chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 11 Packages Projects Releases Wiki Activity

[Minor] update docs (#14212)

Browse Source
This commit is contained in:
Lianmin Zheng
2025-12-01 02:33:58 -08:00
committed by GitHub
parent d815d00248
commit bc3d2a85af
28 changed files with 48 additions and 46 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
docs/basic_usage/deepseek_v3.md
View File

@@ -153,7 +153,7 @@ python3 -m sglang.compile_deep_gemm --model deepseek-ai/DeepSeek-V3 --tp 8 --tru
The precompilation process typically takes around 10 minutes to complete.
### Multi-token Prediction
**Description**: SGLang implements DeepSeek V3 Multi-Token Prediction (MTP) based on [EAGLE speculative decoding](https://docs.sglang.ai/advanced_features/speculative_decoding.html#EAGLE-Decoding). With this optimization, the decoding speed can be improved by **1.8x** for batch size 1 and **1.5x** for batch size 32 respectively on H200 TP8 setting.
**Description**: SGLang implements DeepSeek V3 Multi-Token Prediction (MTP) based on [EAGLE speculative decoding](https://docs.sglang.io/advanced_features/speculative_decoding.html#EAGLE-Decoding). With this optimization, the decoding speed can be improved by **1.8x** for batch size 1 and **1.5x** for batch size 32 respectively on H200 TP8 setting.
**Usage**:
Add arguments `--speculative-algorithm`, `--speculative-num-steps`, `--speculative-eagle-topk` and `--speculative-num-draft-tokens` to enable this feature. For example:
@@ -176,7 +176,7 @@ python3 -m sglang.launch_server \
### Reasoning Content for DeepSeek R1 & V3.1
See [Reasoning Parser](https://docs.sglang.ai/advanced_features/separate_reasoning.html) and [Thinking Parameter for DeepSeek V3.1](https://docs.sglang.ai/basic_usage/openai_api_completions.html#Example:-DeepSeek-V3-Models).
See [Reasoning Parser](https://docs.sglang.io/advanced_features/separate_reasoning.html) and [Thinking Parameter for DeepSeek V3.1](https://docs.sglang.io/basic_usage/openai_api_completions.html#Example:-DeepSeek-V3-Models).
### Function calling for DeepSeek Models

4
docs/basic_usage/deepseek_v32.md
View File

@@ -59,7 +59,7 @@ python -m sglang.launch_server --model deepseek-ai/DeepSeek-V3.2-Exp --tp 8
- B200: `flashmla_auto` prefill attention (short-seq prefill uses MHA via TRT-LLM ragged), `flashmla_kv` decode attention, `fp8_e4m3` kv cache dtype. `flashmla_auto` enables automatic selection of either `flashmla_sparse` or `flashmla_kv` kernel for prefill based on KV cache dtype, hardware, and heuristics. When FP8 KV cache is enabled and `total_kv_tokens < total_q_tokens * 512`, it uses the `flashmla_sparse` kernel; otherwise, it falls back to the `flashmla_kv` kernel. The heuristics may need to be tuned if the performance of either the `flashmla_sparse` or `flashmla_kv` kernel changes significantly.
## Multi-token Prediction
SGLang implements Multi-Token Prediction (MTP) for DeepSeek V3.2 based on [EAGLE speculative decoding](https://docs.sglang.ai/advanced_features/speculative_decoding.html#EAGLE-Decoding). With this optimization, the decoding speed can be improved significantly on small batch sizes. Please look at [this PR](https://github.com/sgl-project/sglang/pull/11652) for more information.
SGLang implements Multi-Token Prediction (MTP) for DeepSeek V3.2 based on [EAGLE speculative decoding](https://docs.sglang.io/advanced_features/speculative_decoding.html#EAGLE-Decoding). With this optimization, the decoding speed can be improved significantly on small batch sizes. Please look at [this PR](https://github.com/sgl-project/sglang/pull/11652) for more information.
Example usage:
```bash
@@ -70,7 +70,7 @@ python -m sglang.launch_server --model deepseek-ai/DeepSeek-V3.2-Exp --tp 8 --dp
## Function Calling and Reasoning Parser
The usage of function calling and reasoning parser is the same as DeepSeek V3.1. Please refer to [Reasoning Parser](https://docs.sglang.ai/advanced_features/separate_reasoning.html) and [Tool Parser](https://docs.sglang.ai/advanced_features/tool_parser.html) documents.
The usage of function calling and reasoning parser is the same as DeepSeek V3.1. Please refer to [Reasoning Parser](https://docs.sglang.io/advanced_features/separate_reasoning.html) and [Tool Parser](https://docs.sglang.io/advanced_features/tool_parser.html) documents.
## PD Disaggregation

2
docs/basic_usage/llama4.md
View File

@@ -36,7 +36,7 @@ python3 -m sglang.launch_server \
### EAGLE Speculative Decoding
**Description**: SGLang has supported Llama 4 Maverick (400B) with [EAGLE speculative decoding](https://docs.sglang.ai/advanced_features/speculative_decoding.html#EAGLE-Decoding).
**Description**: SGLang has supported Llama 4 Maverick (400B) with [EAGLE speculative decoding](https://docs.sglang.io/advanced_features/speculative_decoding.html#EAGLE-Decoding).
**Usage**:
Add arguments `--speculative-draft-model-path`, `--speculative-algorithm`, `--speculative-num-steps`, `--speculative-eagle-topk` and `--speculative-num-draft-tokens` to enable this feature. For example:

2
docs/basic_usage/qwen3.md
View File

@@ -15,7 +15,7 @@ python3 -m sglang.launch_server --model Qwen/Qwen3-Next-80B-A3B-Instruct --tp 4
- `--mamba-ssm-dtype`: `bfloat16` or `float32`, use `bfloat16` to save mamba cache size and `float32` to get more accurate results. The default setting is `float32`.
### EAGLE Speculative Decoding
**Description**: SGLang has supported Qwen3-Next models with [EAGLE speculative decoding](https://docs.sglang.ai/advanced_features/speculative_decoding.html#EAGLE-Decoding).
**Description**: SGLang has supported Qwen3-Next models with [EAGLE speculative decoding](https://docs.sglang.io/advanced_features/speculative_decoding.html#EAGLE-Decoding).
**Usage**:
Add arguments `--speculative-algorithm`, `--speculative-num-steps`, `--speculative-eagle-topk` and `--speculative-num-draft-tokens` to enable this feature. For example: