support llava video (#426)

.gitignore

@@ -177,3 +177,7 @@ tmp*.txt
 # Plots
 *.png
 *.pdf
+
+# personnal
+work_dirs/
+*.csv

README.md

@@ -410,4 +410,4 @@ https://github.com/sgl-project/sglang/issues/157
 }
 ```
 
-We learned from the design and reused some code of the following projects: [Guidance](https://github.com/guidance-ai/guidance), [vLLM](https://github.com/vllm-project/vllm), [LightLLM](https://github.com/ModelTC/lightllm), [FlashInfer](https://github.com/flashinfer-ai/flashinfer), [Outlines](https://github.com/outlines-dev/outlines), [LMQL](https://github.com/eth-sri/lmql).
+We learned from the design and reused some code of the following projects: [Guidance](https://github.com/guidance-ai/guidance), [vLLM](https://github.com/vllm-project/vllm), [LightLLM](https://github.com/ModelTC/lightllm), [FlashInfer](https://github.com/flashinfer-ai/flashinfer), [Outlines](https://github.com/outlines-dev/outlines), [LMQL](https://github.com/eth-sri/lmql).

examples/quick_start/srt_example_llava.py

@@ -14,7 +14,7 @@ def single():
     state = image_qa.run(
         image_path="images/cat.jpeg",
         question="What is this?",
-        max_new_tokens=64)
+        max_new_tokens=128)
     print(state["answer"], "\n")
 
 
@@ -36,7 +36,7 @@ def batch():
             {"image_path": "images/cat.jpeg", "question":"What is this?"},
             {"image_path": "images/dog.jpeg", "question":"What is this?"},
         ],
-        max_new_tokens=64,
+        max_new_tokens=128,
     )
     for s in states:
         print(s["answer"], "\n")

examples/usage/llava_video/srt_example_llava_v.py

@@ -0,0 +1,208 @@
+"""
+Usage: python3 srt_example_llava.py
+"""
+
+import sglang as sgl
+import os
+import csv
+import time
+import argparse
+
+@sgl.function
+def video_qa(s, num_frames, video_path, question):
+    s += sgl.user(sgl.video(video_path,num_frames) + question)
+    s += sgl.assistant(sgl.gen("answer"))
+
+
+def single(path, num_frames=16):
+    state = video_qa.run(
+        num_frames=num_frames,
+        video_path=path,
+        question="Please provide a detailed description of the video, focusing on the main subjects, their actions, the background scenes",
+        temperature=0.0,
+        max_new_tokens=1024,
+    )
+    print(state["answer"], "\n")
+
+
+
+def split_into_chunks(lst, num_chunks):
+    """Split a list into a specified number of chunks."""
+    # Calculate the chunk size using integer division. Note that this may drop some items if not evenly divisible.
+    chunk_size = len(lst) // num_chunks
+
+    if chunk_size == 0:
+        chunk_size = len(lst)
+    # Use list comprehension to generate chunks. The last chunk will take any remainder if the list size isn't evenly divisible.
+    chunks = [lst[i:i + chunk_size] for i in range(0, len(lst), chunk_size)]
+    # Ensure we have exactly num_chunks chunks, even if some are empty
+    chunks.extend([[] for _ in range(num_chunks - len(chunks))])
+    return chunks
+
+
+def save_batch_results(batch_video_files, states, cur_chunk, batch_idx, save_dir):
+    csv_filename = f"{save_dir}/chunk_{cur_chunk}_batch_{batch_idx}.csv"
+    with open(csv_filename, 'w', newline='') as csvfile:
+        writer = csv.writer(csvfile)
+        writer.writerow(['video_name', 'answer'])
+        for video_path, state in zip(batch_video_files, states):
+            video_name = os.path.basename(video_path)
+            writer.writerow([video_name, state["answer"]])
+
+def compile_and_cleanup_final_results(cur_chunk, num_batches, save_dir):
+    final_csv_filename = f"{save_dir}/final_results_chunk_{cur_chunk}.csv"
+    with open(final_csv_filename, 'w', newline='') as final_csvfile:
+        writer = csv.writer(final_csvfile)
+        writer.writerow(['video_name', 'answer'])
+        for batch_idx in range(num_batches):
+            batch_csv_filename = f"{save_dir}/chunk_{cur_chunk}_batch_{batch_idx}.csv"
+            with open(batch_csv_filename, 'r') as batch_csvfile:
+                reader = csv.reader(batch_csvfile)
+                next(reader)  # Skip header row
+                for row in reader:
+                    writer.writerow(row)
+            os.remove(batch_csv_filename)
+
+def find_video_files(video_dir):
+    # Check if the video_dir is actually a file
+    if os.path.isfile(video_dir):
+        # If it's a file, return it as a single-element list
+        return [video_dir]
+    
+    # Original logic to find video files in a directory
+    video_files = []
+    for root, dirs, files in os.walk(video_dir):
+        for file in files:
+            if file.endswith(('.mp4', '.avi', '.mov')):
+                video_files.append(os.path.join(root, file))
+    return video_files
+
+def batch(video_dir, save_dir, cur_chunk, num_chunks, num_frames=16, batch_size=64):
+    video_files = find_video_files(video_dir)
+    chunked_video_files = split_into_chunks(video_files, num_chunks)[cur_chunk]
+    num_batches = 0
+
+    for i in range(0, len(chunked_video_files), batch_size):
+        batch_video_files = chunked_video_files[i:i + batch_size]
+        print(f"Processing batch of {len(batch_video_files)} video(s)...")
+
+        if not batch_video_files:
+            print("No video files found in the specified directory.")
+            return
+        
+        batch_input = [
+            {   
+                "num_frames": num_frames,
+                "video_path": video_path,
+                "question": "Please provide a detailed description of the video, focusing on the main subjects, their actions, the background scenes.",
+            } for video_path in batch_video_files
+        ]
+
+        start_time = time.time()
+        states = video_qa.run_batch(batch_input, max_new_tokens=512, temperature=0.2)
+        total_time = time.time() - start_time
+        average_time = total_time / len(batch_video_files)
+        print(f"Number of videos in batch: {len(batch_video_files)}. Average processing time per video: {average_time:.2f} seconds. Total time for this batch: {total_time:.2f} seconds")
+
+        save_batch_results(batch_video_files, states, cur_chunk, num_batches, save_dir)
+        num_batches += 1
+
+    compile_and_cleanup_final_results(cur_chunk, num_batches, save_dir)
+
+
+if __name__ == "__main__":
+
+    # Create the parser
+    parser = argparse.ArgumentParser(description='Run video processing with specified port.')
+
+    # Add an argument for the port
+    parser.add_argument('--port', type=int, default=30000, help='The master port for distributed serving.')
+    parser.add_argument('--chunk-idx', type=int, default=0, help='The index of the chunk to process.')
+    parser.add_argument('--num-chunks', type=int, default=8, help='The number of chunks to process.')
+    parser.add_argument('--save-dir', type=str, default="./work_dirs/llava_video", help='The directory to save the processed video files.')
+    parser.add_argument('--video-dir', type=str, default="./videos/Q98Z4OTh8RwmDonc.mp4", help='The directory or path for the processed video files.')
+    parser.add_argument('--model-path', type=str, default="lmms-lab/LLaVA-NeXT-Video-7B", help='The model path for the video processing.')
+    parser.add_argument('--num-frames', type=int, default=16, help='The number of frames to process in each video.' )
+    parser.add_argument("--mm_spatial_pool_stride", type=int, default=2)
+
+    # Parse the arguments
+    args = parser.parse_args()
+
+    cur_port = args.port
+
+    cur_chunk = args.chunk_idx
+
+    num_chunks = args.num_chunks
+
+    num_frames = args.num_frames
+
+    if "34b" in args.model_path.lower():
+        tokenizer_path = "liuhaotian/llava-v1.6-34b-tokenizer"
+    elif "7b" in args.model_path.lower():
+        tokenizer_path = "llava-hf/llava-1.5-7b-hf"
+    else:
+        print("Invalid model path. Please specify a valid model path.")
+        exit()
+
+    model_overide_args = {}
+
+    model_overide_args["mm_spatial_pool_stride"] = args.mm_spatial_pool_stride
+    model_overide_args["architectures"] = ["LlavaVidForCausalLM"]
+    model_overide_args["num_frames"] = args.num_frames
+    model_overide_args["model_type"] = "llava"
+
+    if "34b" in args.model_path.lower():
+        model_overide_args["image_token_index"] = 64002
+
+
+    if args.num_frames == 32:
+        model_overide_args["rope_scaling"] = {"factor": 2.0, "type": "linear"}
+        model_overide_args["max_sequence_length"] = 4096 * 2
+        model_overide_args["tokenizer_model_max_length"] = 4096 * 2
+    elif args.num_frames < 32:
+        pass
+    else:
+        print("The maximum number of frames to process is 32. Please specify a valid number of frames.")
+        exit()
+
+
+    runtime = sgl.Runtime(
+        model_path=args.model_path, #"liuhaotian/llava-v1.6-vicuna-7b",
+        tokenizer_path=tokenizer_path,
+        port=cur_port,
+        additional_ports=[cur_port+1,cur_port+2,cur_port+3,cur_port+4],
+        model_overide_args=model_overide_args,
+        tp_size=1
+    )
+    sgl.set_default_backend(runtime)
+    print(f"chat template: {runtime.endpoint.chat_template.name}")
+
+
+    # Run a single request
+    # try:
+    print("\n========== single ==========\n")
+    root = args.video_dir
+    if os.path.isfile(root):
+        video_files = [root]
+    else:
+        video_files = [os.path.join(root, f) for f in os.listdir(root) if f.endswith(('.mp4', '.avi', '.mov'))]  # Add more extensions if needed
+    start_time = time.time()  # Start time for processing a single video
+    for cur_video in video_files[:1]:
+        print(cur_video)
+        single(cur_video, num_frames)
+    end_time = time.time()  # End time for processing a single video
+    total_time = end_time - start_time
+    average_time = total_time / len(video_files)  # Calculate the average processing time
+    print(f"Average processing time per video: {average_time:.2f} seconds")
+    runtime.shutdown()
+    # except Exception as e:
+    #     print(e)
+    runtime.shutdown()
+
+
+    # # # Run a batch of requests
+    # print("\n========== batch ==========\n")
+    # if not os.path.exists(args.save_dir):
+    #     os.makedirs(args.save_dir)
+    # batch(args.video_dir,args.save_dir,cur_chunk, num_chunks, num_frames, num_chunks)
+    # runtime.shutdown()

examples/usage/llava_video/srt_example_llava_v.sh

@@ -0,0 +1,130 @@
+#!/bin/bash
+
+##### USAGE #####
+#    - First node:
+#      ```sh
+#      bash examples/quick_start/srt_example_llava_v.sh K 0 YOUR_VIDEO_PATH YOUR_MODEL_PATH FRAMES_PER_VIDEO
+#      ```
+#    - Second node:
+#      ```sh
+#      bash examples/quick_start/srt_example_llava_v.sh K 1 YOUR_VIDEO_PATH YOUR_MODEL_PATH FRAMES_PER_VIDEO
+#      ```
+#    - The K node:
+#      ```sh
+#      bash examples/quick_start/srt_example_llava_v.sh K K-1 YOUR_VIDEO_PATH YOUR_MODEL_PATH FRAMES_PER_VIDEO
+#      ```
+
+
+# Replace `K`, `YOUR_VIDEO_PATH`, `YOUR_MODEL_PATH`, and `FRAMES_PER_VIDEO` with your specific details.
