sglang 0.3.1.post3__py3-none-any.whl → 0.3.3__py3-none-any.whl

sglang/test/test_programs.py

@@ -72,7 +72,7 @@ def test_select(check_answer):
         statement="The capital of Germany is Berlin.",
     )
     if check_answer:
-        assert ret["answer"] == "True", ret.text
+        assert ret["answer"] == "True", ret.text()
     else:
         assert ret["answer"] in ["True", "False", "Unknown"]
 
@@ -80,7 +80,7 @@ def test_select(check_answer):
         statement="The capital of Canada is Tokyo.",
     )
     if check_answer:
-        assert ret["answer"] == "False", ret.text
+        assert ret["answer"] == "False", ret.text()
     else:
         assert ret["answer"] in ["True", "False", "Unknown"]
 
@@ -88,7 +88,7 @@ def test_select(check_answer):
         statement="Purple is a better color than green.",
     )
     if check_answer:
-        assert ret["answer"] == "Unknown", ret.text
+        assert ret["answer"] == "Unknown", ret.text()
     else:
         assert ret["answer"] in ["True", "False", "Unknown"]
 
@@ -100,8 +100,8 @@ def test_decode_int():
         s += "The number of days in a year is " + sgl.gen_int("days") + "\n"
 
     ret = decode_int.run(temperature=0.1)
-    assert int(ret["hours"]) == 24, ret.text
-    assert int(ret["days"]) == 365, ret.text
+    assert int(ret["hours"]) == 24, ret.text()
+    assert int(ret["days"]) == 365, ret.text()
 
 
 def test_decode_json_regex():
@@ -517,3 +517,36 @@ def test_hellaswag_select():
     accuracy = np.mean(np.array(preds) == np.array(labels))
 
     return accuracy, latency
+
+
+def test_gen_min_new_tokens():
+    """
+    Validate sgl.gen(min_tokens) functionality.
+
+    The test asks a question where, without a min_tokens constraint, the generated answer is expected to be short.
+    By enforcing the min_tokens parameter, we ensure the generated answer has at least the specified number of tokens.
+    We verify that the number of tokens in the answer is >= the min_tokens threshold.
+    """
+    import sglang as sgl
+    from sglang.srt.hf_transformers_utils import get_tokenizer
+
+    model_path = sgl.global_config.default_backend.endpoint.get_model_name()
+    MIN_TOKENS, MAX_TOKENS = 64, 128
+
+    @sgl.function
+    def convo_1(s):
+        s += sgl.user("What is the capital of the United States?")
+        s += sgl.assistant(
+            sgl.gen("answer", min_tokens=MIN_TOKENS, max_tokens=MAX_TOKENS)
+        )
+
+    def assert_min_tokens(tokenizer, text):
+        token_ids = tokenizer.encode(text)
+        assert (
+            len(token_ids) >= MIN_TOKENS
+        ), f"Generated {len(token_ids)} tokens, min required: {MIN_TOKENS}. Text: {text}"
+
+    tokenizer = get_tokenizer(model_path)
+
+    state = convo_1.run()
+    assert_min_tokens(tokenizer, state["answer"])

sglang/test/test_utils.py

@@ -23,12 +23,13 @@ from sglang.srt.utils import kill_child_process
 from sglang.utils import get_exception_traceback
 
 DEFAULT_FP8_MODEL_NAME_FOR_TEST = "neuralmagic/Meta-Llama-3.1-8B-FP8"
-DEFAULT_MODEL_NAME_FOR_TEST = "meta-llama/Meta-Llama-3.1-8B-Instruct"
+DEFAULT_MODEL_NAME_FOR_TEST = "meta-llama/Llama-3.1-8B-Instruct"
 DEFAULT_MOE_MODEL_NAME_FOR_TEST = "mistralai/Mixtral-8x7B-Instruct-v0.1"
 DEFAULT_MLA_MODEL_NAME_FOR_TEST = "deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct"
+DEFAULT_MLA_FP8_MODEL_NAME_FOR_TEST = "neuralmagic/DeepSeek-Coder-V2-Lite-Instruct-FP8"
 DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH = 600
-DEFAULT_MODEL_NAME_FOR_NIGHTLY_EVAL_TP1 = "meta-llama/Meta-Llama-3.1-8B-Instruct,mistralai/Mistral-7B-Instruct-v0.3,deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct,google/gemma-2-27b-it"
-DEFAULT_MODEL_NAME_FOR_NIGHTLY_EVAL_TP2 = "meta-llama/Meta-Llama-3.1-70B-Instruct,mistralai/Mixtral-8x7B-Instruct-v0.1,Qwen/Qwen2-57B-A14B-Instruct,deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct"
+DEFAULT_MODEL_NAME_FOR_NIGHTLY_EVAL_TP1 = "meta-llama/Llama-3.1-8B-Instruct,mistralai/Mistral-7B-Instruct-v0.3,deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct,google/gemma-2-27b-it"
+DEFAULT_MODEL_NAME_FOR_NIGHTLY_EVAL_TP2 = "meta-llama/Llama-3.1-70B-Instruct,mistralai/Mixtral-8x7B-Instruct-v0.1,Qwen/Qwen2-57B-A14B-Instruct,deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct"
 DEFAULT_MODEL_NAME_FOR_NIGHTLY_EVAL_FP8_TP1 = "neuralmagic/Meta-Llama-3.1-8B-Instruct-FP8,neuralmagic/Mistral-7B-Instruct-v0.3-FP8,neuralmagic/DeepSeek-Coder-V2-Lite-Instruct-FP8,neuralmagic/gemma-2-2b-it-FP8"
 DEFAULT_MODEL_NAME_FOR_NIGHTLY_EVAL_FP8_TP2 = "neuralmagic/Meta-Llama-3.1-70B-Instruct-FP8,neuralmagic/Mixtral-8x7B-Instruct-v0.1-FP8,neuralmagic/Qwen2-72B-Instruct-FP8,neuralmagic/Qwen2-57B-A14B-Instruct-FP8,neuralmagic/DeepSeek-Coder-V2-Lite-Instruct-FP8"
 DEFAULT_MODEL_NAME_FOR_NIGHTLY_EVAL_QUANT_TP1 = "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4,hugging-quants/Meta-Llama-3.1-8B-Instruct-GPTQ-INT4"
@@ -84,7 +85,7 @@ def call_generate_vllm(prompt, temperature, max_tokens, stop=None, n=1, url=None
 
