adapt to sglang v0.5.2rc1 on dcu

examples/runtime/lora.py

+# launch server
+# python -m sglang.launch_server --model mistralai/Mistral-7B-Instruct-v0.3 --lora-paths /home/ying/test_lora lora1=/home/ying/test_lora_1 lora2=/home/ying/test_lora_2 --disable-radix --disable-cuda-graph --max-loras-per-batch 4
+
+# send requests
+# lora_path[i] specifies the LoRA used for text[i], so make sure they have the same length
+# use None to specify base-only prompt, e.x. "lora_path": [None, "/home/ying/test_lora"]
+import json
+
+import requests
+
+url = "http://127.0.0.1:30000"
+json_data = {
+    "text": [
+        "prompt 1",
+        "prompt 2",
+        "prompt 3",
+        "prompt 4",
+        "prompt 5",
+        "prompt 6",
+        "prompt 7",
+    ],
+    "sampling_params": {"max_new_tokens": 32},
+    "lora_path": [
+        "/home/ying/test_lora",
+        "lora1",
+        "lora2",
+        "lora1",
+        "lora2",
+        None,
+        None,
+    ],
+}
+response = requests.post(
+    url + "/generate",
+    json=json_data,
+)
+print(json.dumps(response.json()))

examples/runtime/multimodal/llama3_llava_server.py

+"""
+Usage:
+# Installing latest llava-next: pip install git+https://github.com/LLaVA-VL/LLaVA-NeXT.git
+# Installing latest sglang.
+
+# Endpoint Service CLI:
+python -m sglang.launch_server --model-path lmms-lab/llama3-llava-next-8b --port=30000
+
+python3 llama3_llava_server.py
+
+Output:
+"Friends posing for a fun photo with a life-sized teddy bear, creating a playful and memorable moment."
+"""
+
+import argparse
+import asyncio
+import copy
+import json
+
+import aiohttp
+import requests
+from llava.conversation import conv_llava_llama_3
+
+
+async def send_request(url, data, delay=0):
+    await asyncio.sleep(delay)
+    async with aiohttp.ClientSession() as session:
+        async with session.post(url, json=data) as resp:
+            output = await resp.json()
+    return output
+
+
+async def test_concurrent(args):
+    url = f"{args.host}:{args.port}"
+
+    prompt = "<image>\nPlease generate caption towards this image."
+    conv_template = copy.deepcopy(conv_llava_llama_3)
+    conv_template.append_message(role=conv_template.roles[0], message=prompt)
+    conv_template.append_message(role=conv_template.roles[1], message=None)
+    prompt_with_template = conv_template.get_prompt()
+    response = []
+    for i in range(1):
+        response.append(
+            send_request(
+                url + "/generate",
+                {
+                    "text": prompt_with_template,
+                    "image_data": "https://farm4.staticflickr.com/3175/2653711032_804ff86d81_z.jpg",
+                    "sampling_params": {
+                        "max_new_tokens": 1024,
+                        "temperature": 0,
+                        "top_p": 1.0,
+                        "presence_penalty": 2,
+                        "frequency_penalty": 2,
+                        "stop": "<|eot_id|>",
+                    },
+                },
+            )
+        )
+
+    rets = await asyncio.gather(*response)
+    for ret in rets:
+        print(ret["text"])
+
+
+def test_streaming(args):
+    url = f"{args.host}:{args.port}"
+    prompt = "<image>\nPlease generate caption towards this image."
+    conv_template = copy.deepcopy(conv_llava_llama_3)
+    conv_template.append_message(role=conv_template.roles[0], message=prompt)
+    conv_template.append_message(role=conv_template.roles[1], message=None)
+    prompt_with_template = conv_template.get_prompt()
+    pload = {
+        "text": prompt_with_template,
+        "sampling_params": {
+            "max_new_tokens": 1024,
+            "temperature": 0,
+            "top_p": 1.0,
+            "presence_penalty": 2,
+            "frequency_penalty": 2,
+            "stop": "<|eot_id|>",
+        },
+        "image_data": "https://farm4.staticflickr.com/3175/2653711032_804ff86d81_z.jpg",
+        "stream": True,
+    }
+    response = requests.post(
+        url + "/generate",
+        json=pload,
+        stream=True,
+    )
+
+    prev = 0
+    for chunk in response.iter_lines(decode_unicode=False):
+        chunk = chunk.decode("utf-8")
+        if chunk and chunk.startswith("data:"):
+            if chunk == "data: [DONE]":
+                break
+            data = json.loads(chunk[5:].strip("\n"))
+            output = data["text"].strip()
+            print(output[prev:], end="", flush=True)
+            prev = len(output)
+    print("")
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--host", type=str, default="http://127.0.0.1")
+    parser.add_argument("--port", type=int, default=30000)
+    args = parser.parse_args()
+    asyncio.run(test_concurrent(args))
+    test_streaming(args)

examples/runtime/multimodal/llava_onevision_server.py

+"""
+Usage:
+
+python3 -m sglang.launch_server --model-path lmms-lab/llava-onevision-qwen2-72b-ov --port=30000 --tp-size=8
+
+python3 llava_onevision_server.py
+"""
+
+import base64
+import io
+import os
+import sys
+import time
+
+import numpy as np
+import openai
+import requests
+from decord import VideoReader, cpu
+from PIL import Image
+
+# pip install httpx==0.23.3
+# pip install decord
+# pip install protobuf==3.20.0
+
+
+def download_video(url, cache_dir):
+    file_path = os.path.join(cache_dir, "jobs.mp4")
+    os.makedirs(cache_dir, exist_ok=True)
+
+    response = requests.get(url)
+    response.raise_for_status()
+
+    with open(file_path, "wb") as f:
+        f.write(response.content)
+
+    print(f"File downloaded and saved to: {file_path}")
+    return file_path
+
+
+def create_openai_client(base_url):
+    return openai.Client(api_key="EMPTY", base_url=base_url)
+
+
+def image_stream_request_test(client):
+    print("----------------------Image Stream Request Test----------------------")
+    stream_request = client.chat.completions.create(
+        model="default",
+        messages=[
+            {
+                "role": "user",
+                "content": [
+                    {
+                        "type": "image_url",
+                        "image_url": {
+                            "url": "https://raw.githubusercontent.com/sgl-project/sglang/main/assets/logo.png"
+                        },
+                    },
+                    {
+                        "type": "text",
+                        "text": "Please describe this image. Please list the benchmarks and the models.",
+                    },
+                ],
+            },
+        ],
+        temperature=0.7,
+        max_tokens=1024,
+        stream=True,
+    )
+    stream_response = ""
+
+    for chunk in stream_request:
+        if chunk.choices[0].delta.content is not None:
+            content = chunk.choices[0].delta.content
+            stream_response += content
+            sys.stdout.write(content)
+            sys.stdout.flush()
+
+    print("-" * 30)
+
+
+def multi_image_stream_request_test(client):
+    print(
+        "----------------------Multi-Images Stream Request Test----------------------"
+    )
+    stream_request = client.chat.completions.create(
+        model="default",
+        messages=[
+            {
+                "role": "user",
+                "content": [
+                    {
+                        "type": "image_url",
+                        "image_url": {
+                            "url": "https://raw.githubusercontent.com/sgl-project/sglang/main/assets/logo.png"
+                        },
+                        "modalities": "multi-images",
+                    },
+                    {
+                        "type": "image_url",
+                        "image_url": {
+                            "url": "https://raw.githubusercontent.com/sgl-project/sglang/main/test/lang/example_image.png"
+                        },
+                        "modalities": "multi-images",
+                    },
+                    {
+                        "type": "text",
+                        "text": "I have shown you two images. Please describe the two images to me.",
+                    },
+                ],
+            },
+        ],
+        temperature=0.7,
+        max_tokens=1024,
+        stream=True,
+    )
+    stream_response = ""
+
+    for chunk in stream_request:
+        if chunk.choices[0].delta.content is not None:
+            content = chunk.choices[0].delta.content
+            stream_response += content
+            sys.stdout.write(content)
+            sys.stdout.flush()
+
+    print("-" * 30)
+
+
+def video_stream_request_test(client, video_path):
+    print("------------------------Video Stream Request Test----------------------")
+    messages = prepare_video_messages(video_path)
+
+    video_request = client.chat.completions.create(
+        model="default",
+        messages=messages,
+        temperature=0,
+        max_tokens=1024,
+        stream=True,
+    )
+    print("-" * 30)
+    video_response = ""
+
+    for chunk in video_request:
+        if chunk.choices[0].delta.content is not None:
+            content = chunk.choices[0].delta.content
+            video_response += content
+            sys.stdout.write(content)
+            sys.stdout.flush()
+    print("-" * 30)
+
+
+def image_speed_test(client):
+    print("----------------------Image Speed Test----------------------")
+    start_time = time.perf_counter()
+    request = client.chat.completions.create(
+        model="default",
+        messages=[
+            {
+                "role": "user",
+                "content": [
+                    {
+                        "type": "image_url",
+                        "image_url": {
+                            "url": "https://raw.githubusercontent.com/sgl-project/sglang/main/assets/logo.png"
+                        },
+                    },
+                    {
+                        "type": "text",
+                        "text": "Please describe this image. Please list the benchmarks and the models.",
+                    },
+                ],
+            },
+        ],
+        temperature=0,
+        max_tokens=1024,
+    )
+    end_time = time.perf_counter()
+    response = request.choices[0].message.content
+    print(response)
+    print("-" * 30)
+    print_speed_test_results(request, start_time, end_time)
+
+
+def video_speed_test(client, video_path):
+    print("------------------------Video Speed Test------------------------")
+    messages = prepare_video_messages(video_path)
+
+    start_time = time.perf_counter()
+    video_request = client.chat.completions.create(
+        model="default",
+        messages=messages,
+        temperature=0,
+        max_tokens=1024,
+    )
+    end_time = time.perf_counter()
+    video_response = video_request.choices[0].message.content
+    print(video_response)
+    print("-" * 30)
+    print_speed_test_results(video_request, start_time, end_time)
+
+
+def prepare_video_messages(video_path):
+    max_frames_num = 32
+    vr = VideoReader(video_path, ctx=cpu(0))
+    total_frame_num = len(vr)
+    uniform_sampled_frames = np.linspace(
+        0, total_frame_num - 1, max_frames_num, dtype=int
+    )
+    frame_idx = uniform_sampled_frames.tolist()
+    frames = vr.get_batch(frame_idx).asnumpy()
+
+    base64_frames = []
+    for frame in frames:
+        pil_img = Image.fromarray(frame)
+        buff = io.BytesIO()
+        pil_img.save(buff, format="JPEG")
+        base64_str = base64.b64encode(buff.getvalue()).decode("utf-8")
+        base64_frames.append(base64_str)
+
+    messages = [{"role": "user", "content": []}]
+
+    for base64_frame in base64_frames:
+        frame_format = {
+            "type": "image_url",
+            "image_url": {"url": f"data:image/jpeg;base64,{base64_frame}"},
+            "modalities": "video",
+        }
+        messages[0]["content"].append(frame_format)
+
+    prompt = {"type": "text", "text": "Please describe the video in detail."}
+    messages[0]["content"].append(prompt)
+
+    return messages
+
+
+def print_speed_test_results(request, start_time, end_time):
+    total_tokens = request.usage.total_tokens
+    completion_tokens = request.usage.completion_tokens
+    prompt_tokens = request.usage.prompt_tokens
+
+    print(f"Total tokens: {total_tokens}")
+    print(f"Completion tokens: {completion_tokens}")
+    print(f"Prompt tokens: {prompt_tokens}")
+    print(f"Time taken: {end_time - start_time} seconds")
+    print(f"Token per second: {total_tokens / (end_time - start_time)}")
+    print(f"Completion token per second: {completion_tokens / (end_time - start_time)}")
+    print(f"Prompt token per second: {prompt_tokens / (end_time - start_time)}")
+
+
+def main():
+    url = "https://raw.githubusercontent.com/EvolvingLMMs-Lab/sglang/dev/onevision_local/assets/jobs.mp4"
+    cache_dir = os.path.expanduser("~/.cache")
+    video_path = download_video(url, cache_dir)
+
+    client = create_openai_client("http://127.0.0.1:30000/v1")
+
+    image_stream_request_test(client)
+    multi_image_stream_request_test(client)
+    video_stream_request_test(client, video_path)
+    image_speed_test(client)
+    video_speed_test(client, video_path)
+
+
+if __name__ == "__main__":
+    main()

