sglang 0.4.1.post3__py3-none-any.whl → 0.4.1.post4__py3-none-any.whl

sglang/srt/speculative/eagle_worker.py

@@ -0,0 +1,170 @@
+from typing import List, Optional, Union
+
+import torch
+
+from sglang.srt.layers.logits_processor import LogitsProcessorOutput
+from sglang.srt.managers.schedule_batch import Req, ScheduleBatch
+from sglang.srt.managers.tp_worker import TpModelWorker
+from sglang.srt.model_executor.forward_batch_info import (
+    CaptureHiddenMode,
+    ForwardBatch,
+    ForwardMode,
+)
+from sglang.srt.model_executor.model_runner import ModelRunner
+from sglang.srt.server_args import ServerArgs
+from sglang.srt.speculative.eagle_utils import EAGLEDraftInput
+
+
+class EAGLEWorker(TpModelWorker):
+
+    def __init__(
+        self,
+        server_args: ServerArgs,
+        gpu_id: int,
+        tp_rank: int,
+        dp_rank: Optional[int],
+        nccl_port: int,
+        target_worker: TpModelWorker,
+    ):
+        # Do not capture cuda graph in `super().__init__()`
+        # We will capture it later
+        backup_disable_cuda_graph = server_args.disable_cuda_graph
+        server_args.disable_cuda_graph = True
+        super().__init__(
+            gpu_id=gpu_id,
+            tp_rank=tp_rank,
+            server_args=server_args,
+            nccl_port=nccl_port,
+            dp_rank=dp_rank,
+            is_draft_worker=True,
+        )
+        self.target_worker = target_worker
+        self.server_args = server_args
+
+        # Share the embedding and lm_head
+        embed, head = self.target_worker.model_runner.model.get_embed_and_head()
+        self.model_runner.model.set_embed_and_head(embed, head)
+        self.model_runner.server_args.disable_cuda_graph = backup_disable_cuda_graph
+        self.model_runner.init_cuda_graphs()
+
+    def forward_draft_decode(self, batch: ScheduleBatch):
+        batch.spec_info.prepare_for_decode(batch)
+        model_worker_batch = batch.get_model_worker_batch()
+        forward_batch = ForwardBatch.init_new(model_worker_batch, self.model_runner)
+        forward_batch.spec_info.capture_hidden_mode = CaptureHiddenMode.LAST
+        logits_output = self.model_runner.forward(forward_batch)
+        self.capture_for_decode(logits_output, forward_batch)
+
+    def forward_draft_extend(self, batch: ScheduleBatch):
+        self._swap_mem_pool(batch, self.model_runner)
+        batch.spec_info.prepare_for_extend(batch)
+        model_worker_batch = batch.get_model_worker_batch()
+        forward_batch = ForwardBatch.init_new(model_worker_batch, self.model_runner)
+        forward_batch.spec_info.capture_hidden_mode = CaptureHiddenMode.LAST
+        logits_output = self.model_runner.forward(forward_batch)
+        self.capture_for_decode(logits_output, forward_batch)
+        self._swap_mem_pool(batch, self.target_worker.model_runner)
+
+    def forward_batch_speculative_generation(self, batch: ScheduleBatch):
+        if batch.forward_mode.is_decode():
+            prev_spec_info = batch.spec_info
+            self._swap_mem_pool(batch, self.model_runner)
+            for i in range(self.server_args.speculative_num_steps):
+                self.forward_draft_decode(batch)
+            batch.spec_info.clear_draft_cache(batch)
+            self._swap_mem_pool(batch, self.target_worker.model_runner)
+            (
+                next_draft_input,
+                logits_output,
+                verified_id,
+                self.finish_extend_len,
+                model_worker_batch,
+            ) = self.verify(batch)
+            next_draft_input.init(self.server_args)
+            batch.spec_info = next_draft_input
+            # if it is None, means all requsets are finished
+            if batch.spec_info.verified_id is not None:
+                self.forward_extend_after_decode(batch)
+            batch.spec_info = prev_spec_info
+            return logits_output, verified_id, model_worker_batch, next_draft_input
+
+        else:
+            spec_info = EAGLEDraftInput()
+            spec_info.init(self.server_args)
+            model_worker_batch = batch.get_model_worker_batch()
+            model_worker_batch.spec_info = spec_info
+            spec_info.capture_hidden_mode = CaptureHiddenMode.FULL
+            logits_output, next_token_ids = self.target_worker.forward_batch_generation(
+                model_worker_batch
+            )
+            model_worker_batch.spec_info.verified_id = next_token_ids
+            model_worker_batch.spec_info.hidden_states = logits_output.hidden_states
+            batch.spec_info = spec_info
+            self.forward_draft_extend(batch)
+            batch.spec_info = None
+            return logits_output, next_token_ids, model_worker_batch, spec_info
+
+    def verify(self, batch: ScheduleBatch):
+        verify_input = batch.spec_info.prepare_for_verify(batch)
+        batch.forward_mode = ForwardMode.TARGET_VERIFY
+        verify_input.prepare_for_verify(batch)
+        batch.spec_info = verify_input
+        batch.spec_info.capture_hidden_mode = CaptureHiddenMode.FULL
+        model_worker_batch = batch.get_model_worker_batch()
+        logits_output, _ = self.target_worker.forward_batch_generation(
+            model_worker_batch, skip_sample=True
+        )
+        verify_input.hidden_states = logits_output.hidden_states
+        res = verify_input.verify(batch, logits_output)
+        batch.forward_mode = ForwardMode.DECODE
+        return res + (model_worker_batch,)
+
+    def _swap_mem_pool(self, batch: ScheduleBatch, runner: ModelRunner):
+        batch.token_to_kv_pool = runner.token_to_kv_pool
+        batch.req_to_token_pool = runner.req_to_token_pool
+
+    def forward_extend_after_decode(self, batch: ScheduleBatch):
+        self._swap_mem_pool(batch, self.model_runner)
+        batch.forward_mode = ForwardMode.DRAFT_EXTEND
+        if batch.spec_info.has_finished:
+            index = batch.spec_info.unfinished_index
+            seq_lens = batch.seq_lens
+            batch.seq_lens = batch.seq_lens[index]
+        batch.spec_info.prepare_extend_after_decode(batch)
+        model_worker_batch = batch.get_model_worker_batch()
+        forward_batch = ForwardBatch.init_new(model_worker_batch, self.model_runner)
+        forward_batch.spec_info.capture_hidden_mode = CaptureHiddenMode.LAST
+        logits_output = self.model_runner.forward(forward_batch)
+        batch.spec_info.hidden_states = logits_output.hidden_states
+        self.capture_for_decode(logits_output, forward_batch)
+        batch.forward_mode = ForwardMode.DECODE
+        if batch.spec_info.has_finished:
+            batch.seq_lens = seq_lens
+        self._swap_mem_pool(batch, self.target_worker.model_runner)
+
+    def capture_for_decode(self, logits_output, forward_batch):
+        if isinstance(logits_output, LogitsProcessorOutput):
+            logits = logits_output.next_token_logits
+        sample_output = torch.softmax(
+            logits, dim=-1
+        )  # TODO: Support more sampling method @kavioyu
+        forward_batch.spec_info.capture_for_decode(
+            sample_output, logits_output.hidden_states, forward_batch.forward_mode
+        )
+
+    # Don't support prefix share now.
+    def finish_request(self, reqs: Union[Req, List[Req]]):
+        if not isinstance(reqs, List):
+            reqs = [reqs]
+        for req in reqs:
+            req_len = (
+                len(req.origin_input_ids)
+                + len(req.output_ids)
+                - self.finish_extend_len[req.rid]
+                - 1
+            )
+            kv_indices = self.model_runner.req_to_token_pool.req_to_token[
+                req.req_pool_idx
+            ][:req_len]
+            self.model_runner.token_to_kv_pool.free(kv_indices)
+            self.model_runner.req_to_token_pool.free(req.req_pool_idx)