 
 def call_generate_outlines(
-    prompt, temperature, max_tokens, stop=[], regex=None, n=1, url=None
+    prompt, temperature, max_tokens, stop=None, regex=None, n=1, url=None
 ):
     assert url is not None
 
@@ -513,7 +514,16 @@ def get_similarities(vec1, vec2):
     return F.cosine_similarity(torch.tensor(vec1), torch.tensor(vec2), dim=0)
 
 
-def run_bench_serving(model, num_prompts, request_rate, other_server_args):
+def run_bench_serving(
+    model,
+    num_prompts,
+    request_rate,
+    other_server_args,
+    dataset_name="random",
+    random_input_len=4096,
+    random_output_len=2048,
+    disable_stream=False,
+):
     # Launch the server
     base_url = DEFAULT_URL_FOR_TEST
     process = popen_launch_server(
@@ -529,21 +539,21 @@ def run_bench_serving(model, num_prompts, request_rate, other_server_args):
         base_url=base_url,
         host=None,
         port=None,
-        dataset_name="random",
+        dataset_name=dataset_name,
         dataset_path="",
         model=None,
         tokenizer=None,
         num_prompts=num_prompts,
         sharegpt_output_len=None,
-        random_input_len=4096,
-        random_output_len=2048,
+        random_input_len=random_input_len,
+        random_output_len=random_output_len,
         random_range_ratio=0.0,
         request_rate=request_rate,
         multi=None,
         seed=0,
         output_file=None,
         disable_tqdm=False,
-        disable_stream=False,
+        disable_stream=disable_stream,
         disable_ignore_eos=False,
         extra_request_body=None,
     )
@@ -587,3 +597,37 @@ def run_bench_latency(model, other_args):
         kill_child_process(process.pid)
 
     return output_throughput
+
+
+def lcs(X, Y):
+    m = len(X)
+    n = len(Y)
+    L = [[0] * (n + 1) for _ in range(m + 1)]
+
+    for i in range(m + 1):
+        for j in range(n + 1):
+            if i == 0 or j == 0:
+                L[i][j] = 0
+            elif X[i - 1] == Y[j - 1]:
+                L[i][j] = L[i - 1][j - 1] + 1
+            else:
+                L[i][j] = max(L[i - 1][j], L[i][j - 1])
+
+    return L[m][n]
+
+
+def calculate_rouge_l(output_strs_list1, output_strs_list2):
+    """calculate the ROUGE-L score"""
+    rouge_l_scores = []
+
+    for s1, s2 in zip(output_strs_list1, output_strs_list2):
+        lcs_len = lcs(s1, s2)
+        precision = lcs_len / len(s1) if len(s1) > 0 else 0
+        recall = lcs_len / len(s2) if len(s2) > 0 else 0
+        if precision + recall > 0:
+            fmeasure = (2 * precision * recall) / (precision + recall)
+        else:
+            fmeasure = 0.0
+        rouge_l_scores.append(fmeasure)
+
+    return rouge_l_scores

sglang/version.py

@@ -1 +1 @@
-__version__ = "0.3.1.post3"
+__version__ = "0.3.3"

{sglang-0.3.1.post3.dist-info → sglang-0.3.3.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: sglang
-Version: 0.3.1.post3
+Version: 0.3.3
 Summary: SGLang is yet another fast serving framework for large language models and vision language models.
 License: Apache License
                                    Version 2.0, January 2004
@@ -248,6 +248,7 @@ Requires-Dist: uvloop; extra == "srt"
 Requires-Dist: zmq; extra == "srt"
 Requires-Dist: vllm==0.5.5; extra == "srt"
 Requires-Dist: outlines>=0.0.44; extra == "srt"
+Requires-Dist: modelscope; extra == "srt"
 Provides-Extra: test
 Requires-Dist: jsonlines; extra == "test"
 Requires-Dist: matplotlib; extra == "test"
@@ -269,16 +270,11 @@ Requires-Dist: peft; extra == "test"
 
 --------------------------------------------------------------------------------
 
-| [**Blog**](https://lmsys.org/blog/2024-07-25-sglang-llama3/) | [**Paper**](https://arxiv.org/abs/2312.07104) | [**Join Slack**](https://join.slack.com/t/sgl-fru7574/shared_invite/zt-2ngly9muu-t37XiH87qvD~6rVBTkTEHw) | [**Join Weekly Development Meeting**](https://calendar.app.google/v2Tw3kuHkKYyp8VV7) |
+| [**Blog**](https://lmsys.org/blog/2024-07-25-sglang-llama3/) | [**Paper**](https://arxiv.org/abs/2312.07104) | [**Join Slack**](https://join.slack.com/t/sgl-fru7574/shared_invite/zt-2ngly9muu-t37XiH87qvD~6rVBTkTEHw) | [**Join Bi-Weekly Development Meeting (Oct. 19)**](https://calendar.app.google/GYW7S8QGoanCuaxW6) |
 