+CURRENT_ROOT="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )"
+
+echo ${CURRENT_ROOT}
+
+cd ${CURRENT_ROOT}
+
+export PYTHONWARNINGS=ignore
+
+START_TIME=$(date +%s)  # Capture start time
+
+NUM_NODES=$1
+
+CUR_NODES_IDX=$2
+
+VIDEO_DIR=$3
+
+MODEL_PATH=$4   
+
+NUM_FRAMES=$5
+
+
+# FRAME_FORMAT=$6
+
+# FRAME_FORMAT=$(echo $FRAME_FORMAT | tr '[:lower:]' '[:upper:]')
+
+# # Check if FRAME_FORMAT is either JPEG or PNG
+# if [[ "$FRAME_FORMAT" != "JPEG" && "$FRAME_FORMAT" != "PNG" ]]; then
+#     echo "Error: FRAME_FORMAT must be either JPEG or PNG."
+#     exit 1
+# fi
+
+# export TARGET_FRAMES=$TARGET_FRAMES
+
+echo "Each video you will sample $NUM_FRAMES frames"
+
+# export FRAME_FORMAT=$FRAME_FORMAT
+
+# echo "The frame format is $FRAME_FORMAT"
+
+# Assuming GPULIST is a bash array containing your GPUs
+GPULIST=(0 1 2 3 4 5 6 7)
+LOCAL_CHUNKS=${#GPULIST[@]}
+
+echo "Number of GPUs in GPULIST: $LOCAL_CHUNKS"
+
+ALL_CHUNKS=$((NUM_NODES * LOCAL_CHUNKS))
+
+# Calculate GPUs per chunk
+GPUS_PER_CHUNK=8
+
+echo $GPUS_PER_CHUNK
+
+for IDX in $(seq 1 $LOCAL_CHUNKS); do
+    (
+        START=$(((IDX-1) * GPUS_PER_CHUNK))
+        LENGTH=$GPUS_PER_CHUNK # Length for slicing, not the end index
+        
+        CHUNK_GPUS=(${GPULIST[@]:$START:$LENGTH})
+        
+        # Convert the chunk GPUs array to a comma-separated string
+        CHUNK_GPUS_STR=$(IFS=,; echo "${CHUNK_GPUS[*]}")
+
+        LOCAL_IDX=$((CUR_NODES_IDX * LOCAL_CHUNKS + IDX))
+
+        echo "Chunk $(($LOCAL_IDX - 1)) will run on GPUs $CHUNK_GPUS_STR"
+        
+        # Calculate the port for this chunk. Ensure it's incremented by 5 for each chunk.
+        PORT=$((10000 + RANDOM % 55536))
+
+        MAX_RETRIES=10
+        RETRY_COUNT=0
+        COMMAND_STATUS=1  # Initialize as failed
+
+        while [ $RETRY_COUNT -lt $MAX_RETRIES ] && [ $COMMAND_STATUS -ne 0 ]; do
+            echo "Running chunk $(($LOCAL_IDX - 1)) on GPUs $CHUNK_GPUS_STR with port $PORT. Attempt $(($RETRY_COUNT + 1))"
+            
+#!/bin/bash
+            CUDA_VISIBLE_DEVICES=$CHUNK_GPUS_STR python3 examples/usage/llava_video/srt_example_llava_v.py \
+            --port $PORT \
+            --num-chunks $ALL_CHUNKS \
+            --chunk-idx $(($LOCAL_IDX - 1)) \
+            --save-dir work_dirs/llava_next_video_inference_results \
+            --video-dir $VIDEO_DIR \
+            --model-path $MODEL_PATH \
+            --num-frames $NUM_FRAMES #&
+            
+            wait $!  # Wait for the process to finish and capture its exit status
+            COMMAND_STATUS=$?
+            
+            if [ $COMMAND_STATUS -ne 0 ]; then
+                echo "Execution failed for chunk $(($LOCAL_IDX - 1)), attempt $(($RETRY_COUNT + 1)). Retrying..."
+                RETRY_COUNT=$(($RETRY_COUNT + 1))
+                sleep 180  # Wait a bit before retrying
+            else
+                echo "Execution succeeded for chunk $(($LOCAL_IDX - 1))."
+            fi
+        done
+
+        if [ $COMMAND_STATUS -ne 0 ]; then
+            echo "Execution failed for chunk $(($LOCAL_IDX - 1)) after $MAX_RETRIES attempts."
+        fi
+    ) #&
+    sleep 2  # Slight delay to stagger the start times
+done
+
+wait
+
+cat work_dirs/llava_next_video_inference_results/final_results_chunk_*.csv > work_dirs/llava_next_video_inference_results/final_results_node_${CUR_NODES_IDX}.csv   
+
+END_TIME=$(date +%s)  # Capture end time
+ELAPSED_TIME=$(($END_TIME - $START_TIME))
+echo "Total execution time: $ELAPSED_TIME seconds."

python/pyproject.toml

@@ -20,7 +20,7 @@ dependencies = [
 
 [project.optional-dependencies]
 srt = ["aiohttp", "fastapi", "psutil", "rpyc", "torch", "uvloop", "uvicorn",
-       "zmq", "vllm>=0.4.2", "interegular", "pydantic", "pillow", "outlines>=0.0.27", "packaging"]
+       "zmq", "vllm>=0.4.2", "interegular", "pydantic", "pillow", "packaging", "huggingface_hub", "hf_transfer", "outlines>=0.0.34"]
 openai = ["openai>=1.0", "numpy", "tiktoken"]
 anthropic = ["anthropic>=0.20.0", "numpy"]
 all = ["sglang[srt]", "sglang[openai]", "sglang[anthropic]"]
@@ -33,4 +33,4 @@ all = ["sglang[srt]", "sglang[openai]", "sglang[anthropic]"]
 exclude = ["assets*", "benchmark*", "docs*", "dist*", "playground*", "scripts*", "tests*"]
 
 [tool.wheel]
-exclude = ["assets*", "benchmark*", "docs*", "dist*", "playground*", "scripts*", "tests*"]
+exclude = ["assets*", "benchmark*", "docs*", "dist*", "playground*", "scripts*", "tests*"]

python/sglang/__init__.py

@@ -19,6 +19,7 @@ from sglang.api import (
     user,
     user_begin,
     user_end,
+    video,
 )
 
 # SGL Backends
@@ -46,6 +47,7 @@ __all__ = [
     "gen_int",
     "gen_string",
     "image",
+    "video",
     "select",
     "system",
     "user",

python/sglang/api.py

@@ -15,6 +15,7 @@ from sglang.lang.ir import (
     SglRoleBegin,
     SglRoleEnd,
     SglSelect,
+    SglVideo,
 )
 
 
@@ -151,6 +152,10 @@ def image(expr: SglExpr):
     return SglImage(expr)
 
 
+def video(path: str, num_frames: int):
+    return SglVideo(path, num_frames)
+
+
 def select(
     name: Optional[str] = None,
     choices: List[str] = None,

python/sglang/lang/chat_template.py

@@ -259,6 +259,8 @@ def match_vicuna(model_path: str):
         return get_chat_template("vicuna_v1.1")
     if "llava-v1.5" in model_path.lower():
         return get_chat_template("vicuna_v1.1")
+    if "llava-next-video-7b" in model_path.lower():
+        return get_chat_template("vicuna_v1.1")
 
 
 @register_chat_template_matching_function
@@ -283,19 +285,24 @@ def match_llama3_instruct(model_path: str):
 
 @register_chat_template_matching_function
 def match_chat_ml(model_path: str):
+    # import pdb;pdb.set_trace()
     model_path = model_path.lower()
     if "tinyllama" in model_path:
         return get_chat_template("chatml")
     if "qwen" in model_path and "chat" in model_path:
         return get_chat_template("chatml")
-    if "llava-v1.6-34b" in model_path:
+    if (
+        "llava-v1.6-34b" in model_path
+        or "llava-v1.6-yi-34b" in model_path
+        or "llava-next-video-34b" in model_path
+    ):
         return get_chat_template("chatml-llava")
 
 
 @register_chat_template_matching_function
 def match_chat_yi(model_path: str):
     model_path = model_path.lower()
-    if "yi" in model_path:
+    if "yi" in model_path and "llava" not in model_path:
         return get_chat_template("yi")
 
 

python/sglang/lang/interpreter.py

@@ -28,8 +28,9 @@ from sglang.lang.ir import (
     SglVariable,
     SglVarScopeBegin,
     SglVarScopeEnd,
+    SglVideo,
 )
-from sglang.utils import encode_image_base64, get_exception_traceback
+from sglang.utils import encode_image_base64, encode_video_base64, get_exception_traceback
 
 
 def run_internal(state, program, func_args, func_kwargs, sync):
@@ -361,6 +362,8 @@ class StreamExecutor:
             self._execute_role_end(other)
         elif isinstance(other, SglImage):
             self._execute_image(other)
+        elif isinstance(other, SglVideo):
+            self._execute_video(other)
         elif isinstance(other, SglVariable):
             self._execute_variable(other)
         elif isinstance(other, SglVarScopeBegin):
@@ -397,6 +400,16 @@ class StreamExecutor:
         self.cur_images.append((path, base64_data))
         self.text_ += self.chat_template.image_token
 
+    def _execute_video(self, expr: SglVideo):
+        path = expr.path
+        num_frames = expr.num_frames
+
+        base64_data = encode_video_base64(path, num_frames)
+
+        self.images_.append((path, base64_data))
+        self.cur_images.append((path, base64_data))
+        self.text_ += self.chat_template.image_token
+
         # if global_config.eager_fill_image:
         #     self.backend.fill_image(self)
 

python/sglang/lang/ir.py

@@ -330,6 +330,15 @@ class SglImage(SglExpr):
         return f"SglImage({self.path})"
 
 
+class SglVideo(SglExpr):
+    def __init__(self, path, num_frames):
+        self.path = path
+        self.num_frames = num_frames
+
+    def __repr__(self) -> str:
+        return f"SglVideo({self.path}, {self.num_frames})"
+
+
 class SglGen(SglExpr):
     def __init__(
         self,

python/sglang/lang/tracer.py

@@ -110,7 +110,7 @@ class TracerProgramState(ProgramState):
     ##################################
 
     def fork(self, size: int = 1, position_ids_offset: Optional[List[int]] = None):
-        assert (size >= 1)
+        assert size >= 1
 
         if self.only_trace_prefix:
             raise StopTracing()

python/sglang/launch_server.py

@@ -2,11 +2,10 @@ import argparse
 
 from sglang.srt.server import ServerArgs, launch_server
 
-
 if __name__ == "__main__":
     parser = argparse.ArgumentParser()
     ServerArgs.add_cli_args(parser)
     args = parser.parse_args()
     server_args = ServerArgs.from_cli_args(args)
 
-    launch_server(server_args, None)
+    launch_server(server_args, None)

python/sglang/launch_server_llavavid.py

@@ -0,0 +1,31 @@
+import argparse
+import multiprocessing as mp
+
+from sglang.srt.server import ServerArgs, launch_server
+
+if __name__ == "__main__":
+
+    model_overide_args = {}
+
+    model_overide_args["mm_spatial_pool_stride"] = 2
+    model_overide_args["architectures"] = ["LlavaVidForCausalLM"]
+    model_overide_args["num_frames"] = 16
+    model_overide_args["model_type"] = "llavavid"
+    if model_overide_args["num_frames"] == 32:
+        model_overide_args["rope_scaling"] = {"factor": 2.0, "type": "linear"}
+        model_overide_args["max_sequence_length"] = 4096 * 2
+        model_overide_args["tokenizer_model_max_length"] = 4096 * 2
+        model_overide_args["model_max_length"] = 4096 * 2
+
+    parser = argparse.ArgumentParser()
+    ServerArgs.add_cli_args(parser)
+    args = parser.parse_args()
+
+    if "34b" in args.model_path.lower():
+        model_overide_args["image_token_index"] = 64002
+
+    server_args = ServerArgs.from_cli_args(args)
+
+    pipe_reader, pipe_writer = mp.Pipe(duplex=False)
+
+    launch_server(server_args, pipe_writer, model_overide_args)

python/sglang/srt/hf_transformers_utils.py

@@ -30,10 +30,17 @@ def get_config_json(model_path: str):
     return config
 
 
-def get_config(model: str, trust_remote_code: bool, revision: Optional[str] = None):
