sglang 0.1.17__tar.gz → 0.1.21__tar.gz

{sglang-0.1.17/sglang.egg-info → sglang-0.1.21}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: sglang
-Version: 0.1.17
+Version: 0.1.21
 Summary: A structured generation langauge for LLMs.
 License:                                  Apache License
                                    Version 2.0, January 2004
@@ -213,30 +213,29 @@ Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: requests
 Requires-Dist: tqdm
+Requires-Dist: numpy
 Provides-Extra: srt
 Requires-Dist: aiohttp; extra == "srt"
 Requires-Dist: fastapi; extra == "srt"
+Requires-Dist: hf_transfer; extra == "srt"
+Requires-Dist: huggingface_hub; extra == "srt"
+Requires-Dist: interegular; extra == "srt"
+Requires-Dist: packaging; extra == "srt"
+Requires-Dist: pillow; extra == "srt"
 Requires-Dist: psutil; extra == "srt"
+Requires-Dist: pydantic; extra == "srt"
 Requires-Dist: rpyc; extra == "srt"
 Requires-Dist: torch; extra == "srt"
-Requires-Dist: uvloop; extra == "srt"
 Requires-Dist: uvicorn; extra == "srt"
+Requires-Dist: uvloop; extra == "srt"
 Requires-Dist: zmq; extra == "srt"
-Requires-Dist: vllm==0.4.3; extra == "srt"
-Requires-Dist: interegular; extra == "srt"
-Requires-Dist: pydantic; extra == "srt"
-Requires-Dist: pillow; extra == "srt"
-Requires-Dist: packaging; extra == "srt"
-Requires-Dist: huggingface_hub; extra == "srt"
-Requires-Dist: hf_transfer; extra == "srt"
-Requires-Dist: outlines>=0.0.34; extra == "srt"
+Requires-Dist: vllm==0.5.1; extra == "srt"
+Requires-Dist: outlines>=0.0.44; extra == "srt"
 Provides-Extra: openai
 Requires-Dist: openai>=1.0; extra == "openai"
-Requires-Dist: numpy; extra == "openai"
 Requires-Dist: tiktoken; extra == "openai"
 Provides-Extra: anthropic
 Requires-Dist: anthropic>=0.20.0; extra == "anthropic"
-Requires-Dist: numpy; extra == "anthropic"
 Provides-Extra: litellm
 Requires-Dist: litellm>=1.0.0; extra == "litellm"
 Provides-Extra: all
@@ -257,8 +256,8 @@ SGLang is a structured generation language designed for large language models (L
 It makes your interaction with LLMs faster and more controllable by co-designing the frontend language and the runtime system.
 
 The core features include:
-- **A Flexible Front-End Language**: This allows for easy programming of LLM applications with multiple chained generation calls, advanced prompting techniques, control flow, multiple modalities, parallelism, and external interaction.
-- **A High-Performance Runtime with RadixAttention**: This feature significantly accelerates the execution of complex LLM programs by automatically reusing the KV cache across multiple calls. It can also be used as a standalone serving engine with all common techniques implemented, such as continuous batching and tensor parallelism.
+- **Flexible Frontend Language**: Enables easy programming of LLM applications with chained generation calls, advanced prompting, control flow, multiple modalities, parallelism, and external interactions.
+- **High-Performance Backend Runtime**: Features RadixAttention for accelerating complex LLM programs by reusing the KV cache across multiple calls. It can also serve as a standalone inference engine with all common techniques implemented (e.g., continuous batching and tensor parallelism).
 