-SGLang is a fast serving framework for large language models and vision language models.
-It makes your interaction with models faster and more controllable by co-designing the backend runtime and frontend language.
-The core features include:
-
-- **Fast Backend Runtime**: Provides efficient serving with RadixAttention for prefix caching, jump-forward constrained decoding, continuous batching, token attention (paged attention), tensor parallelism, FlashInfer kernels, chunked prefill, and quantization (INT4/FP8/AWQ/GPTQ).
-- **Flexible Frontend Language**: Offers an intuitive interface for programming LLM applications, including chained generation calls, advanced prompting, control flow, multi-modal inputs, parallelism, and external interactions.
-- **Extensive Model Support**: Supports a wide range of generative models (Llama 3, Gemma 2, Mistral, QWen, DeepSeek, LLaVA, etc.) and embedding models (e5-mistral), with easy extensibility for integrating new models.
-- **Active Community**: SGLang is open-source and backed by an active community with industry adoption.
+## Upcoming Events
+- [Oct. 11, 2024] Invited talks at [AMD Advancing AI](https://www.amd.com/en/corporate/events/advancing-ai.html) Developer Day.
+- [Oct. 16, 2024] Online meetup for efficient LLM deployment and serving, co-hosted by SGLang, FlashInfer, and MLC LLM! Fill out the [Google form](https://forms.gle/B3YeedLxmrrhL1NM8) to receive the invite link.
 
 ## News
 - [2024/09] 🔥 SGLang v0.3 Release: 7x Faster DeepSeek MLA, 1.5x Faster torch.compile, Multi-Image/Video LLaVA-OneVision ([blog](https://lmsys.org/blog/2024-09-04-sglang-v0-3/)).
@@ -294,6 +290,16 @@ The core features include:
 
 </details>
 
+## About
+SGLang is a fast serving framework for large language models and vision language models.
+It makes your interaction with models faster and more controllable by co-designing the backend runtime and frontend language.
+The core features include:
+
+- **Fast Backend Runtime**: Provides efficient serving with RadixAttention for prefix caching, jump-forward constrained decoding, continuous batching, token attention (paged attention), tensor parallelism, FlashInfer kernels, chunked prefill, and quantization (INT4/FP8/AWQ/GPTQ).
+- **Flexible Frontend Language**: Offers an intuitive interface for programming LLM applications, including chained generation calls, advanced prompting, control flow, multi-modal inputs, parallelism, and external interactions.
+- **Extensive Model Support**: Supports a wide range of generative models (Llama, Gemma, Mistral, QWen, DeepSeek, LLaVA, etc.) and embedding models (e5-mistral), with easy extensibility for integrating new models.
+- **Active Community**: SGLang is open-source and backed by an active community with industry adoption.
+
 ## Contents
 - [Install](#install)
 - [Backend: SGLang Runtime (SRT)](#backend-sglang-runtime-srt)
@@ -318,7 +324,7 @@ pip install flashinfer -i https://flashinfer.ai/whl/cu121/torch2.4/
 ### Method 2: From source
 ```
 # Use the last release branch
-git clone -b v0.3.1.post3 https://github.com/sgl-project/sglang.git
+git clone -b v0.3.3 https://github.com/sgl-project/sglang.git
 cd sglang
 
 pip install --upgrade pip
@@ -339,7 +345,7 @@ docker run --gpus all \
     --env "HF_TOKEN=<secret>" \
     --ipc=host \
     lmsysorg/sglang:latest \
-    python3 -m sglang.launch_server --model-path meta-llama/Meta-Llama-3.1-8B-Instruct --host 0.0.0.0 --port 30000
+    python3 -m sglang.launch_server --model-path meta-llama/Llama-3.1-8B-Instruct --host 0.0.0.0 --port 30000
 ```
 
 ### Method 4: Using docker compose
@@ -348,9 +354,9 @@ docker run --gpus all \
 <summary>More</summary>
 
 > This method is recommended if you plan to serve it as a service.
-> A better approach is to use the [k8s-sglang-service.yaml](./docker/k8s-sglang-service.yaml).
+> A better approach is to use the [k8s-sglang-service.yaml](docker/k8s-sglang-service.yaml).
 
-1. Copy the [compose.yml](./docker/compose.yaml) to your local machine
+1. Copy the [compose.yml](docker/compose.yaml) to your local machine
 2. Execute the command `docker compose up -d` in your terminal.
 </details>
 
@@ -379,7 +385,7 @@ resources:
 run: |
   conda deactivate
   python3 -m sglang.launch_server \
-    --model-path meta-llama/Meta-Llama-3.1-8B-Instruct \
+    --model-path meta-llama/Llama-3.1-8B-Instruct \
     --host 0.0.0.0 \
     --port 30000
 ```
@@ -421,7 +427,8 @@ curl http://localhost:30000/generate \
     }
   }'
 ```
-Learn more about the argument format [here](docs/en/sampling_params.md).
+
+Learn more about the argument specification, streaming, and multi-modal support [here](docs/en/sampling_params.md).
 