+def get_config(
+    model: str,
+    trust_remote_code: bool,
+    revision: Optional[str] = None,
+    model_overide_args: Optional[dict] = None,
+):
     config = AutoConfig.from_pretrained(
         model, trust_remote_code=trust_remote_code, revision=revision
     )
+    if model_overide_args:
+        config.update(model_overide_args)
     return config
 
 

python/sglang/srt/managers/router/manager.py

@@ -60,9 +60,7 @@ class RouterManager:
 
 
 def start_router_process(
-    server_args: ServerArgs,
-    port_args: PortArgs,
-    pipe_writer,
+    server_args: ServerArgs, port_args: PortArgs, pipe_writer, model_overide_args
 ):
     logging.basicConfig(
         level=getattr(logging, server_args.log_level.upper()),
@@ -70,7 +68,7 @@ def start_router_process(
     )
 
     try:
-        model_client = ModelRpcClient(server_args, port_args)
+        model_client = ModelRpcClient(server_args, port_args, model_overide_args)
         router = RouterManager(model_client, port_args)
     except Exception:
         pipe_writer.send(get_exception_traceback())

python/sglang/srt/managers/router/model_rpc.py

@@ -4,12 +4,13 @@ import multiprocessing
 import time
 import warnings
 from concurrent.futures import ThreadPoolExecutor
-from typing import List
+from typing import Any, Dict, List, Optional, Tuple, Union
 
 import rpyc
 import torch
 from rpyc.utils.classic import obtain
 from rpyc.utils.server import ThreadedServer
+
 try:
     from vllm.logger import _default_handler as vllm_default_logger
 except ImportError:
@@ -48,6 +49,7 @@ class ModelRpcServer:
         tp_rank: int,
         server_args: ServerArgs,
         port_args: PortArgs,
+        model_overide_args: Optional[dict] = None,
     ):
         server_args, port_args = [obtain(x) for x in [server_args, port_args]]
 
@@ -62,6 +64,7 @@ class ModelRpcServer:
             server_args.model_path,
             server_args.trust_remote_code,
             context_length=server_args.context_length,
+            model_overide_args=model_overide_args,
         )
 
         # For model end global settings
@@ -673,13 +676,15 @@ class ModelRpcService(rpyc.Service):
 
 
 class ModelRpcClient:
-    def __init__(self, server_args: ServerArgs, port_args: PortArgs):
+    def __init__(
+        self, server_args: ServerArgs, port_args: PortArgs, model_overide_args
+    ):
         tp_size = server_args.tp_size
 
         if tp_size == 1:
             # Init model
             self.model_server = ModelRpcService().exposed_ModelRpcServer(
-                0, server_args, port_args
+                0, server_args, port_args, model_overide_args
             )
 
             # Wrap functions
@@ -700,7 +705,7 @@ class ModelRpcClient:
                 # Init model
                 def init_model(i):
                     return self.remote_services[i].ModelRpcServer(
-                        i, server_args, port_args
+                        i, server_args, port_args, model_overide_args
                     )
 
                 self.model_servers = executor.map(init_model, range(tp_size))
@@ -723,7 +728,11 @@ def _init_service(port):
     t = ThreadedServer(
         ModelRpcService(),
         port=port,
-        protocol_config={"allow_pickle": True, "sync_request_timeout": 1800},
+        protocol_config={
+            "allow_public_attrs": True,
+            "allow_pickle": True,
+            "sync_request_timeout": 1800,
+        },
     )
     t.start()
 
@@ -739,7 +748,11 @@ def start_model_process(port):
             con = rpyc.connect(
                 "localhost",
                 port,
-                config={"allow_pickle": True, "sync_request_timeout": 1800},
+                config={
+                    "allow_public_attrs": True,
+                    "allow_pickle": True,
+                    "sync_request_timeout": 1800,
+                },
             )
             break
         except ConnectionRefusedError:

python/sglang/srt/managers/router/model_runner.py

@@ -9,11 +9,11 @@ from typing import List
 
 import numpy as np
 import torch
+from vllm.distributed import initialize_model_parallel
 from vllm.model_executor.layers.quantization.awq import AWQConfig
 from vllm.model_executor.layers.quantization.gptq import GPTQConfig
 from vllm.model_executor.layers.quantization.marlin import MarlinConfig
 from vllm.model_executor.model_loader.utils import set_default_torch_dtype
-from vllm.distributed import initialize_model_parallel
 
 from sglang.srt.managers.router.infer_batch import Batch, ForwardMode
 from sglang.srt.memory_pool import ReqToTokenPool, TokenToKVPool
@@ -143,7 +143,7 @@ class InputMetadata:
                 self.kv_last_page_len,
                 self.model_runner.model_config.num_attention_heads // tp_size,
                 self.model_runner.model_config.num_key_value_heads // tp_size,
-                self.model_runner.model_config.head_dim
+                self.model_runner.model_config.head_dim,
             ]
 
             self.prefill_wrapper.begin_forward(*args)

python/sglang/srt/managers/tokenizer_manager.py

@@ -60,21 +60,29 @@ def get_pixel_values(
 ):
     try:
         processor = processor or global_processor
-        image = load_image(image_data)
-        image_hash = hash(image_data)
-        if image_aspect_ratio == "pad":
-            image = expand2square(
-                image, tuple(int(x * 255) for x in processor.image_processor.image_mean)
-            )
-            pixel_values = processor.image_processor(image)["pixel_values"][0]
-        elif image_aspect_ratio == "anyres":
-            pixel_values = process_anyres_image(
-                image, processor.image_processor, image_grid_pinpoints
-            )
+        image, image_size = load_image(image_data)
+        if image_size != None:
+            image_hash = hash(image_data)
+            pixel_values = processor.image_processor(image)["pixel_values"]
+            for _ in range(len(pixel_values)):
+                pixel_values[_] = pixel_values[_].astype(np.float16)
+            pixel_values = np.stack(pixel_values, axis=0)
+            return pixel_values, image_hash, image_size
         else:
-            pixel_values = processor.image_processor(image)["pixel_values"][0]
-        pixel_values = pixel_values.astype(np.float16)
-        return pixel_values, image_hash, image.size
+            image_hash = hash(image_data)
+            if image_aspect_ratio == "pad":
+                image = expand2square(
+                    image, tuple(int(x * 255) for x in processor.image_processor.image_mean)
+                )
+                pixel_values = processor.image_processor(image)["pixel_values"][0]
+            elif image_aspect_ratio == "anyres":
+                pixel_values = process_anyres_image(
+                    image, processor.image_processor, image_grid_pinpoints
+                )
+            else:
+                pixel_values = processor.image_processor(image)["pixel_values"][0]
+            pixel_values = pixel_values.astype(np.float16)
+            return pixel_values, image_hash, image.size
     except Exception:
         print("Exception in TokenizerManager:\n" + get_exception_traceback())
 
@@ -84,6 +92,7 @@ class TokenizerManager:
         self,
         server_args: ServerArgs,
         port_args: PortArgs,
+        model_overide_args: dict = None,
     ):
         self.server_args = server_args
 
@@ -96,7 +105,9 @@ class TokenizerManager:
 
         self.model_path = server_args.model_path
         self.hf_config = get_config(
-            self.model_path, trust_remote_code=server_args.trust_remote_code
+            self.model_path,
+            trust_remote_code=server_args.trust_remote_code,
+            model_overide_args=model_overide_args,
         )
         self.context_len = get_context_length(self.hf_config)
 

python/sglang/srt/model_config.py

@@ -10,12 +10,16 @@ class ModelConfig:
         trust_remote_code: bool = True,
         revision: Optional[str] = None,
         context_length: Optional[int] = None,
+        model_overide_args: Optional[dict] = None,
     ) -> None:
         self.path = path
         self.trust_remote_code = trust_remote_code
         self.revision = revision
         self.hf_config = get_config(self.path, trust_remote_code, revision)
 
+        if model_overide_args is not None:
+            self.hf_config.update(model_overide_args)
+
         if context_length is not None:
             self.context_len = context_length
         else:

python/sglang/srt/models/commandr.py

@@ -27,29 +27,25 @@ import torch.utils.checkpoint
 from torch import nn
 from torch.nn.parameter import Parameter
 from transformers import PretrainedConfig
+from vllm.distributed import (
+    get_tensor_model_parallel_rank,
+    get_tensor_model_parallel_world_size,
+)
 from vllm.model_executor.layers.activation import SiluAndMul
 from vllm.model_executor.layers.linear import (
     MergedColumnParallelLinear,
     QKVParallelLinear,
     RowParallelLinear,
 )
-from vllm.model_executor.layers.quantization.base_config import (
-    QuantizationConfig)
+from vllm.model_executor.layers.quantization.base_config import QuantizationConfig
 from vllm.model_executor.layers.rotary_embedding import get_rope
 from vllm.model_executor.layers.vocab_parallel_embedding import VocabParallelEmbedding
-from vllm.distributed import (
-    get_tensor_model_parallel_rank,
-    get_tensor_model_parallel_world_size,
-)
 from vllm.model_executor.utils import set_weight_attrs
-from sglang.srt.weight_utils import (
-    default_weight_loader,
-    hf_model_weights_iterator,
-)
 
 from sglang.srt.layers.logits_processor import LogitsProcessor
 from sglang.srt.layers.radix_attention import RadixAttention
 from sglang.srt.managers.router.model_runner import InputMetadata
+from sglang.srt.weight_utils import default_weight_loader, hf_model_weights_iterator
 
 
 @torch.compile

python/sglang/srt/models/dbrx.py

@@ -5,37 +5,31 @@ from typing import Optional
 
 import torch
 import torch.nn as nn
+from vllm.distributed import (
+    get_tensor_model_parallel_rank,
+    get_tensor_model_parallel_world_size,
+    tensor_model_parallel_all_reduce,
+)
 from vllm.model_executor.layers.fused_moe import fused_moe
