[docs] Update guidance on how to implement and register new models #1126
+247
−0
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
github-actions
bot
added
the
documentation
|
mv this doc to developer guide |
linfeng-yuan
force-pushed
the
model_docs
branch
2 times, most recently
from
20e817a to
cc2016c
Compare
… vLLM-Ascend Signed-off-by: linfeng-yuan <1102311262@qq.com>
|
@shen-shanshan could help review. |
| This guide demonstrates how to integrate novel or customized models into vLLM-Ascend. For foundational concepts, it is highly recommended to refer to: | ||
| [Adding a New Model - vLLM Documentation](https://docs.vllm.ai/en/stable/contributing/model/) | ||
|
|
||
| ### 1. Implementing Models Using PyTorch and Ascend Extension for PyTorch |
There was a problem hiding this comment.
Please fix the CI failure:
WARNING: Non-consecutive header level increase; H1 to H3 [myst.header]
Use ## before ###.
| - `*Model` (main architecture) | ||
| - `*DecoderLayer` (transformer block) | ||
| - `*Attention` & `*MLP` (specific computation unit) | ||
| `*` denotes your model's unique identifier |
There was a problem hiding this comment.
It's better to add this line into a NOTE block.
|
|
||
|
|
||
| (4) **Attention Backend Integration**: | ||
| Import attention via `from vllm.attention import Attention` can automatically leverage vLLM-Ascend's attention backend routing (see: `get_attn_backend_cls()` in [vllm_ascend/platform.py](https://github.com/vllm-project/vllm-ascend/blob/main/vllm_ascend/platform.py)) |
There was a problem hiding this comment.
Please make sure adding a . at the end of each sentences.
| Import attention via `from vllm.attention import Attention` can automatically leverage vLLM-Ascend's attention backend routing (see: `get_attn_backend_cls()` in [vllm_ascend/platform.py](https://github.com/vllm-project/vllm-ascend/blob/main/vllm_ascend/platform.py)) | ||
|
|
||
| (5) **Tensor Parallelism**: | ||
| Use vLLM's parallel layers (`ColumnParallelLinear`, `VocabParallelEmbedding`, etc.), but note Ascend-specific customizations implemented in [vllm_ascend/ops/](https://github.com/vllm-project/vllm-ascend/tree/main/vllm_ascend/ops) directory (RMSNorm, VocabParallelEmbedding, etc.). |
There was a problem hiding this comment.
I think this is a little bit confused, please make it clear of when to use vLLM's parallel layers and when to use ascend-specific customizations.
|
|
||
| For a complete implementation reference, see: [vllm_ascend/models/deepseek_v2.py](https://github.com/vllm-project/vllm-ascend/blob/main/vllm_ascend/models/deepseek_v2.py) | ||
|
|
||
| ### 2. Registering Custom Models as Out-of-Tree Plugins in vLLM |
There was a problem hiding this comment.
Registering Custom Out-of-Tree Models using ModelRegistry Plugins in vLLM
Comment on lines
+200
to
+216
| **Key Note** | ||
| The first argument of `vllm.ModelRegistry.register_model()` indicates the unique architecture identifier which must match 'architectures' in `config.json` of the model. | ||
|
|
||
| ```json | ||
| { | ||
| "architectures": [ | ||
| "CustomModelForCausalLM" | ||
| ], | ||
| } | ||
| ``` | ||
|
|
||
| ```json | ||
| { | ||
| "architectures": [ | ||
| "DeepseekV2ForCausalLM" | ||
| ], | ||
| } |
There was a problem hiding this comment.
It's better to add this line into a NOTE block.
| If you're registering a novel model architecture not present in vLLM (creating a completely new class), current logs won't provide explicit confirmation by default. It's recommended to add the following logging statement at the end of the `register_model` method in [vllm/models\_executor/models/registry.py](https://github.com/vllm-project/vllm/blob/main/vllm/model_executor/models/registry.py) : | ||
|
|
||
| ``` | ||
| logger.warning(f"model_arch: {model_arch} has been registered here!") |
There was a problem hiding this comment.
Why use warning? I think use info may be better.
|
@linfeng-yuan I think it's better to add more contents like:
|
|
developer_guide/index should be updated as well |
|
@shen-shanshan it's good to create a new one with co-author, if the path is not updated for a long time. |
OK. |
wangxiyuan
pushed a commit
that referenced
this pull request
Jun 27, 2025
### What this PR does / why we need it? Add guidance on how to implement and register new models. Modified based on PR #1126, thanks for the contribution of @linfeng-yuan. --------- Signed-off-by: shen-shanshan <467638484@qq.com>
zhanghw0354
pushed a commit
to zhanghw0354/vllm-ascend
that referenced
this pull request
Jun 30, 2025
…project#1426) ### What this PR does / why we need it? Add guidance on how to implement and register new models. Modified based on PR vllm-project#1126, thanks for the contribution of @linfeng-yuan. --------- Signed-off-by: shen-shanshan <467638484@qq.com> Signed-off-by: zhanghw0354 <zhanghaiwen_yewu@cmss.chinamobile.com>
weijinqian0
pushed a commit
to weijinqian0/vllm-ascend
that referenced
this pull request
Jun 30, 2025
…project#1426) ### What this PR does / why we need it? Add guidance on how to implement and register new models. Modified based on PR vllm-project#1126, thanks for the contribution of @linfeng-yuan. --------- Signed-off-by: shen-shanshan <467638484@qq.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
… vLLM-Ascend
What this PR does / why we need it?
Update guidance on implementing and registering models for users. [Only documentation]
Does this PR introduce any user-facing change?
No.
How was this patch tested?
Not involve codes and functions.