 ## News
 - [2024/02] 🔥 SGLang enables **3x faster JSON decoding** with compressed finite state machine ([blog](https://lmsys.org/blog/2024-02-05-compressed-fsm/)).
@@ -279,19 +278,33 @@ The core features include:
 ### Method 1: With pip
 ```
 pip install "sglang[all]"
+
+# Install FlashInfer CUDA kernels
+pip install flashinfer -i https://flashinfer.ai/whl/cu121/torch2.3/
 ```
 
 ### Method 2: From source
 ```
-git clone git@github.com:sgl-project/sglang.git
+git clone https://github.com/sgl-project/sglang.git
 cd sglang
 
-pip install --upgrade pip
 pip install -e "python[all]"
+
+# Install FlashInfer CUDA kernels
+pip install flashinfer -i https://flashinfer.ai/whl/cu121/torch2.3/
 ```
 
-### Notes
-- If you only need to use the OpenAI backend, you can avoid installing other dependencies by using `pip install "sglang[openai]"`
+### Method 3: Using docker
+The docker images are available on Docker Hub as [lmsysorg/sglang](https://hub.docker.com/r/lmsysorg/sglang/tags).
+
+### Common Notes
+- If you see errors from the Triton compiler, please install the [Triton Nightly](https://triton-lang.org/main/getting-started/installation.html) by
+```
+pip uninstall -y triton triton-nightly
+pip install -U --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/Triton-Nightly/pypi/simple/ triton-nightly
+```
+- If you cannot install FlashInfer, check out its [installation](https://docs.flashinfer.ai/installation.html#) page. If you still cannot install it, you can use the slower Triton kernels by adding `--disable-flashinfer` when launching the server.
+- If you only need to use the OpenAI backend, you can avoid installing other dependencies by using `pip install "sglang[openai]"`.
 
 ## Quick Start
 The example below shows how to use sglang to answer a mulit-turn question.
@@ -511,8 +524,8 @@ for out in state.text_iter():
 ```
 
 ### Tips and Implementation Details
-- The `choices` argument in `sgl.gen` is implemented by computing the normalized log probabilities of all choices and selecting the one with the highest probability.
-- The `regex` argument in `sgl.gen` is implemented through autoregressive decoding with logit bias masking, according to the constraints set by the regex.
+- The `choices` argument in `sgl.gen` is implemented by computing the [token-length normalized log probabilities](https://blog.eleuther.ai/multiple-choice-normalization/) of all choices and selecting the one with the highest probability.
+- The `regex` argument in `sgl.gen` is implemented through autoregressive decoding with logit bias masking, according to the constraints set by the regex. It is compatible with `temperature=0` and `temperature != 0`.
 
 ## Backend: SGLang Runtime (SRT)
 The SGLang Runtime (SRT) is designed to work best with the SGLang frontend.
@@ -569,7 +582,6 @@ response = client.chat.completions.create(
 print(response)
 ```
 
-
 By default, the server uses the chat template specified in the model tokenizer from Hugging Face. It should just work for most official models such as Llama-2/Llama-3.
 