 from vllm.model_executor.layers.linear import (
     QKVParallelLinear,
     ReplicatedLinear,
     RowParallelLinear,
 )
-from vllm.model_executor.layers.quantization.base_config import (
-    QuantizationConfig)
+from vllm.model_executor.layers.quantization.base_config import QuantizationConfig
 from vllm.model_executor.layers.rotary_embedding import get_rope
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     DEFAULT_VOCAB_PADDING_SIZE,
     ParallelLMHead,
     VocabParallelEmbedding,
 )
-from vllm.distributed import (
-    tensor_model_parallel_all_reduce,
-)
-from vllm.distributed import (
-    get_tensor_model_parallel_rank,
-    get_tensor_model_parallel_world_size,
-)
 from vllm.model_executor.utils import set_weight_attrs
-from sglang.srt.weight_utils import (
-    default_weight_loader,
-    hf_model_weights_iterator,
-)
 
 from sglang.srt.layers.logits_processor import LogitsProcessor
 from sglang.srt.layers.radix_attention import RadixAttention
 from sglang.srt.managers.router.model_runner import InputMetadata
 from sglang.srt.models.dbrx_config import DbrxConfig
+from sglang.srt.weight_utils import default_weight_loader, hf_model_weights_iterator
 
 
 class DbrxRouter(nn.Module):
@@ -291,7 +285,9 @@ class DbrxBlock(nn.Module):
         quant_config: Optional[QuantizationConfig] = None,
     ):
         super().__init__()
-        self.norm_attn_norm = DbrxFusedNormAttention(config, layer_id, quant_config=quant_config)
+        self.norm_attn_norm = DbrxFusedNormAttention(
+            config, layer_id, quant_config=quant_config
+        )
         self.ffn = DbrxExperts(config, quant_config=quant_config)
 
     def forward(
@@ -322,7 +318,10 @@ class DbrxModel(nn.Module):
             config.d_model,
         )
         self.blocks = nn.ModuleList(
-            [DbrxBlock(config, i, quant_config=quant_config) for i in range(config.n_layers)]
+            [
+                DbrxBlock(config, i, quant_config=quant_config)
+                for i in range(config.n_layers)
+            ]
         )
         self.norm_f = nn.LayerNorm(config.d_model, eps=1e-5)
         for module in self.modules():

python/sglang/srt/models/gemma.py

@@ -7,6 +7,7 @@ import torch
 from torch import nn
 from transformers import PretrainedConfig
 from vllm.config import LoRAConfig
+from vllm.distributed import get_tensor_model_parallel_world_size
 from vllm.model_executor.layers.activation import GeluAndMul
 from vllm.model_executor.layers.layernorm import RMSNorm
 from vllm.model_executor.layers.linear import (
@@ -14,21 +15,14 @@ from vllm.model_executor.layers.linear import (
     QKVParallelLinear,
     RowParallelLinear,
 )
-from vllm.model_executor.layers.quantization.base_config import (
-    QuantizationConfig)
+from vllm.model_executor.layers.quantization.base_config import QuantizationConfig
 from vllm.model_executor.layers.rotary_embedding import get_rope
 from vllm.model_executor.layers.vocab_parallel_embedding import VocabParallelEmbedding
-from vllm.distributed import (
-    get_tensor_model_parallel_world_size,
-)
-from sglang.srt.weight_utils import (
-    default_weight_loader,
-    hf_model_weights_iterator,
-)
 
 from sglang.srt.layers.logits_processor import LogitsProcessor
 from sglang.srt.layers.radix_attention import RadixAttention
 from sglang.srt.managers.router.model_runner import InputMetadata
+from sglang.srt.weight_utils import default_weight_loader, hf_model_weights_iterator
 
 
 class GemmaMLP(nn.Module):
@@ -46,7 +40,10 @@ class GemmaMLP(nn.Module):
             quant_config=quant_config,
         )
         self.down_proj = RowParallelLinear(
-            intermediate_size, hidden_size, bias=False, quant_config=quant_config,
+            intermediate_size,
+            hidden_size,
+            bias=False,
+            quant_config=quant_config,
         )
         self.act_fn = GeluAndMul()
 

python/sglang/srt/models/llama2.py

@@ -6,6 +6,7 @@ from typing import Any, Dict, Optional, Tuple
 import torch
 from torch import nn
 from transformers import LlamaConfig
+from vllm.distributed import get_tensor_model_parallel_world_size
 from vllm.model_executor.layers.activation import SiluAndMul
 from vllm.model_executor.layers.layernorm import RMSNorm
 from vllm.model_executor.layers.linear import (
@@ -13,24 +14,17 @@ from vllm.model_executor.layers.linear import (
     QKVParallelLinear,
     RowParallelLinear,
 )
-from vllm.model_executor.layers.quantization.base_config import (
-    QuantizationConfig)
+from vllm.model_executor.layers.quantization.base_config import QuantizationConfig
 from vllm.model_executor.layers.rotary_embedding import get_rope
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     ParallelLMHead,
     VocabParallelEmbedding,
 )
-from vllm.distributed import (
-    get_tensor_model_parallel_world_size,
-)
-from sglang.srt.weight_utils import (
-    default_weight_loader,
-    hf_model_weights_iterator,
-)
 
 from sglang.srt.layers.logits_processor import LogitsProcessor
 from sglang.srt.layers.radix_attention import RadixAttention
 from sglang.srt.managers.router.model_runner import InputMetadata
+from sglang.srt.weight_utils import default_weight_loader, hf_model_weights_iterator
 
 
 class LlamaMLP(nn.Module):
@@ -49,7 +43,10 @@ class LlamaMLP(nn.Module):
             quant_config=quant_config,
         )
         self.down_proj = RowParallelLinear(
-            intermediate_size, hidden_size, bias=False, quant_config=quant_config,
+            intermediate_size,
+            hidden_size,
+            bias=False,
+            quant_config=quant_config,
         )
         if hidden_act != "silu":
             raise ValueError(

python/sglang/srt/models/llava.py

@@ -7,12 +7,7 @@ import torch
 from torch import nn
 from transformers import CLIPVisionModel, LlavaConfig
 from transformers.models.llava.modeling_llava import LlavaMultiModalProjector
-from vllm.model_executor.layers.quantization.base_config import (
-    QuantizationConfig)
-from sglang.srt.weight_utils import (
-    default_weight_loader,
-    hf_model_weights_iterator,
-)
+from vllm.model_executor.layers.quantization.base_config import QuantizationConfig
 
 from sglang.srt.managers.router.infer_batch import ForwardMode
 from sglang.srt.managers.router.model_runner import InputMetadata
@@ -22,6 +17,7 @@ from sglang.srt.mm_utils import (
     unpad_image_shape,
 )
 from sglang.srt.models.llama2 import LlamaForCausalLM
+from sglang.srt.weight_utils import default_weight_loader, hf_model_weights_iterator
 
 
 class LlavaLlamaForCausalLM(nn.Module):

python/sglang/srt/models/llavavid.py

@@ -0,0 +1,307 @@
+"""Inference-only LLaVa video model compatible with HuggingFace weights."""
+
+import os
+from typing import List, Optional
+
+import numpy as np
+import torch
+from torch import nn
+from transformers import CLIPVisionModel, LlamaConfig, LlavaConfig
+from transformers.models.llava.modeling_llava import LlavaMultiModalProjector
+from vllm.model_executor.layers.quantization.base_config import QuantizationConfig
+
+from sglang.srt.managers.router.infer_batch import ForwardMode
+from sglang.srt.managers.router.model_runner import InputMetadata
+from sglang.srt.mm_utils import (
+    get_anyres_image_grid_shape,
+    unpad_image,
+    unpad_image_shape,
+)
+from sglang.srt.models.llama2 import LlamaForCausalLM
+from sglang.srt.weight_utils import default_weight_loader, hf_model_weights_iterator
+
+
+class LlavaVidForCausalLM(nn.Module):
+    def __init__(
+        self,
+        config: LlavaConfig,
+        quant_config: Optional[QuantizationConfig] = None,
+    ) -> None:
+        super().__init__()
+        self.config = config
+        self.vision_tower = None
+        self.config.vision_config.hidden_size = config.mm_hidden_size
+        self.config.text_config.hidden_size = config.hidden_size
+        self.multi_modal_projector = LlavaMultiModalProjector(config)
+        self.mm_spatial_pool_stride = getattr(self.config, "mm_spatial_pool_stride", 2)
+        self.resampler = nn.AvgPool2d(
+            kernel_size=self.mm_spatial_pool_stride, stride=self.mm_spatial_pool_stride
+        )
+        self.language_model = LlamaForCausalLM(config, quant_config=quant_config)
+        self.num_frames = getattr(self.config, "num_frames", 16)
+        if "unpad" in getattr(config, "mm_patch_merge_type", ""):
+            self.language_model.model.image_newline = nn.Parameter(
+                torch.empty(config.text_config.hidden_size, dtype=torch.float16)
+            )
+
+    def pad_input_ids(self, input_ids, pad_value, pt_shape=None, image_size=None):
+        new_image_feature_len = self.image_feature_len
+        # now only support spatial_unpad + anyres
+        # if self.mm_patch_merge_type.startswith("spatial"):
+        #     height = width = self.num_patches_per_side
+        #     if pt_shape[0] > 1:
+        #         if self.image_aspect_ratio == "anyres":
+        #             num_patch_width, num_patch_height = get_anyres_image_grid_shape(
+        #                 image_size,
+        #                 self.image_grid_pinpoints,
+        #                 self.vision_tower.config.image_size,
+        #             )
+        #         if "unpad" in self.mm_patch_merge_type:
+        #             h = num_patch_height * height
+        #             w = num_patch_width * width
+        #             new_h, new_w = unpad_image_shape(h, w, image_size)
+        #             new_image_feature_len += new_h * (new_w + 1)
+
+        pad_ids = pad_value * (
+            (new_image_feature_len + len(pad_value)) // len(pad_value)
+        )
+        # print(input_ids)
+        offset = input_ids.index(self.config.image_token_index)
+        # old_len + pad_len - 1, because we need to remove image_token_id
+        new_input_ids = (
+            input_ids[:offset]
+            + pad_ids[:new_image_feature_len]
+            + input_ids[offset + 1 :]
+        )
+        return new_input_ids, offset
+
+    def encode_images(self, pixel_values: torch.Tensor) -> torch.Tensor:
+        image_outputs = self.vision_tower(pixel_values, output_hidden_states=True)
+        # NOTE: This is not memory efficient. (output_hidden_states=True) will save all the hidden stated.
