sglang 0.4.7__tar.gz → 0.4.8__tar.gz

{sglang-0.4.7/sglang.egg-info → sglang-0.4.8}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sglang
-Version: 0.4.7
+Version: 0.4.8
 Summary: SGLang is yet another fast serving framework for large language models and vision language models.
 License:                                  Apache License
                                    Version 2.0, January 2004
@@ -230,6 +230,7 @@ Requires-Dist: modelscope; extra == "runtime-common"
 Requires-Dist: msgspec; extra == "runtime-common"
 Requires-Dist: ninja; 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"
@@ -248,14 +249,13 @@ Requires-Dist: uvloop; extra == "runtime-common"
 Requires-Dist: xgrammar==0.1.19; extra == "runtime-common"
 Provides-Extra: srt
 Requires-Dist: sglang[runtime_common]; extra == "srt"
-Requires-Dist: sgl-kernel==0.1.7; extra == "srt"
-Requires-Dist: flashinfer_python==0.2.6.post1; extra == "srt"
+Requires-Dist: sgl-kernel==0.1.9; extra == "srt"
 Requires-Dist: torch==2.7.1; extra == "srt"
 Requires-Dist: torchaudio==2.7.1; extra == "srt"
 Requires-Dist: torchvision==0.22.1; extra == "srt"
 Requires-Dist: cuda-python; extra == "srt"
-Requires-Dist: outlines<=0.1.11,>=0.0.44; extra == "srt"
 Requires-Dist: einops; extra == "srt"
+Requires-Dist: flashinfer_python==0.2.6.post1; extra == "srt"
 Provides-Extra: blackwell
 Requires-Dist: sglang[runtime_common]; extra == "blackwell"
 Requires-Dist: sgl-kernel; extra == "blackwell"
@@ -263,27 +263,21 @@ Requires-Dist: torch==2.7.1; extra == "blackwell"
 Requires-Dist: torchaudio==2.7.1; extra == "blackwell"
 Requires-Dist: torchvision==0.22.1; extra == "blackwell"
 Requires-Dist: cuda-python; extra == "blackwell"
-Requires-Dist: outlines<=0.1.11,>=0.0.44; extra == "blackwell"
 Requires-Dist: einops; extra == "blackwell"
 Requires-Dist: flashinfer_python==0.2.6.post1; extra == "blackwell"
 Provides-Extra: srt-hip
 Requires-Dist: sglang[runtime_common]; extra == "srt-hip"
 Requires-Dist: torch; extra == "srt-hip"
 Requires-Dist: vllm==0.6.7.dev2; extra == "srt-hip"
-Requires-Dist: outlines==0.1.11; extra == "srt-hip"
 Provides-Extra: srt-xpu
 Requires-Dist: sglang[runtime_common]; extra == "srt-xpu"
-Requires-Dist: outlines<=0.1.11,>=0.0.44; extra == "srt-xpu"
 Provides-Extra: srt-hpu
 Requires-Dist: sglang[runtime_common]; extra == "srt-hpu"
-Requires-Dist: outlines<=0.1.11,>=0.0.44; extra == "srt-hpu"
 Provides-Extra: srt-cpu
 Requires-Dist: sglang[runtime_common]; extra == "srt-cpu"
-Requires-Dist: outlines<=0.1.11,>=0.0.44; extra == "srt-cpu"
 Requires-Dist: einops; extra == "srt-cpu"
 Provides-Extra: srt-npu
 Requires-Dist: sglang[runtime_common]; extra == "srt-npu"
-Requires-Dist: outlines<=0.1.11,>=0.0.44; extra == "srt-npu"
 Provides-Extra: openai
 Requires-Dist: openai>=1.0; extra == "openai"
 Requires-Dist: tiktoken; extra == "openai"
@@ -292,7 +286,7 @@ 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>=0.0.4; 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
@@ -371,7 +365,7 @@ Dynamic: license-file
 
