sglang 0.3.1.post2__tar.gz → 0.3.1.post3__tar.gz

{sglang-0.3.1.post2/sglang.egg-info → sglang-0.3.1.post3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: sglang
-Version: 0.3.1.post2
+Version: 0.3.1.post3
 Summary: SGLang is yet another fast serving framework for large language models and vision language models.
 License:                                  Apache License
                                    Version 2.0, January 2004
@@ -318,7 +318,7 @@ 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.post2 https://github.com/sgl-project/sglang.git
+git clone -b v0.3.1.post3 https://github.com/sgl-project/sglang.git
 cd sglang
 
 pip install --upgrade pip
@@ -499,6 +499,7 @@ 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
 - [LLaVA-OneVision](https://llava-vl.github.io/blog/2024-08-05-llava-onevision/)

{sglang-0.3.1.post2 → sglang-0.3.1.post3}/README.md

@@ -60,7 +60,7 @@ 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.post2 https://github.com/sgl-project/sglang.git
+git clone -b v0.3.1.post3 https://github.com/sgl-project/sglang.git
 cd sglang
 
 pip install --upgrade pip
@@ -241,6 +241,7 @@ 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
 - [LLaVA-OneVision](https://llava-vl.github.io/blog/2024-08-05-llava-onevision/)

{sglang-0.3.1.post2 → sglang-0.3.1.post3}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "sglang"
-version = "0.3.1.post2"
+version = "0.3.1.post3"
 description = "SGLang is yet another fast serving framework for large language models and vision language models."
 readme = "README.md"
 requires-python = ">=3.8"

{sglang-0.3.1.post2 → sglang-0.3.1.post3}/sglang/bench_latency.py

@@ -64,8 +64,13 @@ from sglang.srt.hf_transformers_utils import get_tokenizer
 from sglang.srt.managers.schedule_batch import Req, ScheduleBatch
 from sglang.srt.model_executor.model_runner import ModelRunner
 from sglang.srt.sampling.sampling_params import SamplingParams
+from sglang.srt.server import _set_envs_and_config
 from sglang.srt.server_args import ServerArgs
-from sglang.srt.utils import kill_child_process, suppress_other_loggers
+from sglang.srt.utils import (
+    configure_logger,
+    kill_child_process,
+    suppress_other_loggers,
+)
 
 
 @dataclasses.dataclass
@@ -341,6 +346,8 @@ def latency_test(
     bench_args,
     tp_rank,
 ):
+    configure_logger(server_args, prefix=f" TP{tp_rank}")
+    _set_envs_and_config(server_args)
     rank_print = print if tp_rank == 0 else lambda *args, **kwargs: None
 
     # Load the model

{sglang-0.3.1.post2 → sglang-0.3.1.post3}/sglang/srt/layers/activation.py

@@ -31,8 +31,9 @@ from vllm.distributed import (
     get_tensor_model_parallel_world_size,
 )
 from vllm.model_executor.custom_op import CustomOp
-from vllm.model_executor.layers.quantization import QuantizationConfig
-from vllm.model_executor.utils import set_weight_attrs
+
+from sglang.srt.layers.quantization.base_config import QuantizationConfig
+from sglang.srt.utils import set_weight_attrs
 
 logger = logging.getLogger(__name__)
 

{sglang-0.3.1.post2 → sglang-0.3.1.post3}/sglang/srt/layers/attention_backend.py

@@ -346,7 +346,9 @@ class TritonAttnBackend(AttentionBackend):
 
         self.decode_attention_fwd = decode_attention_fwd
         self.extend_attention_fwd = extend_attention_fwd
-        self.num_head = model_runner.model_config.num_attention_heads
+        self.num_head = (
+            model_runner.model_config.num_attention_heads // model_runner.tp_size
+        )
 
         if global_server_args_dict.get("triton_attention_reduce_in_fp32", False):
             self.reduce_dtype = torch.float32