+
+        selected_image_feature = image_outputs.hidden_states[self.vision_feature_layer]
+        if self.vision_feature_select_strategy in ["default", "patch"]:
+            selected_image_feature = selected_image_feature[:, 1:]
+        elif self.vision_feature_select_strategy == "full":
+            selected_image_feature = selected_image_feature
+        else:
+            raise ValueError(
+                f"Unexpected select feature strategy: {self.config.vision_feature_select_strategy}"
+            )
+
+        height = width = self.num_patches_per_side
+        num_of_frames = selected_image_feature.shape[0]
+        selected_image_feature = selected_image_feature.view(
+            num_of_frames, height, width, -1
+        )
+        selected_image_feature = selected_image_feature.permute(0, 3, 1, 2).contiguous()
+        selected_image_feature = (
+            self.resampler(selected_image_feature)
+            .flatten(2)
+            .transpose(1, 2)
+            .contiguous()
+        )
+
+        image_features = self.multi_modal_projector(selected_image_feature)
+
+        return image_features
+
+    def forward(
+        self,
+        input_ids: torch.LongTensor,
+        positions: torch.Tensor,
+        input_metadata: InputMetadata,
+        pixel_values: Optional[List[Optional[np.array]]] = None,
+        image_sizes: Optional[List[List[int]]] = None,
+        image_offsets: Optional[List[int]] = None,
+    ) -> torch.Tensor:
+        if input_metadata.forward_mode == ForwardMode.EXTEND:
+            bs = input_metadata.batch_size
+
+            # Embed text input
+            input_embeds = self.language_model.model.embed_tokens(input_ids)
+
+            # Embed vision input
+            need_vision = (
+                (positions[input_metadata.extend_start_loc] < self.image_feature_len)
+                .cpu()
+                .numpy()
+            )
+            # FIXME: We need to substract the length of the system prompt
+            has_pixel = np.array([pixel_values[i] is not None for i in range(bs)])
+            need_vision = need_vision & has_pixel
+
+            if need_vision.any():
+                pixel_values = [pixel_values[i] for i in range(bs) if need_vision[i]]
+                image_sizes = [image_sizes[i] for i in range(bs) if need_vision[i]]
+
+                ########## Encode Image ########
+
+                if pixel_values[0].ndim == 4:
+                    # llava-hd: BS, num_patch, C=3, H=336, W=336, num_patch obtained from process_images
+                    np.concatenate(pixel_values, axis=0)
+                    # ndim=4
+                    concat_images = torch.tensor(
+                        np.concatenate(pixel_values, axis=0),
+                        device=self.vision_tower.device,
+                    )
+                    # image_features = self.encode_images(concat_images)
+                    # split_sizes = [image.shape[0] for image in pixel_values]
+                    # image_features = torch.split(image_features, split_sizes, dim=0)
+                    image_features = self.encode_images(
+                        concat_images
+                    )  # , prompts)#, image_counts, long_video=long_video)
+                    split_sizes = [image.shape[0] for image in pixel_values]
+                    image_features = torch.split(image_features, split_sizes, dim=0)
+
+                    # hd image_features: BS, num_patch, 576, 4096
+                else:
+                    # normal pixel: BS, C=3, H=336, W=336
+                    pixel_values = torch.tensor(
+                        np.array(pixel_values), device=self.vision_tower.device
+                    )
+                    image_features = self.encode_images(pixel_values)
+                    # image_features: BS, 576, 4096
+
+                new_image_features = []
+                for image_idx, image_feature in enumerate(image_features):
+                    new_image_features.append(image_feature.flatten(0, 1))
+                image_features = new_image_features
+
+                extend_start_loc_cpu = input_metadata.extend_start_loc.cpu().numpy()
+                pt = 0
+                for i in range(bs):
+                    if not need_vision[i]:
+                        continue
+
+                    start_idx = extend_start_loc_cpu[i]
+                    pad_len, pad_dim = image_features[pt].shape  # 576, 4096
+                    dim = input_embeds.shape[1]
+                    assert (
+                        pad_dim == dim
+                    ), "invalid pad_dim={}, input_embed_dim={}!".format(pad_dim, dim)
+                    # Fill in the placeholder for the image
+                    try:
+                        input_embeds[
+                            start_idx
+                            + image_offsets[i] : start_idx
+                            + image_offsets[i]
+                            + pad_len
+                        ] = image_features[pt]
+                    except RuntimeError as e:
+                        print(f"RuntimeError in llava image encoding: {e}")
+                        print(input_embeds.shape)
+                        print(start_idx, image_offsets[i])
+                    pt += 1
+
+            return self.language_model(
+                input_ids, positions, input_metadata, input_embeds=input_embeds
+            )
+        elif input_metadata.forward_mode == ForwardMode.DECODE:
+            return self.language_model(input_ids, positions, input_metadata)
+
+    def load_weights(
+        self,
+        model_name_or_path: str,
+        cache_dir: Optional[str] = None,
+        load_format: str = "auto",
+        revision: Optional[str] = None,
+    ):
+        # load clip vision model by cfg['mm_vision_tower']:
+        #   huggingface_name or path_of_clip_relative_to_llava_model_dir
+        vision_path = self.config.mm_vision_tower
+        self.vision_tower = CLIPVisionModel.from_pretrained(
+            vision_path, torch_dtype=torch.float16
+        ).cuda()
+        self.vision_tower.eval()
+
+        self.vision_feature_layer = self.config.mm_vision_select_layer
+        self.vision_feature_select_strategy = self.config.mm_vision_select_feature
+        self.image_size = self.vision_tower.config.image_size
+        self.patch_size = self.vision_tower.config.patch_size
+
+        self.mm_patch_merge_type = getattr(self.config, "mm_patch_merge_type", "flat")
+        self.image_aspect_ratio = getattr(self.config, "image_aspect_ratio", "square")
+        self.image_grid_pinpoints = getattr(self.config, "image_grid_pinpoints", None)
+
+        print(f"target_frames: {self.num_frames}")
+        self.image_feature_len = self.num_frames * int(
+            (self.image_size / self.patch_size / self.mm_spatial_pool_stride) ** 2
+        )
+        if self.vision_feature_select_strategy == "patch":
+            pass
+        elif self.vision_feature_select_strategy == "cls_patch":
+            self.image_feature_len += 1
+        else:
+            raise ValueError(f"Unexpected select feature: {self.select_feature}")
+
+        # load mm_projector
+        projector_weights = {
+            "model.mm_projector.0": "multi_modal_projector.linear_1",
+            "model.mm_projector.2": "multi_modal_projector.linear_2",
+            "model.vision_resampler.mm_projector.0": "multi_modal_projector.linear_1",
+            "model.vision_resampler.mm_projector.2": "multi_modal_projector.linear_2",
+            "model.vision_tower.vision_tower": "vision_tower",  # Update the vision tower weights if we find them in the checkpoint (it may be finetuned).
+        }
+        params_dict = dict(self.named_parameters())
+        for name, loaded_weight in hf_model_weights_iterator(
+            model_name_or_path, cache_dir, load_format, revision
+        ):
+            # FIXME: why projector weights read two times?
+            if "projector" in name or "vision_tower" in name:
+                for weight_name, param_name in projector_weights.items():
+                    if weight_name in name:
+                        name = name.replace(weight_name, param_name)
+                if name in params_dict:
+                    param = params_dict[name]
+                else:
+                    print(f"Warning: {name} not found in the model")
+                    continue
+                weight_loader = getattr(param, "weight_loader", default_weight_loader)
+                weight_loader(param, loaded_weight)
+
+        # load language model
+        self.language_model.load_weights(
+            model_name_or_path, cache_dir, load_format, revision
+        )
+
+        monkey_path_clip_vision_embed_forward()
+
+    @property
+    def num_patches_per_side(self):
+        return self.image_size // self.patch_size
+
+
+first_call = True
+
+
+def clip_vision_embed_forward(self, pixel_values: torch.FloatTensor) -> torch.Tensor:
+    batch_size = pixel_values.shape[0]
+
+    # Move this conv layer to CPU to avoid a bug in torch >= 2.1 on A10G.