 If needed, you can also override the chat template when launching the server:
@@ -598,7 +610,7 @@ python -m sglang.launch_server --model-path meta-llama/Llama-2-7b-chat-hf --port
 ```
 
 ### Additional Arguments
-- Add `--tp 2` to enable tensor parallelism.
+- Add `--tp 2` to enable tensor parallelism. If it indicates `peer access is not supported between these two devices`, add `--enable-p2p-check` option.
 ```
 python -m sglang.launch_server --model-path meta-llama/Llama-2-7b-chat-hf --port 30000 --tp 2
 ```
@@ -610,16 +622,22 @@ python -m sglang.launch_server --model-path meta-llama/Llama-2-7b-chat-hf --port
 ```
 python -m sglang.launch_server --model-path meta-llama/Llama-2-7b-chat-hf --port 30000 --mem-fraction-static 0.7
 ```
-- See [flashinfer.md](docs/flashinfer.md) on accelerating inference using highly optimized CUDA kernels.
 - See [hyperparameter_tuning.md](docs/hyperparameter_tuning.md) on tuning hyperparameters for better performance.
+- Add `--nnodes 2` to run tensor parallelism on multiple nodes. If you have two nodes with two GPUs on each node and want to run TP=4, let `sgl-dev-1` be the hostname of the first node and `50000` be an available port.
+```
+# Node 0
+python -m sglang.launch_server --model-path meta-llama/Llama-2-7b-chat-hf --tp 4 --nccl-init sgl-dev-1:50000 --nnodes 2 --node-rank 0
+
+# Node 1
+python -m sglang.launch_server --model-path meta-llama/Llama-2-7b-chat-hf --tp 4 --nccl-init sgl-dev-1:50000 --nnodes 2 --node-rank 1
+```
 
 ### Supported Models
 - Llama
 - Mistral
 - Mixtral
-- Qwen / Qwen 2
-- Gemma
-  - Please add a new flag `--attention-reduce-in-fp32` to avoid some precision errors.
+- Qwen / Qwen 2 / Qwen 2 MoE
+- Gemma / Gemma 2
   - `python -m sglang.launch_server --model-path google/gemma-7b-it --port 30000 --attention-reduce-in-fp32`
 - LLaVA
   - `python3 -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`
@@ -632,6 +650,8 @@ python -m sglang.launch_server --model-path meta-llama/Llama-2-7b-chat-hf --port
 - StableLM
 - Command-R
 - DBRX
+- Grok
+- ChatGLM
 - AWQ/GPTQ/Marlin quantization
 
 Instructions for supporting a new model are [here](https://github.com/sgl-project/sglang/blob/main/docs/model_support.md).
@@ -643,17 +663,18 @@ Instructions for supporting a new model are [here](https://github.com/sgl-projec
 - Mixtral-8x7B on NVIDIA A10G, FP16, Tensor Parallelism=8
 ![mixtral_8x7b](assets/mixtral_8x7b.jpg)
 
-Learn more [here](docs/benchmark_results.md).
+- Learn more about the above [results](docs/benchmark_results.md).
+- Synthetic latency and throughput benchmark [scripts](https://github.com/sgl-project/sglang/tree/main/benchmark/latency_throughput).
 
 ## Roadmap
 https://github.com/sgl-project/sglang/issues/157
 
 ## Citation And Acknowledgment
 ```
-@misc{zheng2023efficiently,
-      title={Efficiently Programming Large Language Models using SGLang},
-      author={Lianmin Zheng and Liangsheng Yin and Zhiqiang Xie and Jeff Huang and Chuyue Sun and Cody Hao Yu and Shiyi Cao and Christos Kozyrakis and Ion Stoica and Joseph E. Gonzalez and Clark Barrett and Ying Sheng},
-      year={2023},
+@misc{zheng2024sglang,
+      title={SGLang: Efficient Execution of Structured Language Model Programs},
+      author={Lianmin Zheng and Liangsheng Yin and Zhiqiang Xie and Chuyue Sun and Jeff Huang and Cody Hao Yu and Shiyi Cao and Christos Kozyrakis and Ion Stoica and Joseph E. Gonzalez and Clark Barrett and Ying Sheng},
+      year={2024},
       eprint={2312.07104},
       archivePrefix={arXiv},
       primaryClass={cs.AI}

{sglang-0.1.17 → sglang-0.1.21}/README.md

@@ -10,8 +10,8 @@ SGLang is a structured generation language designed for large language models (L
 It makes your interaction with LLMs faster and more controllable by co-designing the frontend language and the runtime system.
 
 The core features include:
-- **A Flexible Front-End Language**: This allows for easy programming of LLM applications with multiple chained generation calls, advanced prompting techniques, control flow, multiple modalities, parallelism, and external interaction.
-- **A High-Performance Runtime with RadixAttention**: This feature significantly accelerates the execution of complex LLM programs by automatically reusing the KV cache across multiple calls. It can also be used as a standalone serving engine with all common techniques implemented, such as continuous batching and tensor parallelism.
+- **Flexible Frontend Language**: Enables easy programming of LLM applications with chained generation calls, advanced prompting, control flow, multiple modalities, parallelism, and external interactions.
+- **High-Performance Backend Runtime**: Features RadixAttention for accelerating complex LLM programs by reusing the KV cache across multiple calls. It can also serve as a standalone inference engine with all common techniques implemented (e.g., continuous batching and tensor parallelism).
 