 ### OpenAI Compatible API
 In addition, the server supports OpenAI-compatible APIs.
@@ -460,7 +467,7 @@ response = client.embeddings.create(
 print(response)
 ```
 
-It supports streaming, vision, and most features of the Chat/Completions/Models/Batch endpoints specified by the [OpenAI API Reference](https://platform.openai.com/docs/api-reference/).
+It supports streaming, vision, and almost all features of the Chat/Completions/Models/Batch endpoints specified by the [OpenAI API Reference](https://platform.openai.com/docs/api-reference/).
 
 ### Additional Server Arguments
 - To enable multi-GPU tensor parallelism, add `--tp 2`. If it reports the error "peer access is not supported between these two devices", add `--enable-p2p-check` to the server launch command.
@@ -481,10 +488,11 @@ python -m sglang.launch_server --model-path meta-llama/Meta-Llama-3-8B-Instruct
 python -m sglang.launch_server --model-path meta-llama/Meta-Llama-3-8B-Instruct --chunked-prefill-size 4096
 ```
 - To enable torch.compile acceleration, add `--enable-torch-compile`. It accelerates small models on small batch sizes.
+- To enable torchao quantization, add `--torchao-config int4wo-128`. It supports various quantization strategies.
 - To enable fp8 weight quantization, add `--quantization fp8` on a fp16 checkpoint or directly load a fp8 checkpoint without specifying any arguments.
 - To enable fp8 kv cache quantization, add `--kv-cache-dtype fp8_e5m2`.
 - If the model does not have a chat template in the Hugging Face tokenizer, you can specify a [custom chat template](docs/en/custom_chat_template.md).
-- To run tensor parallelism on multiple nodes, add `--nnodes 2`. If you have two nodes with two GPUs on each node and want to run TP=4, let `sgl-dev-0` be the hostname of the first node and `50000` be an available port.
+- To run tensor parallelism on multiple nodes, add `--nnodes 2`. If you have two nodes with two GPUs on each node and want to run TP=4, let `sgl-dev-0` be the hostname of the first node and `50000` be an available port, you can use the following commands. If you meet deadlock, please try to add `--disable-cuda-graph`
 ```
 # Node 0
 python -m sglang.launch_server --model-path meta-llama/Meta-Llama-3-8B-Instruct --tp 4 --nccl-init sgl-dev-0:50000 --nnodes 2 --node-rank 0
@@ -499,9 +507,9 @@ python -m sglang.launch_server --model-path meta-llama/Meta-Llama-3-8B-Instruct
 - Llama / Llama 2 / Llama 3 / Llama 3.1
 - Mistral / Mixtral / Mistral NeMo
 - Gemma / Gemma 2
-- OLMoE
 - Qwen / Qwen 2 / Qwen 2 MoE
 - DeepSeek / DeepSeek 2
+- OLMoE
 - [LLaVA-OneVision](https://llava-vl.github.io/blog/2024-08-05-llava-onevision/)
   - `python3 -m sglang.launch_server --model-path lmms-lab/llava-onevision-qwen2-7b-ov --port=30000 --chat-template=chatml-llava`
   - `python3 -m sglang.launch_server --model-path lmms-lab/llava-onevision-qwen2-72b-ov --port=30000 --tp-size=8 --chat-template=chatml-llava`
@@ -521,7 +529,7 @@ python -m sglang.launch_server --model-path meta-llama/Meta-Llama-3-8B-Instruct
 - BaiChuan2
 - MiniCPM / MiniCPM 3
 - XVERSE / XVERSE MoE
-
+- SmolLM
 
 **Embedding Models**
 
@@ -529,7 +537,7 @@ python -m sglang.launch_server --model-path meta-llama/Meta-Llama-3-8B-Instruct
 - gte-Qwen2
   - `python -m sglang.launch_server --model-path Alibaba-NLP/gte-Qwen2-7B-instruct --is-embedding`
 
-Instructions for supporting a new model are [here](https://github.com/sgl-project/sglang/blob/main/docs/en/model_support.md).
+Instructions for supporting a new model are [here](docs/en/model_support.md).
 
