sglang 0.2.4__py3-none-any.whl → 0.2.6__py3-none-any.whl

sglang/srt/server.py

@@ -65,9 +65,6 @@ asyncio.set_event_loop_policy(uvloop.EventLoopPolicy())
 app = FastAPI()
 tokenizer_manager = None
 
-# Put some args for easily access
-global_server_args_dict = {}
-
 
 @app.get("/health")
 async def health() -> Response:
@@ -150,14 +147,6 @@ def available_models():
     return ModelList(data=model_cards)
 
 
-def _set_global_server_args(server_args: ServerArgs):
-    global global_server_args_dict
-    global_server_args_dict = {
-        "disable_flashinfer": server_args.disable_flashinfer,
-        "attention_reduce_in_fp32": server_args.attention_reduce_in_fp32,
-    }
-
-
 def _set_torch_compile_config():
     # The following configurations are for torch compile optimizations
     import torch._dynamo.config
@@ -176,6 +165,8 @@ def launch_server(
     model_overide_args: Optional[dict] = None,
     pipe_finish_writer: Optional[mp.connection.Connection] = None,
 ):
+    server_args.check_server_args()
+
     """Launch an HTTP server."""
     global tokenizer_manager
 
@@ -211,8 +202,6 @@ def launch_server(
     if server_args.enable_torch_compile:
         _set_torch_compile_config()
 
-    _set_global_server_args(server_args)
-
     # Allocate ports
     server_args.port, server_args.additional_ports = allocate_init_ports(
         server_args.port,
@@ -230,8 +219,6 @@ def launch_server(
 
     # Handle multi-node tensor parallelism
     if server_args.nnodes > 1:
-        assert server_args.dp_size == 1, "Multi-node dp is not supported."
-
         if server_args.node_rank != 0:
             tp_size_local = server_args.tp_size // server_args.nnodes
             gpu_ids = [

sglang/srt/server_args.py

@@ -28,6 +28,7 @@ class ServerArgs:
     mem_fraction_static: Optional[float] = None
     max_prefill_tokens: Optional[int] = None
     max_running_requests: Optional[int] = None
+    max_num_reqs: Optional[int] = None
     schedule_heuristic: str = "lpm"
     schedule_conservativeness: float = 1.0
 
@@ -51,13 +52,14 @@ class ServerArgs:
 
     # Optimization/debug options
     disable_flashinfer: bool = False
+    disable_flashinfer_sampling: bool = False
     disable_radix_cache: bool = False
     disable_regex_jump_forward: bool = False
     disable_cuda_graph: bool = False
     disable_disk_cache: bool = False
     enable_torch_compile: bool = False
-    attention_reduce_in_fp32: bool = False
     enable_p2p_check: bool = False
+    attention_reduce_in_fp32: bool = False
     efficient_weight_load: bool = False
 
     # Distributed args
@@ -203,6 +205,12 @@ class ServerArgs:
             default=ServerArgs.max_running_requests,
             help="The maximum number of running requests.",
         )
+        parser.add_argument(
+            "--max-num-reqs",
+            type=int,
+            default=None,
+            help="The maximum number of requests to serve in the memory pool. If the model have a large context length, you may need to decrease this value to avoid out-of-memory errors.",
+        )
         parser.add_argument(
             "--schedule-heuristic",
             type=str,
@@ -296,7 +304,12 @@ class ServerArgs:
         parser.add_argument(
             "--disable-flashinfer",
             action="store_true",
-            help="Disable flashinfer inference kernels.",
+            help="Disable flashinfer attention kernels.",
+        )
+        parser.add_argument(
+            "--disable-flashinfer-sampling",
+            action="store_true",
+            help="Disable flashinfer sampling kernels.",
         )
         parser.add_argument(
             "--disable-radix-cache",
@@ -324,15 +337,15 @@ class ServerArgs:
             help="Optimize the model with torch.compile, experimental feature.",
         )
         parser.add_argument(
-            "--attention-reduce-in-fp32",
+            "--enable-p2p-check",
             action="store_true",
-            help="Cast the intermidiate attention results to fp32 to avoid possible crashes related to fp16."
-            "This only affects Triton attention kernels",
+            help="Enable P2P check for GPU access, otherwise the p2p access is allowed by default.",
         )
         parser.add_argument(
-            "--enable-p2p-check",
+            "--attention-reduce-in-fp32",
             action="store_true",
-            help="Enable P2P check for GPU access, otherwise the p2p access is allowed by default.",
+            help="Cast the intermidiate attention results to fp32 to avoid possible crashes related to fp16."
+            "This only affects Triton attention kernels",
         )
         parser.add_argument(
             "--efficient-weight-load",
@@ -357,6 +370,14 @@ class ServerArgs:
             f"disable_disk_cache={self.disable_disk_cache}, "
         )
 