 --------------------------------------------------------------------------------
 
-| [**Blog**](https://lmsys.org/blog/2024-07-25-sglang-llama3/)
+| [**Blog**](https://lmsys.org/blog/2025-05-05-large-scale-ep/)
 | [**Documentation**](https://docs.sglang.ai/)
 | [**Join Slack**](https://slack.sglang.ai/)
 | [**Join Bi-Weekly Development Meeting**](https://meeting.sglang.ai/)
@@ -403,7 +397,7 @@ SGLang is a fast serving framework for large language models and vision language
 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, zero-overhead CPU scheduler, continuous batching, token attention (paged attention), speculative decoding, tensor parallelism, chunked prefill, structured outputs, quantization (FP8/INT4/AWQ/GPTQ), and multi-lora batching.
+- **Fast Backend Runtime**: Provides efficient serving with RadixAttention for prefix caching, zero-overhead CPU scheduler, prefill-decode disaggregation, speculative decoding, continuous batching, paged attention, tensor parallelism, pipeline parallelism, expert parallelism, structured outputs, chunked prefill, quantization (FP8/INT4/AWQ/GPTQ), and multi-lora batching.
 - **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.), embedding models (e5-mistral, gte, mcdse) and reward models (Skywork), with easy extensibility for integrating new models.
 - **Active Community**: SGLang is open-source and backed by an active community with industry adoption.
@@ -422,7 +416,7 @@ Learn more in the release blogs: [v0.2 blog](https://lmsys.org/blog/2024-07-25-s
 [Development Roadmap (2025 H1)](https://github.com/sgl-project/sglang/issues/4042)
 
 ## Adoption and Sponsorship
-SGLang has been deployed at large scale, serving trillions of tokens in production every day. It is trusted and adopted by a broad range of leading enterprises and institutions, including xAI, NVIDIA, AMD, Google Cloud, Oracle Cloud, LinkedIn, Cursor, Voltage Park, Atlas Cloud, DataCrunch, Baseten, Nebius, Novita, InnoMatrix, RunPod, Stanford, UC Berkeley, UCLA, ETCHED, Jam & Tea Studios, Hyperbolic, as well as major technology organizations across North America and Asia. As an open-source LLM inference engine, SGLang has become the de facto standard in the industry, with production deployments running on over 100,000 GPUs worldwide.
+SGLang has been deployed at large scale, generating trillions of tokens in production every day. It is trusted and adopted by a broad range of leading enterprises and institutions, including xAI, NVIDIA, AMD, Google Cloud, Oracle Cloud, LinkedIn, Cursor, Voltage Park, Atlas Cloud, DataCrunch, Baseten, Nebius, Novita, InnoMatrix, RunPod, Stanford, UC Berkeley, UCLA, ETCHED, Jam & Tea Studios, Hyperbolic, as well as major technology organizations across North America and Asia. As an open-source LLM inference engine, SGLang has become the de facto standard in the industry, with production deployments running on over 100,000 GPUs worldwide.
 
 <img src="https://raw.githubusercontent.com/sgl-project/sgl-learning-materials/refs/heads/main/slides/adoption.png" alt="logo" width="800" margin="10px"></img>
 