examples/runtime/multimodal/pixtral_server.py

+"""
+Usage:
+# Run a Pixtral model with SGLang:
+# HuggingFace:
+python -m sglang.launch_server --model-path mistral-community/pixtral-12b --port=30000
+# ModelScope:
+python -m sglang.launch_server --model-path AI-ModelScope/pixtral-12b --port=30000
+
+# Then test it with:
+python pixtral_server.py
+
+This script tests Pixtral model with both single and multiple images.
+"""
+
+import argparse
+import asyncio
+import json
+
+import aiohttp
+import requests
+
+IMAGE_TOKEN_SEP = "\n[IMG]"
+ROUTE = "/generate"
+
+
+async def send_request(url, data, delay=0):
+    await asyncio.sleep(delay)
+    async with aiohttp.ClientSession() as session:
+        async with session.post(url, json=data) as resp:
+            output = await resp.json()
+    return output
+
+
+async def test_concurrent(args):
+    url = f"{args.host}:{args.port}{ROUTE}"
+
+    # Single image test
+    if args.single_image:
+        prompt = f"<s>[INST]Describe this image in detail.{IMAGE_TOKEN_SEP}[/INST]"
+        image_url = "https://picsum.photos/id/237/400/300"
+        modality = ["image"]
+    # Multiple images test
+    else:
+        image_urls = [
+            "https://picsum.photos/id/237/400/300",
+            "https://picsum.photos/id/27/500/500",
+        ]
+        prompt = f"<s>[INST]How many photos are there? Describe each in a very short sentence.{IMAGE_TOKEN_SEP * len(image_urls)}[/INST]"
+        image_url = image_urls
+        modality = ["multi-images"]
+
+    response = await send_request(
+        url,
+        {
+            "text": prompt,
+            "image_data": image_url,
+            "sampling_params": {
+                "max_new_tokens": 100,
+                "temperature": 0.7,
+                "top_p": 0.9,
+            },
+            "modalities": modality,
+        },
+    )
+
+    print(f"Response: {response}")
+    if "text" in response:
+        print("\nOutput text:", response["text"])
+
+
+def test_streaming(args):
+    url = f"{args.host}:{args.port}/generate"
+
+    # Single image test
+    if args.single_image:
+        prompt = f"<s>[INST]Describe this image in detail.{IMAGE_TOKEN_SEP}[/INST]"
+        image_data = "https://picsum.photos/id/237/400/300"
+        modality = ["image"]
+    # Multiple images test
+    else:
+        image_urls = [
+            "https://picsum.photos/id/237/400/300",
+            "https://picsum.photos/id/27/500/500",
+        ]
+        prompt = f"<s>[INST]How many photos are there? Describe each in a very short sentence.{IMAGE_TOKEN_SEP * len(image_urls)}[/INST]"
+        image_data = image_urls
+        modality = ["multi-images"]
+
+    pload = {
+        "text": prompt,
+        "image_data": image_data,
+        "sampling_params": {"max_new_tokens": 100, "temperature": 0.7, "top_p": 0.9},
+        "modalities": modality,
+        "stream": True,
+    }
+
+    response = requests.post(url, json=pload, stream=True)
+
+    print("Streaming response:")
+    prev = 0
+    for chunk in response.iter_lines(decode_unicode=False):
+        chunk = chunk.decode("utf-8")
+        if chunk and chunk.startswith("data:"):
+            if chunk == "data: [DONE]":
+                break
+            data = json.loads(chunk[5:].strip("\n"))
+            output = data["text"].strip()
+            print(output[prev:], end="", flush=True)
+            prev = len(output)
+    print("\n")
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--host", type=str, default="http://127.0.0.1")
+    parser.add_argument("--port", type=int, default=30000)
+    parser.add_argument(
+        "--single-image",
+        action="store_true",
+        help="Test with single image instead of multiple images",
+    )
+    parser.add_argument("--no-stream", action="store_true", help="Don't test streaming")
+    args = parser.parse_args()
+
+    asyncio.run(test_concurrent(args))
+    if not args.no_stream:
+        test_streaming(args)

examples/runtime/multimodal/qwen_llava_server.py

+"""
+Usage:
+# Installing latest llava-next: pip install git+https://github.com/LLaVA-VL/LLaVA-NeXT.git
+# Installing latest sglang.
+
+# Endpoint Service CLI:
+python -m sglang.launch_server --model-path lmms-lab/llava-next-72b --port=30000 --tp-size=8
+
+python3 qwen_llava_server.py
+
+Output:
+"Two children pose with a large teddy bear, one holding a smaller stuffed bear, in a room with an American flag and potted plants."
+"""
+
+import argparse
+import asyncio
+import copy
+import json
+
+import aiohttp
+import requests
+from llava.conversation import conv_qwen
+
+
+async def send_request(url, data, delay=0):
+    await asyncio.sleep(delay)
+    async with aiohttp.ClientSession() as session:
+        async with session.post(url, json=data) as resp:
+            output = await resp.json()
+    return output
+
+
+async def test_concurrent(args):
+    url = f"{args.host}:{args.port}"
+
+    prompt = "<image>\nPlease generate caption towards this image."
+    conv_template = copy.deepcopy(conv_qwen)
+    conv_template.append_message(role=conv_template.roles[0], message=prompt)
+    conv_template.append_message(role=conv_template.roles[1], message=None)
+    prompt_with_template = conv_template.get_prompt()
+    response = []
+    for i in range(1):
+        response.append(
+            send_request(
+                url + "/generate",
+                {
+                    "text": prompt_with_template,
+                    "image_data": "https://farm4.staticflickr.com/3175/2653711032_804ff86d81_z.jpg",
+                    "sampling_params": {
+                        "max_new_tokens": 1024,
+                        "temperature": 0,
+                        "top_p": 1.0,
+                        "presence_penalty": 2,
+                        "frequency_penalty": 2,
+                        "stop": "<|im_end|>",
+                    },
+                },
+            )
+        )
+
+    rets = await asyncio.gather(*response)
+    for ret in rets:
+        print(ret["text"])
+
+
+def test_streaming(args):
+    url = f"{args.host}:{args.port}"
+    prompt = "<image>\nPlease generate caption towards this image."
+    conv_template = copy.deepcopy(conv_qwen)
+    conv_template.append_message(role=conv_template.roles[0], message=prompt)
+    conv_template.append_message(role=conv_template.roles[1], message=None)
+    prompt_with_template = conv_template.get_prompt()
+    pload = {
+        "text": prompt_with_template,
+        "sampling_params": {
+            "max_new_tokens": 1024,
+            "temperature": 0,
+            "top_p": 1.0,
+            "presence_penalty": 2,
+            "frequency_penalty": 2,
+            "stop": "<|im_end|>",
+        },
+        "image_data": "https://farm4.staticflickr.com/3175/2653711032_804ff86d81_z.jpg",
+        "stream": True,
+    }
+    response = requests.post(
+        url + "/generate",
+        json=pload,
+        stream=True,
+    )
+
+    prev = 0
+    for chunk in response.iter_lines(decode_unicode=False):
+        chunk = chunk.decode("utf-8")
+        if chunk and chunk.startswith("data:"):
+            if chunk == "data: [DONE]":
+                break
+            data = json.loads(chunk[5:].strip("\n"))
+            output = data["text"].strip()
+            print(output[prev:], end="", flush=True)
+            prev = len(output)
+    print("")
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--host", type=str, default="http://127.0.0.1")
+    parser.add_argument("--port", type=int, default=30000)
+    args = parser.parse_args()
+    asyncio.run(test_concurrent(args))
+    test_streaming(args)

examples/runtime/multimodal_embedding.py

+# launch server
+# python -m sglang.launch_server --model-path Alibaba-NLP/gme-Qwen2-VL-2B-Instruct --is-embedding
+
+import requests
+
+url = "http://127.0.0.1:30000"
+
+text_input = "Represent this image in embedding space."
+image_path = "https://huggingface.co/datasets/liuhaotian/llava-bench-in-the-wild/resolve/main/images/023.jpg"
+
+payload = {
+    "model": "gme-qwen2-vl",
+    "input": [{"text": text_input}, {"image": image_path}],
+}
+
+response = requests.post(url + "/v1/embeddings", json=payload).json()
+
+print("Embeddings:", [x.get("embedding") for x in response.get("data", [])])

examples/runtime/openai_chat_with_response_prefill.py

+"""
+Usage:
+1) Launch the server in one terminal:
+   python -m sglang.launch_server --model-path meta-llama/Llama-3.1-8B-Instruct --port 30000
+
+2) Run this script in another terminal:
+   python openai_chat_with_response_prefill.py
+
+This example demonstrates two chat completion calls:
+- One with continue_final_message enabled (the final assistant message is used as a prefill).
+- One without continue_final_message (the final assistant message remains, starting a new turn).
+"""
+
+import openai
+
+client = openai.Client(base_url="http://127.0.0.1:30000/v1", api_key="EMPTY")
+
+messages = [
+    {"role": "system", "content": "You are a helpful AI assistant."},
+    {
+        "role": "user",
+        "content": """
+Extract the name, size, price, and color from this product description as a JSON object:
+
+<description>
+The SmartHome Mini is a compact smart home assistant available in black or white for only $49.99.
+At just 5 inches wide, it lets you control lights, thermostats, and other connected devices via voice or app—
+no matter where you place it in your home.
+This affordable little hub brings convenient hands-free control to your smart devices.
+</description>
+""",
+    },
+    {"role": "assistant", "content": "{\n"},
+]
+
+# Calling the API with continue_final_message enabled.
+print("=== Prefill with continue_final_messagem ===")
+response_with = client.chat.completions.create(
+    model="meta-llama/Llama-3.1-8B-Instruct",
+    messages=messages,
+    temperature=0,
+    extra_body={"continue_final_message": True},
+)
+print(response_with.choices[0].message.content)
+
+# Calling the API without continue_final_message (using default behavior).
+print("\n=== Prefill without continue_final_message ===")
+response_without = client.chat.completions.create(
+    model="meta-llama/Llama-3.1-8B-Instruct",
+    messages=messages,
+    temperature=0,
+)
+print(response_without.choices[0].message.content)

examples/runtime/reward_model.py

+# launch server
+# python -m sglang.launch_server --model LxzGordon/URM-LLaMa-3.1-8B --is-embedding
+
+import requests
+
+url = "http://127.0.0.1:30000"
+
+PROMPT = (
+    "What is the range of the numeric output of a sigmoid node in a neural network?"
+)
+RESPONSE1 = "The output of a sigmoid node is bounded between -1 and 1."
+RESPONSE2 = "The output of a sigmoid node is bounded between 0 and 1."
+
+json_data = {
+    "conv": [
+        [
+            {"role": "user", "content": PROMPT},
+            {"role": "assistant", "content": RESPONSE1},
+        ],
+        [
+            {"role": "user", "content": PROMPT},
+            {"role": "assistant", "content": RESPONSE2},
+        ],
+    ],
+}
+response = requests.post(
+    url + "/classify",
+    json=json_data,
+).json()
+
+print(response)
+print("scores:", [x["embedding"] for x in response])

