Skip to content

Commit c5a9254

Browse files
author
angazenn
committed
change doc description
Signed-off-by: angazenn <zengyanjia@huawei.com>
1 parent 72249ab commit c5a9254

File tree

2 files changed

+5
-5
lines changed

2 files changed

+5
-5
lines changed

docs/source/user_guide/additional_config.md

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -40,14 +40,14 @@ The details of each config option are as follows:
4040

4141
| Name | Type | Default | Description |
4242
| ---- | ---- | ------- | ----------- |
43-
| `enabled` | bool | `False` | Whether to enable torchair graph mode |
44-
| `enable_multistream_mla`| bool | `False` | Whether to put vector ops of MLA to another stream |
45-
| `enable_multistream_moe`| bool | `False` | Whether to enable multistream shared expert |
43+
| `enabled` | bool | `False` | Whether to enable torchair graph mode. Currently only DeepSeek series models and PanguProMoE are supported to use torchair graph mode |
44+
| `enable_multistream_mla`| bool | `False` | Whether to put vector ops of MLA to another stream. This option only takes effects on models using MLA (e.g., DeepSeek). |
45+
| `enable_multistream_moe`| bool | `False` | Whether to enable multistream shared expert. This option only takes effects on DeepSeek moe models. |
4646
| `enable_view_optimize` | bool | `True` | Whether to enable torchair view optimization |
4747
| `use_cached_graph` | bool | `False` | Whether to use cached graph |
4848
| `graph_batch_sizes` | list[int] | `[]` | The batch size for torchair graph cache |
4949
| `graph_batch_sizes_init` | bool | `False` | Init graph batch size dynamically if `graph_batch_sizes` is empty |
50-
| `enable_kv_nz`| bool | `False` | Whether to enable kvcache NZ layout |
50+
| `enable_kv_nz`| bool | `False` | Whether to enable kvcache NZ layout. This option only takes effects on models using MLA (e.g., DeepSeek). |
5151

5252
**ascend_scheduler_config**
5353

docs/source/user_guide/graph_mode.md

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -12,7 +12,7 @@ From v0.9.1rc1 with V1 Engine, vLLM Ascend will run models in graph mode by defa
1212

1313
There are two kinds for graph mode supported by vLLM Ascend:
1414
- **ACLGraph**: This is the default graph mode supported by vLLM Ascend. In v0.9.1rc1, only Qwen series models are well tested.
15-
- **TorchAirGraph**: This is the GE graph mode. In v0.9.1rc1, only DeepSeek series models are supported.
15+
- **TorchAirGraph**: This is the GE graph mode. In v0.9.1rc1, only DeepSeek series models are supported. In v0.9.1rc2, we also support PanguProMoe with torchair.
1616

1717
## Using ACLGraph
1818
ACLGraph is enabled by default. Take Qwen series models as an example, just set to use V1 Engine is enough.

0 commit comments

Comments
 (0)