sglang/srt/speculative/spec_info.py

@@ -2,8 +2,12 @@ from enum import IntEnum, auto
 
 
 class SpeculativeAlgorithm(IntEnum):
+    NONE = auto()
     EAGLE = auto()
 
+    def is_none(self):
+        return self == SpeculativeAlgorithm.NONE
+
     def is_eagle(self):
         return self == SpeculativeAlgorithm.EAGLE
 
@@ -11,6 +15,7 @@ class SpeculativeAlgorithm(IntEnum):
     def from_string(name: str):
         name_map = {
             "EAGLE": SpeculativeAlgorithm.EAGLE,
+            None: SpeculativeAlgorithm.NONE,
         }
         return name_map[name]
 

sglang/srt/utils.py

@@ -15,6 +15,7 @@
 
 import base64
 import dataclasses
+import io
 import ipaddress
 import itertools
 import json
@@ -34,6 +35,7 @@ import warnings
 from functools import lru_cache
 from importlib.metadata import PackageNotFoundError, version
 from io import BytesIO
+from multiprocessing.reduction import ForkingPickler
 from typing import Any, Callable, Dict, List, Optional, Protocol, Tuple, Union
 
 import numpy as np
@@ -60,7 +62,6 @@ from triton.runtime.cache import (
 
 logger = logging.getLogger(__name__)
 
-
 show_time_cost = False
 time_infos = {}
 
@@ -1206,7 +1207,6 @@ def _cuda_device_count_stateless(cuda_visible_devices: Optional[str] = None) ->
     # https://github.com/pytorch/pytorch/blob/
     # c1cd946818442aca8c7f812b16d187ce1586c3bc/
     # torch/cuda/__init__.py#L831C1-L831C17
-    import torch.cuda
     import torch.version
 
     if not torch.cuda._is_compiled():
@@ -1335,3 +1335,16 @@ def parse_tool_response(text, tools, **kwargs):
         for call_info in call_info_list
     ]
     return text, call_info_list
+
+
+class MultiprocessingSerializer:
+    @staticmethod
+    def serialize(obj):
+        buf = io.BytesIO()
+        ForkingPickler(buf).dump(obj)
+        buf.seek(0)
+        return buf.read()
+
+    @staticmethod
+    def deserialize(data):
+        return ForkingPickler.loads(data)

sglang/version.py

@@ -1 +1 @@
-__version__ = "0.4.1.post3"
+__version__ = "0.4.1.post4"

{sglang-0.4.1.post3.dist-info → sglang-0.4.1.post4.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: sglang
-Version: 0.4.1.post3
+Version: 0.4.1.post4
 Summary: SGLang is yet another fast serving framework for large language models and vision language models.
 License:                                  Apache License
                                    Version 2.0, January 2004
@@ -243,11 +243,11 @@ Requires-Dist: torch; extra == "srt"
 Requires-Dist: vllm<=0.6.4.post1,>=0.6.3.post1; extra == "srt"
 Requires-Dist: cuda-python; extra == "srt"
 Requires-Dist: flashinfer==0.1.6; extra == "srt"
-Requires-Dist: sgl-kernel>=0.0.2.post10; extra == "srt"
+Requires-Dist: sgl-kernel>=0.0.2.post11; extra == "srt"
 Provides-Extra: srt-hip
 Requires-Dist: sglang[runtime_common]; extra == "srt-hip"
 Requires-Dist: torch; extra == "srt-hip"
-Requires-Dist: vllm==0.6.3.dev13; extra == "srt-hip"
+Requires-Dist: vllm==0.6.3.post2.dev1; extra == "srt-hip"
 Provides-Extra: srt-xpu
 Requires-Dist: sglang[runtime_common]; extra == "srt-xpu"
 Provides-Extra: srt-hpu
@@ -315,7 +315,7 @@ Requires-Dist: sglang[test]; extra == "dev-hpu"
 
 | [**Blog**](https://lmsys.org/blog/2024-07-25-sglang-llama3/)
 | [**Documentation**](https://sgl-project.github.io/)
-| [**Join Slack**](https://join.slack.com/t/sgl-fru7574/shared_invite/zt-2tmmp6flg-89dOlJW2TjnBrTRk1I_~GA)
+| [**Join Slack**](https://join.slack.com/t/sgl-fru7574/shared_invite/zt-2um0ad92q-LkU19KQTxCGzlCgRiOiQEw)
 | [**Join Bi-Weekly Development Meeting**](https://docs.google.com/document/d/1xEow4eIM152xNcRxqZz9VEcOiTQo8-CEuuQ5qTmkt-E/edit?usp=sharing)
 | [**Slides**](https://github.com/sgl-project/sgl-learning-materials?tab=readme-ov-file#slides) |
 
@@ -347,9 +347,10 @@ The core features include:
 
 ## Getting Started
 - [Install SGLang](https://sgl-project.github.io/start/install.html)
-- [Send requests](https://sgl-project.github.io/start/send_request.html)
-- [Backend: SGLang Runtime (SRT)](https://sgl-project.github.io/backend/backend.html)
-- [Frontend: Structured Generation Language (SGLang)](https://sgl-project.github.io/frontend/frontend.html)
+- [Quick Start](https://sgl-project.github.io/start/send_request.html)
+- [Backend Tutorial](https://sgl-project.github.io/backend/openai_api_completions.html)
+- [Frontend Tutorial](https://sgl-project.github.io/frontend/frontend.html)
+- [Contribution Guide](https://sgl-project.github.io/references/contribution_guide.html)
 
 ## Benchmark and Performance
 Learn more in our release blogs: [v0.2 blog](https://lmsys.org/blog/2024-07-25-sglang-llama3/), [v0.3 blog](https://lmsys.org/blog/2024-09-04-sglang-v0-3/), [v0.4 blog](https://lmsys.org/blog/2024-12-04-sglang-v0-4/)
@@ -361,5 +362,5 @@ Learn more in our release blogs: [v0.2 blog](https://lmsys.org/blog/2024-07-25-s
 The project is supported by (alphabetically): AMD, Baseten, DataCrunch, Etched, Hyperbolic, Jam & Tea Studios, LinkedIn, LMSYS.org, Meituan, NVIDIA, RunPod, Stanford, UC Berkeley, UCLA, xAI, 01.AI.
 