examples/runtime/token_in_token_out/token_in_token_out_llm_engine.py

+"""
+This example demonstrates how to provide tokenized ids to LLM as input instead of text prompt, i.e. a token-in-token-out workflow.
+"""
+
+import sglang as sgl
+from sglang.srt.hf_transformers_utils import get_tokenizer
+
+MODEL_PATH = "meta-llama/Llama-3.1-8B-Instruct"
+
+
+def main():
+    # Sample prompts.
+    prompts = [
+        "Hello, my name is",
+        "The president of the United States is",
+        "The capital of France is",
+        "The future of AI is",
+    ]
+    # Create a sampling params object.
+    sampling_params = {"temperature": 0.8, "top_p": 0.95}
+
+    # Tokenize inputs
+    tokenizer = get_tokenizer(MODEL_PATH)
+    token_ids_list = [tokenizer.encode(prompt) for prompt in prompts]
+
+    # Create an LLM.
+    llm = sgl.Engine(model_path=MODEL_PATH, skip_tokenizer_init=True)
+
+    outputs = llm.generate(input_ids=token_ids_list, sampling_params=sampling_params)
+    # Print the outputs.
+    for prompt, output in zip(prompts, outputs):
+        decode_output = tokenizer.decode(output["output_ids"])
+        print("===============================")
+        print(
+            f"Prompt: {prompt}\nGenerated token ids: {output['output_ids']}\nGenerated text: {decode_output}"
+        )
+        print()
+
+
+# The __main__ condition is necessary here because we use "spawn" to create subprocesses
+# Spawn starts a fresh program every time, if there is no __main__, it will run into infinite loop to keep spawning processes from sgl.Engine
+if __name__ == "__main__":
+    main()

examples/runtime/token_in_token_out/token_in_token_out_llm_server.py

+"""
+Usage:
+
+python token_in_token_out_llm_server.py
+
+"""
+
+import requests
+
+from sglang.srt.hf_transformers_utils import get_tokenizer
+from sglang.test.test_utils import is_in_ci
+from sglang.utils import terminate_process, wait_for_server
+
+if is_in_ci():
+    from docs.backend.patch import launch_server_cmd
+else:
+    from sglang.utils import launch_server_cmd
+
+
+MODEL_PATH = "meta-llama/Llama-3.1-8B-Instruct"
+
+
+def main():
+    # Launch the server
+    server_process, port = launch_server_cmd(
+        f"python -m sglang.launch_server --model-path {MODEL_PATH} --skip-tokenizer-init --host 0.0.0.0"
+    )
+    wait_for_server(f"http://localhost:{port}")
+
+    # Sample prompts.
+    prompts = [
+        "Hello, my name is",
+        "The president of the United States is",
+        "The capital of France is",
+        "The future of AI is",
+    ]
+
+    # Create a sampling params object.
+    sampling_params = {"temperature": 0.8, "top_p": 0.95}
+
+    # Tokenize inputs
+    tokenizer = get_tokenizer(MODEL_PATH)
+    token_ids_list = [tokenizer.encode(prompt) for prompt in prompts]
+
+    json_data = {
+        "input_ids": token_ids_list,
+        "sampling_params": sampling_params,
+    }
+
+    response = requests.post(
+        f"http://localhost:{port}/generate",
+        json=json_data,
+    )
+
+    outputs = response.json()
+    for prompt, output in zip(prompts, outputs):
+        print("===============================")
+        decode_output = tokenizer.decode(output["output_ids"])
+        print(
+            f"Prompt: {prompt}\nGenerated token ids: {output['output_ids']}\nGenerated text: {decode_output}"
+        )
+        print()
+
+    terminate_process(server_process)
+
+
+if __name__ == "__main__":
+    main()

examples/runtime/token_in_token_out/token_in_token_out_vlm_engine.py

+import argparse
+import dataclasses
+from typing import Tuple
+
+from transformers import AutoProcessor
+
+from sglang import Engine
+from sglang.lang.chat_template import get_chat_template_by_model_path
+from sglang.srt.configs.model_config import ModelConfig
+from sglang.srt.server_args import ServerArgs
+from sglang.test.test_utils import DEFAULT_IMAGE_URL
+
+
+def get_input_ids(
+    server_args: ServerArgs, model_config: ModelConfig
+) -> Tuple[list[int], list]:
+    chat_template = get_chat_template_by_model_path(model_config.model_path)
+    text = f"{chat_template.image_token}What is in this picture?"
+    image_data = [DEFAULT_IMAGE_URL]
+
+    processor = AutoProcessor.from_pretrained(
+        model_config.model_path, trust_remote_code=server_args.trust_remote_code
+    )
+
+    input_ids = (
+        processor.tokenizer(
+            text=[text],
+            return_tensors="pt",
+        )
+        .input_ids[0]
+        .tolist()
+    )
+
+    return input_ids, image_data
+
+
+def token_in_out_example(
+    server_args: ServerArgs,
+):
+    input_ids, image_data = get_input_ids(
+        server_args,
+        ModelConfig(
+            server_args.model_path,
+            trust_remote_code=server_args.trust_remote_code,
+            model_override_args=server_args.json_model_override_args,
+        ),
+    )
+    backend = Engine(**dataclasses.asdict(server_args))
+
+    output = backend.generate(
+        input_ids=input_ids,
+        image_data=image_data,
+        sampling_params={
+            "temperature": 0.8,
+            "max_new_tokens": 32,
+        },
+    )
+
+    print("===============================")
+    print(f"Output token ids: ", output["output_ids"])
+
+    backend.shutdown()
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    ServerArgs.add_cli_args(parser)
+    args = [
+        "--model-path=Qwen/Qwen2-VL-2B",
+    ]
+    args = parser.parse_args(args=args)
+    server_args = ServerArgs.from_cli_args(args)
+    server_args.skip_tokenizer_init = True
+    token_in_out_example(server_args)

examples/runtime/token_in_token_out/token_in_token_out_vlm_server.py

+"""
+Usage:
+
+python token_in_token_out_vlm_server.py
+
+"""
+
+from typing import Tuple
+
+import requests
+from transformers import AutoProcessor
+
+from sglang.lang.chat_template import get_chat_template_by_model_path
+from sglang.test.test_utils import DEFAULT_IMAGE_URL, is_in_ci
+from sglang.utils import terminate_process, wait_for_server
+
+if is_in_ci():
+    from docs.backend.patch import launch_server_cmd
+else:
+    from sglang.utils import launch_server_cmd
+
+
+MODEL_PATH = "Qwen/Qwen2-VL-2B"
+
+
+def get_input_ids() -> Tuple[list[int], list]:
+    chat_template = get_chat_template_by_model_path(MODEL_PATH)
+    text = f"{chat_template.image_token}What is in this picture?"
+    image_data = [DEFAULT_IMAGE_URL]
+
+    processor = AutoProcessor.from_pretrained(MODEL_PATH)
+
+    input_ids = (
+        processor.tokenizer(
+            text=[text],
+            return_tensors="pt",
+        )
+        .input_ids[0]
+        .tolist()
+    )
+
+    return input_ids, image_data
+
+
+def main():
+    # Launch the server
+    server_process, port = launch_server_cmd(
+        f"python -m sglang.launch_server --model-path {MODEL_PATH} --skip-tokenizer-init --host 0.0.0.0"
+    )
+    wait_for_server(f"http://localhost:{port}")
+
+    input_ids, image_data = get_input_ids()
+
+    sampling_params = {
+        "temperature": 0.8,
+        "max_new_tokens": 32,
+    }
+
+    json_data = {
+        "input_ids": input_ids,
+        "image_data": image_data,
+        "sampling_params": sampling_params,
+    }
+
+    response = requests.post(
+        f"http://localhost:{port}/generate",
+        json=json_data,
+    )
+
+    output = response.json()
+    print("===============================")
+    print(f"Output token ids: ", output["output_ids"])
+
+    terminate_process(server_process)
+
+
+if __name__ == "__main__":
+    main()

examples/runtime/vertex_predict.py

+"""
+Usage:
+python -m sglang.launch_server --model meta-llama/Llama-2-7b-hf --port 30000
+python vertex_predict.py
+
+This example shows the request and response formats of the prediction route for
+Google Cloud Vertex AI Online Predictions.
+
+Vertex AI SDK for Python is recommended for deploying models to Vertex AI
+instead of a local server. After deploying the model to a Vertex AI Online
+Prediction Endpoint, send requests via the Python SDK:
+
+response = endpoint.predict(
+    instances=[
+        {"text": "The capital of France is"},
+        {"text": "What is a car?"},
+    ],
+    parameters={"sampling_params": {"max_new_tokens": 16}},
+)
+print(response.predictions)
+
+More details about get online predictions from Vertex AI can be found at
+https://cloud.google.com/vertex-ai/docs/predictions/get-online-predictions.
+"""
+
+from dataclasses import dataclass
+from typing import List, Optional
+
+import requests
+
+
+@dataclass
+class VertexPrediction:
+    predictions: List
+
+
+class LocalVertexEndpoint:
+    def __init__(self) -> None:
+        self.base_url = "http://127.0.0.1:30000"
+
+    def predict(self, instances: List[dict], parameters: Optional[dict] = None):
+        response = requests.post(
+            self.base_url + "/vertex_generate",
+            json={
+                "instances": instances,
+                "parameters": parameters,
+            },
+        )
+        return VertexPrediction(predictions=response.json()["predictions"])
+
+
+endpoint = LocalVertexEndpoint()
+
+# Predict with a single prompt.
+response = endpoint.predict(instances=[{"text": "The capital of France is"}])
+print(response.predictions)
+
+# Predict with multiple prompts and parameters.
+response = endpoint.predict(
+    instances=[
+        {"text": "The capital of France is"},
+        {"text": "What is a car?"},
+    ],
+    parameters={"sampling_params": {"max_new_tokens": 16}},
+)
+print(response.predictions)

package-lock.json

+{
+  "name": "sglang",
+  "lockfileVersion": 3,
+  "requires": true,
+  "packages": {}
+}