+    def check_server_args(self):
+        assert (
+            self.tp_size % self.nnodes == 0
+        ), "tp_size must be divisible by number of nodes"
+        assert not (
+            self.dp_size > 1 and self.node_rank is not None
+        ), "multi-node data parallel is not supported"
+
 
 @dataclasses.dataclass
 class PortArgs:

sglang/test/test_programs.py

@@ -118,7 +118,11 @@ def test_decode_json_regex():
             s += "}"
 
     ret = decode_json.run()
-    js_obj = json.loads(ret["json_output"])
+    try:
+        js_obj = json.loads(ret["json_output"])
+    except json.decoder.JSONDecodeError:
+        print(ret["json_output"])
+        raise
     assert isinstance(js_obj["name"], str)
     assert isinstance(js_obj["population"], int)
 

sglang/version.py

@@ -1 +1 @@
-__version__ = "0.2.4"
+__version__ = "0.2.6"

{sglang-0.2.4.dist-info → sglang-0.2.6.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: sglang
-Version: 0.2.4
+Version: 0.2.6
 Summary: SGLang is yet another fast serving framework for large language models and vision language models.
 License: Apache License
                                    Version 2.0, January 2004
@@ -249,7 +249,7 @@ Requires-Dist: outlines >=0.0.44 ; extra == 'srt'
 
 --------------------------------------------------------------------------------
 
-| [**Blog**](https://lmsys.org/blog/2024-07-25-sglang-llama3/) | [**Paper**](https://arxiv.org/abs/2312.07104) |
+| [**Blog**](https://lmsys.org/blog/2024-07-25-sglang-llama3/) | [**Paper**](https://arxiv.org/abs/2312.07104) | [**Slack**](https://join.slack.com/t/sgl-fru7574/shared_invite/zt-2ngly9muu-t37XiH87qvD~6rVBTkTEHw) |
 
 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.
@@ -404,16 +404,17 @@ python -m sglang.launch_server --model-path meta-llama/Meta-Llama-3-8B-Instruct
 ### Run Llama 3.1 405B
 
 ```bash
-# 2 nodes run 405B fp16
+## Run 405B (fp8) on a single node
+python -m sglang.launch_server --model-path meta-llama/Meta-Llama-3.1-405B-Instruct-FP8 --tp 8
+
+## Run 405B (fp16) on two nodes
 # replace the `172.16.4.52:20000` with your own first node ip address and port, disable CUDA Graph temporarily
+
 # on the first node
 GLOO_SOCKET_IFNAME=eth0 python3 -m sglang.launch_server --model-path meta-llama/Meta-Llama-3.1-405B-Instruct --tp 16 --nccl-init-addr 172.16.4.52:20000 --nnodes 2 --node-rank 0 --disable-cuda-graph --mem-frac 0.75
 
 # on the second
 GLOO_SOCKET_IFNAME=eth0 python3 -m sglang.launch_server --model-path meta-llama/Meta-Llama-3.1-405B-Instruct --tp 16 --nccl-init-addr 172.16.4.52:20000 --nnodes 2 --node-rank 1 --disable-cuda-graph --mem-frac 0.75
-
-# single node run 405B fp8
-python -m sglang.launch_server --model-path meta-llama/Meta-Llama-3.1-405B-Instruct-FP8 --tp 8
 ```
 
 ### Supported Models
@@ -422,6 +423,7 @@ python -m sglang.launch_server --model-path meta-llama/Meta-Llama-3.1-405B-Instr
 - Mistral / Mixtral
 - Gemma / Gemma 2
 - Qwen / Qwen 2 / Qwen 2 MoE
+- DeepSeek / DeepSeek 2
 - LLaVA 1.5 / 1.6
   - `python -m sglang.launch_server --model-path liuhaotian/llava-v1.5-7b --tokenizer-path llava-hf/llava-1.5-7b-hf --chat-template vicuna_v1.1 --port 30000`
   - `python -m sglang.launch_server --model-path liuhaotian/llava-v1.6-vicuna-7b --tokenizer-path llava-hf/llava-1.5-7b-hf --chat-template vicuna_v1.1 --port 30000`
@@ -442,7 +444,7 @@ Instructions for supporting a new model are [here](https://github.com/sgl-projec
 
 ### Benchmark Performance
 
-- Benchmark a single static batch. Run the following command without launching a server. The arguments are the same as those for `launch_server.py`.
+- Benchmark a single static batch by running the following command without launching a server. The arguments are the same as those for `launch_server.py`. This is not a dynamic batching server, so it may run out of memory for a batch size that can run successfully with a real server. This is because a real server will truncate the prefill into several batches/chunks, while this unit test does not do this.
   ```
   python -m sglang.bench_latency --model-path meta-llama/Meta-Llama-3-8B-Instruct --batch 32 --input-len 256 --output-len 32
   ```

{sglang-0.2.4.dist-info → sglang-0.2.6.dist-info}/RECORD

@@ -7,11 +7,11 @@ sglang/global_config.py,sha256=CyhGL7PE-KlMcg7IHWykzImU1y4NQlpeIlh9lHA77uo,1749
 sglang/launch_server.py,sha256=Gg8CwNlTCCfg1dF65ZT9ePLxOT9LKtY79GhIPG6PCrU,358
 sglang/launch_server_llavavid.py,sha256=40uaazMsavKuk6YXFa5v37kdUpFGuealgJJeph1g8gU,1025
 sglang/utils.py,sha256=arJuwOAEX445M2NL9SAOi6jBNu0-cfU04PLAr-hIH3U,8168
-sglang/version.py,sha256=SBl2EPFW-ltPvQ7vbVWItyAsz3aKYIpjO7vcfr84GkU,22
+sglang/version.py,sha256=Oz5HbwHMyE87nmwV80AZzpkJPf-wBg7eDuJr_BXZkhU,22
 sglang/lang/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 sglang/lang/chat_template.py,sha256=psIlhaDo70twgLrx5Lgln03metLEA3-FZuixeI0Y7Ao,13309
 sglang/lang/compiler.py,sha256=UiXUmPR9wBAPtnORrLcyQX8Uh0ZL0nKeV8ZgBozAJPw,7531
-sglang/lang/interpreter.py,sha256=27j7H9p7TY4uUfF9f5E17FxK1xCNeNju4aut_PaWCrQ,29693
+sglang/lang/interpreter.py,sha256=fbPrKF_SDpVPsiV2WbmlMfwRA7C9T9_IyVmGnpaXa0A,29687
 sglang/lang/ir.py,sha256=5VVK2JnbspdysrhcGgkmp_JlAprd2XqqRnS_GfP_XWc,16645
 sglang/lang/tracer.py,sha256=borJmlSJOhg1RUndGRnilnR60eEZz2Y9aU7BpftsOxU,8287
 sglang/lang/backend/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -19,55 +19,56 @@ sglang/lang/backend/anthropic.py,sha256=EXRX7xJgA5KZszX7toSLVnKzFQ5EO0Loj-YjHFtx
 sglang/lang/backend/base_backend.py,sha256=APiMht4WYECLCOGRPCEUF6lX-an1vjVe2dWoMSgymWY,1831
 sglang/lang/backend/litellm.py,sha256=QsaLRh0KVyuaxRZGAvLOdCCSStIMs-V0XyMX0PR6y0w,2452
 sglang/lang/backend/openai.py,sha256=-ScfI2TFALB_FTYBur9ab0gNYxK1ogHkhdLxX19t6-Y,14808
-sglang/lang/backend/runtime_endpoint.py,sha256=TZ0NV89or5_3MIZZFnc1JXAAjnv7tCfeQmHDla8R0e0,9208
+sglang/lang/backend/runtime_endpoint.py,sha256=6iW1S62KmYyQGiWsHJFhZidK01vlIE55IsYN2tP38WQ,9202
 sglang/lang/backend/vertexai.py,sha256=O-iBLD-y3vq80UxnrAoJri7bxpgd-_eakZ88Cf8bEGA,4855
 sglang/srt/conversation.py,sha256=Il7JJuu4o42k2xdBWVfONNmstTsAM-4idX6AcEOnrXQ,15526
 sglang/srt/flush_cache.py,sha256=SJsbZnmDhH-gb9ch3hIwnI_nuwaOLlKvlXADyLBGENk,403
-sglang/srt/hf_transformers_utils.py,sha256=94mOI93B2xOmXKqfJfEoGxqHgwwlWNbPHgsA47AQJK8,11245
+sglang/srt/hf_transformers_utils.py,sha256=RnyxC1_OmOf-QzdPBziqAUOIQXyRzrb4RNlqFB1ArEc,11354
 sglang/srt/memory_pool.py,sha256=FhJk5GtYortO3MJIsMMQ-o49agwDHVX1aEQH2LITq6c,3949
 sglang/srt/mm_utils.py,sha256=OptgAHDX-73Bk4jAdr2BOAJtiEXJNzPrMhaM-dy275c,8889
-sglang/srt/model_config.py,sha256=lZu1D-XLVMETHS6FBMoPn8Uowa9QFGe95d3SuWrr2q8,5282
-sglang/srt/sampling_params.py,sha256=OI11asr1Bd_E5soDjih614v4flgWxdMZU9HAF0aBafQ,3062