{sglang-0.4.7 → sglang-0.4.8}/README.md

@@ -12,7 +12,7 @@
 
 --------------------------------------------------------------------------------
 
-| [**Blog**](https://lmsys.org/blog/2024-07-25-sglang-llama3/)
+| [**Blog**](https://lmsys.org/blog/2025-05-05-large-scale-ep/)
 | [**Documentation**](https://docs.sglang.ai/)
 | [**Join Slack**](https://slack.sglang.ai/)
 | [**Join Bi-Weekly Development Meeting**](https://meeting.sglang.ai/)
@@ -44,7 +44,7 @@ SGLang is a fast serving framework for large language models and vision language
 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, zero-overhead CPU scheduler, continuous batching, token attention (paged attention), speculative decoding, tensor parallelism, chunked prefill, structured outputs, quantization (FP8/INT4/AWQ/GPTQ), and multi-lora batching.
+- **Fast Backend Runtime**: Provides efficient serving with RadixAttention for prefix caching, zero-overhead CPU scheduler, prefill-decode disaggregation, speculative decoding, continuous batching, paged attention, tensor parallelism, pipeline parallelism, expert parallelism, structured outputs, chunked prefill, quantization (FP8/INT4/AWQ/GPTQ), and multi-lora batching.
 - **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.), embedding models (e5-mistral, gte, mcdse) and reward models (Skywork), with easy extensibility for integrating new models.
 - **Active Community**: SGLang is open-source and backed by an active community with industry adoption.
@@ -63,7 +63,7 @@ Learn more in the release blogs: [v0.2 blog](https://lmsys.org/blog/2024-07-25-s
 [Development Roadmap (2025 H1)](https://github.com/sgl-project/sglang/issues/4042)
 
 ## Adoption and Sponsorship
-SGLang has been deployed at large scale, serving trillions of tokens in production every day. It is trusted and adopted by a broad range of leading enterprises and institutions, including xAI, NVIDIA, AMD, Google Cloud, Oracle Cloud, LinkedIn, Cursor, Voltage Park, Atlas Cloud, DataCrunch, Baseten, Nebius, Novita, InnoMatrix, RunPod, Stanford, UC Berkeley, UCLA, ETCHED, Jam & Tea Studios, Hyperbolic, as well as major technology organizations across North America and Asia. As an open-source LLM inference engine, SGLang has become the de facto standard in the industry, with production deployments running on over 100,000 GPUs worldwide.
+SGLang has been deployed at large scale, generating trillions of tokens in production every day. It is trusted and adopted by a broad range of leading enterprises and institutions, including xAI, NVIDIA, AMD, Google Cloud, Oracle Cloud, LinkedIn, Cursor, Voltage Park, Atlas Cloud, DataCrunch, Baseten, Nebius, Novita, InnoMatrix, RunPod, Stanford, UC Berkeley, UCLA, ETCHED, Jam & Tea Studios, Hyperbolic, as well as major technology organizations across North America and Asia. As an open-source LLM inference engine, SGLang has become the de facto standard in the industry, with production deployments running on over 100,000 GPUs worldwide.
 
 <img src="https://raw.githubusercontent.com/sgl-project/sgl-learning-materials/refs/heads/main/slides/adoption.png" alt="logo" width="800" margin="10px"></img>
 

{sglang-0.4.7 → sglang-0.4.8}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "sglang"
-version = "0.4.7"
+version = "0.4.8"
 description = "SGLang is yet another fast serving framework for large language models and vision language models."
 readme = "README.md"
 requires-python = ">=3.8"
@@ -29,6 +29,7 @@ runtime_common = [
     "msgspec",
     "ninja",
     "orjson",
+    "outlines==0.1.11",
     "packaging",
     "partial_json_parser",
     "pillow",
@@ -49,14 +50,13 @@ runtime_common = [
 
 srt = [
     "sglang[runtime_common]",
-    "sgl-kernel==0.1.7",
-    "flashinfer_python==0.2.6.post1",
+    "sgl-kernel==0.1.9",
     "torch==2.7.1",
     "torchaudio==2.7.1",
     "torchvision==0.22.1",
     "cuda-python",
-    "outlines>=0.0.44,<=0.1.11",
     "einops",
+    "flashinfer_python==0.2.6.post1",
 ]
 
 blackwell = [
@@ -66,7 +66,6 @@ blackwell = [
     "torchaudio==2.7.1",
     "torchvision==0.22.1",
     "cuda-python",
-    "outlines>=0.0.44,<=0.1.11",
     "einops",
     "flashinfer_python==0.2.6.post1",
 ]
@@ -77,28 +76,27 @@ srt_hip = [
     "sglang[runtime_common]",
     "torch",
     "vllm==0.6.7.dev2",
-    "outlines==0.1.11"
 ]
 
 # 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]", "outlines>=0.0.44,<=0.1.11"]
+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]", "outlines>=0.0.44,<=0.1.11"]
+srt_hpu = ["sglang[runtime_common]"]
 
 # CPU: currently, there are no pre-built vllm wheels for CPU.
 # To install vllm for CPU, please follow the instruction here:
 # https://docs.vllm.ai/en/latest/getting_started/installation/cpu/index.html
-srt_cpu = ["sglang[runtime_common]", "outlines>=0.0.44,<=0.1.11", "einops"]
+srt_cpu = ["sglang[runtime_common]", "einops"]
 # https://vllm-ascend.readthedocs.io/en/latest/installation.html
-srt_npu = ["sglang[runtime_common]", "outlines>=0.0.44,<=0.1.11"]
+srt_npu = ["sglang[runtime_common]"]
 
 openai = ["openai>=1.0", "tiktoken"]
 anthropic = ["anthropic>=0.20.0"]
 litellm = ["litellm>=1.0.0"]
-torch_memory_saver = ["torch_memory_saver>=0.0.4"]
+torch_memory_saver = ["torch_memory_saver>=0.0.8"]
 decord = ["decord"]
 test = [
     "accelerate",

{sglang-0.4.7 → sglang-0.4.8}/sglang/__init__.py

@@ -15,6 +15,7 @@ from sglang.api import (
     get_server_info,
     image,
     select,
+    separate_reasoning,
     set_default_backend,
     system,
     system_begin,
@@ -54,6 +55,7 @@ __all__ = [
     "get_server_info",
     "image",
     "select",
+    "separate_reasoning",
     "set_default_backend",
     "system",
     "system_begin",

{sglang-0.4.7 → sglang-0.4.8}/sglang/api.py

@@ -15,6 +15,7 @@ from sglang.lang.ir import (
     SglRoleBegin,
     SglRoleEnd,
     SglSelect,
+    SglSeparateReasoning,
     SglVideo,
 )
 
@@ -277,3 +278,9 @@ def assistant_begin():
 
 def assistant_end():
     return SglRoleEnd("assistant")
+
+
+def separate_reasoning(
+    expr: Optional[SglExpr] = None, model_type: Optional[str] = None
+):
+    return SglExprList([expr, SglSeparateReasoning(model_type, expr=expr)])

{sglang-0.4.7 → sglang-0.4.8}/sglang/bench_one_batch.py

@@ -71,6 +71,8 @@ from sglang.srt.utils import (
     configure_logger,
     get_bool_env_var,
     kill_process_tree,
+    require_mlp_sync,
+    require_mlp_tp_gather,
     set_gpu_proc_affinity,
     suppress_other_loggers,
 )
@@ -243,7 +245,7 @@ def extend(reqs, model_runner):
         enable_custom_logit_processor=False,
     )
     batch.prepare_for_extend()
-    _maybe_prepare_dp_attn_batch(batch, model_runner)
+    _maybe_prepare_mlp_sync_batch(batch, model_runner)
     model_worker_batch = batch.get_model_worker_batch()
     forward_batch = ForwardBatch.init_new(model_worker_batch, model_runner)
     logits_output, _ = model_runner.forward(forward_batch)
@@ -255,7 +257,7 @@ def extend(reqs, model_runner):
 def decode(input_token_ids, batch, model_runner):
     batch.output_ids = input_token_ids
     batch.prepare_for_decode()
-    _maybe_prepare_dp_attn_batch(batch, model_runner)
+    _maybe_prepare_mlp_sync_batch(batch, model_runner)
     model_worker_batch = batch.get_model_worker_batch()
     forward_batch = ForwardBatch.init_new(model_worker_batch, model_runner)
     logits_output, _ = model_runner.forward(forward_batch)
@@ -263,18 +265,18 @@ def decode(input_token_ids, batch, model_runner):
     return next_token_ids, logits_output.next_token_logits
 