python/pyproject.toml

+[build-system]
+requires = ["setuptools>=61.0", "wheel"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "sglang"
+version = "0.5.2rc1"
+description = "SGLang is yet another fast serving framework for large language models and vision language models."
+readme = "README.md"
+requires-python = ">=3.10"
+license = { file = "LICENSE" }
+classifiers = [
+    "Programming Language :: Python :: 3",
+    "License :: OSI Approved :: Apache Software License",
+]
+dependencies = ["aiohttp", "requests", "tqdm", "numpy", "IPython", "setproctitle"]
+
+[project.optional-dependencies]
+runtime_common = [
+    "blobfile==3.0.0",
+    "build",
+    "compressed-tensors",
+    "datasets",
+    "einops",
+    "fastapi",
+    "hf_transfer",
+    "huggingface_hub",
+    "interegular",
+    "llguidance>=0.7.11,<0.8.0",
+    "modelscope",
+    "msgspec",
+    "ninja",
+    "openai==1.99.1",
+    "openai-harmony==0.0.4",
+    "orjson",
+    "outlines==0.1.11",
+    "packaging",
+    "partial_json_parser",
+    "pillow",
+    "prometheus-client>=0.20.0",
+    "psutil",
+    "pybase64",
+    "pydantic",
+    "pynvml",
+    "python-multipart",
+    "pyzmq>=25.1.2",
+    "sentencepiece",
+    "soundfile==0.13.1",
+    "scipy",
+    "timm==1.0.16",
+    "tiktoken",
+    "torchao==0.9.0",
+    "transformers==4.56.0",
+    "uvicorn",
+    "uvloop",
+    "xgrammar==0.1.23",
+]
+
+srt = [
+    "sglang[runtime_common]",
+    "sgl-kernel==0.3.8",
+    "torch==2.8.0",
+    "torchaudio==2.8.0",
+    "torchvision",
+    "cuda-python",
+    "flashinfer_python==0.3.0",
+]
+
+blackwell = [
+    "sglang[runtime_common]",
+    "sgl-kernel",
+    "torch==2.8.0",
+    "torchaudio==2.8.0",
+    "torchvision",
+    "cuda-python",
+    "flashinfer_python==0.3.0",
+]
+
+# HIP (Heterogeneous-computing Interface for Portability) for AMD
+# => base docker rocm/vllm-dev:20250114, not from public vllm whl
+srt_hip = [
+    "sglang[runtime_common]",
+    "torch",
+    "petit_kernel==0.0.2",
+    "wave-lang==1.0.1",
+]
+
+# https://docs.sglang.ai/platforms/cpu_server.html
+srt_cpu = ["sglang[runtime_common]"]
+
+# https://docs.sglang.ai/platforms/ascend_npu.html
+srt_npu = ["sglang[runtime_common]"]
+
+# xpu is not enabled in public vllm and torch whl,
+# need to follow https://docs.vllm.ai/en/latest/getting_started/xpu-installation.htmlinstall vllm
+srt_xpu = ["sglang[runtime_common]"]
+
+# For Intel Gaudi(device : hpu) follow the installation guide
+# https://docs.vllm.ai/en/latest/getting_started/gaudi-installation.html
+srt_hpu = ["sglang[runtime_common]"]
+
+openai = ["openai==1.99.1", "tiktoken"]
+anthropic = ["anthropic>=0.20.0"]
+litellm = ["litellm>=1.0.0"]
+torch_memory_saver = ["torch_memory_saver==0.0.8"]
+decord = ["decord"]
+test = [
+    "accelerate",
+    "expecttest",
+    "jsonlines",
+    "matplotlib",
+    "pandas",
+    "peft",
+    "sentence_transformers",
+    "pytest",
+    "tabulate",
+]
+all = ["sglang[srt]", "sglang[openai]", "sglang[anthropic]", "sglang[torch_memory_saver]", "sglang[decord]"]
+all_hip = ["sglang[srt_hip]", "sglang[openai]", "sglang[anthropic]", "sglang[decord]"]
+all_xpu = ["sglang[srt_xpu]", "sglang[openai]", "sglang[anthropic]", "sglang[decord]"]
+all_hpu = ["sglang[srt_hpu]", "sglang[openai]", "sglang[anthropic]", "sglang[decord]"]
+all_cpu = ["sglang[srt_cpu]", "sglang[openai]", "sglang[anthropic]", "sglang[decord]"]
+all_npu = ["sglang[srt_npu]", "sglang[openai]", "sglang[anthropic]", "sglang[decord]"]
+
+dev = ["sglang[all]", "sglang[test]"]
+dev_hip = ["sglang[all_hip]", "sglang[test]"]
+dev_xpu = ["sglang[all_xpu]", "sglang[test]"]
+dev_hpu = ["sglang[all_hpu]", "sglang[test]"]
+dev_cpu = ["sglang[all_cpu]", "sglang[test]"]
+
+[project.urls]
+"Homepage" = "https://github.com/sgl-project/sglang"
+"Bug Tracker" = "https://github.com/sgl-project/sglang/issues"
+
+[tool.setuptools.package-data]
+"sglang" = [
+    "srt/layers/moe/fused_moe_triton/configs/*/*.json",
+    "srt/layers/quantization/configs/*.json",
+    "srt/mem_cache/storage/hf3fs/hf3fs_utils.cpp",
+]
+
+[tool.setuptools.packages.find]
+exclude = [
+    "assets*",
+    "benchmark*",
+    "docs*",
+    "dist*",
+    "playground*",
+    "scripts*",
+    "tests*",
+]
+
+[tool.wheel]
+exclude = [
+    "assets*",
+    "benchmark*",
+    "docs*",
+    "dist*",
+    "playground*",
+    "scripts*",
+    "tests*",
+]
+
+[tool.codespell]
+ignore-words-list = "ans, als, hel, boostrap, childs, te, vas, hsa, ment"
+skip = "*.json,*.jsonl,*.patch,*.txt"

python/sglang.egg-info/PKG-INFO

+Metadata-Version: 2.4
+Name: sglang
+Version: 0.5.2rc1
+Summary: SGLang is yet another fast serving framework for large language models and vision language models.
+Project-URL: Homepage, https://github.com/sgl-project/sglang
+Project-URL: Bug Tracker, https://github.com/sgl-project/sglang/issues
+Classifier: Programming Language :: Python :: 3
+Classifier: License :: OSI Approved :: Apache Software License
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+Requires-Dist: aiohttp
+Requires-Dist: requests
+Requires-Dist: tqdm
+Requires-Dist: numpy
+Requires-Dist: IPython
+Requires-Dist: setproctitle
+Provides-Extra: runtime-common
+Requires-Dist: blobfile==3.0.0; extra == "runtime-common"
+Requires-Dist: build; extra == "runtime-common"
+Requires-Dist: compressed-tensors; extra == "runtime-common"
+Requires-Dist: datasets; extra == "runtime-common"
+Requires-Dist: einops; extra == "runtime-common"
+Requires-Dist: fastapi; extra == "runtime-common"
+Requires-Dist: hf_transfer; extra == "runtime-common"
+Requires-Dist: huggingface_hub; extra == "runtime-common"
+Requires-Dist: interegular; extra == "runtime-common"
+Requires-Dist: llguidance<0.8.0,>=0.7.11; extra == "runtime-common"
+Requires-Dist: modelscope; extra == "runtime-common"
+Requires-Dist: msgspec; extra == "runtime-common"
+Requires-Dist: ninja; extra == "runtime-common"
+Requires-Dist: openai==1.99.1; extra == "runtime-common"
+Requires-Dist: openai-harmony==0.0.4; extra == "runtime-common"
+Requires-Dist: orjson; extra == "runtime-common"
+Requires-Dist: outlines==0.1.11; extra == "runtime-common"
+Requires-Dist: packaging; extra == "runtime-common"
+Requires-Dist: partial_json_parser; extra == "runtime-common"
+Requires-Dist: pillow; extra == "runtime-common"
+Requires-Dist: prometheus-client>=0.20.0; extra == "runtime-common"
+Requires-Dist: psutil; extra == "runtime-common"
+Requires-Dist: pybase64; extra == "runtime-common"
+Requires-Dist: pydantic; extra == "runtime-common"
+Requires-Dist: pynvml; extra == "runtime-common"
+Requires-Dist: python-multipart; extra == "runtime-common"
+Requires-Dist: pyzmq>=25.1.2; extra == "runtime-common"
+Requires-Dist: sentencepiece; extra == "runtime-common"
+Requires-Dist: soundfile==0.13.1; extra == "runtime-common"
+Requires-Dist: scipy; extra == "runtime-common"
+Requires-Dist: timm==1.0.16; extra == "runtime-common"
+Requires-Dist: tiktoken; extra == "runtime-common"
+Requires-Dist: torchao==0.9.0; extra == "runtime-common"
+Requires-Dist: transformers==4.56.0; extra == "runtime-common"
+Requires-Dist: uvicorn; extra == "runtime-common"
+Requires-Dist: uvloop; extra == "runtime-common"
+Requires-Dist: xgrammar==0.1.23; extra == "runtime-common"
+Provides-Extra: srt
+Requires-Dist: sglang[runtime_common]; extra == "srt"
+Requires-Dist: sgl-kernel==0.3.8; extra == "srt"
+Requires-Dist: torch==2.8.0; extra == "srt"
+Requires-Dist: torchaudio==2.8.0; extra == "srt"
+Requires-Dist: torchvision; extra == "srt"
+Requires-Dist: cuda-python; extra == "srt"
+Requires-Dist: flashinfer_python==0.3.0; extra == "srt"
+Provides-Extra: blackwell
+Requires-Dist: sglang[runtime_common]; extra == "blackwell"
+Requires-Dist: sgl-kernel; extra == "blackwell"
+Requires-Dist: torch==2.8.0; extra == "blackwell"
+Requires-Dist: torchaudio==2.8.0; extra == "blackwell"
+Requires-Dist: torchvision; extra == "blackwell"
+Requires-Dist: cuda-python; extra == "blackwell"
+Requires-Dist: flashinfer_python==0.3.0; extra == "blackwell"
+Provides-Extra: srt-hip
+Requires-Dist: sglang[runtime_common]; extra == "srt-hip"
+Requires-Dist: torch; extra == "srt-hip"
+Requires-Dist: petit_kernel==0.0.2; extra == "srt-hip"
+Requires-Dist: wave-lang==1.0.1; extra == "srt-hip"
+Provides-Extra: srt-cpu
+Requires-Dist: sglang[runtime_common]; extra == "srt-cpu"
+Provides-Extra: srt-npu
+Requires-Dist: sglang[runtime_common]; extra == "srt-npu"
+Provides-Extra: srt-xpu
+Requires-Dist: sglang[runtime_common]; extra == "srt-xpu"
+Provides-Extra: srt-hpu
+Requires-Dist: sglang[runtime_common]; extra == "srt-hpu"
+Provides-Extra: openai
+Requires-Dist: openai==1.99.1; extra == "openai"
+Requires-Dist: tiktoken; extra == "openai"
+Provides-Extra: anthropic
+Requires-Dist: anthropic>=0.20.0; extra == "anthropic"
+Provides-Extra: litellm
+Requires-Dist: litellm>=1.0.0; extra == "litellm"
+Provides-Extra: torch-memory-saver
+Requires-Dist: torch_memory_saver==0.0.8; extra == "torch-memory-saver"
+Provides-Extra: decord
+Requires-Dist: decord; extra == "decord"
+Provides-Extra: test
+Requires-Dist: accelerate; extra == "test"
+Requires-Dist: expecttest; extra == "test"
+Requires-Dist: jsonlines; extra == "test"
+Requires-Dist: matplotlib; extra == "test"
+Requires-Dist: pandas; extra == "test"
+Requires-Dist: peft; extra == "test"
+Requires-Dist: sentence_transformers; extra == "test"
+Requires-Dist: pytest; extra == "test"
+Requires-Dist: tabulate; extra == "test"
+Provides-Extra: all
+Requires-Dist: sglang[srt]; extra == "all"
+Requires-Dist: sglang[openai]; extra == "all"
+Requires-Dist: sglang[anthropic]; extra == "all"
+Requires-Dist: sglang[torch_memory_saver]; extra == "all"
+Requires-Dist: sglang[decord]; extra == "all"
+Provides-Extra: all-hip
+Requires-Dist: sglang[srt_hip]; extra == "all-hip"
+Requires-Dist: sglang[openai]; extra == "all-hip"
+Requires-Dist: sglang[anthropic]; extra == "all-hip"
+Requires-Dist: sglang[decord]; extra == "all-hip"
+Provides-Extra: all-xpu
+Requires-Dist: sglang[srt_xpu]; extra == "all-xpu"
+Requires-Dist: sglang[openai]; extra == "all-xpu"
+Requires-Dist: sglang[anthropic]; extra == "all-xpu"
+Requires-Dist: sglang[decord]; extra == "all-xpu"
+Provides-Extra: all-hpu
+Requires-Dist: sglang[srt_hpu]; extra == "all-hpu"
+Requires-Dist: sglang[openai]; extra == "all-hpu"
+Requires-Dist: sglang[anthropic]; extra == "all-hpu"
+Requires-Dist: sglang[decord]; extra == "all-hpu"
+Provides-Extra: all-cpu
+Requires-Dist: sglang[srt_cpu]; extra == "all-cpu"
+Requires-Dist: sglang[openai]; extra == "all-cpu"
+Requires-Dist: sglang[anthropic]; extra == "all-cpu"
+Requires-Dist: sglang[decord]; extra == "all-cpu"
+Provides-Extra: all-npu
+Requires-Dist: sglang[srt_npu]; extra == "all-npu"
+Requires-Dist: sglang[openai]; extra == "all-npu"
+Requires-Dist: sglang[anthropic]; extra == "all-npu"
+Requires-Dist: sglang[decord]; extra == "all-npu"
+Provides-Extra: dev
+Requires-Dist: sglang[all]; extra == "dev"
+Requires-Dist: sglang[test]; extra == "dev"
+Provides-Extra: dev-hip
+Requires-Dist: sglang[all_hip]; extra == "dev-hip"
+Requires-Dist: sglang[test]; extra == "dev-hip"
+Provides-Extra: dev-xpu
+Requires-Dist: sglang[all_xpu]; extra == "dev-xpu"
+Requires-Dist: sglang[test]; extra == "dev-xpu"
+Provides-Extra: dev-hpu
+Requires-Dist: sglang[all_hpu]; extra == "dev-hpu"
+Requires-Dist: sglang[test]; extra == "dev-hpu"
+Provides-Extra: dev-cpu
+Requires-Dist: sglang[all_cpu]; extra == "dev-cpu"
+Requires-Dist: sglang[test]; extra == "dev-cpu"

python/sglang.egg-info/SOURCES.txt

+pyproject.toml
+sglang/__init__.py
+sglang/bench_offline_throughput.py
+sglang/bench_one_batch.py
+sglang/bench_one_batch_server.py
+sglang/bench_serving.py
+sglang/check_env.py
+sglang/compile_deep_gemm.py
+sglang/global_config.py
+sglang/launch_server.py
+sglang/profiler.py
+sglang/utils.py
+sglang/version.py
+sglang.egg-info/PKG-INFO
+sglang.egg-info/SOURCES.txt
+sglang.egg-info/dependency_links.txt
+sglang.egg-info/requires.txt
+sglang.egg-info/top_level.txt
+sglang/eval/llama3_eval.py
+sglang/eval/loogle_eval.py
+sglang/lang/api.py
+sglang/lang/chat_template.py
+sglang/lang/choices.py
+sglang/lang/compiler.py
+sglang/lang/interpreter.py
+sglang/lang/ir.py
+sglang/lang/tracer.py
+sglang/lang/backend/anthropic.py
+sglang/lang/backend/base_backend.py
+sglang/lang/backend/litellm.py
+sglang/lang/backend/openai.py
+sglang/lang/backend/runtime_endpoint.py
+sglang/lang/backend/vertexai.py
+sglang/srt/_custom_ops.py
+sglang/srt/aio_rwlock.py
+sglang/srt/bench_utils.py
+sglang/srt/constants.py
+sglang/srt/custom_op.py
+sglang/srt/hf_transformers_utils.py
+sglang/srt/host_shared_memory.py
+sglang/srt/offloader.py
+sglang/srt/operations.py
+sglang/srt/operations_strategy.py
+sglang/srt/patch_torch.py
+sglang/srt/poll_based_barrier.py
+sglang/srt/server_args.py
+sglang/srt/torch_memory_saver_adapter.py
+sglang/srt/two_batch_overlap.py
+sglang/srt/utils.py
+sglang/srt/warmup.py
+sglang/srt/configs/__init__.py
+sglang/srt/configs/chatglm.py
+sglang/srt/configs/dbrx.py
+sglang/srt/configs/deepseekvl2.py
+sglang/srt/configs/device_config.py
+sglang/srt/configs/exaone.py
+sglang/srt/configs/internvl.py
+sglang/srt/configs/janus_pro.py
+sglang/srt/configs/kimi_vl.py
+sglang/srt/configs/kimi_vl_moonvit.py
+sglang/srt/configs/load_config.py
+sglang/srt/configs/longcat_flash.py
+sglang/srt/configs/model_config.py
+sglang/srt/configs/step3_vl.py
+sglang/srt/configs/update_config.py
+sglang/srt/configs/utils.py
+sglang/srt/connector/__init__.py
+sglang/srt/connector/base_connector.py
+sglang/srt/connector/redis.py
+sglang/srt/connector/s3.py
+sglang/srt/connector/utils.py
+sglang/srt/connector/serde/__init__.py
+sglang/srt/connector/serde/safe_serde.py
+sglang/srt/connector/serde/serde.py
+sglang/srt/constrained/base_grammar_backend.py
+sglang/srt/constrained/llguidance_backend.py
+sglang/srt/constrained/outlines_backend.py
+sglang/srt/constrained/outlines_jump_forward.py
+sglang/srt/constrained/reasoner_grammar_backend.py
+sglang/srt/constrained/xgrammar_backend.py
+sglang/srt/constrained/triton_ops/bitmask_ops.py
+sglang/srt/debug_utils/__init__.py
+sglang/srt/debug_utils/dump_comparator.py
+sglang/srt/debug_utils/dumper.py
+sglang/srt/debug_utils/text_comparator.py
+sglang/srt/disaggregation/decode.py
+sglang/srt/disaggregation/decode_schedule_batch_mixin.py
+sglang/srt/disaggregation/kv_events.py
+sglang/srt/disaggregation/launch_lb.py
+sglang/srt/disaggregation/mini_lb.py
+sglang/srt/disaggregation/prefill.py
+sglang/srt/disaggregation/utils.py
+sglang/srt/disaggregation/ascend/__init__.py
+sglang/srt/disaggregation/ascend/conn.py
+sglang/srt/disaggregation/ascend/transfer_engine.py
+sglang/srt/disaggregation/base/__init__.py
+sglang/srt/disaggregation/base/conn.py
+sglang/srt/disaggregation/common/__init__.py
+sglang/srt/disaggregation/common/conn.py
+sglang/srt/disaggregation/common/utils.py
+sglang/srt/disaggregation/fake/__init__.py
+sglang/srt/disaggregation/fake/conn.py
+sglang/srt/disaggregation/mooncake/__init__.py
+sglang/srt/disaggregation/mooncake/conn.py
+sglang/srt/disaggregation/mooncake/transfer_engine.py
+sglang/srt/disaggregation/nixl/__init__.py
+sglang/srt/disaggregation/nixl/conn.py
+sglang/srt/distributed/__init__.py
+sglang/srt/distributed/communication_op.py
+sglang/srt/distributed/naive_distributed.py
+sglang/srt/distributed/parallel_state.py
+sglang/srt/distributed/utils.py
+sglang/srt/distributed/device_communicators/cuda_wrapper.py
+sglang/srt/distributed/device_communicators/custom_all_reduce.py
+sglang/srt/distributed/device_communicators/custom_all_reduce_utils.py
+sglang/srt/distributed/device_communicators/hpu_communicator.py
+sglang/srt/distributed/device_communicators/npu_communicator.py
+sglang/srt/distributed/device_communicators/pymscclpp.py
+sglang/srt/distributed/device_communicators/pynccl.py
+sglang/srt/distributed/device_communicators/pynccl_allocator.py
+sglang/srt/distributed/device_communicators/pynccl_wrapper.py
+sglang/srt/distributed/device_communicators/quick_all_reduce.py
+sglang/srt/distributed/device_communicators/shm_broadcast.py
+sglang/srt/distributed/device_communicators/xpu_communicator.py
+sglang/srt/entrypoints/EngineBase.py
+sglang/srt/entrypoints/context.py
+sglang/srt/entrypoints/engine.py
+sglang/srt/entrypoints/harmony_utils.py
+sglang/srt/entrypoints/http_server.py
+sglang/srt/entrypoints/http_server_engine.py
+sglang/srt/entrypoints/tool.py
+sglang/srt/entrypoints/openai/__init__.py
+sglang/srt/entrypoints/openai/protocol.py
+sglang/srt/entrypoints/openai/serving_base.py
+sglang/srt/entrypoints/openai/serving_chat.py
+sglang/srt/entrypoints/openai/serving_completions.py
+sglang/srt/entrypoints/openai/serving_embedding.py
+sglang/srt/entrypoints/openai/serving_rerank.py
+sglang/srt/entrypoints/openai/serving_responses.py
+sglang/srt/entrypoints/openai/serving_score.py
+sglang/srt/entrypoints/openai/tool_server.py
+sglang/srt/entrypoints/openai/usage_processor.py
+sglang/srt/entrypoints/openai/utils.py
+sglang/srt/eplb/__init__.py
+sglang/srt/eplb/eplb_manager.py
+sglang/srt/eplb/expert_distribution.py
+sglang/srt/eplb/expert_location.py
+sglang/srt/eplb/expert_location_dispatch.py
+sglang/srt/eplb/expert_location_updater.py
+sglang/srt/eplb/eplb_algorithms/__init__.py
+sglang/srt/eplb/eplb_algorithms/deepseek.py
+sglang/srt/eplb/eplb_algorithms/deepseek_vec.py
+sglang/srt/eplb/eplb_simulator/__init__.py
+sglang/srt/eplb/eplb_simulator/reader.py
+sglang/srt/function_call/base_format_detector.py
+sglang/srt/function_call/core_types.py
+sglang/srt/function_call/deepseekv31_detector.py
+sglang/srt/function_call/deepseekv3_detector.py
+sglang/srt/function_call/ebnf_composer.py
+sglang/srt/function_call/function_call_parser.py
+sglang/srt/function_call/glm4_moe_detector.py
+sglang/srt/function_call/gpt_oss_detector.py
+sglang/srt/function_call/kimik2_detector.py
+sglang/srt/function_call/llama32_detector.py
+sglang/srt/function_call/mistral_detector.py
+sglang/srt/function_call/pythonic_detector.py
+sglang/srt/function_call/qwen25_detector.py
+sglang/srt/function_call/qwen3_coder_detector.py
+sglang/srt/function_call/step3_detector.py
+sglang/srt/function_call/utils.py
+sglang/srt/layers/activation.py
+sglang/srt/layers/amx_utils.py
+sglang/srt/layers/communicator.py
+sglang/srt/layers/dp_attention.py
+sglang/srt/layers/elementwise.py
+sglang/srt/layers/flashinfer_comm_fusion.py
+sglang/srt/layers/layernorm.py
+sglang/srt/layers/linear.py
+sglang/srt/layers/logits_processor.py
+sglang/srt/layers/model_parallel.py
+sglang/srt/layers/multimodal.py
+sglang/srt/layers/parameter.py
+sglang/srt/layers/pooler.py
+sglang/srt/layers/radix_attention.py
+sglang/srt/layers/rotary_embedding.py
+sglang/srt/layers/sampler.py
+sglang/srt/layers/torchao_utils.py
+sglang/srt/layers/utils.py
+sglang/srt/layers/vocab_parallel_embedding.py
+sglang/srt/layers/attention/aiter_backend.py
+sglang/srt/layers/attention/ascend_backend.py
+sglang/srt/layers/attention/base_attn_backend.py
+sglang/srt/layers/attention/cutlass_mla_backend.py
+sglang/srt/layers/attention/double_sparsity_backend.py
+sglang/srt/layers/attention/dual_chunk_flashattention_backend.py
+sglang/srt/layers/attention/flashattention_backend.py
+sglang/srt/layers/attention/flashinfer_backend.py
+sglang/srt/layers/attention/flashinfer_mla_backend.py
+sglang/srt/layers/attention/flashmla_backend.py
+sglang/srt/layers/attention/hybrid_attn_backend.py
+sglang/srt/layers/attention/intel_amx_backend.py
+sglang/srt/layers/attention/merge_state.py
+sglang/srt/layers/attention/tbo_backend.py
+sglang/srt/layers/attention/torch_native_backend.py
+sglang/srt/layers/attention/triton_backend.py
+sglang/srt/layers/attention/trtllm_mha_backend.py
+sglang/srt/layers/attention/trtllm_mla_backend.py
+sglang/srt/layers/attention/utils.py
+sglang/srt/layers/attention/vision.py
+sglang/srt/layers/attention/vision_utils.py
+sglang/srt/layers/attention/wave_backend.py
+sglang/srt/layers/attention/triton_ops/decode_attention.py
+sglang/srt/layers/attention/triton_ops/double_sparsity_attention.py
+sglang/srt/layers/attention/triton_ops/extend_attention.py
+sglang/srt/layers/attention/triton_ops/merge_state.py
+sglang/srt/layers/attention/triton_ops/prefill_attention.py
+sglang/srt/layers/attention/triton_ops/rocm_mla_decode_rope.py
+sglang/srt/layers/attention/wave_ops/decode_attention.py
+sglang/srt/layers/attention/wave_ops/extend_attention.py
+sglang/srt/layers/attention/wave_ops/prefill_attention.py
+sglang/srt/layers/moe/__init__.py
+sglang/srt/layers/moe/cutlass_moe.py
+sglang/srt/layers/moe/cutlass_moe_params.py
+sglang/srt/layers/moe/cutlass_w4a8_moe.py
+sglang/srt/layers/moe/fused_moe_native.py
+sglang/srt/layers/moe/rocm_moe_utils.py
+sglang/srt/layers/moe/router.py
+sglang/srt/layers/moe/topk.py
+sglang/srt/layers/moe/utils.py
+sglang/srt/layers/moe/ep_moe/__init__.py
+sglang/srt/layers/moe/ep_moe/kernels.py
+sglang/srt/layers/moe/ep_moe/layer.py
+sglang/srt/layers/moe/fused_moe_triton/__init__.py
+sglang/srt/layers/moe/fused_moe_triton/fused_moe.py
+sglang/srt/layers/moe/fused_moe_triton/fused_moe_triton_config.py
+sglang/srt/layers/moe/fused_moe_triton/fused_moe_triton_kernels.py
+sglang/srt/layers/moe/fused_moe_triton/layer.py
+sglang/srt/layers/moe/fused_moe_triton/moe_align_block_size.py
+sglang/srt/layers/moe/fused_moe_triton/triton_kernels_moe.py
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=1,N=14336,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=1,N=14336,device_name=NVIDIA_A100-SXM4-80GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=1,N=1792,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=1,N=1792,device_name=NVIDIA_A100-SXM4-80GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=1,N=3072,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=1,N=3072,device_name=NVIDIA_H100_80GB_HBM3,dtype=int8_w8a16.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=1,N=3072,device_name=NVIDIA_H100_80GB_HBM3.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=1,N=3584,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=1,N=3584,device_name=NVIDIA_A100-SXM4-80GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=1,N=7168,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=1,N=7168,device_name=NVIDIA_A100-SXM4-80GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=144,N=512,device_name=NVIDIA_H100_80GB_HBM3.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=16,N=1024,device_name=NVIDIA_H100_80GB_HBM3.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=16,N=1024,device_name=NVIDIA_H200.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=16,N=1344,device_name=NVIDIA_A100-SXM4-40GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=16,N=1344,device_name=NVIDIA_A100-SXM4-80GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=16,N=1344,device_name=NVIDIA_H100_80GB_HBM3.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=16,N=14336,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=16,N=14336,device_name=NVIDIA_A100-SXM4-80GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=16,N=1792,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=16,N=1792,device_name=NVIDIA_A100-SXM4-80GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=16,N=2048,device_name=NVIDIA_H100_80GB_HBM3.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=16,N=2688,device_name=NVIDIA_A100-SXM4-80GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=16,N=2688,device_name=NVIDIA_H100_80GB_HBM3.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=16,N=3072,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=16,N=3072,device_name=NVIDIA_H100_80GB_HBM3,dtype=int8_w8a16.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=16,N=3200,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=16,N=3584,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=16,N=3584,device_name=NVIDIA_A100-SXM4-80GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=16,N=6400,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=16,N=7168,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=16,N=7168,device_name=NVIDIA_A100-SXM4-80GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=16,N=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=int8_w8a16.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=16,N=800,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=160,N=192,device_name=NVIDIA_A800-SXM4-80GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=20,N=2048,device_name=NVIDIA_H100_80GB_HBM3.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=24,N=1024,device_name=NVIDIA_H100_80GB_HBM3.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=256,N=128,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=256,N=128,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=256,N=128,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=256,N=128,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=256,N=128,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=256,N=128,device_name=NVIDIA_H20,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=256,N=128,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=256,N=128,device_name=NVIDIA_L20Y,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=256,N=256,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=256,N=256,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=256,N=256,device_name=AMD_Radeon_Graphics,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=256,N=256,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=256,N=256,device_name=NVIDIA_H20,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=256,N=256,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=256,N=64,device_name=NVIDIA_A800-SXM4-80GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=256,N=64,device_name=NVIDIA_L20,dtype=int8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=256,N=64,device_name=NVIDIA_L40S,dtype=int8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=64,N=1024,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=64,N=1280,device_name=NVIDIA_A100-SXM4-80GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=64,N=1280,device_name=NVIDIA_A800-SXM4-80GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=64,N=1280,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=64,N=1280,device_name=NVIDIA_H100_80GB_HBM3.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=64,N=1280,device_name=NVIDIA_H200,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=64,N=1280,device_name=NVIDIA_H200.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=64,N=2560,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=64,N=2560,device_name=NVIDIA_H200,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=64,N=2560,device_name=NVIDIA_H200.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=64,N=320,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=64,N=320,device_name=NVIDIA_H100_80GB_HBM3.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=64,N=320,device_name=NVIDIA_H200,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=64,N=320,device_name=NVIDIA_H200.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=64,N=512,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=64,N=640,device_name=NVIDIA_A100-SXM4-80GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=64,N=640,device_name=NVIDIA_A800-SXM4-80GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=64,N=640,device_name=NVIDIA_GeForce_RTX_4090,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=64,N=640,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=64,N=640,device_name=NVIDIA_H100_80GB_HBM3.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=64,N=640,device_name=NVIDIA_H200,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=64,N=640,device_name=NVIDIA_H200.