-sglang/srt/server.py,sha256=IUed6vnXCx7-xbrpEMAaJZ_aa4UubPAQ5pXvcv-xNoY,14607
-sglang/srt/server_args.py,sha256=aF6L35mEB-FU3BL_ooKuCIcOXLhYLxA9-MjpaOTQRCo,13189
+sglang/srt/model_config.py,sha256=9VF7ET0CGKEY-zdiU7kGv8Cg7H_9Q1fmqtI3C0z22S0,5458
+sglang/srt/sampling_params.py,sha256=WjJ_sOhbJVMKIBH8gJWQKhzeK5Ipu9XRNV7soWnLtak,3122
+sglang/srt/server.py,sha256=IKSTgp6FJN6TE9anog47zh9GJYXoyMjEKBNXUZ89Cuk,14197
+sglang/srt/server_args.py,sha256=RfWoipSUURmv5NqT4L_YF9qJ6gOkZ8omRUFC_5fmgts,14043
 sglang/srt/utils.py,sha256=HvKkGbut8sOxMpGIzYsJ9NEZJg48LOnxyGESaGZmANs,22385
 sglang/srt/constrained/__init__.py,sha256=5LB3_mDTMW6wcRkFA5J2Rd5HPHHEKRyiELhe4gtlBYM,1472
 sglang/srt/constrained/base_cache.py,sha256=QQjmFEiT8jlOskJoZobhrDl2TKB-B4b1LPQo9JQCP_w,1405
-sglang/srt/constrained/fsm_cache.py,sha256=P4qNDHHxpKpTnYL_8V1R6OFXlUwbM6ZcBdzddpcBgb4,1135
+sglang/srt/constrained/fsm_cache.py,sha256=HlzFs9TXvMFmeZhTpXmJU3UNQ_Kix4Ir-SwpqXGhX8k,2061
 sglang/srt/constrained/jump_forward.py,sha256=s60jZ7Ue8zaodgQm7gDpN6pSedpvpUck_waJALUMj60,5615
 sglang/srt/layers/context_flashattention_nopad.py,sha256=7ps_9W_ia9zikL9HqsSUwWHyBVotywosE-dOiPtaGY8,4615
 sglang/srt/layers/extend_attention.py,sha256=aYAAL9HZJpaSASp-ulMvbmSmyMcqdYUsgVQC-Lbm7_U,12008
 sglang/srt/layers/fused_moe.py,sha256=uyrbCaIHioq3G00xQUrCo53hYDoHzk5rep3Eji3oQiQ,20258
 sglang/srt/layers/linear.py,sha256=qLwFkOiRAljzE7LkAkLRdcCdVMk-t7b56jEjwQAuYDM,33953
-sglang/srt/layers/logits_processor.py,sha256=KyRYANCiq9Cfu_VPjrIbSBAlqN_clcAgF3JrG9waU5k,9674
-sglang/srt/layers/radix_attention.py,sha256=A3J_wOlysjblFXHgehAqRHBQmpYAHLyUovyLFsrMJ7A,6386
-sglang/srt/layers/token_attention.py,sha256=EJ4gjbVLfshOZ_vr1iB-Eq8_B-4F26n_wPDj6e1Zrww,7386
+sglang/srt/layers/logits_processor.py,sha256=VjP6T582K64X0mfyPUkhcIEZxsqJNu6ziqR3V82N_jE,10118
+sglang/srt/layers/radix_attention.py,sha256=to6w0kIq6dtaOYJtqIZcqR3t1yf05qBH1LWnFlE-jEQ,6374
+sglang/srt/layers/token_attention.py,sha256=uBtk3I6KeFjBRKRuQoG5BEZtVJsX4p7UOtJoej6ILZI,7411
 sglang/srt/layers/quantization/__init__.py,sha256=PQFzdPpul98DvywBA6YMBOnrMjtHE1LMlMpJ7FM8J3I,1971
 sglang/srt/layers/quantization/fp8.py,sha256=jaqgRFnHC--IL8iqB6Qygi-KXYPYBKKqt_j4Rk55_h4,24946
 sglang/srt/managers/detokenizer_manager.py,sha256=8rN2cdMr61LWy07lingEqLnNy0W5Rebdn14IsTQ9PCs,5049
-sglang/srt/managers/io_struct.py,sha256=VHy9wdZ3sfZA7fS6iq8lqbxdHL5WkBZNqxpacyZ8_8c,5483
-sglang/srt/managers/tokenizer_manager.py,sha256=SbivhFhZUR9HU9pLTe93MlYprAFAHzOU3KMBA2piQUk,19308