 
-def _maybe_prepare_dp_attn_batch(batch: ScheduleBatch, model_runner):
-    if model_runner.server_args.enable_dp_attention:
-        Scheduler.prepare_dp_attn_batch_raw(
+def _maybe_prepare_mlp_sync_batch(batch: ScheduleBatch, model_runner):
+    if require_mlp_sync(model_runner.server_args):
+        Scheduler.prepare_mlp_sync_batch_raw(
             batch,
             dp_size=model_runner.server_args.dp_size,
             attn_tp_size=1,
-            moe_dense_tp_size=model_runner.server_args.moe_dense_tp_size,
             tp_cpu_group=model_runner.tp_group.cpu_group,
             get_idle_batch=None,
             disable_cuda_graph=model_runner.server_args.disable_cuda_graph,
             spec_algorithm=SpeculativeAlgorithm.NONE,
             speculative_num_draft_tokens=None,
+            require_mlp_tp_gather=require_mlp_tp_gather(model_runner.server_args),
         )
 
 

{sglang-0.4.7 → sglang-0.4.8}/sglang/bench_serving.py

@@ -399,7 +399,7 @@ async def async_request_sglang_generate(
                             # NOTE: Some completion API might have a last
                             # usage summary response without a token so we
                             # want to check a token was generated
-                            if data["text"]:
+                            if "text" in data and data["text"]:
                                 timestamp = time.perf_counter()
                                 generated_text = data["text"]
                                 output_len = data["meta_info"]["completion_tokens"]

{sglang-0.4.7 → sglang-0.4.8}/sglang/lang/interpreter.py

@@ -26,6 +26,7 @@ from sglang.lang.ir import (
     SglRoleBegin,
     SglRoleEnd,
     SglSelect,
+    SglSeparateReasoning,
     SglVariable,
     SglVarScopeBegin,
     SglVarScopeEnd,
@@ -472,6 +473,8 @@ class StreamExecutor:
                 self._execute_concatenate_and_append_kv_cache(other)
             else:
                 self._execute_concatenate_and_append_text(other)
+        elif isinstance(other, SglSeparateReasoning):
+            self._execute_separate_reasoning(other)
         else:
             raise ValueError(f"Unknown type: {type(other)}")
 
@@ -724,8 +727,44 @@ class StreamExecutor:
         src_rids = [state.stream_executor.sid for state in expr.states]
         self.backend.concatenate_and_append(src_rids, self.sid)
 
+    def _execute_separate_reasoning(self, expr: SglSeparateReasoning):
+        if self.stream:
+            # separate reasoning for stream is not supported
+            return
+
+        if (
+            self.cur_role == "assistant"
+            and self.num_api_spec_tokens is not None
+            and self.backend.is_chat_model
+        ):
+            # Execute the stored lazy generation calls
+            self.backend.role_end_generate(self)
+
+        from sglang.srt.reasoning_parser import ReasoningParser
+
+        reasoning_parser = ReasoningParser(expr.model_type)
+        other = expr.expr
+        if not other:
+            return
+        elif isinstance(other, SglGen) or isinstance(other, SglSelect):
+            cur_text = self.get_var(other.name)
+            reasoning, normal_text = reasoning_parser.parse_non_stream(cur_text)
+            reasoning_name = expr.process_name_for_reasoning(other.name)
+            self.set_var(other.name, normal_text)
+            self.set_var(reasoning_name, reasoning)
+            # the variable is ready to be used
+            self.variable_event[reasoning_name].set()
+            self.text_ = self.text_[: self.cur_role_begin_pos] + normal_text
+        elif isinstance(other, SglExprList):
+            for x in other.expr_list:
+                self._execute_separate_reasoning(
+                    SglSeparateReasoning(expr.model_type, x)
+                )
+
     def _init_var_event(self, expr):
-        if isinstance(expr, (SglGen, SglSelect, SglVarScopeBegin)):
+        if isinstance(
+            expr, (SglGen, SglSelect, SglVarScopeBegin, SglSeparateReasoning)
+        ):
             self.variable_event[expr.name] = threading.Event()
             if self.stream:
                 self.stream_var_event[expr.name] = threading.Event()