 #### Use Models From ModelScope
 <details>
@@ -543,6 +551,17 @@ Launch [Qwen2-7B-Instruct](https://www.modelscope.cn/models/qwen/qwen2-7b-instru
 ```
 SGLANG_USE_MODELSCOPE=true python -m sglang.launch_server --model-path qwen/Qwen2-7B-Instruct --port 30000
 ```
+
+Or start it by docker.
+```bash
+docker run --gpus all \
+    -p 30000:30000 \
+    -v ~/.cache/modelscope:/root/.cache/modelscope \
+    --env "SGLANG_USE_MODELSCOPE=true" \
+    --ipc=host \
+    lmsysorg/sglang:latest \
+    python3 -m sglang.launch_server --model-path Qwen/Qwen2.5-7B-Instruct --host 0.0.0.0 --port 30000
+```
   
 </details>
 
@@ -581,7 +600,7 @@ GLOO_SOCKET_IFNAME=eth0 python3 -m sglang.launch_server --model-path meta-llama/
 The frontend language can be used with local models or API models. It is an alternative to the OpenAI API. You may found it easier to use for complex prompting workflow.
 
 ### Quick Start
-The example below shows how to use sglang to answer a mulit-turn question.
+The example below shows how to use sglang to answer a multi-turn question.
 
 #### Using Local Models
 First, launch a server with
@@ -824,7 +843,7 @@ def chat_example(s):
 Learn more at this [blog](https://lmsys.org/blog/2024-07-25-sglang-llama3/).
 
 ## Roadmap
-[Development Roadmap (2024 Q3)](https://github.com/sgl-project/sglang/issues/634)
+[Development Roadmap (2024 Q4)](https://github.com/sgl-project/sglang/issues/1487)
 
 ## Citation And Acknowledgment
 Please cite our paper, [SGLang: Efficient Execution of Structured Language Model Programs](https://arxiv.org/abs/2312.07104), if you find the project useful.

sglang-0.3.3.dist-info/RECORD

@@ -0,0 +1,139 @@
+sglang/__init__.py,sha256=b_pqO9bR2fjK9En_tigfzKTiQzE8b_hUizY0DAKVk1M,1616
+sglang/api.py,sha256=5x591S4rLbmNPs75qPwGKVu1sonVGDyjPAJlHTyWw50,6956
+sglang/bench_latency.py,sha256=NkaL4YFWqDnochwaLd8o2pyZGqu6TeURbFB3TGyZHr4,17893
+sglang/bench_server_latency.py,sha256=rRSDqjJ5jan9AzppOGx75KRUjZCU2dUG2h06CQOdJgk,5377
+sglang/bench_serving.py,sha256=1AQzkQ8ci9-rMZEM7wap8I09oPP4AZd93RfXMQRgVro,36386
+sglang/check_env.py,sha256=rGRABCgt-0SfUrow4px28b2P59aMn8eVTnN5eZc_a8s,5397
+sglang/global_config.py,sha256=38id86i3tRGCSOFZlN1LM01a3xt-V98xuNgKGG9boCk,1058
+sglang/launch_server.py,sha256=UnjNjYuZ8TtvmRtgYEsFImkbvCwvn_tQjk0V7cHy67E,450
+sglang/launch_server_llavavid.py,sha256=olPKyhozi1coCwoRMwBRYWsTFByrgus9CwPSeNmskgc,1002
+sglang/utils.py,sha256=NA_4xUrTI7KICQ3PEACfNWKE3nxSA5QvQZJNd4TQrDc,9395
+sglang/version.py,sha256=8KcCYTXH99C2-gCLuPILJvtT9YftRWJsartIx6TQ2ZY,22
+sglang/lang/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+sglang/lang/chat_template.py,sha256=uqI_I9zIKXGXg7-W-yjqvx1ZeS_TuwFCms6wkmC2QmY,13411
+sglang/lang/choices.py,sha256=-W1DVw9N9ZliVpvmWrzIXG4cswAah8eMQrHWzkS3D8o,6234
+sglang/lang/compiler.py,sha256=o1C6G3TzhjSlsH-doTPy5oiVehr57dxNTa5oZw5TTAI,7639
+sglang/lang/interpreter.py,sha256=zakc6IkzATaMqVDWKWvqDRrqnRykxFawajA7aUHUDbI,30640
+sglang/lang/ir.py,sha256=F_9ac10OjktxR7KhOV07wiJXV20s79cRfh9d4koExJc,18262
+sglang/lang/tracer.py,sha256=borJmlSJOhg1RUndGRnilnR60eEZz2Y9aU7BpftsOxU,8287
+sglang/lang/backend/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+sglang/lang/backend/anthropic.py,sha256=EXRX7xJgA5KZszX7toSLVnKzFQ5EO0Loj-YjHFtxSxg,2081
+sglang/lang/backend/base_backend.py,sha256=Q5HdiDtyBewQeoYH0kDtBRVL8KFiEPNq9dw7XmauHQ8,1985
+sglang/lang/backend/litellm.py,sha256=ugmL7sfUxkUHVbHtwNzHgdQAEd4UCjNQboFuE3KThcY,2450
+sglang/lang/backend/openai.py,sha256=qM7eVH_kMxnDd2rpxOH0v76KxtOJFlAwgLgWIKvFGCI,15060
