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 15, 11:00 UTC
[Doc] Update several links in reasoning_outputs.md (#17846)
Signed-off-by: windsonsea <haifeng.yao@daocloud.io>
Michael Yao committed
May 9, 2025 at 16:20 UTC
89a0315f4cdf34da40d7634a069068c09082b7dc
1 file changed
+8
-6
docs/source/features/reasoning_outputs.md
+8
-6
@@ -17,7 +17,9 @@ vLLM currently supports the following reasoning models:
17
| [IBM Granite 3.2 language models](https://huggingface.co/collections/ibm-granite/granite-32-language-models-67b3bc8c13508f6d064cff9a) | `granite` | ❌ | ❌ |
18
| [Qwen3 series](https://huggingface.co/collections/Qwen/qwen3-67dd247413f0e2e4f653967f) | `qwen3` | `guided_json`, `guided_regex` | ✅ |
19
20
-- IBM Granite 3.2 reasoning is disabled by default; to enable it, you must also pass `thinking=True` in your `chat_template_kwargs`.
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`.
22
+:::
23
24
## Quickstart
25
@@ -83,7 +85,7 @@ Streaming chat completions are also supported for reasoning models. The `reasoni
85
}
86
```
87
86
-OpenAI Python client library does not officially support `reasoning_content` attribute for streaming output. But the client support extra attributes in the response. You can use `hasattr` to check if the `reasoning_content` attribute is present in the response. For example:
88
+OpenAI Python client library does not officially support `reasoning_content` attribute for streaming output. But the client supports extra attributes in the response. You can use `hasattr` to check if the `reasoning_content` attribute is present in the response. For example:
89
90
```python
91
from openai import OpenAI
@@ -221,7 +223,7 @@ print(f"Function called: {tool_call.name}")
223
print(f"Arguments: {tool_call.arguments}")
224
```
225
224
-For more examples, please refer to <gh-file:examples/online_serving/openai_chat_completion_tool_calls_with_reasoning.py> .
226
+For more examples, please refer to <gh-file:examples/online_serving/openai_chat_completion_tool_calls_with_reasoning.py>.
227
228
## Limitations
229
@@ -229,7 +231,7 @@ For more examples, please refer to <gh-file:examples/online_serving/openai_chat_
231
232
## How to support a new reasoning model
233
232
-You can add a new `ReasoningParser` similar to `vllm/entrypoints/openai/reasoning_parsers/deepseek_r1_reasoning_parser.py`.
234
+You can add a new `ReasoningParser` similar to <gh-file:vllm/entrypoints/openai/reasoning_parsers/deepseek_r1_reasoning_parser.py>.
235
236
```python
237
# import the required packages
@@ -286,7 +288,7 @@ class ExampleParser(ReasoningParser):
288
"""
289
```
290
289
-Additionally, to enable structured output, you'll need to create a new `Reasoner` similar to the one in `vllm/model_executor/guided_decoding/reasoner/deepseek_reasoner.py`.
291
+Additionally, to enable structured output, you'll need to create a new `Reasoner` similar to the one in <gh-file:vllm/model_executor/guided_decoding/reasoner/deepseek_reasoner.py>.
292
293
```python
294
@dataclass
@@ -312,7 +314,7 @@ class DeepSeekReasoner(Reasoner):
314
...
315
```
316
315
-The structured output engine like `xgrammar` will use `end_token_id` to check if the reasoning content is present in the model output and skip the structured output if it is the case.
317
+The structured output engine like [xgrammar](https://github.com/mlc-ai/xgrammar) will use `end_token_id` to check if the reasoning content is present in the model output and skip the structured output if it is the case.
318
319
Finally, you can enable reasoning for the model by using the `--reasoning-parser` flags.
320