 ## Acknowledgment and Citation
-We learned from the design and reused code from the following projects: [Guidance](https://github.com/guidance-ai/guidance), [vLLM](https://github.com/vllm-project/vllm), [LightLLM](https://github.com/ModelTC/lightllm), [FlashInfer](https://github.com/flashinfer-ai/flashinfer), [Outlines](https://github.com/outlines-dev/outlines), and [LMQL](https://github.com/eth-sri/lmql).
+We learned the design and reused code from the following projects: [Guidance](https://github.com/guidance-ai/guidance), [vLLM](https://github.com/vllm-project/vllm), [LightLLM](https://github.com/ModelTC/lightllm), [FlashInfer](https://github.com/flashinfer-ai/flashinfer), [Outlines](https://github.com/outlines-dev/outlines), and [LMQL](https://github.com/eth-sri/lmql).
 Please cite the paper, [SGLang: Efficient Execution of Structured Language Model Programs](https://arxiv.org/abs/2312.07104), if you find the project useful.

{sglang-0.4.1.post3.dist-info → sglang-0.4.1.post4.dist-info}/RECORD

@@ -2,7 +2,7 @@ sglang/__init__.py,sha256=b2oIdWzp5P8SzieeOs2TzJoN3Do3tfJbV8gZS_imVcs,1619
 sglang/api.py,sha256=NdO6cYnklnEBQBKqQjlqI8-P1EownKQ71t5ibCGhEVo,6953
 sglang/bench_latency.py,sha256=oZjSAzX7dUiSu-zdz0dkyUPo-qAX_lsXFH1gf03akgI,76
 sglang/bench_offline_throughput.py,sha256=r-uBvpnx-30mAnVwQB4WlqiXxy2fn5a1NUARwZcaIo4,12533
-sglang/bench_one_batch.py,sha256=jkyMhK0lqn5dRCYgAh30qZrNHP4gAbXODymBMNXK86I,15859
+sglang/bench_one_batch.py,sha256=uw__0H3e3lY_6EDz4IAZUoYxq9kQIOPbbcyguYxttSA,15975
 sglang/bench_one_batch_server.py,sha256=-fV9FTLNNcSIy0pgYeggXedPVK0fVsXZqVQswT8OMOY,5945
 sglang/bench_serving.py,sha256=YQiCZreejCPBTqMmZsCB99RMi1N-Jx-dZtaafcQ8-14,53377
 sglang/check_env.py,sha256=4OqpZaEJOfBM6-vtPILto5kqDmgiZM1Koc7lK78A7CI,8427
@@ -11,7 +11,7 @@ sglang/launch_server.py,sha256=4y2QeSj0wVNB9MJQZeahD4ahTDU6gwqo7MPUytyFop0,403
 sglang/launch_server_llavavid.py,sha256=tGc17S1vUfLwbi1GB26oOdXxTWr7gjlqpTrPnrMRNO8,1007
 sglang/llama3_eval.py,sha256=gWSboDchIGybIce88bJlrCG0yiLZ513mw4gcutJlzGM,10017
 sglang/utils.py,sha256=23jf4Mz8E5p5a6JOkjnfYZixdjZUk88F_mZ8rZcby5Q,11597
-sglang/version.py,sha256=FT2VkJCvJQmaJgb_t19PhogLhJaJvLV2NK6x3Bt9CeQ,28
+sglang/version.py,sha256=efEbFOIgkOX__fKbqiqjj6UK2e0KofwnPDZo0VFdehs,28
 sglang/lang/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 sglang/lang/chat_template.py,sha256=cnfjjxIIcYRGRxXlJlOGnpFxFuhMHut7DS52LsOMKcA,15826
 sglang/lang/choices.py,sha256=-W1DVw9N9ZliVpvmWrzIXG4cswAah8eMQrHWzkS3D8o,6234
@@ -32,9 +32,9 @@ sglang/srt/conversation.py,sha256=u9zFU8aMYzwHUbQRKU76B_T-jfLlPoxUcWG_nRbDM2I,21
 sglang/srt/hf_transformers_utils.py,sha256=38Ms0H2-VMerOS6jnczcFtZMS6lhw9B5rSWKAfxVUfQ,7945
 sglang/srt/mm_utils.py,sha256=1ScBunw_x4W8ebM_AcJ62-1T2mfT8NlMJqdAhkF1lb0,12367
 sglang/srt/model_parallel.py,sha256=eLXZhvJ4wG6dh0FontNCIdVZvHYdWgaeY-5cu7TD9tE,6078
-sglang/srt/server.py,sha256=sDERAZlRa6OTaUk-SfW5aKJbPui1COpPG34HDlMHMNc,34916
-sglang/srt/server_args.py,sha256=wkafWcLHqm-p52psy75WYvTZ1-fIlkfB0YUr_g-vgjY,34210
-sglang/srt/utils.py,sha256=i8MjcaSQjPPfPZ0txufTtqLr4Q7YhHQ86L1i9j-y5yY,44131
+sglang/srt/server.py,sha256=2HPaIwN8-KijB44ADUnpWD4A2mTKwW9CPl4NJBu9yzE,35068
+sglang/srt/server_args.py,sha256=oIwBCTwSoj6qyKJ5zD38dPfgdXcNol64wV7l6UjmJNo,36208
+sglang/srt/utils.py,sha256=Xn5Zf_HzRVSsGF6_lvd85hS-0I2MdQXbe0Yo46WsTOc,44447
 sglang/srt/configs/__init__.py,sha256=_usVIXHQjft4PAJ1Y-yGQOn2QNOv501GYMlQwpGXbns,208
 sglang/srt/configs/device_config.py,sha256=dResqHjkg_dq10v6rnVpbXpvABZRB0jylOm-2_JAnx0,428
 sglang/srt/configs/exaone.py,sha256=Duxd4yQoKy8GWEzZD_kCY_OzmN_67CTJL_Kgn0eXk3g,10731
@@ -63,19 +63,19 @@ sglang/srt/layers/activation.py,sha256=EboMjT9HV2tNHQ6rzpojtlkzev1lAFbhQlxMg9hwx
 sglang/srt/layers/custom_op_util.py,sha256=0vu-yX2wwonmO1L_o5G7SA6C-8XuhDIh9rPDvNeLhoc,922
 sglang/srt/layers/layernorm.py,sha256=nRQ1w1xSUcU-zlqVC61BnGG6otS5W1w9VaSzeXizrx4,4037
 sglang/srt/layers/linear.py,sha256=KyRFU0VcoNuN-hnQB9QQcBN9NCpeqPtLzzufIHUpV6w,47064
-sglang/srt/layers/logits_processor.py,sha256=Imh-qY1D9J80DZVSVV0LfTiHMEw6oQ3JbY9lXxPZAXE,15656
+sglang/srt/layers/logits_processor.py,sha256=Yd7GisSfEgSq3cLMzz5lYiB5Cv-YgE4AMmVcACMBBZ4,12991
 sglang/srt/layers/pooler.py,sha256=rj2lygvleBnyLCBZ8I11HGMgpfIDsT0l3PIkshJwdu4,1606
 sglang/srt/layers/radix_attention.py,sha256=E4cmvkcCdCtb6VyLNrCKy1D6VwHQ063oH3JQXPaRy6w,2178
 sglang/srt/layers/rotary_embedding.py,sha256=29tx3JNR40AoXqBa2cFGBjva9vU2xgFipETlpMaaZas,3985
-sglang/srt/layers/sampler.py,sha256=k4Op_HMkQfT7t9wgQwBVotfTUXEocrzRyQqEFnff1pc,5511
