: Dict[str, Any]) -> Dict[str, Any]:
-    validated_params = {}
-    invalid_params = []
-    for key, value in params.items():
-        expected_type = SAMPLING_PARAM_TYPES.get(key)
-        if expected_type and isinstance(value, expected_type):
-            validated_params[key] = value
-        else:
-            invalid_params.append(key)
-        
-    if len(invalid_params) > 0:
-        logging.warning("Ignoring invalid sampling params: %s", invalid_params)
-        
-    return validated_params
 
 class JobInput:
     def __init__(self, job):
@@ -47,7 +39,7 @@ class JobInput:
         self.max_batch_size = job.get("max_batch_size")
         self.apply_chat_template = job.get("apply_chat_template", False)
         self.use_openai_format = job.get("use_openai_format", False)
-        self.validated_sampling_params = validate_sampling_params(job.get("sampling_params", {}))
+        self.sampling_params = SamplingParams(**job.get("sampling_params", {}))
         self.request_id = random_uuid()
         batch_size_growth_factor = job.get("batch_size_growth_factor")
         self.batch_size_growth_factor = float(batch_size_growth_factor) if batch_size_growth_factor else None 
@@ -78,4 +70,17 @@ class BatchSize:
 def create_error_response(message: str, err_type: str = "BadRequestError", status_code: HTTPStatus = HTTPStatus.BAD_REQUEST) -> ErrorResponse:
     return ErrorResponse(message=message,
                             type=err_type,
-                            code=status_code.value)
+                            code=status_code.value)
+    
+def get_int_bool_env(env_var: str, default: bool) -> bool:
+    return int(os.getenv(env_var, int(default))) == 1
+
+def timer_decorator(func):
+    @wraps(func)
+    def wrapper(*args, **kwargs):
+        start = time()
+        result = func(*args, **kwargs)
+        end = time()
+        logging.info(f"{func.__name__} completed in {end - start:.2f} seconds")
+        return result
+    return wrapper

vllm-base/Dockerfile

@@ -1,109 +0,0 @@
-################### vLLM Base Dockerfile ###################
-# This Dockerfile is for building the image that the  
-# vLLM worker container will use as its base image. 
-# If your changes are outside of the vLLM source code, you
-# do not need to build this image.
-##########################################################
-
-# Define the CUDA version for the build
-ARG WORKER_CUDA_VERSION=11.8.0
-
-FROM nvidia/cuda:${WORKER_CUDA_VERSION}-devel-ubuntu22.04 AS dev
-
-# Re-declare ARG after FROM
-ARG WORKER_CUDA_VERSION
-
-# Update and install dependencies
-RUN apt-get update -y \
-    && apt-get install -y python3-pip git
-
-RUN if [ "${WORKER_CUDA_VERSION}" = "12.1.0" ]; then \
-        ldconfig /usr/local/cuda-12.1/compat/; \
-    fi
-
-# Set working directory
-WORKDIR /vllm-installation
-
-# Install build and runtime dependencies
-COPY vllm-${WORKER_CUDA_VERSION}/requirements.txt requirements.txt
-RUN --mount=type=cache,target=/root/.cache/pip \
-    pip install -r requirements.txt
-
-RUN --mount=type=cache,target=/root/.cache/pip \
-    if [ "${WORKER_CUDA_VERSION}" = "11.8.0" ]; then \
-        pip install -U --force-reinstall torch==2.1.2 xformers==0.0.23.post1 --index-url https://download.pytorch.org/whl/cu118; \
-    fi
-
-# Install development dependencies
-COPY vllm-${WORKER_CUDA_VERSION}/requirements-dev.txt requirements-dev.txt
-RUN --mount=type=cache,target=/root/.cache/pip \
-    pip install -r requirements-dev.txt
-
-FROM dev AS build
-
-# Re-declare ARG after FROM
-ARG WORKER_CUDA_VERSION
-
-# Install build dependencies
-COPY vllm-${WORKER_CUDA_VERSION}/requirements-build.txt requirements-build.txt
-RUN --mount=type=cache,target=/root/.cache/pip \
-    pip install -r requirements-build.txt
-
-# Copy necessary files
-COPY vllm-${WORKER_CUDA_VERSION}/csrc csrc
-COPY vllm-${WORKER_CUDA_VERSION}/setup.py setup.py
-COPY vllm-12.1.0/pyproject.toml pyproject.toml
-COPY vllm-${WORKER_CUDA_VERSION}/vllm/__init__.py vllm/__init__.py
-
-# Conditional installation based on CUDA version
-RUN --mount=type=cache,target=/root/.cache/pip \
-    if [ "${WORKER_CUDA_VERSION}" = "11.8.0" ]; then \
-        pip install -U --force-reinstall torch==2.1.2 xformers==0.0.23.post1 --index-url https://download.pytorch.org/whl/cu118; \
-        rm pyproject.toml; \
-    elif [ "${WORKER_CUDA_VERSION}" != "12.1.0" ]; then \
-        echo "WORKER_CUDA_VERSION not supported"; \
-        exit 1; \
-    fi
-
-# Set environment variables for building extensions
-ARG torch_cuda_arch_list='7.0 7.5 8.0 8.6 8.9 9.0+PTX'
-ENV TORCH_CUDA_ARCH_LIST=${torch_cuda_arch_list}
-ARG max_jobs=48
-ENV MAX_JOBS=${max_jobs}
-ARG nvcc_threads=1024
-ENV NVCC_THREADS=${nvcc_threads}
-
-# Build extensions
-RUN python3 setup.py build_ext --inplace
-
-FROM nvidia/cuda:${WORKER_CUDA_VERSION}-runtime-ubuntu22.04 AS vllm-base
-
-# Re-declare ARG after FROM
-ARG WORKER_CUDA_VERSION
-
-# Update and install necessary libraries
-RUN apt-get update -y \
-    && apt-get install -y python3-pip
-
-# Set working directory
-WORKDIR /vllm-installation
-
-# Install runtime dependencies
-COPY vllm-${WORKER_CUDA_VERSION}/requirements.txt requirements.txt
-RUN --mount=type=cache,target=/root/.cache/pip \
-    pip install -r requirements.txt
-
-RUN --mount=type=cache,target=/root/.cache/pip \
-    if [ "${WORKER_CUDA_VERSION}" = "11.8.0" ]; then \
-        pip install -U --force-reinstall torch==2.1.2 xformers==0.0.23.post1 --index-url https://download.pytorch.org/whl/cu118; \
-    fi
-
-# Copy built files from the build stage
-COPY --from=build /vllm-installation/vllm/*.so /vllm-installation/vllm/
-COPY vllm-${WORKER_CUDA_VERSION}/vllm vllm
-
-# Set PYTHONPATH environment variable
-ENV PYTHONPATH="/"
-
-# Validate the installation
-RUN python3 -c "import sys; print(sys.path); import vllm; print(vllm.__file__)"

vllm-base/README.md

@@ -1 +0,0 @@
-This directory is for building the vllm-base image utilized by the worker.

vllm-base/download_required_files.sh

@@ -1,12 +0,0 @@
-#!/bin/bash
-
-git clone https://github.com/runpod/vllm-fork-for-sls-worker.git
-
-cp -r vllm-fork-for-sls-worker vllm-12.1.0
-cp -r vllm-fork-for-sls-worker vllm-11.8.0
-rm -rf vllm-fork-for-sls-worker
-
-cd vllm-11.8.0
-git checkout cuda-11.8
-
-echo "vLLM Base Image Builder Setup Complete."