-sglang/srt/managers/controller/cuda_graph_runner.py,sha256=0aRqA1_34oJ557Zn8PjpJecex5bBWJdnCmBlcDVvYO0,8509
-sglang/srt/managers/controller/infer_batch.py,sha256=SKwCwhnZ_CNlG0mVCEc4X0e4HNjJFke-c8zdWP3TzjQ,34186
+sglang/srt/managers/io_struct.py,sha256=WmBGrWR8R6X2zh2p1FkfPZtJzuGSlNW8cmIDm0EEqMA,5528
+sglang/srt/managers/tokenizer_manager.py,sha256=2it1o4dKd7nFzfZflOw1cT03gFktqC2sVPICbBSR4c0,19594
+sglang/srt/managers/controller/cuda_graph_runner.py,sha256=KEqX4Tc1yEWW52LzzFb4THb-guYIaft2pxxH8rWchSA,8808
+sglang/srt/managers/controller/infer_batch.py,sha256=3DixMdSW0odH5I6p7h8_xtRlHx4q76ArR6YZW8Gkqzg,35888
 sglang/srt/managers/controller/manager_multi.py,sha256=DT8Y9RF5OyTxlrLEZYz4claNWir3UrVztdOZaVPiA6g,6077
 sglang/srt/managers/controller/manager_single.py,sha256=2xO_iWK6tWvc0B31nKbe2N3klxwQBJmPTnFhNjzhVSI,4566
-sglang/srt/managers/controller/model_runner.py,sha256=4-nBd9_MgIlamjEdLZDepBEykYNR8nL-65Sf1EYsnx0,14371
+sglang/srt/managers/controller/model_runner.py,sha256=9o4xWnfI9-FJU6-S7WfEFlGMjWA2YesAhUKpuq8urhk,14854
 sglang/srt/managers/controller/radix_cache.py,sha256=tx8LEQpqLxipw9UUVj4D1YQLMMDmWnjDYv8oDlOl-co,8210
 sglang/srt/managers/controller/schedule_heuristic.py,sha256=SQAGzPS3aB_TPj7rnPBhewwyR6W1sVwW4D3zG3JUY00,2714
-sglang/srt/managers/controller/tp_worker.py,sha256=yjz-Xzl0zEy4QSU-EYneZH5vi3oHtBuXTtYe4VuDp2g,30517
+sglang/srt/managers/controller/tp_worker.py,sha256=VYhO3xcJrcDQwonGLWSWKHq4T7BvFmb6-L5LxY3-fhE,30607
 sglang/srt/model_loader/model_loader.py,sha256=VS8VQL5ITN3akZ9eU_-uHWMan1axLMNG2_O12HzGysA,10132
 sglang/srt/model_loader/utils.py,sha256=I2PS5HIH5Cg-p7xKO_Cw_foK2vQ61xVc3zQv7CbeGEw,10120
 sglang/srt/models/chatglm.py,sha256=pH8g2Dj8qQLGPYpWVTb-IONfXsdfmpWi0-IEYNdSi4s,13296
 sglang/srt/models/commandr.py,sha256=hHsNQWi0X8rNL7_gpcoUxQxdhxtvx5_RVx8u6cLzqYQ,13606
 sglang/srt/models/dbrx.py,sha256=rRxOusGPu670ommeqXg62AllwB1apzE4yZoWc1fcr2M,14095
 sglang/srt/models/deepseek.py,sha256=YtoPmv4fKmiH_jsRMSab9Wxq3aOZga9pCPGnkCs3Vvs,15457
+sglang/srt/models/deepseek_v2.py,sha256=1FqLe6tSENFpYgcEkmMr2-M4qksgne2glU3kZhSBB0Q,19527
 sglang/srt/models/gemma.py,sha256=DweoalfWYhLL-ZWLAO5gl4SCZflWmejVeDG3Vky_WNo,11719
 sglang/srt/models/gemma2.py,sha256=x3Dua-TVwRm5fJjo5UDekdoWqwt9xYbMuB-ogfXyiT8,15860
 sglang/srt/models/gpt_bigcode.py,sha256=XHO1naPdXfiKYQRQ6uZe1fN3PBDhKH3-bchsaaZvfE4,9637
 sglang/srt/models/grok.py,sha256=611zrlIchvFaVfztRdBY7z97oU3KB-anykbOZy1hK6M,27295