{sglang-0.4.7 → sglang-0.4.8}/sglang/lang/ir.py

@@ -606,3 +606,30 @@ class SglCommitLazy(SglExpr):
 
     def __repr__(self):
         return "CommitLazy()"
+
+
+class SglSeparateReasoning(SglExpr):
+    def __init__(self, model_type: str, expr: SglExpr):
+        super().__init__()
+        self.model_type = model_type
+
+        self.expr = expr
+        self.name = None
+        self._process_expr(expr)
+
+    def process_name_for_reasoning(self, name):
+        if not name:
+            raise ValueError("name must be provided")
+        return f"{name}_reasoning_content"
+
+    def _process_expr(self, expr):
+        if isinstance(expr, SglGen):
+            self.name = self.process_name_for_reasoning(expr.name)
+        elif isinstance(expr, SglSelect):
+            self.name = self.process_name_for_reasoning(expr.name)
+        elif isinstance(expr, SglExprList):
+            for x in expr.expr_list:
+                self._process_expr(x)
+
+    def __repr__(self):
+        return f"SeparateReasoning(model_type={self.model_type}, name={self.name})"

sglang-0.4.8/sglang/math_utils.py

@@ -0,0 +1,8 @@
+# COPIED FROM DeepGEMM
+def align(x: int, y: int) -> int:
+    return ceil_div(x, y) * y
+
+
+# COPIED FROM DeepGEMM
+def ceil_div(x: int, y: int) -> int:
+    return (x + y - 1) // y

{sglang-0.4.7 → sglang-0.4.8}/sglang/srt/_custom_ops.py

@@ -4,7 +4,7 @@ from typing import List, Tuple
 
 import torch
 
-from sglang.srt.utils import get_bool_env_var, is_hip, is_hpu
+from sglang.srt.utils import get_bool_env_var, is_hip, is_hpu, is_npu
 
 logger = logging.getLogger(__name__)
 use_vllm_custom_allreduce = get_bool_env_var(
@@ -25,7 +25,7 @@ if not is_hpu():
             logger.warning("Failed to import from custom_ar with %r", e)
 
 
-if not is_hip():
+if not is_hip() and not is_npu():
     if use_vllm_custom_allreduce:
         custom_op = torch.ops._C_custom_ar
     else:

{sglang-0.4.7 → sglang-0.4.8}/sglang/srt/code_completion_parser.py

@@ -15,12 +15,10 @@
 
 
 import dataclasses
-import json
 import logging
-import os
 from enum import auto
 
-from sglang.srt.openai_api.protocol import ChatCompletionRequest
+from sglang.srt.entrypoints.openai.protocol import CompletionRequest
 
 logger = logging.getLogger(__name__)
 completion_template_name = None
@@ -57,46 +55,6 @@ class CompletionTemplate:
 completion_templates: dict[str, CompletionTemplate] = {}
 