+    global first_call
+    if first_call:
+        self.patch_embedding.cpu().float()
+        first_call = False
+    pixel_values = pixel_values.to(dtype=torch.float32, device="cpu")
+    patch_embeds = self.patch_embedding(pixel_values).cuda().half()
+
+    patch_embeds = patch_embeds.flatten(2).transpose(1, 2)
+
+    class_embeds = self.class_embedding.expand(batch_size, 1, -1)
+    embeddings = torch.cat([class_embeds, patch_embeds], dim=1)
+    embeddings = embeddings + self.position_embedding(self.position_ids)
+    return embeddings
+
+
+def monkey_path_clip_vision_embed_forward():
+    import transformers
+
+    setattr(
+        transformers.models.clip.modeling_clip.CLIPVisionEmbeddings,
+        "forward",
+        clip_vision_embed_forward,
+    )
+
+
+EntryClass = LlavaVidForCausalLM

python/sglang/srt/models/mixtral.py

@@ -8,34 +8,28 @@ import torch
 import torch.nn.functional as F
 from torch import nn
 from transformers import MixtralConfig
+from vllm.distributed import (
+    get_tensor_model_parallel_rank,
+    get_tensor_model_parallel_world_size,
+    tensor_model_parallel_all_reduce,
+)
 from vllm.model_executor.layers.layernorm import RMSNorm
 from vllm.model_executor.layers.linear import (
     QKVParallelLinear,
     ReplicatedLinear,
     RowParallelLinear,
 )
-from vllm.model_executor.layers.quantization.base_config import (
-    QuantizationConfig)
+from vllm.model_executor.layers.quantization.base_config import QuantizationConfig
 from vllm.model_executor.layers.rotary_embedding import get_rope
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     ParallelLMHead,
     VocabParallelEmbedding,
 )
-from vllm.distributed import (
-    tensor_model_parallel_all_reduce,
-)
-from vllm.distributed import (
-    get_tensor_model_parallel_rank,
-    get_tensor_model_parallel_world_size,
-)
-from sglang.srt.weight_utils import (
-    default_weight_loader,
-    hf_model_weights_iterator,
-)
 
 from sglang.srt.layers.logits_processor import LogitsProcessor
 from sglang.srt.layers.radix_attention import RadixAttention
 from sglang.srt.managers.router.model_runner import InputMetadata
+from sglang.srt.weight_utils import default_weight_loader, hf_model_weights_iterator
 
 
 class MixtralMLP(nn.Module):

python/sglang/srt/models/qwen.py

@@ -3,6 +3,7 @@ from typing import Any, Dict, Optional
 import torch
 from torch import nn
 from transformers import PretrainedConfig
+from vllm.distributed import get_tensor_model_parallel_world_size
 from vllm.model_executor.layers.activation import SiluAndMul
 from vllm.model_executor.layers.layernorm import RMSNorm
 from vllm.model_executor.layers.linear import (
@@ -10,24 +11,17 @@ from vllm.model_executor.layers.linear import (
     QKVParallelLinear,
     RowParallelLinear,
 )
-from vllm.model_executor.layers.quantization.base_config import (
-    QuantizationConfig)
+from vllm.model_executor.layers.quantization.base_config import QuantizationConfig
 from vllm.model_executor.layers.rotary_embedding import get_rope
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     ParallelLMHead,
     VocabParallelEmbedding,
 )
-from vllm.distributed import (
-    get_tensor_model_parallel_world_size,
-)
-from sglang.srt.weight_utils import (
-    default_weight_loader,
-    hf_model_weights_iterator,
-)
 
 from sglang.srt.layers.logits_processor import LogitsProcessor
 from sglang.srt.layers.radix_attention import RadixAttention
 from sglang.srt.managers.router.model_runner import InputMetadata
+from sglang.srt.weight_utils import default_weight_loader, hf_model_weights_iterator
 
 
 class QWenMLP(nn.Module):
@@ -132,7 +126,12 @@ class QWenAttention(nn.Module):
 
 
 class QWenBlock(nn.Module):
-    def __init__(self, config: PretrainedConfig, layer_id, quant_config: Optional[QuantizationConfig] = None,):
+    def __init__(
+        self,
+        config: PretrainedConfig,
+        layer_id,
+        quant_config: Optional[QuantizationConfig] = None,
+    ):
         super().__init__()
         self.ln_1 = RMSNorm(config.hidden_size, eps=config.layer_norm_epsilon)
 
@@ -181,7 +180,11 @@ class QWenBlock(nn.Module):
 
 
 class QWenModel(nn.Module):
-    def __init__(self, config: PretrainedConfig, quant_config: Optional[QuantizationConfig] = None,):
+    def __init__(
+        self,
+        config: PretrainedConfig,
+        quant_config: Optional[QuantizationConfig] = None,
+    ):
         super().__init__()
         self.config = config
         self.vocab_size = config.vocab_size
@@ -218,7 +221,11 @@ class QWenModel(nn.Module):
 
 
 class QWenLMHeadModel(nn.Module):
-    def __init__(self, config: PretrainedConfig, quant_config: Optional[QuantizationConfig] = None,):
+    def __init__(
+        self,
+        config: PretrainedConfig,
+        quant_config: Optional[QuantizationConfig] = None,
+    ):
         super().__init__()
         self.config = config
         self.transformer = QWenModel(config, quant_config=quant_config)
@@ -276,4 +283,4 @@ class QWenLMHeadModel(nn.Module):
                 weight_loader(param, loaded_weight)
 
 
-EntryClass = QWenLMHeadModel
+EntryClass = QWenLMHeadModel

python/sglang/srt/models/qwen2.py

@@ -5,6 +5,7 @@ from typing import Any, Dict, Optional, Tuple
 
 import torch
 from torch import nn
+from vllm.distributed import get_tensor_model_parallel_world_size
 from vllm.model_executor.layers.activation import SiluAndMul
 from vllm.model_executor.layers.layernorm import RMSNorm
 from vllm.model_executor.layers.linear import (
@@ -12,24 +13,17 @@ from vllm.model_executor.layers.linear import (
     QKVParallelLinear,
     RowParallelLinear,
 )
-from vllm.model_executor.layers.quantization.base_config import (
-    QuantizationConfig)
+from vllm.model_executor.layers.quantization.base_config import QuantizationConfig
 from vllm.model_executor.layers.rotary_embedding import get_rope
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     ParallelLMHead,
     VocabParallelEmbedding,
 )
-from vllm.distributed import (
-    get_tensor_model_parallel_world_size,
-)
-from sglang.srt.weight_utils import (
-    default_weight_loader,
-    hf_model_weights_iterator,
-)
 
 from sglang.srt.layers.logits_processor import LogitsProcessor
 from sglang.srt.layers.radix_attention import RadixAttention
 from sglang.srt.managers.router.model_runner import InputMetadata
+from sglang.srt.weight_utils import default_weight_loader, hf_model_weights_iterator
 
 Qwen2Config = None
 
@@ -50,7 +44,10 @@ class Qwen2MLP(nn.Module):
             quant_config=quant_config,
         )
         self.down_proj = RowParallelLinear(
-            intermediate_size, hidden_size, bias=False, quant_config=quant_config,
+            intermediate_size,
+            hidden_size,
+            bias=False,
+            quant_config=quant_config,
         )
         if hidden_act != "silu":
             raise ValueError(

python/sglang/srt/models/stablelm.py

@@ -7,35 +7,31 @@ from typing import Optional, Tuple
 import torch
 from torch import nn
 from transformers import PretrainedConfig
+from vllm.distributed import get_tensor_model_parallel_world_size
 from vllm.model_executor.layers.activation import SiluAndMul
 from vllm.model_executor.layers.linear import (
     MergedColumnParallelLinear,
     QKVParallelLinear,
     RowParallelLinear,
 )
-from vllm.model_executor.layers.quantization.base_config import (
-    QuantizationConfig)
+from vllm.model_executor.layers.quantization.base_config import QuantizationConfig
 from vllm.model_executor.layers.rotary_embedding import get_rope
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     ParallelLMHead,
     VocabParallelEmbedding,
 )
-from vllm.distributed import (
-    get_tensor_model_parallel_world_size,
-)
-from sglang.srt.weight_utils import (
-    default_weight_loader,
-    hf_model_weights_iterator,
-)
 
 from sglang.srt.layers.logits_processor import LogitsProcessor
 from sglang.srt.layers.radix_attention import RadixAttention
 from sglang.srt.managers.router.model_runner import InputMetadata
+from sglang.srt.weight_utils import default_weight_loader, hf_model_weights_iterator
 
 
 class StablelmMLP(nn.Module):
     def __init__(
-        self, config: PretrainedConfig, quant_config: Optional[QuantizationConfig] = None,
+        self,
+        config: PretrainedConfig,
+        quant_config: Optional[QuantizationConfig] = None,
     ) -> None:
         super().__init__()
         self.config = config
@@ -48,7 +44,10 @@ class StablelmMLP(nn.Module):
             quant_config=quant_config,
         )
         self.down_proj = RowParallelLinear(
-            config.intermediate_size, config.hidden_size, bias=False, quant_config=quant_config,
+            config.intermediate_size,
+            config.hidden_size,
+            bias=False,
+            quant_config=quant_config,
         )
         self.act_fn = SiluAndMul()
 
@@ -181,7 +180,9 @@ class StablelmDecoderLayer(nn.Module):
 
 class StableLMEpochModel(nn.Module):
     def __init__(
-        self, config: PretrainedConfig, quant_config: Optional[QuantizationConfig] = None,
+        self,
+        config: PretrainedConfig,
+        quant_config: Optional[QuantizationConfig] = None,
     ) -> None:
         super().__init__()
         self.embed_tokens = VocabParallelEmbedding(

python/sglang/srt/models/yivl.py

@@ -6,16 +6,13 @@ from typing import List, Optional
 import torch
 import torch.nn as nn
 from transformers import CLIPVisionModel, LlavaConfig
-from sglang.srt.weight_utils import (
-    default_weight_loader,
-    hf_model_weights_iterator,
-)
 
 from sglang.srt.models.llava import (
     LlavaLlamaForCausalLM,
     clip_vision_embed_forward,
     monkey_path_clip_vision_embed_forward,
 )
+from sglang.srt.weight_utils import default_weight_loader, hf_model_weights_iterator
 
 
 class YiVLForCausalLM(LlavaLlamaForCausalLM):

python/sglang/srt/server.py

@@ -107,7 +107,7 @@ async def openai_v1_chat_completions(raw_request: Request):
     return await v1_chat_completions(tokenizer_manager, raw_request)
 
 
-def launch_server(server_args: ServerArgs, pipe_finish_writer):
+def launch_server(server_args: ServerArgs, pipe_finish_writer, model_overide_args=None):
     global tokenizer_manager
 
     logging.basicConfig(
@@ -140,17 +140,13 @@ def launch_server(server_args: ServerArgs, pipe_finish_writer):
     )
 
     # Launch processes
-    tokenizer_manager = TokenizerManager(server_args, port_args)
+    tokenizer_manager = TokenizerManager(server_args, port_args, model_overide_args)
     pipe_router_reader, pipe_router_writer = mp.Pipe(duplex=False)
     pipe_detoken_reader, pipe_detoken_writer = mp.Pipe(duplex=False)
 
     proc_router = mp.Process(
         target=start_router_process,
-        args=(
-            server_args,
-            port_args,
-            pipe_router_writer,
-        ),
+        args=(server_args, port_args, pipe_router_writer, model_overide_args),
     )
     proc_router.start()
     proc_detoken = mp.Process(
@@ -170,8 +166,13 @@ def launch_server(server_args: ServerArgs, pipe_finish_writer):
     if router_init_state != "init ok" or detoken_init_state != "init ok":
         proc_router.kill()
         proc_detoken.kill()
-        print(f"Initialization failed. router_init_state: {router_init_state}", flush=True)
-        print(f"Initialization failed. detoken_init_state: {detoken_init_state}", flush=True)
+        print(
+            f"Initialization failed. router_init_state: {router_init_state}", flush=True
+        )
+        print(
+            f"Initialization failed. detoken_init_state: {detoken_init_state}",
+            flush=True,
+        )
         sys.exit(1)
     assert proc_router.is_alive() and proc_detoken.is_alive()
 
@@ -189,6 +190,7 @@ def launch_server(server_args: ServerArgs, pipe_finish_writer):
             time.sleep(0.5)
             try:
                 requests.get(url + "/get_model_info", timeout=5, headers=headers)
+                success = True  # Set flag to True if request succeeds
                 break
             except requests.exceptions.RequestException as e:
                 pass
@@ -205,7 +207,7 @@ def launch_server(server_args: ServerArgs, pipe_finish_writer):
                     },
                 },
                 headers=headers,
-                timeout=60,
+                timeout=600,
             )
             assert res.status_code == 200
         except Exception as e:
@@ -235,7 +237,8 @@ def launch_server(server_args: ServerArgs, pipe_finish_writer):
 class Runtime:
     def __init__(
         self,
-        log_evel="error",
+        log_evel: str = "error",
+        model_overide_args: Optional[dict] = None,
         *args,
         **kwargs,
     ):
@@ -244,7 +247,10 @@ class Runtime:
 
         # Pre-allocate ports
         self.server_args.port, self.server_args.additional_ports = allocate_init_ports(
-            self.server_args.port, self.server_args.additional_ports, self.server_args.tp_size)
+            self.server_args.port,
+            self.server_args.additional_ports,
+            self.server_args.tp_size,
+        )
 
         self.url = self.server_args.url()
         self.generate_url = (
@@ -253,7 +259,10 @@ class Runtime:
 
         self.pid = None
         pipe_reader, pipe_writer = mp.Pipe(duplex=False)
-        proc = mp.Process(target=launch_server, args=(self.server_args, pipe_writer))
+        proc = mp.Process(
+            target=launch_server,
+            args=(self.server_args, pipe_writer, model_overide_args),
+        )
         proc.start()
         pipe_writer.close()
         self.pid = proc.pid
@@ -265,7 +274,9 @@ class Runtime:
 
         if init_state != "init ok":
             self.shutdown()
-            raise RuntimeError("Initialization failed. Please see the error messages above.")