+sglang/lang/backend/runtime_endpoint.py,sha256=iVb7SlrpJ1ic92QG5kQUphZUb2EaVWY43dkmAO5pju4,10514
+sglang/lang/backend/vertexai.py,sha256=O-iBLD-y3vq80UxnrAoJri7bxpgd-_eakZ88Cf8bEGA,4855
+sglang/srt/conversation.py,sha256=S5w5V6G1xigNxa3UQoSxRcMpQLWWDT9EPBoHBvHkSAk,19663
+sglang/srt/hf_transformers_utils.py,sha256=rt6flb6BoYTO8fw7AKCXmQLJx5XuSUuRmZX-VJHmuLQ,6064
+sglang/srt/mm_utils.py,sha256=zox644S3IHUWmADdK4MnIbdTS2DWHOy0_Dq0gCU38QQ,12273
+sglang/srt/server.py,sha256=SKV6IxR8w0AmuwgHSEOfag_t-f6hAEq9Xg49iBioi2U,22224
+sglang/srt/server_args.py,sha256=LI8ehxs0sfI0EDhON-OhNGbDx0-oo9QhfnpYjYwnH54,24405
+sglang/srt/utils.py,sha256=amDWXIu1syU-kvdV8bUkNfYaMfpcN22BKZm_2xp59jI,22202
+sglang/srt/configs/__init__.py,sha256=292SuEorST-lAq2Uvsv2M7yC28uYZlssVvRDsF-bZCQ,86
+sglang/srt/configs/exaone.py,sha256=Duxd4yQoKy8GWEzZD_kCY_OzmN_67CTJL_Kgn0eXk3g,10731
+sglang/srt/configs/model_config.py,sha256=36My-o44trhWY3KYDeSFMGvv9XuUtIVI5e7F8VlOTWo,6723
+sglang/srt/constrained/__init__.py,sha256=ze8awDPvwAzdeMwzJ-25kXOQ4nVWoaP55jBDt5UOS_4,2070
+sglang/srt/constrained/base_tool_cache.py,sha256=5sazBMHHDpHMoqOjuY6itCxwTmIFCflIWEDXMtmrPVs,2006
+sglang/srt/constrained/fsm_cache.py,sha256=9GtliIN55Ov8Q9MSFfQC5rKrz3qTsB7Cm5OkhivKngY,3271
+sglang/srt/constrained/jump_forward.py,sha256=o-CzJu3DEs0eFKlLzsQVYMSo4vBKpffs25sXLOJd6jc,6997
+sglang/srt/layers/activation.py,sha256=7VEkCrx2dvl629Lz0fkJcJfVoZA-ykEdkpTzKEc_drQ,5225
+sglang/srt/layers/layernorm.py,sha256=HCj8Y_X6MNNdtQU2sWKgyjIqVERxl9dqrmjbBbyJjpE,3796
+sglang/srt/layers/linear.py,sha256=9rjCiSb_QOn5RgpVjIhEKdReRvSYVfcTSjbWBEbApLI,45173
+sglang/srt/layers/logits_processor.py,sha256=Fq7VHwjP4iSzl_OBLo8qw_HVbIDbYB-0MGmfiD3Jk_E,12521
+sglang/srt/layers/pooler.py,sha256=rj2lygvleBnyLCBZ8I11HGMgpfIDsT0l3PIkshJwdu4,1606
+sglang/srt/layers/radix_attention.py,sha256=i07VRXPDHj-zJ1TSrXEqCxumQwYSHwAvc8DoIg-Irtg,1964
+sglang/srt/layers/sampler.py,sha256=J5vd0CcLpLfgtLniCoe2VF6hjM_ld76hbDG4p1qoAMc,4010
+sglang/srt/layers/torchao_utils.py,sha256=1nzZkSzbF4qCAMeBKAeeDpMl_mK8imiY2RL3xFEgvAw,3340
+sglang/srt/layers/attention/__init__.py,sha256=zLLwinbYLAQHfVEz0jZiVa_cYNgSYoy4wYD_0y-ErHQ,1798
+sglang/srt/layers/attention/flashinfer_backend.py,sha256=DOvm-d3XLjE6XJDD3a8aCnlpuAJZZ946YFDH_Ec4lqc,10150
+sglang/srt/layers/attention/flashinfer_utils.py,sha256=9YMt7ab6F0gEVkxdVm8vDB0LVBRYRL0XIKVrmndp4n8,7571
+sglang/srt/layers/attention/triton_backend.py,sha256=I_kw0LXdgziHAFC8Qv5n5PDFJRLvZyzVsXwjmFZ0KSc,6041
+sglang/srt/layers/attention/triton_ops/decode_attention.py,sha256=XCQTX0kUttT1AG5FRMgfQbiXgvoempYD0UR2r6D_vJg,16711
+sglang/srt/layers/attention/triton_ops/extend_attention.py,sha256=oyqon1KG5-ICHcCANAbrglXLYKvWHFML-4tIQI9M5VI,11063
+sglang/srt/layers/attention/triton_ops/prefill_attention.py,sha256=QkXPcT02c13zha2M4mBm2S5dh_sS-Gc4FkkrcywRqvc,5377
+sglang/srt/layers/fused_moe/__init__.py,sha256=bWCrDdOy2ANEXTb8CHYO63O3Iu3eZnn0PJbgl0z5vvE,75
+sglang/srt/layers/fused_moe/fused_moe.py,sha256=1WM2cObWXcFWtqh_utGJFPnrT344rORwuQ9hJDaH2s0,23104
+sglang/srt/layers/fused_moe/layer.py,sha256=raFyvPzjYz-Fv8B3IcOxQYKKCWqXis5mXwg1GFE61y4,22243
+sglang/srt/layers/fused_moe/patch.py,sha256=B9cDtHqHfnWE0QqZAffvUi6cVRKcMBMKDGJWGIaKh3U,3898
+sglang/srt/layers/quantization/__init__.py,sha256=QilMNqgu3eOFUkEjXLSDa1NvoNdi_CAvC8a1hprOgN8,2979
+sglang/srt/layers/quantization/base_config.py,sha256=vlpSPvSrFmUe65ETg4SoPocQ9bVNY6As3QuHdr_3Dr4,4023
+sglang/srt/lora/lora.py,sha256=a5j_Yy0s95msVPFgOuH5PCe7sMu0AyZFQ5wL0H-YIg8,14913
+sglang/srt/lora/lora_config.py,sha256=paVB7F7SIuxr_vodvKf8zzAlH2fdVYHhXxcXV62D0Vo,1411
