@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / 36cacd0958

Fork 0

A high-throughput and memory-efficient inference and serving engine for LLMs

Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git

Synced Nov 14, 05:30 UTC

[Doc] Add documentation for GLM-4.5 series models: tool-calling and reasoning parser (#24589)

Signed-off-by: WangErXiao <863579016@qq.com>

Robin committed Sep 10, 2025 at 22:50 UTC 36cacd09584cb16668487105b04b3ae8193fa28b

2 files changed +10

docs/features/reasoning_outputs.md

@@ -15,6 +15,7 @@ vLLM currently supports the following reasoning models:

15 | [IBM Granite 3.2 language models](https://huggingface.co/collections/ibm-granite/granite-32-language-models-67b3bc8c13508f6d064cff9a) | `granite` | ❌ | ❌ |

18 +| [GLM-4.5 series](https://huggingface.co/collections/zai-org/glm-45-687c621d34bda8c9e4bf503b) | `glm45` | `guided_json`, `guided_regex` | ✅ |

20 !!! note

21 IBM Granite 3.2 reasoning is disabled by default; to enable it, you must also pass `thinking=True` in your `chat_template_kwargs`.

docs/features/tool_calling.md

@@ -311,6 +311,15 @@ Flags:

311 * For non-reasoning: `--tool-call-parser hunyuan_a13b`

312 * For reasoning: `--tool-call-parser hunyuan_a13b --reasoning-parser hunyuan_a13b --enable_reasoning`

313

314 +### GLM-4.5 Models (`glm45`)

315 +

316 +Supported models:

317 +

318 +* `ZhipuAI/GLM-4.5`

319 +* `ZhipuAI/GLM-4.5-Air`

320 +

321 +Flags: `--tool-call-parser glm45`

322 +

323 ### Models with Pythonic Tool Calls (`pythonic`)

324

325 A growing number of models output a python list to represent tool calls instead of using JSON. This has the advantage of inherently supporting parallel tool calls and removing ambiguity around the JSON schema required for tool calls. The `pythonic` tool parser can support such models.