 ## News
 - [2024/02] 🔥 SGLang enables **3x faster JSON decoding** with compressed finite state machine ([blog](https://lmsys.org/blog/2024-02-05-compressed-fsm/)).
@@ -32,19 +32,33 @@ The core features include:
 ### Method 1: With pip
 ```
 pip install "sglang[all]"
+
+# Install FlashInfer CUDA kernels
+pip install flashinfer -i https://flashinfer.ai/whl/cu121/torch2.3/
 ```
 
 ### Method 2: From source
 ```
-git clone git@github.com:sgl-project/sglang.git
+git clone https://github.com/sgl-project/sglang.git
 cd sglang
 
-pip install --upgrade pip
 pip install -e "python[all]"
+
+# Install FlashInfer CUDA kernels
+pip install flashinfer -i https://flashinfer.ai/whl/cu121/torch2.3/
 ```
 
-### Notes
-- If you only need to use the OpenAI backend, you can avoid installing other dependencies by using `pip install "sglang[openai]"`
+### Method 3: Using docker
+The docker images are available on Docker Hub as [lmsysorg/sglang](https://hub.docker.com/r/lmsysorg/sglang/tags).
+
+### Common Notes
+- If you see errors from the Triton compiler, please install the [Triton Nightly](https://triton-lang.org/main/getting-started/installation.html) by
+```
+pip uninstall -y triton triton-nightly
+pip install -U --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/Triton-Nightly/pypi/simple/ triton-nightly
+```
+- If you cannot install FlashInfer, check out its [installation](https://docs.flashinfer.ai/installation.html#) page. If you still cannot install it, you can use the slower Triton kernels by adding `--disable-flashinfer` when launching the server.
+- If you only need to use the OpenAI backend, you can avoid installing other dependencies by using `pip install "sglang[openai]"`.
 
 ## Quick Start
 The example below shows how to use sglang to answer a mulit-turn question.
@@ -264,8 +278,8 @@ for out in state.text_iter():
 ```
 
 ### Tips and Implementation Details
-- The `choices` argument in `sgl.gen` is implemented by computing the normalized log probabilities of all choices and selecting the one with the highest probability.
-- The `regex` argument in `sgl.gen` is implemented through autoregressive decoding with logit bias masking, according to the constraints set by the regex.
+- The `choices` argument in `sgl.gen` is implemented by computing the [token-length normalized log probabilities](https://blog.eleuther.ai/multiple-choice-normalization/) of all choices and selecting the one with the highest probability.
+- The `regex` argument in `sgl.gen` is implemented through autoregressive decoding with logit bias masking, according to the constraints set by the regex. It is compatible with `temperature=0` and `temperature != 0`.
 
 ## Backend: SGLang Runtime (SRT)
 The SGLang Runtime (SRT) is designed to work best with the SGLang frontend.
@@ -322,7 +336,6 @@ response = client.chat.completions.create(
 print(response)
 ```
 
-
 By default, the server uses the chat template specified in the model tokenizer from Hugging Face. It should just work for most official models such as Llama-2/Llama-3.
 
 If needed, you can also override the chat template when launching the server:
@@ -351,7 +364,7 @@ python -m sglang.launch_server --model-path meta-llama/Llama-2-7b-chat-hf --port
 ```
 
 ### Additional Arguments
-- Add `--tp 2` to enable tensor parallelism.
+- Add `--tp 2` to enable tensor parallelism. If it indicates `peer access is not supported between these two devices`, add `--enable-p2p-check` option.
 ```
 python -m sglang.launch_server --model-path meta-llama/Llama-2-7b-chat-hf --port 30000 --tp 2
 ```
@@ -363,16 +376,22 @@ python -m sglang.launch_server --model-path meta-llama/Llama-2-7b-chat-hf --port
 ```
 python -m sglang.launch_server --model-path meta-llama/Llama-2-7b-chat-hf --port 30000 --mem-fraction-static 0.7
 ```
-- See [flashinfer.md](docs/flashinfer.md) on accelerating inference using highly optimized CUDA kernels.
 - See [hyperparameter_tuning.md](docs/hyperparameter_tuning.md) on tuning hyperparameters for better performance.
+- Add `--nnodes 2` to run tensor parallelism on multiple nodes. If you have two nodes with two GPUs on each node and want to run TP=4, let `sgl-dev-1` be the hostname of the first node and `50000` be an available port.
+```
+# Node 0
+python -m sglang.launch_server --model-path meta-llama/Llama-2-7b-chat-hf --tp 4 --nccl-init sgl-dev-1:50000 --nnodes 2 --node-rank 0
+
+# Node 1
+python -m sglang.launch_server --model-path meta-llama/Llama-2-7b-chat-hf --tp 4 --nccl-init sgl-dev-1:50000 --nnodes 2 --node-rank 1
+```
 
 ### Supported Models
 - Llama
 - Mistral
 - Mixtral
-- Qwen / Qwen 2
-- Gemma
-  - Please add a new flag `--attention-reduce-in-fp32` to avoid some precision errors.
+- Qwen / Qwen 2 / Qwen 2 MoE
+- Gemma / Gemma 2
   - `python -m sglang.launch_server --model-path google/gemma-7b-it --port 30000 --attention-reduce-in-fp32`
 - LLaVA
   - `python3 -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`
@@ -385,6 +404,8 @@ python -m sglang.launch_server --model-path meta-llama/Llama-2-7b-chat-hf --port
 - StableLM
 - Command-R
 - DBRX
+- Grok
+- ChatGLM
 - AWQ/GPTQ/Marlin quantization
 
 Instructions for supporting a new model are [here](https://github.com/sgl-project/sglang/blob/main/docs/model_support.md).
@@ -396,17 +417,18 @@ Instructions for supporting a new model are [here](https://github.com/sgl-projec
 - Mixtral-8x7B on NVIDIA A10G, FP16, Tensor Parallelism=8
 ![mixtral_8x7b](assets/mixtral_8x7b.jpg)
 
-Learn more [here](docs/benchmark_results.md).
+- Learn more about the above [results](docs/benchmark_results.md).
+- Synthetic latency and throughput benchmark [scripts](https://github.com/sgl-project/sglang/tree/main/benchmark/latency_throughput).
 
 ## Roadmap
 https://github.com/sgl-project/sglang/issues/157
 
 ## Citation And Acknowledgment
 ```
-@misc{zheng2023efficiently,
-      title={Efficiently Programming Large Language Models using SGLang},
-      author={Lianmin Zheng and Liangsheng Yin and Zhiqiang Xie and Jeff Huang and Chuyue Sun and Cody Hao Yu and Shiyi Cao and Christos Kozyrakis and Ion Stoica and Joseph E. Gonzalez and Clark Barrett and Ying Sheng},
-      year={2023},
+@misc{zheng2024sglang,
+      title={SGLang: Efficient Execution of Structured Language Model Programs},
+      author={Lianmin Zheng and Liangsheng Yin and Zhiqiang Xie and Chuyue Sun and Jeff Huang and Cody Hao Yu and Shiyi Cao and Christos Kozyrakis and Ion Stoica and Joseph E. Gonzalez and Clark Barrett and Ying Sheng},
+      year={2024},
       eprint={2312.07104},
       archivePrefix={arXiv},
       primaryClass={cs.AI}

{sglang-0.1.17 → sglang-0.1.21}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "sglang"
-version = "0.1.17"
+version = "0.1.21"
 description = "A structured generation langauge for LLMs."
 readme = "README.md"
 requires-python = ">=3.8"
@@ -16,13 +16,14 @@ classifiers = [
 dependencies = [
     "requests",
     "tqdm",
+    "numpy",
 ]
 
 [project.optional-dependencies]
-srt = ["aiohttp", "fastapi", "psutil", "rpyc", "torch", "uvloop", "uvicorn",
-       "zmq", "vllm==0.4.3", "interegular", "pydantic", "pillow", "packaging", "huggingface_hub", "hf_transfer", "outlines>=0.0.34"]
-openai = ["openai>=1.0", "numpy", "tiktoken"]
-anthropic = ["anthropic>=0.20.0", "numpy"]
+srt = ["aiohttp", "fastapi", "hf_transfer", "huggingface_hub", "interegular", "packaging", "pillow",
+       "psutil", "pydantic", "rpyc", "torch", "uvicorn", "uvloop", "zmq", "vllm==0.5.1", "outlines>=0.0.44"]
+openai = ["openai>=1.0", "tiktoken"]
+anthropic = ["anthropic>=0.20.0"]
 litellm = ["litellm>=1.0.0"]
 all = ["sglang[srt]", "sglang[openai]", "sglang[anthropic]", "sglang[litellm]"]
 

{sglang-0.1.17 → sglang-0.1.21}/sglang/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "0.1.17"
+__version__ = "0.1.21"
 
 # SGL API Components
 from sglang.api import (
@@ -24,10 +24,10 @@ from sglang.api import (
 
 # SGL Backends
 from sglang.backend.anthropic import Anthropic
+from sglang.backend.litellm import LiteLLM
 from sglang.backend.openai import OpenAI
 from sglang.backend.runtime_endpoint import RuntimeEndpoint
 from sglang.backend.vertexai import VertexAI
-from sglang.backend.litellm import LiteLLM
 
 # Global Configurations
 from sglang.global_config import global_config

{sglang-0.1.17 → sglang-0.1.21}/sglang/api.py

@@ -1,4 +1,4 @@
-"""Some Public API Definitions"""
+"""Public APIs of the language."""
 
 import os
 import re
@@ -43,14 +43,14 @@ def set_default_backend(backend: BaseBackend):
     global_config.default_backend = backend
 
 
-def flush_cache(backend: BaseBackend = None):
+def flush_cache(backend: Optional[BaseBackend] = None):
     backend = backend or global_config.default_backend
     if backend is None:
         return False
     return backend.flush_cache()
 
 
-def get_server_args(backend: BaseBackend = None):
+def get_server_args(backend: Optional[BaseBackend] = None):
     backend = backend or global_config.default_backend
     if backend is None:
         return None
@@ -67,10 +67,16 @@ def gen(
     frequency_penalty: Optional[float] = None,
     presence_penalty: Optional[float] = None,
     ignore_eos: Optional[bool] = None,
+    return_logprob: Optional[bool] = None,
+    logprob_start_len: Optional[int] = None,
+    top_logprobs_num: Optional[int] = None,
+    return_text_in_logprobs: Optional[bool] = None,
     dtype: Optional[type] = None,
     choices: Optional[List[str]] = None,
     regex: Optional[str] = None,
 ):
+    """Call the model to generate. See the meaning of the arguments in docs/sampling_params.md"""
+
     if choices:
         return SglSelect(name, choices, 0.0 if temperature is None else temperature)
 
@@ -91,6 +97,10 @@ def gen(
         frequency_penalty,
         presence_penalty,
         ignore_eos,
+        return_logprob,
+        logprob_start_len,
+        top_logprobs_num,
+        return_text_in_logprobs,
         dtype,
         regex,
     )
@@ -106,6 +116,10 @@ def gen_int(
     frequency_penalty: Optional[float] = None,
     presence_penalty: Optional[float] = None,
     ignore_eos: Optional[bool] = None,
+    return_logprob: Optional[bool] = None,
+    logprob_start_len: Optional[int] = None,
+    top_logprobs_num: Optional[int] = None,
+    return_text_in_logprobs: Optional[bool] = None,
 ):
     return SglGen(
         name,
@@ -117,6 +131,10 @@ def gen_int(
         frequency_penalty,
         presence_penalty,
         ignore_eos,
+        return_logprob,
+        logprob_start_len,
+        top_logprobs_num,
+        return_text_in_logprobs,
         int,
         None,
     )
@@ -132,6 +150,10 @@ def gen_string(
     frequency_penalty: Optional[float] = None,
     presence_penalty: Optional[float] = None,
     ignore_eos: Optional[bool] = None,
+    return_logprob: Optional[bool] = None,
+    logprob_start_len: Optional[int] = None,
+    top_logprobs_num: Optional[int] = None,
+    return_text_in_logprobs: Optional[bool] = None,
 ):
     return SglGen(
         name,
@@ -143,6 +165,10 @@ def gen_string(
         frequency_penalty,
         presence_penalty,
         ignore_eos,
+        return_logprob,
+        logprob_start_len,
+        top_logprobs_num,
+        return_text_in_logprobs,
         str,
         None,
     )
@@ -158,7 +184,7 @@ def video(path: str, num_frames: int):
 