+sglang/srt/lora/lora_manager.py,sha256=gzBwYXZEPYj56PkGTshTbWRfl_370wb6uTcRhDaLiF8,12801
+sglang/srt/managers/detokenizer_manager.py,sha256=iCLPdHkL6lAp_-Qew1u4Tyt3jYRkJ8i-Bj3l8TC-uaA,7278
+sglang/srt/managers/image_processor.py,sha256=9Y9RqyLdbt4uOK7pnJCJIhY77791klskSrEg8U6pyS4,6910
+sglang/srt/managers/io_struct.py,sha256=rPyQk5y-jJu4eyoqUVh4M8B14PifjkE8B3K5yI0NX24,12185
+sglang/srt/managers/schedule_batch.py,sha256=mqdMg1QB6PNLbBjxkXoP_Ld82R1w34g_13YH82DGMh8,31216
+sglang/srt/managers/schedule_policy.py,sha256=PiTKvsAFwoNWNsv_SFkghIHCL452MdboRc2cmN6ITcU,11935
+sglang/srt/managers/scheduler.py,sha256=N9GQnp2SXd8-uN49KmQO-144N27M6h3dxRZuFZ-9AmY,39132
+sglang/srt/managers/tokenizer_manager.py,sha256=BAvLW_cRtIgjL0_cwrvDAb7g740fgEddyqaT3JtofR4,24548
+sglang/srt/managers/tp_worker.py,sha256=fcaW-u7AAX49kQCNn_AEtdRPykRdT6Z6lx1O9LHA15E,4833
+sglang/srt/mem_cache/base_prefix_cache.py,sha256=qEQwEkG4E5rab2ZoTqcesf5pR_J4nV2jBxIHsBJHtIM,924
+sglang/srt/mem_cache/chunk_cache.py,sha256=CjZZYlqQzq7mYOiBMLWA5XNb6HIyh5lIMdY-K0OUZEc,2368
+sglang/srt/mem_cache/flush_cache.py,sha256=pTLKPRB17U6vl5RFJJvuJ4jCL2SyomgkUBNlkDpGRqo,978
+sglang/srt/mem_cache/memory_pool.py,sha256=L-5drUt7vlyvple4OcjH1jJRzt2qhVrpc9klZn-bQfE,7125
+sglang/srt/mem_cache/radix_cache.py,sha256=00bghOihUm7lA1i4gxxMYQLept9LaHg2ZSXZryuFZZI,10121
+sglang/srt/model_executor/cuda_graph_runner.py,sha256=iheZYErwFT_W4kJUE1dgbGoQQx7hyOSKa-Yv8guq0DI,10479
+sglang/srt/model_executor/forward_batch_info.py,sha256=FIQ8XIIP724mIL2l7w7mSEFH452qw-TPpqm43J4YeHM,5822
+sglang/srt/model_executor/model_runner.py,sha256=KyglHFIMb5TC-NszN2D85_k7oVQLhbwhUYa7u3RFkoc,22874
+sglang/srt/models/baichuan.py,sha256=50m43kIVo-YamHFwxyiLGG_pCbF7mzUJfhEyuuSmVC8,15100
+sglang/srt/models/chatglm.py,sha256=XaS_6-ZvRw7X-56sk9xQogqT0NzGEMVpiAdQnC5qbBY,13333
+sglang/srt/models/commandr.py,sha256=2urK7u2FiwPBl60hMmt-wfaJ8V-ilv6l1B37MUlvSxk,14121
+sglang/srt/models/dbrx.py,sha256=qTpyA1Iv56VI-ksPKt4JryX2Pn7T5FXAa0n0ZoT4qbw,14615
+sglang/srt/models/deepseek.py,sha256=4sl4YYoxqe-vif7KJKcMjMA3KgvzYHqpQBgM58lzLHc,15973
+sglang/srt/models/deepseek_v2.py,sha256=dt0FGAgW3jd7OJJnKfH-LIU13U0I9b7R9shYmAEins4,28390
+sglang/srt/models/exaone.py,sha256=9JfFhYbpcHMXIaBNn8rc_GOlkItkIgbGNslNyFD7gvU,13054
+sglang/srt/models/gemma.py,sha256=gui46inEJsrmppEMTUIQuzMxGPEBx_TjiZ5-PacjuSk,12240
+sglang/srt/models/gemma2.py,sha256=V0GjEdTqxyXvBqjgyiyONipohjOqw0pLITmZZRb2kIE,14890
+sglang/srt/models/gpt_bigcode.py,sha256=LgSm-8oxBfnzMAC4Jqqg-RJGge4E_wgJ1br7ylbTPZ0,10162
+sglang/srt/models/grok.py,sha256=lUR_SmD_KhIiZx5OVUPZp8VVdrAga6WWTdMKJ5PCFbw,14896
+sglang/srt/models/internlm2.py,sha256=4SUaeJl2dZlUowahfv7kLbz3jLXtmvdBPGURmhAeX6Q,12169
+sglang/srt/models/llama.py,sha256=5j66LmvFhOKgFZiE75mJ80XBjZ2dNx7e8Yea5lsD0P0,15828
+sglang/srt/models/llama_classification.py,sha256=Yhabu9FuBxjNo74crMsK0FqpD53ehOx_zcHgIXjvlvQ,3379
+sglang/srt/models/llama_embedding.py,sha256=4j3WNLB-x7XQnJvohdRs7VSSEabbhiE2BRHmnG5IZRU,3453
+sglang/srt/models/llama_reward.py,sha256=qQOPfn-9oqhsD0EaffXtk-EXKRdSZL1X7CYAGCDoG9A,5383
+sglang/srt/models/llava.py,sha256=zbJs1P4_Bjh2_dSbyoheJZ1wGXuKHGz6BpV766G7ZUY,25094
+sglang/srt/models/llavavid.py,sha256=qhBGHTxzGAOMgqMiwOc3mUbaK6qeXsEYSlNmlEEIdeM,12198
+sglang/srt/models/minicpm.py,sha256=5vc-Lq7ggHrRxxkciVMdZ5Vq6ThLwnhFS62UCokFC2g,13792
+sglang/srt/models/minicpm3.py,sha256=hhhgZTKQApUZpH_MYQZTk3K1Ox-xpJRxGCemoUw8x4U,25184
+sglang/srt/models/mistral.py,sha256=tiYoKjyYVzlQl52QUZ33odD2yCxj9dxcqln474VuZOw,744
+sglang/srt/models/mixtral.py,sha256=BonqX_rSB_UuBDQe3uy8-NOxB4Q4s2mTxTQItvFB9ZQ,13864