+            raise RuntimeError(
+                "Initialization failed. Please see the error messages above."
+            )
 
         self.endpoint = RuntimeEndpoint(self.url)
 
@@ -317,4 +328,4 @@ class Runtime:
                         pos += len(cur)
 
     def __del__(self):
-        self.shutdown()
+        self.shutdown()

python/sglang/srt/server_args.py

@@ -80,10 +80,12 @@ class ServerArgs:
             default=ServerArgs.tokenizer_path,
             help="The path of the tokenizer.",
         )
-        parser.add_argument("--host", type=str, default=ServerArgs.host,
-                            help="The host of the server.")
-        parser.add_argument("--port", type=int, default=ServerArgs.port,
-                            help="The port of the server.")
+        parser.add_argument(
+            "--host", type=str, default=ServerArgs.host, help="The host of the server."
+        )
+        parser.add_argument(
+            "--port", type=int, default=ServerArgs.port, help="The port of the server."
+        )
         parser.add_argument(
             "--additional-ports",
             type=int,
@@ -261,4 +263,4 @@ class PortArgs:
     router_port: int
     detokenizer_port: int
     nccl_port: int
-    model_rpc_ports: List[int]
+    model_rpc_ports: List[int]

python/sglang/srt/utils.py

@@ -131,11 +131,13 @@ def alloc_usable_network_port(num, used_list=()):
             continue
 
         with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
+            s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
             try:
                 s.bind(("", port))
+                s.listen(1)  # Attempt to listen on the port
                 port_list.append(port)
             except socket.error:
-                pass
+                pass  # If any error occurs, this port is not usable
 
             if len(port_list) == num:
                 return port_list
@@ -265,20 +267,102 @@ def wrap_kernel_launcher(kernel):
 
 
 def is_multimodal_model(model):
-    if isinstance(model, str):
-        return "llava" in model or "yi-vl" in model
     from sglang.srt.model_config import ModelConfig
 
+    if isinstance(model, str):
+        model = model.lower()
+        return "llava" in model or "yi-vl" in model or "llava-next" in model
+
     if isinstance(model, ModelConfig):
         model_path = model.path.lower()
-        return "llava" in model_path or "yi-vl" in model_path
-    raise Exception("unrecognized type")
+        return "llava" in model_path or "yi-vl" in model_path or "llava-next" in model_path
+
+    raise ValueError("unrecognized type")
+
+
+def decode_video_base64(video_base64):
+    from PIL import Image
+
+    # Decode the base64 string
+    video_bytes = base64.b64decode(video_base64)
+
+    # Placeholder for the start indices of each PNG image
+    img_starts = []
+
+    frame_format = "PNG"  # str(os.getenv('FRAME_FORMAT', "JPEG"))
+
+    assert frame_format in [
+        "PNG",
+        "JPEG",
+    ], "FRAME_FORMAT must be either 'PNG' or 'JPEG'"
+
+    if frame_format == "PNG":
+        # Find each PNG start signature to isolate images
+        i = 0
+        while i < len(video_bytes) - 7:  # Adjusted for the length of the PNG signature
+            # Check if we found the start of a PNG file
+            if (
+                video_bytes[i] == 0x89
+                and video_bytes[i + 1] == 0x50
+                and video_bytes[i + 2] == 0x4E
+                and video_bytes[i + 3] == 0x47
+                and video_bytes[i + 4] == 0x0D
+                and video_bytes[i + 5] == 0x0A
+                and video_bytes[i + 6] == 0x1A
+                and video_bytes[i + 7] == 0x0A
+            ):
+                img_starts.append(i)
+                i += 8  # Skip the PNG signature
+            else:
+                i += 1
+    else:
+        # Find each JPEG start (0xFFD8) to isolate images
+        i = 0
+        while (
+            i < len(video_bytes) - 1
+        ):  # Adjusted for the length of the JPEG SOI signature
+            # Check if we found the start of a JPEG file
+            if video_bytes[i] == 0xFF and video_bytes[i + 1] == 0xD8:
+                img_starts.append(i)
+                # Move to the next byte to continue searching for the next image start
+                i += 2
+            else:
+                i += 1
+
+    frames = []
+    for start_idx in img_starts:
+        # Assuming each image is back-to-back, the end of one image is the start of another
+        # The last image goes until the end of the byte string
+        end_idx = (
+            img_starts[img_starts.index(start_idx) + 1]
+            if img_starts.index(start_idx) + 1 < len(img_starts)
+            else len(video_bytes)
+        )
+        img_bytes = video_bytes[start_idx:end_idx]
+
+        # Convert bytes to a PIL Image
+        img = Image.open(BytesIO(img_bytes))
+
+        # Convert PIL Image to a NumPy array
+        frame = np.array(img)
+
+        # Append the frame to the list of frames
+        frames.append(frame)
+
+    # Ensure there's at least one frame to avoid errors with np.stack
+    if frames:
+        return np.stack(frames, axis=0), img.size
+    else:
+        return np.array([]), (
+            0,
+            0,
+        )  # Return an empty array and size tuple if no frames were found
 
 
 def load_image(image_file):
     from PIL import Image
 
-    image = None
+    image = image_size = None
 
     if image_file.startswith("http://") or image_file.startswith("https://"):
         timeout = int(os.getenv("REQUEST_TIMEOUT", "3"))
@@ -289,10 +373,13 @@ def load_image(image_file):
     elif image_file.startswith("data:"):
         image_file = image_file.split(",")[1]
         image = Image.open(BytesIO(base64.b64decode(image_file)))
+    elif image_file.startswith("video:"):
+        image_file = image_file.replace("video:", "")
+        image, image_size = decode_video_base64(image_file)
     else:
         image = Image.open(BytesIO(base64.b64decode(image_file)))
 
-    return image
+    return image, image_size
 
 
 def assert_pkg_version(pkg: str, min_version: str):
@@ -304,7 +391,9 @@ def assert_pkg_version(pkg: str, min_version: str):
                 f"is less than the minimum required version {min_version}"
             )
     except PackageNotFoundError:
-        raise Exception(f"{pkg} with minimum required version {min_version} is not installed")
+        raise Exception(
+            f"{pkg} with minimum required version {min_version} is not installed"
+        )
 
 
 API_KEY_HEADER_NAME = "X-API-Key"

python/sglang/srt/weight_utils.py

@@ -19,11 +19,12 @@ import torch
 from huggingface_hub import HfFileSystem, snapshot_download
 from safetensors.torch import load_file, safe_open, save_file
 from tqdm.auto import tqdm
-
 from vllm.config import ModelConfig
 from vllm.logger import init_logger
-from vllm.model_executor.layers.quantization import (QuantizationConfig,
-                                                     get_quantization_config)
+from vllm.model_executor.layers.quantization import (
+    QuantizationConfig,
+    get_quantization_config,
+)
 from vllm.model_executor.layers.quantization.schema import QuantParamSchema
 
 logger = init_logger(__name__)
@@ -32,17 +33,21 @@ logger = init_logger(__name__)
 # can share the same lock without error.
 # lock files in the temp directory will be automatically deleted when the
 # system reboots, so users will not complain about annoying lock files
-temp_dir = os.environ.get('TMPDIR') or os.environ.get(
-    'TEMP') or os.environ.get('TMP') or "/tmp/"
+temp_dir = (
+    os.environ.get("TMPDIR")
+    or os.environ.get("TEMP")
+    or os.environ.get("TMP")
+    or "/tmp/"
+)
 
 
 def enable_hf_transfer():
-    """automatically activates hf_transfer
-    """
+    """automatically activates hf_transfer"""
     if "HF_HUB_ENABLE_HF_TRANSFER" not in os.environ:
         try:
             # enable hf hub transfer if available
             import hf_transfer  # type: ignore # noqa
+
             huggingface_hub.constants.HF_HUB_ENABLE_HF_TRANSFER = True
         except ImportError:
             pass
@@ -65,8 +70,7 @@ def get_lock(model_name_or_path: str, cache_dir: Optional[str] = None):
     # add hash to avoid conflict with old users' lock files
     lock_file_name = hash_name + model_name + ".lock"
     # mode 0o666 is required for the filelock to be shared across users
-    lock = filelock.FileLock(os.path.join(lock_dir, lock_file_name),
-                             mode=0o666)
+    lock = filelock.FileLock(os.path.join(lock_dir, lock_file_name), mode=0o666)
     return lock
 
 
@@ -104,10 +108,12 @@ def convert_bin_to_safetensor_file(
     sf_size = os.stat(sf_filename).st_size
     pt_size = os.stat(pt_filename).st_size
     if (sf_size - pt_size) / pt_size > 0.01:
-        raise RuntimeError(f"""The file size different is more than 1%:
+        raise RuntimeError(
+            f"""The file size different is more than 1%:
          - {sf_filename}: {sf_size}
          - {pt_filename}: {pt_size}
-         """)
+         """
+        )
 
     # check if the tensors are the same
     reloaded = load_file(sf_filename)
@@ -122,8 +128,7 @@ def convert_bin_to_safetensor_file(
 def get_quant_config(model_config: ModelConfig) -> QuantizationConfig:
     quant_cls = get_quantization_config(model_config.quantization)