 
-def load_completion_template_for_openai_api(completion_template_arg):
-    global completion_template_name
-
-    logger.info(
-        f"Use completion template for the OpenAI-compatible API server: {completion_template_arg}"
-    )
-
-    if not completion_template_exists(completion_template_arg):
-        if not os.path.exists(completion_template_arg):
-            raise RuntimeError(
-                f"Completion template {completion_template_arg} is not a built-in template name "
-                "or a valid completion template file path."
-            )
-
-        assert completion_template_arg.endswith(
-            ".json"
-        ), "unrecognized format of completion template file"
-        with open(completion_template_arg, "r") as filep:
-            template = json.load(filep)
-            try:
-                fim_position = FimPosition[template["fim_position"]]
-            except KeyError:
-                raise ValueError(
-                    f"Unknown fim position: {template['fim_position']}"
-                ) from None
-            register_completion_template(
-                CompletionTemplate(
-                    name=template["name"],
-                    fim_begin_token=template["fim_begin_token"],
-                    fim_middle_token=template["fim_middle_token"],
-                    fim_end_token=template["fim_end_token"],
-                    fim_position=fim_position,
-                ),
-                override=True,
-            )
-        completion_template_name = template["name"]
-    else:
-        completion_template_name = completion_template_arg
-
-
 def register_completion_template(template: CompletionTemplate, override: bool = False):
     """Register a new completion template."""
     if not override:
@@ -116,7 +74,7 @@ def is_completion_template_defined() -> bool:
     return completion_template_name is not None
 
 
-def generate_completion_prompt_from_request(request: ChatCompletionRequest) -> str:
+def generate_completion_prompt_from_request(request: CompletionRequest) -> str:
     global completion_template_name
     if request.suffix == "":
         return request.prompt

{sglang-0.4.7 → sglang-0.4.8}/sglang/srt/configs/model_config.py

@@ -550,6 +550,11 @@ def is_generation_model(model_architectures: List[str], is_embedding: bool = Fal
         or "Qwen2ForRewardModel" in model_architectures
         or "Qwen2ForSequenceClassification" in model_architectures
         or "CLIPModel" in model_architectures
+        or "BertModel" in model_architectures
+        or "Contriever" in model_architectures
+        or "BertForSequenceClassification" in model_architectures
+        or "XLMRobertaModel" in model_architectures
+        or "XLMRobertaForSequenceClassification" in model_architectures
     ):
         return False
     else:
@@ -578,6 +583,7 @@ multimodal_model_archs = [
     "KimiVLForConditionalGeneration",
     "InternVLChatModel",
     "Phi4MMForCausalLM",
+    "VILAForConditionalGeneration",
 ]
 
 

sglang-0.4.8/sglang/srt/constants.py

@@ -0,0 +1,3 @@
+# GPU Memory Types
+GPU_MEMORY_TYPE_KV_CACHE = "kv_cache"
+GPU_MEMORY_TYPE_WEIGHTS = "weights"

{sglang-0.4.7 → sglang-0.4.8}/sglang/srt/conversation.py

@@ -11,7 +11,17 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # ==============================================================================
-"""Conversation chat templates."""
+"""Conversation chat templates.
+
+This module provides conversation template definitions, data structures, and utilities
+for managing chat templates across different model types in SGLang.
+
+Key components:
+- Conversation class: Defines the structure and behavior of chat templates
+- SeparatorStyle enum: Different conversation formatting styles
+- Template registry: Functions to register and retrieve templates by name or model path
+- Built-in templates: Pre-defined templates for popular models
+"""
 
 # Adapted from
 # https://github.com/lm-sys/FastChat/blob/main/fastchat/conversation.py
@@ -20,7 +30,7 @@ import re
 from enum import IntEnum, auto
 from typing import Callable, Dict, List, Optional, Tuple, Union
 
-from sglang.srt.openai_api.protocol import ChatCompletionRequest
+from sglang.srt.entrypoints.openai.protocol import ChatCompletionRequest
 from sglang.srt.utils import read_system_prompt_from_file
 
 
@@ -618,7 +628,7 @@ def generate_chat_conv(
 
 
 # llama2 template
-# reference: https://huggingface.co/blog/codellama#conversational-instructions
+# reference: https://github.com/lm-sys/FastChat/blob/main/fastchat/conversation.py
 # reference: https://github.com/facebookresearch/llama/blob/1a240688810f8036049e8da36b073f63d2ac552c/llama/generation.py#L212
 register_conv_template(
     Conversation(
@@ -983,3 +993,9 @@ def match_devstral(model_path: str):
 def match_phi_4_mm(model_path: str):
     if "phi-4-multimodal" in model_path.lower():
         return "phi-4-mm"
+
+
+@register_conv_template_matching_function
+def match_vila(model_path: str):
+    if re.search(r"vila", model_path, re.IGNORECASE):
+        return "chatml"