+sglang/srt/models/mixtral_quant.py,sha256=SAHBIiD5O1TnojCpqTLcPy3TEvfSCKeOe3GC47fdFSg,14039
+sglang/srt/models/olmoe.py,sha256=ghhNpZe4SzaZEpw0APYBbAmLb3LBagRC2N724RkOkH4,15312
+sglang/srt/models/qwen.py,sha256=IrOKHS7b4SL2fnJegq811eeHnAQDya2PujIgKQ9URVY,9921
+sglang/srt/models/qwen2.py,sha256=B7hXnW5uYPmpMgSN7tI3tTvMEmmQLpddsw_iNTiaHJI,12398
+sglang/srt/models/qwen2_moe.py,sha256=MK-9W6FJhXoQYayg_jpXjKKq4n5j3s2b2ZaoCBfVJ2I,17120
+sglang/srt/models/stablelm.py,sha256=ldtlRG1XGdYcjwqb48dpMTfbdh8KHUjcWrrUYNJ0MEk,11326
+sglang/srt/models/torch_native_llama.py,sha256=c5GJ_k9zbSOk0PjLCXAK8YebGEy0RUVYZ9_h6_19A3M,19215
+sglang/srt/models/xverse.py,sha256=i11wEKqqVCoVtH7yo9jfpNyGHxhw7NvTPid3ojmg79s,13634
+sglang/srt/models/xverse_moe.py,sha256=JwkBhsyusP7e_hAMnomkP8cEmKNCLJPRtwaTERQ0D0M,15818
+sglang/srt/models/yivl.py,sha256=N3noJ5M-FiZS-E_zfaJs4prQOu_ineRt11MWloYgOR8,4826
+sglang/srt/openai_api/adapter.py,sha256=ULX1lo23r6semogKcbUOXGSgPJi8NJ7IuC0WVvEbVbs,51458
+sglang/srt/openai_api/protocol.py,sha256=rdSwUAoO5-KLemJOE50xwSUagxY4T1QIiNyCYsTtCi0,9868
+sglang/srt/sampling/sampling_batch_info.py,sha256=mtE_kLC6U-X6Q20BVjPWyDOoGc4kcTdIPpcsNeZcRYo,6462
+sglang/srt/sampling/sampling_params.py,sha256=Xwh4_M6PP4SWyGV-zNyIhp4XbRKbeU4251ao8UOlZlI,5704
+sglang/srt/sampling/penaltylib/__init__.py,sha256=5vQw0Y5DSzmsoFg1IdMIKLwFVhYZ5ArADHVBYbSmOec,513
+sglang/srt/sampling/penaltylib/orchestrator.py,sha256=WkTNeDhj9H9rtp2ZZeX6MS2sdKSGlLboE6FcuKrwUo0,10815
+sglang/srt/sampling/penaltylib/penalizers/frequency_penalty.py,sha256=IvYioX53Vq_ji-0Zhcz_r5mUa3T3GaIydVS6K4FhWfE,2557
+sglang/srt/sampling/penaltylib/penalizers/min_new_tokens.py,sha256=XJZP0C4NFyXgcODbIWXxrgVEjmRgqLdZuVAtoN-LveY,3565
+sglang/srt/sampling/penaltylib/penalizers/presence_penalty.py,sha256=0PlANTrR959foTA3Nj5qBE7ndaOZgG-9X6LhzlmEUc8,2533
+sglang/srt/sampling/penaltylib/penalizers/repetition_penalty.py,sha256=v9jOgA0-I31WcrhIydiFbpy2ZJPLytFLGM98NRPd2sU,2820
+sglang/test/few_shot_gsm8k.py,sha256=To7Sdg-DLF8poIQLwiOBYKbkz-1C_gn6H79vIbyPR-o,3860
+sglang/test/run_eval.py,sha256=NWxeLWmInBgkCvC9Jr_QzF7GfAiBve3Gf1JQrEOlNlU,3899
+sglang/test/runners.py,sha256=VCmtH08FsAq_JTAKfKo0zB4o-osNMAxxwe4aKcSxr4c,13515
+sglang/test/simple_eval_common.py,sha256=r0G-9QLycs2ax3RMc44T_61fzMxlpTzv6pececC7lyY,12379
+sglang/test/simple_eval_gpqa.py,sha256=8Xt9Bw05c7SZTYrCZgB68OZUqUbLo69ywiyx0bTvSUk,3220
+sglang/test/simple_eval_humaneval.py,sha256=7lTi841NT58smNOtRwCedrdX9IWWypdLkOtaQOBy-GI,5687
+sglang/test/simple_eval_math.py,sha256=6kGKNwNbLN-Af3Wj8WTimWhH-Xp3enDmSvvSjsgWUpk,2550
+sglang/test/simple_eval_mgsm.py,sha256=wfbqJW9Rkc66vzq2fEMF6jchmoA8mw1OUiGU55cZ2B0,10261
+sglang/test/simple_eval_mmlu.py,sha256=FkwamjGMjueTixymkedF-YiPloSLiy4ftILFUrKZ9XI,4357
+sglang/test/test_activation.py,sha256=jkdNRzJnbd5OgZliQaIXpxovlcky17UrweomcOcMxoE,1442
+sglang/test/test_layernorm.py,sha256=IacByD5d-stXjzBz8Ypamc7povlcedpKPbb_4JLgo3c,3720
+sglang/test/test_programs.py,sha256=1Z0umrsUu9pagzyGH5SrXl_qhKSyTfUv_kWC2mcn0qo,18208
+sglang/test/test_utils.py,sha256=NkJuezjmonjgC3_i_CTBd8KSqWh6W9CLcgoaqvTNK2U,18684
+sglang/test/srt/sampling/penaltylib/utils.py,sha256=-0p0rV-P4lNo7xAe3rQSBHTubc50a-DFyOQmLGAkgkQ,12515
+sglang-0.3.3.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
+sglang-0.3.3.dist-info/METADATA,sha256=zeY2pmiGPJb52zaHqiRHY4OcZqAHPvG_zPyve5KfANc,39063
+sglang-0.3.3.dist-info/WHEEL,sha256=GV9aMThwP_4oNCtvEC2ec3qUYutgWeAzklro_0m4WJQ,91
+sglang-0.3.3.dist-info/top_level.txt,sha256=yxhh3pYQkcnA7v3Bg889C2jZhvtJdEincysO7PEB09M,7
+sglang-0.3.3.dist-info/RECORD,,