 sglang/srt/models/internlm2.py,sha256=8MNcwxU5Th9IxWa314HqqmbCRlPUFScnfneBDs0riIU,11659
 sglang/srt/models/llama2.py,sha256=OyAf_lun5aZEsT80WmrIYBF8QXTXRpW8sUlylr4AZIc,14204
-sglang/srt/models/llama_classification.py,sha256=foCPvNyP2bTZ0YcRBF-qkmBv-gT24lhLNCXP30Oq4VU,4370
+sglang/srt/models/llama_classification.py,sha256=Z2dvZAdOwCnN-lGFZRcwU0rNreE1gKwLefeWzEH36Uw,4366
 sglang/srt/models/llava.py,sha256=vBI6EEeOG_9o23Shi9h8k58rxTOHZnSKMmPl3B3Q3uc,17924
 sglang/srt/models/llavavid.py,sha256=SrNQ-U2wekHvP_up-ZXRkCSros2NzheHpPfXHrp0YBU,13050
 sglang/srt/models/minicpm.py,sha256=9uE8D-NopAj-sfaKJ7d-0x-PuCTEevQPoHPZvZlwstA,13277
@@ -79,14 +80,14 @@ sglang/srt/models/qwen2.py,sha256=87Tt1Bti-Py3AGudcf7k5ni-OHhtDKPj_Hke44YGw4U,11
 sglang/srt/models/qwen2_moe.py,sha256=oHNoo45myV5kitkls2GWVzuGt1Q4pRHN2nLlXEltFI8,17581
 sglang/srt/models/stablelm.py,sha256=Z_XCDSHY_QMz3lZwwkZdIZjEOizZjLYJU9GDi8o08qQ,10802
 sglang/srt/models/yivl.py,sha256=55KPrQ-dVplI0hh2WCSugjc1luE0J2UAafjZxu_7Xuc,4367
-sglang/srt/openai_api/adapter.py,sha256=A0IG9ZKEMkkYCsLrVEspnVWzZHBUbc1vHv747LrF8ew,15920
-sglang/srt/openai_api/protocol.py,sha256=j7ifIR2SFQxTwaHAd9ksM096vfffcNltzTH4sg7H0RA,5739
+sglang/srt/openai_api/adapter.py,sha256=DVZ2niAEOgE8GQdYnuvwjrGiFRkAu5YtOB-yxOlF_Eg,15868
+sglang/srt/openai_api/protocol.py,sha256=jTb22jv5caB7k7Ub2ltYEbTtDheZjwwWAAUdvjiLTR0,5741
 sglang/test/test_conversation.py,sha256=gF_AyOxQgpPQBPnA57-kq-M0p_zFu-rBDMFgAq655Rw,1596
 sglang/test/test_openai_protocol.py,sha256=DVx3r6hrb8oRqbo5AYIleldxbqMBTtb-gtORM6t_Y1c,1661
-sglang/test/test_programs.py,sha256=uefeHUFKT2NJESOujj-CsnPXdw1aQQN2TzUbPCHJjGs,13654
+sglang/test/test_programs.py,sha256=s4WGpTmYP4Yx5g8JYZpbkeF9RN5iUnlKdi8FGAZovTc,13756
 sglang/test/test_utils.py,sha256=kD_fQe3WroZ9Kc3NBRKPiZOFJ_JD2uEE9XIvPp6AD9Y,11048
-sglang-0.2.4.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
-sglang-0.2.4.dist-info/METADATA,sha256=Fr6JX7mNV5-CjsMH-r3sSwdEFOFXbJWqTjlXG0ox08s,31692
-sglang-0.2.4.dist-info/WHEEL,sha256=Wyh-_nZ0DJYolHNn1_hMa4lM7uDedD_RGVwbmTjyItk,91
-sglang-0.2.4.dist-info/top_level.txt,sha256=yxhh3pYQkcnA7v3Bg889C2jZhvtJdEincysO7PEB09M,7
-sglang-0.2.4.dist-info/RECORD,,
+sglang-0.2.6.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
+sglang-0.2.6.dist-info/METADATA,sha256=g_G_XHbWCNSY9F6RieXV43svnNzq1wonwrArNxX0VNA,32095
+sglang-0.2.6.dist-info/WHEEL,sha256=Wyh-_nZ0DJYolHNn1_hMa4lM7uDedD_RGVwbmTjyItk,91
+sglang-0.2.6.dist-info/top_level.txt,sha256=yxhh3pYQkcnA7v3Bg889C2jZhvtJdEincysO7PEB09M,7
+sglang-0.2.6.dist-info/RECORD,,