{sglang-0.4.7 → sglang-0.4.8}/sglang/srt/custom_op.py

@@ -1,9 +1,11 @@
 from torch import nn
 
-from sglang.srt.utils import is_cuda, is_hip
+from sglang.srt.utils import cpu_has_amx_support, is_cpu, is_cuda, is_hip
 
 _is_cuda = is_cuda()
 _is_hip = is_hip()
+_is_cpu = is_cpu()
+_is_cpu_amx_available = cpu_has_amx_support()
 
 
 class CustomOp(nn.Module):
@@ -75,5 +77,7 @@ class CustomOp(nn.Module):
             return self.forward_cuda
         elif _is_hip:
             return self.forward_hip
+        elif _is_cpu and _is_cpu_amx_available:
+            return self.forward_cpu
         else:
             return self.forward_native

{sglang-0.4.7 → sglang-0.4.8}/sglang/srt/disaggregation/base/__init__.py

@@ -1,4 +1,4 @@
-from .conn import (
+from sglang.srt.disaggregation.base.conn import (
     BaseKVBootstrapServer,
     BaseKVManager,
     BaseKVReceiver,

{sglang-0.4.7 → sglang-0.4.8}/sglang/srt/disaggregation/base/conn.py

@@ -1,23 +1,32 @@
+from __future__ import annotations
+
 from abc import ABC, abstractmethod
-from typing import Optional
+from typing import TYPE_CHECKING, List, Optional
 
 import numpy as np
 import numpy.typing as npt
 
-from sglang.srt.disaggregation.utils import DisaggregationMode
 from sglang.srt.server_args import ServerArgs
 
+if TYPE_CHECKING:
+    from sglang.srt.disaggregation.utils import DisaggregationMode
+
 
 class KVArgs:
     engine_rank: int
-    kv_data_ptrs: list[int]
-    kv_data_lens: list[int]
-    kv_item_lens: list[int]
-    aux_data_ptrs: list[int]
-    aux_data_lens: list[int]
-    aux_item_lens: list[int]
+    kv_data_ptrs: List[int]
+    kv_data_lens: List[int]
+    kv_item_lens: List[int]
+    aux_data_ptrs: List[int]
+    aux_data_lens: List[int]
+    aux_item_lens: List[int]
     ib_device: str
+    ib_traffic_class: str
     gpu_id: int
+    # for different tp
+    decode_tp_size: int
+    # for pp prefill
+    prefill_pp_size: int
 
 
 class KVPoll:
@@ -45,7 +54,12 @@ class BaseKVSender(ABC):
 
     @abstractmethod
     def __init__(
-        self, mgr: BaseKVManager, bootstrap_addr: str, bootstrap_room: int
+        self,
+        mgr: BaseKVManager,
+        bootstrap_addr: str,
+        bootstrap_room: int,
+        dest_tp_ranks: List[int],
+        pp_rank: int,
     ): ...
 
     @abstractmethod
@@ -56,7 +70,7 @@ class BaseKVSender(ABC):
         ...
 
     @abstractmethod
-    def send(self, kv_indices: npt.NDArray[np.int64]):
+    def send(self, kv_indices: npt.NDArray[np.int32]):
         """
         Send the kv cache at the given kv indices to the decoder server
         """
@@ -88,7 +102,7 @@ class BaseKVReceiver(ABC):
     ): ...
 
     @abstractmethod
-    def init(self, kv_indices: npt.NDArray[np.int64], aux_index: Optional[int] = None):
+    def init(self, kv_indices: npt.NDArray[np.int32], aux_index: Optional[int] = None):
         """
         Notify the prefill server about the kv indices and aux index
         """