-sglang/srt/layers/torchao_utils.py,sha256=dQVuWNXxAvOPjr2G5BBMWqC2oKcS2B52rx-fEc_elmc,3545
+sglang/srt/layers/sampler.py,sha256=HQWi1zb1gmD9pHMQyEP3WPjnL8vy-ncZDVMENbjQW7c,6944
+sglang/srt/layers/torchao_utils.py,sha256=8c2vzt106iP_QKbJtfN1GuABW8nCuP5dElQLUeci6qg,3934
 sglang/srt/layers/vocab_parallel_embedding.py,sha256=slGwLiWjuFLCUdRe-GTlfumyZpqVX9VF6No_UGOT-hA,21624
-sglang/srt/layers/attention/__init__.py,sha256=lNLfWqePc5NMej-AcXl97vxVXsxQOgP7dNNb2ibyUWI,2562
-sglang/srt/layers/attention/double_sparsity_backend.py,sha256=RQdEKRykSLf9ilnaHmR6T7RFqh4emH_adfB3aJN2BUU,10920
-sglang/srt/layers/attention/flashinfer_backend.py,sha256=8nH4EIEXvNk9yZVl7mSn78w5Dli5UiWL-ZCeYykG9HI,27280
-sglang/srt/layers/attention/torch_native_backend.py,sha256=nQdeqWEMMH_wrod5wssDCJG-uPKm0uslvkALKqPRPQ8,10509
-sglang/srt/layers/attention/triton_backend.py,sha256=-TobyZHwlbJ5HhbFg-jgCqVOw4Y-opgEuFo-EusASQc,6264
-sglang/srt/layers/attention/triton_ops/decode_attention.py,sha256=oJ_UK1t229zF3hbTDiQe7t-X-IbM2dOxx4U2ch-vmjA,17847
+sglang/srt/layers/attention/__init__.py,sha256=GUoygIsXzDFt9I1w9p0GO7leaDK6l2J3cBRQfpp4oDI,2869
+sglang/srt/layers/attention/double_sparsity_backend.py,sha256=QEDF8tQKMkh-nbt4jHKHZhhgHuV0Fla_BPzzoo9JfT4,9231
+sglang/srt/layers/attention/flashinfer_backend.py,sha256=I8b3Dq1O9PijLN40lEK0Gjj8GNS46WF4K-QVOtVccTg,33218
+sglang/srt/layers/attention/torch_native_backend.py,sha256=KrcAqTLVZLtwgOmB0xhwUUsX32M-5LYZpNxaRNT4VuA,9252
+sglang/srt/layers/attention/triton_backend.py,sha256=44ScKsVs-rFvqsaAZG_mREEpczhGaUBvaflvWqrukVE,6743
+sglang/srt/layers/attention/triton_ops/decode_attention.py,sha256=ltWcZ00ugpglSYvszpGb-UCpGIixdG25cWtSrOOOMik,17943
 sglang/srt/layers/attention/triton_ops/double_sparsity_attention.py,sha256=1pSXfY3EEaM7iRN_uElHnAfsrJMhTFbu9fj8Z0O2PbE,21480
 sglang/srt/layers/attention/triton_ops/extend_attention.py,sha256=DWOZXSTVN5ZbcFjDjcqs-nPdUkxSwum0SVXhVKqwh2g,11688
 sglang/srt/layers/attention/triton_ops/prefill_attention.py,sha256=lojFXRZMLWkzS2Y8uxaolnQhXaWKG19mCAWaF5KQeiI,6087
@@ -85,8 +85,8 @@ sglang/srt/layers/moe/ep_moe/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NM
 sglang/srt/layers/moe/ep_moe/kernels.py,sha256=wb_S2qLxoWWgQu9coXy0XLNGvHzdZSdwXr0PGy4QySg,10940
 sglang/srt/layers/moe/ep_moe/layer.py,sha256=6iQU5ZjQ8IXGoQ8ZlBuJqyQxYTEem9vXI6rbVIWKlZw,22303
 sglang/srt/layers/moe/fused_moe_triton/__init__.py,sha256=h9yMFAL_bagUf-qBED8gSWdCOb7d8IdA-pE-L_nIg8E,842
-sglang/srt/layers/moe/fused_moe_triton/fused_moe.py,sha256=KvOy544x_4nRqg50o5YHQpHvF8TUD7q9LXDAWPGJlAA,31796
-sglang/srt/layers/moe/fused_moe_triton/layer.py,sha256=BclDj5JyCy-8Bfue4broL1-IG6a4dUyggE9WQLa06sg,20575
+sglang/srt/layers/moe/fused_moe_triton/fused_moe.py,sha256=JjSn3fNTHgXFDxcAINZUYAttbYOxl9VSIF305NT73Wg,36255
+sglang/srt/layers/moe/fused_moe_triton/layer.py,sha256=MZF6BHJVjduz-XerTrHvCP3qSZ3NW0pUK2p8zNwDuac,20798
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=1,N=14336,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json",sha256=iNGsE2ZeVnQEnN4A8UJ9Jv0d3hbRF2MJ9oBgjup5Szk,2737
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=1,N=14336,device_name=NVIDIA_A100-SXM4-80GB.json",sha256=JJN0hryyLr5Zv3dSS7C8cPFhAwTT6XxUVnBGMZvV6JA,2752
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=1,N=1792,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json",sha256=ouRyZ5PEMPP2njPftCNhs-1g1y6wueWLmhI7G1SjV1k,4131
@@ -123,43 +123,64 @@ sglang/srt/layers/moe/fused_moe_triton/layer.py,sha256=BclDj5JyCy-8Bfue4broL1-IG
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=1280,device_name=NVIDIA_A800-SXM4-80GB.json",sha256=uv-RwTNZT2n264dLo4eWxUpB3g7QqUyf2MFEGiRvoqQ,3251
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=1280,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json",sha256=XbCRIOmiNqVKh89p-0UxvvspINRDA1iV83f9l5yORwA,3254
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=1280,device_name=NVIDIA_H100_80GB_HBM3.json",sha256=4uQnjGPWokscrxiXDIvexOA8OkK5vkoIulmvvMFIEog,3250
+"sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=1280,device_name=NVIDIA_H200,dtype=fp8_w8a8.json",sha256=i5AXj26mWI-xEoOnLCZDXUzz8jk2RjDcGuaiT1QYSbY,3263
+"sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=1280,device_name=NVIDIA_H200.json",sha256=fJKk5LEM_LSnq1yc3ekLqAfbUWzPojQA6yX3XgSFo-o,3254
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=2560,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json",sha256=iJVlnCYTNDMb6U1UnV46ZuL_8LcpOv_XFaYWIeRFeNA,3263