     # Read the quantization config from the HF model config, if available.
-    hf_quant_config = getattr(model_config.hf_config, "quantization_config",
-                              None)
+    hf_quant_config = getattr(model_config.hf_config, "quantization_config", None)
     if hf_quant_config is not None:
         return quant_cls.from_config(hf_quant_config)
     model_name_or_path = model_config.model
@@ -131,26 +136,29 @@ def get_quant_config(model_config: ModelConfig) -> QuantizationConfig:
     if not is_local:
         # Download the config files.
         with get_lock(model_name_or_path, model_config.download_dir):
-            hf_folder = snapshot_download(model_name_or_path,
-                                          revision=model_config.revision,
-                                          allow_patterns="*.json",
-                                          cache_dir=model_config.download_dir,
-                                          tqdm_class=Disabledtqdm)
+            hf_folder = snapshot_download(
+                model_name_or_path,
+                revision=model_config.revision,
+                allow_patterns="*.json",
+                cache_dir=model_config.download_dir,
+                tqdm_class=Disabledtqdm,
+            )
     else:
         hf_folder = model_name_or_path
     config_files = glob.glob(os.path.join(hf_folder, "*.json"))
 
     quant_config_files = [
-        f for f in config_files if any(
-            f.endswith(x) for x in quant_cls.get_config_filenames())
+        f
+        for f in config_files
+        if any(f.endswith(x) for x in quant_cls.get_config_filenames())
     ]
     if len(quant_config_files) == 0:
-        raise ValueError(
-            f"Cannot find the config file for {model_config.quantization}")
+        raise ValueError(f"Cannot find the config file for {model_config.quantization}")
     if len(quant_config_files) > 1:
         raise ValueError(
             f"Found multiple config files for {model_config.quantization}: "
-            f"{quant_config_files}")
+            f"{quant_config_files}"
+        )
 
     quant_config_file = quant_config_files[0]
     with open(quant_config_file, "r") as f:
@@ -166,8 +174,7 @@ def prepare_hf_model_weights(
     revision: Optional[str] = None,
 ) -> Tuple[str, List[str], bool]:
     # Download model weights from huggingface.
-    is_local = os.path.isdir(model_name_or_path) \
-               and load_format != "tensorizer"
+    is_local = os.path.isdir(model_name_or_path) and load_format != "tensorizer"
     use_safetensors = False
     # Some quantized models use .pt files for storing the weights.
     if load_format == "auto":
@@ -203,11 +210,13 @@ def prepare_hf_model_weights(
         # Use file lock to prevent multiple processes from
         # downloading the same model weights at the same time.
         with get_lock(model_name_or_path, cache_dir):
-            hf_folder = snapshot_download(model_name_or_path,
-                                          allow_patterns=allow_patterns,
-                                          cache_dir=cache_dir,
-                                          tqdm_class=Disabledtqdm,
-                                          revision=revision)
+            hf_folder = snapshot_download(
+                model_name_or_path,
+                allow_patterns=allow_patterns,
+                cache_dir=cache_dir,
+                tqdm_class=Disabledtqdm,
+                revision=revision,
+            )
     else:
         hf_folder = model_name_or_path
     hf_weights_files: List[str] = []
@@ -228,16 +237,14 @@ def prepare_hf_model_weights(
             "scaler.pt",
         ]
         hf_weights_files = [
-            f for f in hf_weights_files
-            if not any(f.endswith(x) for x in blacklist)
+            f for f in hf_weights_files if not any(f.endswith(x) for x in blacklist)
         ]
 
     if load_format == "tensorizer":
         return hf_folder, hf_weights_files, use_safetensors
 
     if len(hf_weights_files) == 0:
-        raise RuntimeError(
-            f"Cannot find any model weights with `{model_name_or_path}`")
+        raise RuntimeError(f"Cannot find any model weights with `{model_name_or_path}`")
 
     return hf_folder, hf_weights_files, use_safetensors
 
@@ -254,7 +261,8 @@ def hf_model_weights_iterator(
         cache_dir=cache_dir,
         load_format=load_format,
         fall_back_to_pt=fall_back_to_pt,
-        revision=revision)
+        revision=revision,
+    )
 
     if load_format == "npcache":
         # Currently np_cache only support *.bin checkpoints
@@ -289,22 +297,25 @@ def hf_model_weights_iterator(
                 param = np.load(f)
             yield name, torch.from_numpy(param)
     elif load_format == "tensorizer":
-        from vllm.model_executor.tensorizer_loader import (TensorDeserializer,
-                                                           open_stream,
-                                                           tensorizer_warning)
+        from vllm.model_executor.tensorizer_loader import (
+            TensorDeserializer,
+            open_stream,
+            tensorizer_warning,
+        )
+
         tensorizer_args = load_format.params
         tensorizer_warning(
             "Deserializing HuggingFace models is not optimized for "
             "loading on vLLM, as tensorizer is forced to load to CPU. "
             "Consider deserializing a vLLM model instead for faster "
             "load times. See the examples/tensorize_vllm_model.py example "
-            "script for serializing vLLM models.")
+            "script for serializing vLLM models."
+        )
 
         deserializer_args = tensorizer_args.deserializer_params
         stream_params = tensorizer_args.stream_params
         stream = open_stream(tensorizer_args.tensorizer_uri, **stream_params)
-        with TensorDeserializer(stream, **deserializer_args,
-                                device="cpu") as state:
+        with TensorDeserializer(stream, **deserializer_args, device="cpu") as state:
             for name, param in state.items():
                 yield name, param
         del state
@@ -324,8 +335,12 @@ def hf_model_weights_iterator(
 
 
 def kv_cache_scales_loader(
-        filename: str, tp_rank: int, tp_size: int, num_hidden_layers: int,
-        model_type: Optional[str]) -> Iterable[Tuple[int, float]]:
+    filename: str,
+    tp_rank: int,
+    tp_size: int,
+    num_hidden_layers: int,
+    model_type: Optional[str],
+) -> Iterable[Tuple[int, float]]:
     """
     A simple utility to read in KV cache scaling factors that have been
     previously serialized to disk. Used by the model to populate the appropriate
@@ -343,8 +358,7 @@ def kv_cache_scales_loader(
                 "tp_size": tp_size,
             }
             schema_dct = json.load(f)
-            schema = QuantParamSchema.model_validate(schema_dct,
-                                                     context=context)
+            schema = QuantParamSchema.model_validate(schema_dct, context=context)
             layer_scales_map = schema.kv_cache.scaling_factor[tp_rank]
             return layer_scales_map.items()
 
@@ -357,9 +371,11 @@ def kv_cache_scales_loader(
     # This section is reached if and only if any of the excepts are hit
     # Return an empty iterable (list) => no KV cache scales are loaded
     # which ultimately defaults to 1.0 scales
-    logger.warning("Defaulting to KV cache scaling factors = 1.0 "
-                   f"for all layers in TP rank {tp_rank} "
-                   "as an error occurred during loading.")
+    logger.warning(
+        "Defaulting to KV cache scaling factors = 1.0 "
+        f"for all layers in TP rank {tp_rank} "
+        "as an error occurred during loading."
+    )
     return []
 
 
@@ -378,8 +394,7 @@ def convert_pyslice_to_tensor(x: Any) -> torch.Tensor:
     return x
 
 
-def default_weight_loader(param: torch.Tensor,
-                          loaded_weight: torch.Tensor) -> None:
+def default_weight_loader(param: torch.Tensor, loaded_weight: torch.Tensor) -> None:
     """Default weight loader."""
     assert param.size() == loaded_weight.size()
     param.data.copy_(loaded_weight)
@@ -399,4 +414,4 @@ def initialize_dummy_weights(
     """
     for param in model.state_dict().values():
         if torch.is_floating_point(param):
-            param.data.uniform_(low, high)
+            param.data.uniform_(low, high)

python/sglang/utils.py

@@ -2,13 +2,16 @@
 
 import base64
 import json
+import os
 import sys
 import threading
 import traceback
 import urllib.request
+from concurrent.futures import ThreadPoolExecutor
 from io import BytesIO
 from json import dumps
 
+import numpy as np
 import requests
 
 
@@ -110,6 +113,74 @@ def encode_image_base64(image_path):
         return base64.b64encode(buffered.getvalue()).decode("utf-8")
 
 
+def encode_frame(frame):
+    import cv2  # pip install opencv-python-headless
+    from PIL import Image
+
+    # Convert the frame to RGB (OpenCV uses BGR by default)
+    frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
+
+    # Convert the frame to PIL Image to easily convert to bytes
+    im_pil = Image.fromarray(frame)
+
+    # Convert to bytes
+    buffered = BytesIO()
+
+    # frame_format = str(os.getenv('FRAME_FORMAT', "JPEG"))
+
+    im_pil.save(buffered, format="PNG")
+
+    frame_bytes = buffered.getvalue()
+
+    # Return the bytes of the frame
+    return frame_bytes
+
+
+def encode_video_base64(video_path, num_frames=16):
+    import cv2
+    cap = cv2.VideoCapture(video_path)
+    if not cap.isOpened():
+        raise IOError(f"Could not open video file:{video_path}")
+
+    total_frames = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
+    print(f"target_frames: {num_frames}")
+
+    frame_indices = np.linspace(0, total_frames - 1, num_frames, dtype=int)
+
+    frames = []
+    for i in range(total_frames):
+        ret, frame = cap.read()
+        if ret:
+            frames.append(frame)
+        else:
+            # Handle the case where the frame could not be read
+            # print(f"Warning: Could not read frame at index {i}.")
+            pass
+
+    cap.release()
+
+    # Safely select frames based on frame_indices, avoiding IndexError
+    frames = [frames[i] for i in frame_indices if i < len(frames)]
+
+    # If there are not enough frames, duplicate the last frame until we reach the target
+    while len(frames) < num_frames:
+        frames.append(frames[-1])
+
+    # Use ThreadPoolExecutor to process and encode frames in parallel
+    with ThreadPoolExecutor() as executor:
+        encoded_frames = list(executor.map(encode_frame, frames))
+
+    # encoded_frames = list(map(encode_frame, frames))
+
+    # Concatenate all frames bytes
+    video_bytes = b"".join(encoded_frames)
+
+    # Encode the concatenated bytes to base64
+    video_base64 = "video:" + base64.b64encode(video_bytes).decode("utf-8")
+
+    return video_base64
+
+
 def _is_chinese_char(cp):
     """Checks whether CP is the codepoint of a CJK character."""
     # This defines a "chinese character" as anything in the CJK Unicode block:
@@ -170,4 +241,4 @@ def run_with_timeout(func, args=(), kwargs=None, timeout=None):
     if not ret_value:
         raise RuntimeError()
 
-    return ret_value[0]
+    return ret_value[0]