 def select(
     name: Optional[str] = None,
-    choices: List[str] = None,
+    choices: Optional[List[str]] = None,
     temperature: float = 0.0,
 ):
     assert choices is not None

{sglang-0.1.17 → sglang-0.1.21}/sglang/backend/litellm.py

@@ -13,7 +13,6 @@ except ImportError as e:
 
 
 class LiteLLM(BaseBackend):
-
     def __init__(
         self,
         model_name,
@@ -33,7 +32,8 @@ class LiteLLM(BaseBackend):
         self.model_name = model_name
 
         self.chat_template = chat_template or get_chat_template_by_model_path(
-            model_name)
+            model_name
+        )
 
         self.client_params = {
             "api_key": api_key,

{sglang-0.1.17 → sglang-0.1.21}/sglang/backend/openai.py

@@ -1,7 +1,7 @@
+import dataclasses
 import logging
 import time
 import warnings
-import dataclasses
 from typing import Callable, List, Optional, Union
 
 import numpy as np
@@ -105,14 +105,16 @@ class OpenAI(BaseBackend):
     def get_chat_template(self):
         return self.chat_template
 
-    def _prepare_spec_execution(self, sampling_params: SglSamplingParams,
-                                num_api_spec_tokens: int, spec_var_name: str):
+    def _prepare_spec_execution(
+        self,
+        sampling_params: SglSamplingParams,
+        num_api_spec_tokens: int,
+        spec_var_name: str,
+    ):
         if "max_tokens" not in self.spec_kwargs:
             self.spec_kwargs["max_tokens"] = num_api_spec_tokens
         else:
-            assert (
-                self.spec_kwargs["max_tokens"] == num_api_spec_tokens
-            )
+            assert self.spec_kwargs["max_tokens"] == num_api_spec_tokens
 
         params = sampling_params.to_openai_kwargs()
         for key, value in params.items():
@@ -151,8 +153,9 @@ class OpenAI(BaseBackend):
                         )
                     prompt = s.messages_
                 else:
-                    return self._prepare_spec_execution(sampling_params,
-                        s.num_api_spec_tokens, spec_var_name)
+                    return self._prepare_spec_execution(
+                        sampling_params, s.num_api_spec_tokens, spec_var_name
+                    )
             else:
                 prompt = s.text_
 
@@ -325,7 +328,7 @@ class OpenAI(BaseBackend):
             ret_str = ret.choices[0].text
             ret_token = self.tokenizer.encode(ret_str)[0]
             self.token_usage.prompt_tokens += ret.usage.prompt_tokens
-            self.token_usage.completion_tokens= ret.usage.completion_tokens
+            self.token_usage.completion_tokens = ret.usage.completion_tokens
 
             # TODO:
             # 1. return logits as the scores
@@ -355,7 +358,9 @@ class OpenAI(BaseBackend):
         return decision, scores, None, None
 
 
-def openai_completion(client, token_usage, is_chat=None, retries=3, prompt=None, **kwargs):
+def openai_completion(
+    client, token_usage, is_chat=None, retries=3, prompt=None, **kwargs
+):
     for attempt in range(retries):
         try:
             if is_chat:
@@ -385,15 +390,19 @@ def openai_completion(client, token_usage, is_chat=None, retries=3, prompt=None,
     return comp
 