+"sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=2560,device_name=NVIDIA_H200,dtype=fp8_w8a8.json",sha256=KfvSYCMG48vnRb35d9WOxYyZulI-RBrUGXUHQxXi4hk,3264
+"sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=2560,device_name=NVIDIA_H200.json",sha256=60yuOluuk6q88Ze0toPJB8hzMBvF7ZWyMZpriMdQf3g,3252
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=320,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json",sha256=DA4PrCu_BNLSWWVTwOicNfbyqUNW7BTZC2dyFz9DVbU,3265
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=320,device_name=NVIDIA_H100_80GB_HBM3.json",sha256=prj-QxdfS9Ns5WRPvahY_Tr7CyqlaVgNHPT89SS5zzg,3239
+"sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=320,device_name=NVIDIA_H200,dtype=fp8_w8a8.json",sha256=AUgoHK1PmAFehSNmsbxunlBdzM50Q5nFvdnG9FSOjOw,3265
+"sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=320,device_name=NVIDIA_H200.json",sha256=pC5fdtEFc5aVNzpj_REHhz1QPrGvgI9iQCvlodDP7J8,3244
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=640,device_name=NVIDIA_A100-SXM4-80GB.json",sha256=oxOKFDrgmw1YmgxTtRa1uoe3p09ylTLrkj_jOTqNh1Q,3249
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=640,device_name=NVIDIA_A800-SXM4-80GB.json",sha256=bfr70r6PmM95w7raabQOaOOSPiwU2OQCOZh-kKXIehY,3248
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=640,device_name=NVIDIA_GeForce_RTX_4090,dtype=fp8_w8a8.json",sha256=4g9lABUJnB-iVwXfYPYcI05XFPG4jY8o0yJUK7kSPZM,3253
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=640,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json",sha256=RGLqrAGvPCFZ0jMPBCJ0TqsnrSdW-EbUaSZu61cWGN8,3265
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=640,device_name=NVIDIA_H100_80GB_HBM3.json",sha256=sjEVPVTgUAlp4s8tZLGSyeNzbW6zTtUm2IioH3nZsIg,3254
+"sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=640,device_name=NVIDIA_H200,dtype=fp8_w8a8.json",sha256=eD3Y9sOwHFcgVdOfya8KxPhvLx_b4whfEWm4d8Y2HW8,3268
+"sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=640,device_name=NVIDIA_H200.json",sha256=KIfpZydSl31FOEqq0EBfxTyWRj1QTDwTjkPHFjNO3_A,3253
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=14336,device_name=AMD_Instinct_MI300X.json",sha256=OnadAdmDbX17Ni9VPrNXYSsxYhbtBeniCxxhhb0UmUk,4733
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=14336,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json",sha256=V_sgDtEtGEuBsGVa0maYJHhhGqe1NE7l-1ek2ed9WP8,3082
+"sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=14336,device_name=NVIDIA_H200,dtype=fp8_w8a8.json",sha256=LD4Z5MRR5Ivi4bYB5hMgymtvmFyVJwq6gmehA7fzecc,3271
+"sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=14336,device_name=NVIDIA_H200.json",sha256=GLIH4egg-pE-NWU5XqKuJCoRXciHN6GSc3NaE4PaeYg,3261
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=1792,device_name=AMD_Instinct_MI300X.json",sha256=bKsYVJm-IvWHWpxUG-lMPkyNz0nQpDb4UEIv895c9JI,4730
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=1792,device_name=NVIDIA_A100-SXM4-40GB.json",sha256=AffDc0_51ML8HiA3757zbD10TZJdUsUDIYIqO4g0yUw,3250
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=1792,device_name=NVIDIA_A100-SXM4-80GB.json",sha256=IEYBNjt9HGnzoOVSWvL0A0jUqq926QD0_BvVYR4RA1Y,3252
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=1792,device_name=NVIDIA_H100_80GB_HBM3.json",sha256=Ns9Y12aZbJnFhcG3nwb67bDqqiQAo9tdTAIe8K2Ajz4,3255
+"sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=1792,device_name=NVIDIA_H200,dtype=fp8_w8a8.json",sha256=obNtHlqs6l6EBqGm0e0TD2wR9TYoQV_N9Y7om847WJk,3268
+"sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=1792,device_name=NVIDIA_H200.json",sha256=juxJYeWYFHeLb-83_IDgrHEpoeSEursjXD43mTHBdLE,3246
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=2048,device_name=NVIDIA_A100-SXM4-80GB.json",sha256=HOxWmCI2ifHmWc0or2y8nEen86jDeLDov1-tuMzuhxo,3256
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=2048,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json",sha256=csHezh0HGWaNwrblGzMgcE95hqbqjWS8HImLRJYr_ts,3266
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=2048,device_name=NVIDIA_H100_80GB_HBM3.json",sha256=_5weLBinQCDzyV75hHKIT95Y0ce94KWft2_5BC6EkbQ,3254
+"sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=2048,device_name=NVIDIA_H200,dtype=fp8_w8a8.json",sha256=4O4VeMpgFNrqWyWqWgYgcYAgBQnOlAXvt26CRSXK-sY,3270
+"sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=2048,device_name=NVIDIA_H200.json",sha256=qfjbXqbl902TuiyzzomUy2sMvs-Dud8ZphDRY5WIPBM,3260