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=14336,device_name=AMD_Instinct_MI300X.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=14336,device_name=AMD_Instinct_MI325X.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=14336,device_name=AMD_Radeon_Graphics.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=14336,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=14336,device_name=NVIDIA_H200,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=14336,device_name=NVIDIA_H200.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=1792,device_name=AMD_Instinct_MI300X.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=1792,device_name=AMD_Instinct_MI325X.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=1792,device_name=AMD_Radeon_Graphics.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=1792,device_name=NVIDIA_A100-SXM4-40GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=1792,device_name=NVIDIA_A100-SXM4-80GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=1792,device_name=NVIDIA_H100_80GB_HBM3.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=1792,device_name=NVIDIA_H200,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=1792,device_name=NVIDIA_H200.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=2048,device_name=NVIDIA_A100-SXM4-80GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=2048,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=2048,device_name=NVIDIA_H100_80GB_HBM3.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=2048,device_name=NVIDIA_H200,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=2048,device_name=NVIDIA_H200.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=3584,device_name=AMD_Instinct_MI300X.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=3584,device_name=AMD_Instinct_MI325X.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=3584,device_name=AMD_Radeon_Graphics.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=3584,device_name=NVIDIA_A100-SXM4-40GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=3584,device_name=NVIDIA_A100-SXM4-80GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=3584,device_name=NVIDIA_GeForce_RTX_4090,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=3584,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=3584,device_name=NVIDIA_H100_80GB_HBM3.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=3584,device_name=NVIDIA_H200,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=3584,device_name=NVIDIA_H200.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=3584,device_name=NVIDIA_L40S.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=4096,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=4096,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=4096,device_name=AMD_Radeon_Graphics,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=4096,device_name=NVIDIA_A100-SXM4-80GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=4096,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=4096,device_name=NVIDIA_H100_80GB_HBM3.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=4096,device_name=NVIDIA_H200,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=4096,device_name=NVIDIA_H200.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=7168,device_name=AMD_Instinct_MI300X.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=7168,device_name=AMD_Instinct_MI325X.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=7168,device_name=AMD_Radeon_Graphics.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=7168,device_name=NVIDIA_A100-SXM4-80GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=7168,device_name=NVIDIA_H100_80GB_HBM3.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=7168,device_name=NVIDIA_H200.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=8192,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=8192,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=8192,device_name=AMD_Radeon_Graphics,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=8192,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_1_0/E=8,N=8192,device_name=NVIDIA_H200,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=128,N=192,device_name=NVIDIA_A800-SXM4-80GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=128,N=192,device_name=NVIDIA_H100_80GB_HBM3.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=128,N=192,device_name=NVIDIA_H20.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=128,N=192,device_name=NVIDIA_H200.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=128,N=384,device_name=NVIDIA_H100_80GB_HBM3.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=128,N=384,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=128,N=384,device_name=NVIDIA_H20.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=128,N=384,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=128,N=384,device_name=NVIDIA_H200.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=128,N=512,device_name=NVIDIA_H100_80GB_HBM3.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=128,N=768,device_name=NVIDIA_A800-SXM4-80GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=128,N=768,device_name=NVIDIA_H100_80GB_HBM3.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=128,N=768,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=128,N=768,device_name=NVIDIA_H20.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=128,N=768,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=128,N=768,device_name=NVIDIA_H200.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=128,N=96,device_name=NVIDIA_H20.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=129,N=352,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=160,N=320,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=161,N=192,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=257,N=128,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=257,N=128,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=257,N=256,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=257,N=256,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=257,N=256,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=264,N=128,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=264,N=256,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=264,N=256,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=264,N=256,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=272,N=128,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=272,N=128,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=272,N=128,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=272,N=64,device_name=NVIDIA_A800-SXM4-80GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=288,N=64,device_name=NVIDIA_A800-SXM4-80GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_2_0/E=8,N=7168,device_name=NVIDIA_H100_80GB_HBM3.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_3_0/E=16,N=1024,device_name=NVIDIA_B200.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_3_1/E=128,N=352,device_name=NVIDIA_RTX_6000_Ada_Generation,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_3_1/E=128,N=384,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_3_1/E=128,N=384,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_3_1/E=128,N=768,device_name=NVIDIA_H20.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_3_1/E=160,N=192,device_name=NVIDIA_H200,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_3_1/E=160,N=320,device_name=NVIDIA_H20-3e.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_3_1/E=160,N=384,device_name=NVIDIA_H200,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_3_1/E=160,N=640,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_3_1/E=257,N=128,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_3_1/E=257,N=128,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_3_1/E=257,N=256,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_3_1/E=257,N=256,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_3_1/E=257,N=256,device_name=NVIDIA_H20-3e,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_3_1/E=257,N=256,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_3_1/E=384,N=128,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_3_1/E=384,N=128,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_3_1/E=384,N=256,device_name=NVIDIA_H20-3e,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_3_1/E=385,N=128,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_3_1/E=385,N=128,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_3_1/E=8,N=7168,device_name=NVIDIA_H100_80GB_HBM3.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_4_0/E=128,N=384,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_4_0/E=128,N=768,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_4_0/E=129,N=352,device_name=NVIDIA_B200,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_4_0/E=129,N=352,device_name=NVIDIA_RTX_PRO_6000_Blackwell_Max-Q_Workstation_Edition,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_4_0/E=129,N=704,device_name=NVIDIA_B200,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_4_0/E=161,N=384,device_name=NVIDIA_RTX_PRO_6000_Blackwell_Max-Q_Workstation_Edition,dtype=fp8_w8a8.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_4_0/E=256,N=256,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_4_0/E=257,N=128,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_4_0/E=257,N=64,device_name=NVIDIA_A100-SXM4-80GB.json
+sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_4_0/E=384,N=256,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/moe/moe_runner/__init__.py
+sglang/srt/layers/moe/moe_runner/base.py
+sglang/srt/layers/moe/token_dispatcher/__init__.py
+sglang/srt/layers/moe/token_dispatcher/base_dispatcher.py
+sglang/srt/layers/moe/token_dispatcher/deepep.py
+sglang/srt/layers/moe/token_dispatcher/standard.py
+sglang/srt/layers/quantization/__init__.py
+sglang/srt/layers/quantization/awq.py
+sglang/srt/layers/quantization/awq_triton.py
+sglang/srt/layers/quantization/base_config.py
+sglang/srt/layers/quantization/blockwise_int8.py
+sglang/srt/layers/quantization/fp8.py
+sglang/srt/layers/quantization/fp8_kernel.py
+sglang/srt/layers/quantization/fp8_utils.py
+sglang/srt/layers/quantization/fpgemm_fp8.py
+sglang/srt/layers/quantization/gptq.py
+sglang/srt/layers/quantization/int8_kernel.py
+sglang/srt/layers/quantization/int8_utils.py
+sglang/srt/layers/quantization/kv_cache.py
+sglang/srt/layers/quantization/marlin_utils.py
+sglang/srt/layers/quantization/marlin_utils_fp8.py
+sglang/srt/layers/quantization/modelopt_quant.py
+sglang/srt/layers/quantization/moe_wna16.py
+sglang/srt/layers/quantization/mxfp4.py
+sglang/srt/layers/quantization/mxfp4_tensor.py
+sglang/srt/layers/quantization/petit.py
+sglang/srt/layers/quantization/petit_utils.py
+sglang/srt/layers/quantization/qoq.py
+sglang/srt/layers/quantization/unquant.py
+sglang/srt/layers/quantization/utils.py
+sglang/srt/layers/quantization/w4afp8.py
+sglang/srt/layers/quantization/w8a8_fp8.py
+sglang/srt/layers/quantization/w8a8_int8.py
+sglang/srt/layers/quantization/compressed_tensors/__init__.py
+sglang/srt/layers/quantization/compressed_tensors/compressed_tensors.py
+sglang/srt/layers/quantization/compressed_tensors/compressed_tensors_moe.py
+sglang/srt/layers/quantization/compressed_tensors/utils.py
+sglang/srt/layers/quantization/compressed_tensors/schemes/__init__.py
+sglang/srt/layers/quantization/compressed_tensors/schemes/compressed_tensors_scheme.py
+sglang/srt/layers/quantization/compressed_tensors/schemes/compressed_tensors_w8a16_fp8.py
+sglang/srt/layers/quantization/compressed_tensors/schemes/compressed_tensors_w8a8_fp8.py
+sglang/srt/layers/quantization/configs/N=1536,K=1536,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=1536,K=1536,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=1536,K=1536,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=1536,K=1536,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=1536,K=1536,device_name=NVIDIA_L20Y,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=1536,K=7168,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=1536,K=7168,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=1536,K=7168,device_name=AMD_Radeon_Graphics,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=1536,K=7168,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=1536,K=7168,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=1536,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=1536,K=7168,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=1536,K=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=1536,K=7168,device_name=NVIDIA_L20Y,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=2048,K=512,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=2048,K=512,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=2048,K=512,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=2048,K=512,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=2048,K=512,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=2048,K=512,device_name=NVIDIA_L20Y,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=2304,K=7168,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=2304,K=7168,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=2304,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=2304,K=7168,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=2304,K=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=2304,K=7168,device_name=NVIDIA_L20Y,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=24576,K=1536,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=24576,K=1536,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=24576,K=1536,device_name=AMD_Radeon_Graphics,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=24576,K=7168,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=24576,K=7168,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=24576,K=7168,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=24576,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=24576,K=7168,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=24576,K=7168,device_name=NVIDIA_H20,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=24576,K=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=24576,K=7168,device_name=NVIDIA_L20Y,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=256,K=7168,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=256,K=7168,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=256,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=256,K=7168,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=256,K=7168,device_name=NVIDIA_L20Y,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=3072,K=1536,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=3072,K=1536,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=3072,K=1536,device_name=AMD_Radeon_Graphics,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=3072,K=1536,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=3072,K=1536,device_name=NVIDIA_H20,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=3072,K=1536,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=3072,K=7168,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=3072,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=3072,K=7168,device_name=NVIDIA_H20,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=3072,K=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=32768,K=512,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=32768,K=512,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=32768,K=512,device_name=AMD_Radeon_Graphics,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=32768,K=512,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=32768,K=512,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=32768,K=512,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=32768,K=512,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=32768,K=512,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=32768,K=512,device_name=NVIDIA_H20,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=32768,K=512,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