 
-def openai_completion_stream(client, token_usage, is_chat=None, retries=3, prompt=None, **kwargs):
+def openai_completion_stream(
+    client, token_usage, is_chat=None, retries=3, prompt=None, **kwargs
+):
     for attempt in range(retries):
         try:
             if is_chat:
                 if "stop" in kwargs and kwargs["stop"] is None:
                     kwargs.pop("stop")
                 generator = client.chat.completions.create(
-                    messages=prompt, stream=True, stream_options={"include_usage": True},
-                    **kwargs
+                    messages=prompt,
+                    stream=True,
+                    stream_options={"include_usage": True},
+                    **kwargs,
                 )
                 for ret in generator:
                     if len(ret.choices) == 0:
@@ -405,8 +414,10 @@ def openai_completion_stream(client, token_usage, is_chat=None, retries=3, promp
                     yield content or "", {}
             else:
                 generator = client.completions.create(
-                    prompt=prompt, stream=True, stream_options={"include_usage": True},
-                    **kwargs
+                    prompt=prompt,
+                    stream=True,
+                    stream_options={"include_usage": True},
+                    **kwargs,
                 )
                 for ret in generator:
                     if len(ret.choices) == 0:

{sglang-0.1.17 → sglang-0.1.21}/sglang/backend/runtime_endpoint.py

@@ -1,15 +1,14 @@
 import json
-from typing import Callable, List, Optional, Union
+from typing import List, Optional
 
 import numpy as np
-import requests
 
 from sglang.backend.base_backend import BaseBackend
 from sglang.global_config import global_config
 from sglang.lang.chat_template import get_chat_template_by_model_path
 from sglang.lang.interpreter import StreamExecutor
-from sglang.lang.ir import SglArgument, SglSamplingParams
-from sglang.utils import encode_image_base64, find_printable_text, http_request
+from sglang.lang.ir import SglSamplingParams
+from sglang.utils import http_request
 
 
 class RuntimeEndpoint(BaseBackend):
@@ -125,6 +124,16 @@ class RuntimeEndpoint(BaseBackend):
         else:
             raise RuntimeError(f"Invalid dtype: {sampling_params.dtype}")
 
+        for item in [
+            "return_logprob",
+            "logprob_start_len",
+            "top_logprobs_num",
+            "return_text_in_logprobs",
+        ]:
+            value = getattr(sampling_params, item, None)
+            if value is not None:
+                data[item] = value
+
         self._add_images(s, data)
 
         res = http_request(
@@ -167,6 +176,16 @@ class RuntimeEndpoint(BaseBackend):
         else:
             raise RuntimeError(f"Invalid dtype: {sampling_params.dtype}")
 
+        for item in [
+            "return_logprob",
+            "logprob_start_len",
+            "top_logprobs_num",
+            "return_text_in_logprobs",
+        ]:
+            value = getattr(sampling_params, item, None)
+            if value is not None:
+                data[item] = value
+
         data["stream"] = True
         self._add_images(s, data)
 
@@ -181,21 +200,16 @@ class RuntimeEndpoint(BaseBackend):
         self._assert_success(res)
         pos = 0
 
-        incomplete_text = ""
         for chunk in res.iter_lines(decode_unicode=False):
             chunk = chunk.decode("utf-8")
             if chunk and chunk.startswith("data:"):
                 if chunk == "data: [DONE]":
                     break
                 data = json.loads(chunk[5:].strip("\n"))
-                text = find_printable_text(data["text"][pos:])
+                chunk_text = data["text"][pos:]
                 meta_info = data["meta_info"]
-                pos += len(text)
-                incomplete_text = data["text"][pos:]
-                yield text, meta_info
-
-        if len(incomplete_text) > 0:
-            yield incomplete_text, meta_info
+                pos += len(chunk_text)
+                yield chunk_text, meta_info
 
     def select(
         self,