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=3584,device_name=AMD_Instinct_MI300X.json",sha256=_bw1_oads8tz51i4RVQUAjNi8r3b2Q2jPbi50TLFzlY,4732
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=3584,device_name=NVIDIA_A100-SXM4-40GB.json",sha256=Ru460ZgnUP4U8OsJfwF8n-AI-gfcolNR3_qzoxG6DtY,3254
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=3584,device_name=NVIDIA_A100-SXM4-80GB.json",sha256=K6BGrKw_oHTAtHjsZldcjp-BUM1dIecKXrrRn9OpRGs,3254
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=3584,device_name=NVIDIA_GeForce_RTX_4090,dtype=fp8_w8a8.json",sha256=4Q_-yITMfijOMoguUM2n96clARh-DUFsS-4oW_a3Jpc,3252
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=3584,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json",sha256=qqFoMaObuO8pFWcSb9q0wYsdC4eSCO7B-_ruQhR1N9M,3264
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=3584,device_name=NVIDIA_H100_80GB_HBM3.json",sha256=-5nkLIunjG1ghPoUEtt2AXEQw9oGiilP7K3UvQv9CqE,3252
+"sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=3584,device_name=NVIDIA_H200,dtype=fp8_w8a8.json",sha256=BXjSVGdvgP_-7xTvbHOO6ZrXWe0qSXiQChxoHGgWL7o,3263
+"sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=3584,device_name=NVIDIA_H200.json",sha256=Pi2coJlJlpgqXiPRd77B_eCmmi7sCdBuoSGK1RA5YO8,3258
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=3584,device_name=NVIDIA_L40S.json",sha256=p2qlRhTt7owWB8keEmoCrPZpo39IAxsKnULFQ7R38SI,3873
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=4096,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8.json",sha256=AKIX43JVc26ERb862pNOMEfGhsgyk1OGa42EptAfG1s,4409
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=4096,device_name=NVIDIA_A100-SXM4-80GB.json",sha256=DxYu8regZOSFu8ugFGA_QbwWK4g8xwQUZF9a_nNY4Cs,3255
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=4096,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json",sha256=obzfE_9XgsbFNfC9biYOHxR-V_Bgc7PKT8qZZJaiJJc,3262
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=4096,device_name=NVIDIA_H100_80GB_HBM3.json",sha256=qwKy8oaMsd3QrXgQbM_x9xcfYiHK_Ou1CEwDPL5Gbgo,3259
+"sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=4096,device_name=NVIDIA_H200,dtype=fp8_w8a8.json",sha256=rR8b-OuQ3watb8b2zuNlxKDSZpzlAagm9nb-FdKkt7s,3270
+"sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=4096,device_name=NVIDIA_H200.json",sha256=8G_QqV_DhvZ6xSavMSpeE6qcXPVpsVjEtJabydybKqY,3263
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=7168,device_name=AMD_Instinct_MI300X.json",sha256=54KpHTMGt_zDQHqbdopuVHPpiI44ZsN_5LBUBZ_woY4,4733
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=7168,device_name=NVIDIA_A100-SXM4-80GB.json",sha256=BAJnXTZoewwCtzJLUPJ0oYuALv640MvDuLseGcsYaaw,3252
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json",sha256=-Tj7ImS6ZFDof_0VTyq7kVm8XD9B54RD6CUOPSf3Jjg,3265
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=7168,device_name=NVIDIA_H100_80GB_HBM3.json",sha256=tme0ydWzIxdABZLk4tU8G_X2dJUYGGZNkQzNGcmcvUc,3261
+"sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8.json",sha256=3YQakSmUKhpw1KO7Hn-tEc-yyD1fEj01_6JlSYnrrlI,3274
+"sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=7168,device_name=NVIDIA_H200.json",sha256=W2ka_U8pzwjzX62NEGKXR32uuSR_zfHD1XjXYf5bgBs,3262
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=8192,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8.json",sha256=aMP7oZmh8BZnPOrl0MFibcdhTn3VmOSjqoKoK2rMSbU,4323
 "sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=8192,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json",sha256=sY2nWMPh9lsIkhPCjkHO245wpnfFbrHmzdcZDVFPVww,3265
+"sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=8192,device_name=NVIDIA_H200,dtype=fp8_w8a8.json",sha256=Uz5X80VcNBOaxshwVNUEittHk2zqB4HQCfTJ4TPG5aM,3274
 sglang/srt/layers/quantization/__init__.py,sha256=VPYXShHvbvkOgVBlkIqic4RhdJ1y6EZ3r34T-nZMT1k,4606
 sglang/srt/layers/quantization/base_config.py,sha256=daK9p0aijMszLUm1W4Pc33FK87MdqYK1NoWFKif-j80,4599
-sglang/srt/layers/quantization/fp8.py,sha256=k4mw-iKxlaEWRkGgaoxCLzZ_dYydyRj0y1N1B_umMwU,32668
+sglang/srt/layers/quantization/fp8.py,sha256=ypIb8wUN18trzMhot8QKUj9sSdCXgPC1i2Qi-ESToWw,32670
 sglang/srt/layers/quantization/fp8_kernel.py,sha256=cYF4ckqrUyhCO9Ha7zi05R8EhRaqSa8rFpYisz-9Ed0,10743
 sglang/srt/layers/quantization/fp8_utils.py,sha256=qBVJXxbxqmf8-Juq0t-IXWjlaZoePJqFNYcs9-oT5Yo,4150
 "sglang/srt/layers/quantization/configs/N=1536,K=1536,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json",sha256=tkLjwLC_aVXhzuvo-2QHkojXZauPJsf3jNHFn1S7uRA,3244