=32768,K=512,device_name=NVIDIA_L20Y,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=36864,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=36864,K=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=4096,K=512,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=4096,K=512,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=4096,K=512,device_name=AMD_Radeon_Graphics,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=4096,K=512,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=4096,K=512,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=4096,K=512,device_name=NVIDIA_H20,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=4096,K=512,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=4608,K=7168,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=4608,K=7168,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=4608,K=7168,device_name=AMD_Radeon_Graphics,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=4608,K=7168,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=4608,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=4608,K=7168,device_name=NVIDIA_H20,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=4608,K=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=512,K=7168,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=512,K=7168,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=512,K=7168,device_name=AMD_Radeon_Graphics,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=512,K=7168,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=512,K=7168,device_name=NVIDIA_H20,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=512,K=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=576,K=7168,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=576,K=7168,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=576,K=7168,device_name=AMD_Radeon_Graphics,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=576,K=7168,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=576,K=7168,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=576,K=7168,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=576,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=576,K=7168,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=576,K=7168,device_name=NVIDIA_H20,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=576,K=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=576,K=7168,device_name=NVIDIA_L20Y,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=1024,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=1024,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=1024,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=1024,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=1024,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=1024,device_name=NVIDIA_L20Y,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=1152,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=1152,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=1152,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=1152,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=1152,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=1152,device_name=NVIDIA_L20Y,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=128,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=128,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=128,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=128,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=128,device_name=NVIDIA_L20Y,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=16384,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=16384,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=16384,device_name=AMD_Radeon_Graphics,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=16384,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=16384,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=16384,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=16384,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=16384,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=16384,device_name=NVIDIA_H20,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=16384,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=16384,device_name=NVIDIA_L20Y,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=18432,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=18432,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=18432,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=18432,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=18432,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=18432,device_name=NVIDIA_H20,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=18432,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=18432,device_name=NVIDIA_L20Y,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=2048,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=2048,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=2048,device_name=AMD_Radeon_Graphics,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=2048,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=2048,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=2048,device_name=NVIDIA_H20,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=2048,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=2304,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=2304,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=2304,device_name=AMD_Radeon_Graphics,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=2304,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=2304,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=2304,device_name=NVIDIA_H20,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=2304,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=256,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=256,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=256,device_name=AMD_Radeon_Graphics,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=256,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=256,device_name=NVIDIA_H20,dtype=int8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/configs/N=7168,K=256,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+sglang/srt/layers/quantization/deep_gemm_wrapper/__init__.py
+sglang/srt/layers/quantization/deep_gemm_wrapper/compile_utils.py
+sglang/srt/layers/quantization/deep_gemm_wrapper/configurer.py
+sglang/srt/layers/quantization/deep_gemm_wrapper/entrypoint.py
+sglang/srt/layers/quantization/quark/__init__.py
+sglang/srt/layers/quantization/quark/quark.py
+sglang/srt/layers/quantization/quark/quark_moe.py
+sglang/srt/layers/quantization/quark/utils.py
+sglang/srt/layers/quantization/quark/schemes/__init__.py
+sglang/srt/layers/quantization/quark/schemes/quark_scheme.py
+sglang/srt/layers/quantization/quark/schemes/quark_w4a4_mxfp4.py
+sglang/srt/lora/layers.py
+sglang/srt/lora/lora.py
+sglang/srt/lora/lora_config.py
+sglang/srt/lora/lora_manager.py
+sglang/srt/lora/lora_registry.py
+sglang/srt/lora/mem_pool.py
+sglang/srt/lora/utils.py
+sglang/srt/lora/backend/base_backend.py
+sglang/srt/lora/backend/triton_backend.py
+sglang/srt/lora/triton_ops/__init__.py
+sglang/srt/lora/triton_ops/gate_up_lora_b.py
+sglang/srt/lora/triton_ops/qkv_lora_b.py
+sglang/srt/lora/triton_ops/sgemm_lora_a.py
+sglang/srt/lora/triton_ops/sgemm_lora_b.py
+sglang/srt/managers/cache_controller.py
+sglang/srt/managers/configure_logging.py
+sglang/srt/managers/data_parallel_controller.py
+sglang/srt/managers/detokenizer_manager.py
+sglang/srt/managers/io_struct.py
+sglang/srt/managers/mm_utils.py
+sglang/srt/managers/multi_tokenizer_mixin.py
+sglang/srt/managers/multimodal_processor.py
+sglang/srt/managers/schedule_batch.py
+sglang/srt/managers/schedule_policy.py
+sglang/srt/managers/scheduler.py
+sglang/srt/managers/scheduler_input_blocker.py
+sglang/srt/managers/scheduler_metrics_mixin.py
+sglang/srt/managers/scheduler_output_processor_mixin.py
+sglang/srt/managers/scheduler_profiler_mixin.py
+sglang/srt/managers/scheduler_recv_skipper.py
+sglang/srt/managers/scheduler_update_weights_mixin.py
+sglang/srt/managers/session_controller.py
+sglang/srt/managers/template_manager.py
+sglang/srt/managers/tokenizer_manager.py
+sglang/srt/managers/tp_worker.py
+sglang/srt/managers/tp_worker_overlap_thread.py
+sglang/srt/managers/utils.py
+sglang/srt/mem_cache/allocator.py
+sglang/srt/mem_cache/allocator_ascend.py
+sglang/srt/mem_cache/base_prefix_cache.py
+sglang/srt/mem_cache/chunk_cache.py
+sglang/srt/mem_cache/flush_cache.py
+sglang/srt/mem_cache/hicache_storage.py
+sglang/srt/mem_cache/hiradix_cache.py
+sglang/srt/mem_cache/lora_radix_cache.py
+sglang/srt/mem_cache/memory_pool.py
+sglang/srt/mem_cache/memory_pool_host.py
+sglang/srt/mem_cache/multimodal_cache.py
+sglang/srt/mem_cache/radix_cache.py
+sglang/srt/mem_cache/radix_cache_cpp.py
+sglang/srt/mem_cache/swa_radix_cache.py
+sglang/srt/mem_cache/cpp_radix_tree/radix_tree.py
+sglang/srt/mem_cache/storage/hf3fs/client_hf3fs.py
+sglang/srt/mem_cache/storage/hf3fs/hf3fs_utils.cpp
+sglang/srt/mem_cache/storage/hf3fs/mini_3fs_metadata_server.py
+sglang/srt/mem_cache/storage/hf3fs/storage_hf3fs.py
+sglang/srt/mem_cache/storage/hf3fs/test_hf3fs_utils.py
+sglang/srt/mem_cache/storage/mooncake_store/mooncake_store.py
+sglang/srt/mem_cache/storage/mooncake_store/unit_test.py
+sglang/srt/mem_cache/storage/nixl/hicache_nixl.py
+sglang/srt/mem_cache/storage/nixl/nixl_utils.py
+sglang/srt/mem_cache/storage/nixl/test_hicache_nixl_storage.py
+sglang/srt/metrics/collector.py
+sglang/srt/metrics/func_timer.py
+sglang/srt/model_executor/cuda_graph_runner.py
+sglang/srt/model_executor/forward_batch_info.py
+sglang/srt/model_executor/model_runner.py
+sglang/srt/model_executor/npu_graph_runner.py
+sglang/srt/model_loader/__init__.py
+sglang/srt/model_loader/loader.py
+sglang/srt/model_loader/utils.py
+sglang/srt/model_loader/weight_utils.py
+sglang/srt/models/arcee.py
+sglang/srt/models/baichuan.py
+sglang/srt/models/bailing_moe.py
+sglang/srt/models/bert.py
+sglang/srt/models/chatglm.py
+sglang/srt/models/clip.py
+sglang/srt/models/commandr.py
+sglang/srt/models/dbrx.py
+sglang/srt/models/deepseek.py
+sglang/srt/models/deepseek_janus_pro.py
+sglang/srt/models/deepseek_nextn.py
+sglang/srt/models/deepseek_v2.py
+sglang/srt/models/deepseek_vl2.py
+sglang/srt/models/ernie4.py
+sglang/srt/models/ernie4_eagle.py
+sglang/srt/models/exaone.py
+sglang/srt/models/gemma.py
+sglang/srt/models/gemma2.py
+sglang/srt/models/gemma2_reward.py
+sglang/srt/models/gemma3_causal.py
+sglang/srt/models/gemma3_mm.py
+sglang/srt/models/gemma3n_audio.py
+sglang/srt/models/gemma3n_causal.py
+sglang/srt/models/gemma3n_mm.py
+sglang/srt/models/glm4.py
+sglang/srt/models/glm4_moe.py
+sglang/srt/models/glm4_moe_nextn.py
+sglang/srt/models/glm4v.py
+sglang/srt/models/glm4v_moe.py
+sglang/srt/models/gpt2.py
+sglang/srt/models/gpt_bigcode.py
+sglang/srt/models/gpt_oss.py
+sglang/srt/models/granite.py
+sglang/srt/models/granitemoe.py
+sglang/srt/models/grok.py
+sglang/srt/models/hunyuan.py
+sglang/srt/models/idefics2.py
+sglang/srt/models/internlm2.py
+sglang/srt/models/internlm2_reward.py
+sglang/srt/models/interns1.py
+sglang/srt/models/internvl.py
+sglang/srt/models/kimi_vl.py
+sglang/srt/models/kimi_vl_moonvit.py
+sglang/srt/models/llama.py
+sglang/srt/models/llama4.py
+sglang/srt/models/llama_classification.py
+sglang/srt/models/llama_eagle.py
+sglang/srt/models/llama_eagle3.py
+sglang/srt/models/llama_embedding.py
+sglang/srt/models/llama_reward.py
+sglang/srt/models/llava.py
+sglang/srt/models/llavavid.py
+sglang/srt/models/longcat_flash.py
+sglang/srt/models/longcat_flash_nextn.py
+sglang/srt/models/mimo.py
+sglang/srt/models/mimo_mtp.py
+sglang/srt/models/minicpm.py
+sglang/srt/models/minicpm3.py
+sglang/srt/models/minicpmo.py
+sglang/srt/models/minicpmv.py
+sglang/srt/models/mistral.py
+sglang/srt/models/mixtral.py
+sglang/srt/models/mixtral_quant.py
+sglang/srt/models/mllama.py
+sglang/srt/models/mllama4.py
+sglang/srt/models/nemotron_nas.py
+sglang/srt/models/olmo.py
+sglang/srt/models/olmo2.py
+sglang/srt/models/olmoe.py
+sglang/srt/models/persimmon.py
+sglang/srt/models/phi.py
+sglang/srt/models/phi3_small.py
+sglang/srt/models/phi4mm.py
+sglang/srt/models/phi4mm_audio.py
+sglang/srt/models/phi4mm_utils.py
+sglang/srt/models/phimoe.py
+sglang/srt/models/pixtral.py
+sglang/srt/models/qwen.py
+sglang/srt/models/qwen2.py
+sglang/srt/models/qwen2_5_vl.py
+sglang/srt/models/qwen2_audio.py
+sglang/srt/models/qwen2_classification.py
+sglang/srt/models/qwen2_eagle.py
+sglang/srt/models/qwen2_moe.py
+sglang/srt/models/qwen2_rm.py
+sglang/srt/models/qwen2_vl.py
+sglang/srt/models/qwen3.py
+sglang/srt/models/qwen3_classification.py
+sglang/srt/models/qwen3_moe.py
+sglang/srt/models/registry.py
+sglang/srt/models/roberta.py
+sglang/srt/models/siglip.py
+sglang/srt/models/stablelm.py
+sglang/srt/models/step3_vl.py
+sglang/srt/models/torch_native_llama.py
+sglang/srt/models/transformers.py
+sglang/srt/models/vila.py
+sglang/srt/models/xverse.py
+sglang/srt/models/xverse_moe.py
+sglang/srt/models/yivl.py
+sglang/srt/multimodal/mm_utils.py
+sglang/srt/multimodal/processors/base_processor.py
+sglang/srt/multimodal/processors/clip.py
+sglang/srt/multimodal/processors/deepseek_vl_v2.py
+sglang/srt/multimodal/processors/gemma3.py
+sglang/srt/multimodal/processors/gemma3n.py
+sglang/srt/multimodal/processors/glm4v.py
+sglang/srt/multimodal/processors/internvl.py
+sglang/srt/multimodal/processors/janus_pro.py
+sglang/srt/multimodal/processors/kimi_vl.py
+sglang/srt/multimodal/processors/llava.py
+sglang/srt/multimodal/processors/minicpm.py
+sglang/srt/multimodal/processors/mlama.py
+sglang/srt/multimodal/processors/mllama4.py
+sglang/srt/multimodal/processors/phi4mm.py
+sglang/srt/multimodal/processors/pixtral.py
+sglang/srt/multimodal/processors/qwen_audio.py
+sglang/srt/multimodal/processors/qwen_vl.py
+sglang/srt/multimodal/processors/step3_vl.py
+sglang/srt/multimodal/processors/vila.py
+sglang/srt/parser/code_completion_parser.py
+sglang/srt/parser/conversation.py
+sglang/srt/parser/harmony_parser.py
+sglang/srt/parser/jinja_template_utils.py
+sglang/srt/parser/reasoning_parser.py
+sglang/srt/sampling/custom_logit_processor.py
+sglang/srt/sampling/sampling_batch_info.py
+sglang/srt/sampling/sampling_params.py
+sglang/srt/sampling/penaltylib/__init__.py
+sglang/srt/sampling/penaltylib/frequency_penalty.py
+sglang/srt/sampling/penaltylib/min_new_tokens.py
+sglang/srt/sampling/penaltylib/orchestrator.py
+sglang/srt/sampling/penaltylib/presence_penalty.py
+sglang/srt/speculative/build_eagle_tree.py
+sglang/srt/speculative/eagle_draft_cuda_graph_runner.py
+sglang/srt/speculative/eagle_draft_extend_cuda_graph_runner.py
+sglang/srt/speculative/eagle_utils.py
+sglang/srt/speculative/eagle_worker.py
+sglang/srt/speculative/spec_info.py
+sglang/srt/tokenizer/tiktoken_tokenizer.py
+sglang/srt/weight_sync/tensor_bucket.py
+sglang/srt/weight_sync/utils.py
+sglang/test/__init__.py
+sglang/test/doc_patch.py
+sglang/test/few_shot_gsm8k.py
+sglang/test/few_shot_gsm8k_engine.py
+sglang/test/run_eval.py
+sglang/test/runners.py
+sglang/test/send_one.py
+sglang/test/simple_eval_common.py
+sglang/test/simple_eval_gpqa.py
+sglang/test/simple_eval_humaneval.py
+sglang/test/simple_eval_math.py
+sglang/test/simple_eval_mgsm.py
+sglang/test/simple_eval_mmlu.py
+sglang/test/test_activation.py
+sglang/test/test_block_fp8.py
+sglang/test/test_block_fp8_deep_gemm_blackwell.py
+sglang/test/test_block_fp8_ep.py
+sglang/test/test_custom_ops.py
+sglang/test/test_cutlass_moe.py
+sglang/test/test_cutlass_w4a8_moe.py
+sglang/test/test_deepep_utils.py
+sglang/test/test_dynamic_grad_mode.py
+sglang/test/test_fp4_moe.py
+sglang/test/test_layernorm.py
+sglang/test/test_marlin_moe.py
+sglang/test/test_marlin_utils.py
+sglang/test/test_programs.py
+sglang/test/test_utils.py
+sglang/test/attention/__init__.py
+sglang/test/attention/test_flashattn_backend.py
+sglang/test/attention/test_flashattn_mla_backend.py
+sglang/test/attention/test_prefix_chunk_info.py
+sglang/test/attention/test_trtllm_mla_backend.py
\ No newline at end of file

python/sglang.egg-info/dependency_links.txt

+

python/sglang.egg-info/requires.txt

+aiohttp
+requests
+tqdm
+numpy
+IPython
+setproctitle
+
+[all]
+sglang[srt]
+sglang[openai]
+sglang[anthropic]
+sglang[torch_memory_saver]
+sglang[decord]
+
+[all_cpu]
+sglang[srt_cpu]
+sglang[openai]
+sglang[anthropic]
+sglang[decord]
+
+[all_hip]
+sglang[srt_hip]
+sglang[openai]
+sglang[anthropic]
+sglang[decord]
+
+[all_hpu]
+sglang[srt_hpu]
+sglang[openai]
+sglang[anthropic]
+sglang[decord]
+
+[all_npu]
+sglang[srt_npu]
+sglang[openai]
+sglang[anthropic]
+sglang[decord]
+
+[all_xpu]
+sglang[srt_xpu]
+sglang[openai]
+sglang[anthropic]
+sglang[decord]
+
+[anthropic]
+anthropic>=0.20.0
+
+[blackwell]
+sglang[runtime_common]
+sgl-kernel
+torch==2.8.0
+torchaudio==2.8.0
+torchvision
+cuda-python
+flashinfer_python==0.3.0
+
+[decord]
+decord
+
+[dev]
+sglang[all]
+sglang[test]
+
+[dev_cpu]
+sglang[all_cpu]
+sglang[test]
+
+[dev_hip]
+sglang[all_hip]
+sglang[test]
+
+[dev_hpu]
+sglang[all_hpu]
+sglang[test]
+
+[dev_xpu]
+sglang[all_xpu]
+sglang[test]
+
+[litellm]
+litellm>=1.0.0
+
+[openai]
+openai==1.99.1
+tiktoken
+
+[runtime_common]
+blobfile==3.0.0
+build
+compressed-tensors
+datasets
+einops
+fastapi
+hf_transfer
+huggingface_hub
+interegular
+llguidance<0.8.0,>=0.7.11
+modelscope
+msgspec
+ninja
+openai==1.99.1
+openai-harmony==0.0.4
+orjson
+outlines==0.1.11
+packaging
+partial_json_parser
+pillow
+prometheus-client>=0.20.0
+psutil
+pybase64
+pydantic
+pynvml
+python-multipart
+pyzmq>=25.1.2
+sentencepiece
+soundfile==0.13.1
+scipy
+timm==1.0.16
+tiktoken
+torchao==0.9.0
+transformers==4.56.0
+uvicorn
+uvloop
+xgrammar==0.1.23
+
+[srt]
+sglang[runtime_common]
+sgl-kernel==0.3.8
+torch==2.8.0
+torchaudio==2.8.0
+torchvision
+cuda-python
+flashinfer_python==0.3.0
+
+[srt_cpu]
+sglang[runtime_common]
+
+[srt_hip]
+sglang[runtime_common]
+torch
+petit_kernel==0.0.2
+wave-lang==1.0.1
+
+[srt_hpu]
+sglang[runtime_common]
+
+[srt_npu]
+sglang[runtime_common]
+
+[srt_xpu]
+sglang[runtime_common]
+
+[test]
+accelerate
+expecttest
+jsonlines
+matplotlib
+pandas
+peft
+sentence_transformers
+pytest
+tabulate
+
+[torch_memory_saver]
+torch_memory_saver==0.0.8

python/sglang.egg-info/top_level.txt

+sglang