@@ -206,14 +227,14 @@ sglang/srt/lora/lora_manager.py,sha256=DHiqdl0_4wQ5PxZBZtlCpP14515mDV2_H9tzL3Rds
 sglang/srt/managers/data_parallel_controller.py,sha256=psI4FAuBGjtdnEuwagnGdtRqvqSSxOROfNKQqVDqlVA,8382
 sglang/srt/managers/detokenizer_manager.py,sha256=XvyxUhY_SNXlAcVsx9zczrGllpEMzj7p2Vbh6M_yHy8,8555
 sglang/srt/managers/image_processor.py,sha256=Y8RgyrzbJjJTpjbnZDa5qiiG5wWjZ68rOXUPDi6kkFo,13698
-sglang/srt/managers/io_struct.py,sha256=TUCHR9J0BGoN-ybFgINehHqK-x3BGjeu6202AYh-c0k,16166
-sglang/srt/managers/schedule_batch.py,sha256=KnoVuWgINnyard-BOXCo0jm3IMdXN9wIwnftMKcag-s,46097
-sglang/srt/managers/schedule_policy.py,sha256=QxjQ8-le062AMHHxool6CxkhvB4FIwhOQPzTX_JwL6U,15447
-sglang/srt/managers/scheduler.py,sha256=RFQfgP6Wy0DEzksNF7MkevT4ZOLFx_RYxyBlGG37eXE,63873
+sglang/srt/managers/io_struct.py,sha256=02NMBHRCjs9TUSdhKJmvMp3HculMC-50SkCGOEaYEHg,16197
+sglang/srt/managers/schedule_batch.py,sha256=WlJstk0MP4QA434zu6_VZ4FhDByEPjSggFDT8em4GvQ,46851
+sglang/srt/managers/schedule_policy.py,sha256=aHkIL9pZtc4Kdmy8XU9tsjaDzdChVN2dnGKvJkSyqFg,17965
+sglang/srt/managers/scheduler.py,sha256=sTnZDLN8gLLHkiQW4UvMZsdKBMkgKfUk-rAHUZ9zNR0,65272
 sglang/srt/managers/session_controller.py,sha256=3laMRIXEYWDjfytCjPs0vw_Tw__k-nKBY-bYzycYbfc,5482
-sglang/srt/managers/tokenizer_manager.py,sha256=9umteGE5dsE9TE7nIt5Ju_qRCZsXpfv_atkhzFC-8s4,33368
-sglang/srt/managers/tp_worker.py,sha256=8RVBLQaS3TnX7Z4J35RVrFN0M6PVnRBhct3sczBL4dY,7644
-sglang/srt/managers/tp_worker_overlap_thread.py,sha256=JQfrVPeE56ZGJ3nozkhZR-RSb2oePsY7iuedM7XCtdQ,9157
+sglang/srt/managers/tokenizer_manager.py,sha256=Xryex_dgdZzRmPtSe16WUz1p9IlGysjVWmocaHjOJz4,33686
+sglang/srt/managers/tp_worker.py,sha256=-bvUFCo544QQSEHqPPjeOvCWMEFn01Bva6AeO39Qe3o,8043
+sglang/srt/managers/tp_worker_overlap_thread.py,sha256=rdHz2thdGSmceDedrolHOqjNPhrralyDTuNREL56oNI,9067
 sglang/srt/mem_cache/base_prefix_cache.py,sha256=QC8HS8RC5DXu14kyXsxAgEUsn0f932p2DjqzbKjc6Bs,962
 sglang/srt/mem_cache/chunk_cache.py,sha256=R2gHAuqKd5ayQW3NnsgoGUH31---Z5izCDyCqLL0FjQ,2524
 sglang/srt/mem_cache/flush_cache.py,sha256=GYcxmNXh4hsMpFfNOuCTpKilW7guZwTtAg_usVeM3J0,979
@@ -221,9 +242,9 @@ sglang/srt/mem_cache/memory_pool.py,sha256=oxk3UtiiFA3_1iIP6eFsk8HIcRI_8Z1-FE2KO
 sglang/srt/mem_cache/radix_cache.py,sha256=c5voySV5L855c0G9cBEc9iQ4nR7PDDmg0V6fWWJHcq4,10945
 sglang/srt/metrics/collector.py,sha256=ZWoFx_FKN0sNMSZ8RJWUVQ0RFEYhIHxdw0d4TZTluMU,6861
 sglang/srt/metrics/func_timer.py,sha256=VFyNRrbnKVCwnQsrlLin1lITJfjQpf9m8sGPqL5LIsQ,3438
-sglang/srt/model_executor/cuda_graph_runner.py,sha256=1n5WxoE9-0B3unwkkcR355K_D290h2LGt_7EvH02DQM,16246
-sglang/srt/model_executor/forward_batch_info.py,sha256=vqF8XrHQPk3ZL7HqPvvkfP53oqBx0Fajb5lAIkdifBo,13961
-sglang/srt/model_executor/model_runner.py,sha256=TjvAwwr7EqZdmE-5HbuQMeEa0e0FqY6LeqqzEAHXMPU,30012
+sglang/srt/model_executor/cuda_graph_runner.py,sha256=PFK4aRu8ffOBQw0zU_yYofUK_poi5C1vCc3ePixj0JY,18243
+sglang/srt/model_executor/forward_batch_info.py,sha256=Z4VrcyQt3f4jPYvvm1vimMx3hSNNeUEONAFvSziy6N0,14788
+sglang/srt/model_executor/model_runner.py,sha256=MZv0CNevcZ85L2JYVc1MnXRKbvUWoSuCFcShES07YN4,29889
 sglang/srt/model_loader/__init__.py,sha256=zGZkOBz1zx-pkaIy47BasL3fjDlAcxAXUTjInOhXHAE,919
 sglang/srt/model_loader/loader.py,sha256=7OG_8-66vFDFZ9kVKGNK1BFBjZ6ql449dlyvdCbMqvE,43876
 sglang/srt/model_loader/utils.py,sha256=0NaMR67fESFopaklmsleiL27XH1QUrjZW246MUu1EJ0,1369
@@ -233,7 +254,7 @@ sglang/srt/models/chatglm.py,sha256=DOrEhmb0s-yPId88R6nJeLOTUEtogk-vkB69qT2JdWc,
 sglang/srt/models/commandr.py,sha256=PNXgfOZF84h-rSH0edEECUmEGW8YLb44V75Z_oDhFiA,14223
 sglang/srt/models/dbrx.py,sha256=okIpIwdr8Cfrz_thzc1F75XqCUfHhFLvZ1B6BaswKoA,14585
 sglang/srt/models/deepseek.py,sha256=_cVOvR6eSEgRf6TUBpTD5uMdijDWFw4sSt4lGzl8tbg,15697
-sglang/srt/models/deepseek_v2.py,sha256=-v_OJr2c3gJ0NMxQjvT3Jknz1XPGkzKx0TVR3NIiC6A,37284
+sglang/srt/models/deepseek_v2.py,sha256=vbRhgI8yD7EmHUpq5pzI_sVpGLnkeyJ7ew-3Pl6D8F4,38499
 sglang/srt/models/exaone.py,sha256=dkERTZVxrRroqu5AGLP7D4N6n8HvDqlNaDQUIe15mZY,13038
 sglang/srt/models/gemma.py,sha256=ydRqsG-7004r1fAiz01LHUmcj_6XN0Tn4xO1keJnMQk,12126
 sglang/srt/models/gemma2.py,sha256=-bFN-Te3YWAunLCrF-XFk_6fJS7gHM4Ca6h6aesXUTM,16362
@@ -241,7 +262,7 @@ sglang/srt/models/gemma2_reward.py,sha256=nJ01KfqLSJtqMLm3sG8p2mGZFK1xhhjh7I7Ccb
 sglang/srt/models/gpt2.py,sha256=2je1kE09sGcaORWnJuGYAkcwwOrT9EK-KhQaoCKjCSA,9517
 sglang/srt/models/gpt_bigcode.py,sha256=tovyOdJu2x3LkzmkdFXX_iJdkxuyChIDxwgvPBy6UPo,9528
 sglang/srt/models/granite.py,sha256=AeQY9Dxd1ZnwgCYBK0vSXXiMGM-yt9iaOVf_ruOUHXw,20409
-sglang/srt/models/grok.py,sha256=J9lgNbFebvXgF19nfZyHwlGPlGWY_m0LgP506YvOYrU,15668
+sglang/srt/models/grok.py,sha256=PbLmYP-UEbImJgbEpkHhTjkqibIvt0oENPlAIrK0qSE,17751
 sglang/srt/models/internlm2.py,sha256=_xcKtd6YtEFUTozaN-yUb0xbSYckRpomfPSKcAk4j-Y,12127
 sglang/srt/models/internlm2_reward.py,sha256=8K26A9oIFFGx_9U2mF87j7FX8K87HGKMnVL3ht1Uc7I,2398
 sglang/srt/models/llama.py,sha256=4UPKF7erp7qqBD11uvvQkO1Fo_wDs71BmA8Y2csXRcA,20302
@@ -271,17 +292,20 @@ sglang/srt/models/torch_native_llama.py,sha256=YeXHorFm6QfnczLXwPb5TG9a-He0uiA9R
 sglang/srt/models/xverse.py,sha256=Oq--KqvbYu2H4TMVGEHpSnJLEwXBpxlncR9ilsQeckc,13579
 sglang/srt/models/xverse_moe.py,sha256=7E60YIST4ELYwLRgjtHiLRI5Uyc7XqQTM7jQXiWaQs4,15541
 sglang/srt/models/yivl.py,sha256=88OubtuZ38Dxb2LzfV_MTPBI4wKhh4NJqFu--efbhFM,4809
-sglang/srt/openai_api/adapter.py,sha256=HvgeFPWv-v8LOiYF2iNCo-14BIZLAPznNTCUbubB2Rg,57091
-sglang/srt/openai_api/protocol.py,sha256=anWGr2Br8gVYm6Z0yvDwjXLaPCPuvJZ28gr5rV2dhVQ,11613
-sglang/srt/sampling/sampling_batch_info.py,sha256=s--zNjk-LErZ5lMqnZ7KiuJltaziKRbQAU5qYpKIxAc,8564
-sglang/srt/sampling/sampling_params.py,sha256=BkgCJAOSmQXwJrNXg26zSjKfMy0d5mMN6oHRk_ZuESI,5499
+sglang/srt/openai_api/adapter.py,sha256=Yv-rEA0Jd54iFlnkVy-OZM4EnPqkW_NLtDPGCiPWVWo,56386
+sglang/srt/openai_api/protocol.py,sha256=v_YUwH1PF4vIVqSE5rj1ODdSglprTe_vGiXoS99cOV4,11613
+sglang/srt/sampling/sampling_batch_info.py,sha256=4FZIt_w5pDerRbny3uUplQO23xxnU8lmtG91OPCB_4w,9254
+sglang/srt/sampling/sampling_params.py,sha256=KjUhZzRJvNTQZgJul2zSq3U8r352WzMKLbXfhP3V-nU,5685
 sglang/srt/sampling/penaltylib/__init__.py,sha256=5vQw0Y5DSzmsoFg1IdMIKLwFVhYZ5ArADHVBYbSmOec,513
 sglang/srt/sampling/penaltylib/orchestrator.py,sha256=J-DEemZcKm1--o37kf3qDOE8SZ_6H3d5oex49Mgq2ZU,10762
 sglang/srt/sampling/penaltylib/penalizers/frequency_penalty.py,sha256=1Zp2aL6dD60mwD1tCcSG0x5IYo0v4z9ce-q_YwbJ9f8,2490
 sglang/srt/sampling/penaltylib/penalizers/min_new_tokens.py,sha256=_Nxv0XgUPirZjw2SEJYp_Cd9ZcLwmt7h6JE6J4hhFq4,3629
 sglang/srt/sampling/penaltylib/penalizers/presence_penalty.py,sha256=5tOgCg7OvE9kSN9VMCpH1hwqo1YMxt9iS5PVpct9HpU,2468
 sglang/srt/sampling/penaltylib/penalizers/repetition_penalty.py,sha256=m22Rfn1RuB1HpImBDECsiJ2VooBYpsFADAwnk1EPzk0,2751
-sglang/srt/speculative/spec_info.py,sha256=d-82uWEC-QBqAgv3XGDNDW8DlHv4MtUsZghFqzGwV7U,352
+sglang/srt/speculative/build_eagle_tree.py,sha256=SIKuOFUOIzMLyanL5vViPmFBEiUHm_ezwiGuIyLmauE,9886
+sglang/srt/speculative/eagle_utils.py,sha256=VI8P8j9f1R5fRBIjiLXC8iYteMpqqeTQmaZ9OEudFuE,22983
+sglang/srt/speculative/eagle_worker.py,sha256=cAqw91E-5Tl6Sgltf0q2eYwUEH_trWHdibMWqy1rPKQ,7548
+sglang/srt/speculative/spec_info.py,sha256=D7A27UU1iOwIBEjXTgAxZ7jdftbTiVlMCvK8GmYr2zg,488
 sglang/test/few_shot_gsm8k.py,sha256=7yDbEQe49gZeJhz2wFFX-gf_59ThDKsCS1xwfogNc7k,4034
 sglang/test/few_shot_gsm8k_engine.py,sha256=QQbrwOX6-cJDD3RZC_e7zPnt6aSo8JdF8X_lRHSjdDM,3886
 sglang/test/run_eval.py,sha256=9yO0hXZOcn4abEOs96T-XPguDEklK16Ltco0pGF3zCg,4020
@@ -298,8 +322,8 @@ sglang/test/test_layernorm.py,sha256=IacByD5d-stXjzBz8Ypamc7povlcedpKPbb_4JLgo3c
 sglang/test/test_programs.py,sha256=1Z0umrsUu9pagzyGH5SrXl_qhKSyTfUv_kWC2mcn0qo,18208
 sglang/test/test_utils.py,sha256=HJG7kUQOk6n9FBbH89PDtQ41C3kt1cfJODhAEcFT0AQ,23823
 sglang/test/srt/sampling/penaltylib/utils.py,sha256=CjxHgywh0hx_87iynzQt_ztHu6zBVuE-YrZ-XPmW6U4,12906
-sglang-0.4.1.post3.dist-info/LICENSE,sha256=FJXh51fvTQklojUFY89XVLsjxRcBqOxPs8XNy-2uZ0c,11346
-sglang-0.4.1.post3.dist-info/METADATA,sha256=ICKRXupko-hmKxBCtsLUPYN6f48dMrs03J0Q6zWDPqE,22544
-sglang-0.4.1.post3.dist-info/WHEEL,sha256=PZUExdf71Ui_so67QXpySuHtCi3-J3wvF4ORK6k_S8U,91
-sglang-0.4.1.post3.dist-info/top_level.txt,sha256=yxhh3pYQkcnA7v3Bg889C2jZhvtJdEincysO7PEB09M,7
-sglang-0.4.1.post3.dist-info/RECORD,,
+sglang-0.4.1.post4.dist-info/LICENSE,sha256=FJXh51fvTQklojUFY89XVLsjxRcBqOxPs8XNy-2uZ0c,11346
+sglang-0.4.1.post4.dist-info/METADATA,sha256=nI0C5ivDIygS7D_lOrLwV_xqvHAnlthIEA6zXmf_-54,22601
+sglang-0.4.1.post4.dist-info/WHEEL,sha256=A3WOREP4zgxI0fKrHUG8DC8013e3dK3n7a6HDbcEIwE,91
+sglang-0.4.1.post4.dist-info/top_level.txt,sha256=yxhh3pYQkcnA7v3Bg889C2jZhvtJdEincysO7PEB09M,7
+sglang-0.4.1.post4.dist-info/RECORD,,

{sglang-0.4.1.post3.dist-info → sglang-0.4.1.post4.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: setuptools (75.6.0)
+Generator: setuptools (75.7.0)
 Root-Is-Purelib: true
 Tag: py3-none-any