Skip to content

Commit 7d84862

Browse files
ckl117ckl117
committed
Update eb45-0.3B cuda memory
1 parent de20e5a commit 7d84862

File tree

2 files changed

+43
-42
lines changed

2 files changed

+43
-42
lines changed

‎docs/supported_models.md

Lines changed: 23 additions & 22 deletions
Original file line numberDiff line numberDiff line change
@@ -2,34 +2,35 @@
22

33
FastDeploy currently supports the following models, which can be downloaded via three methods:
44

5-
- 1. During FastDeploy deployment, specify the ```model``` parameter as the model name in the table below to automatically download model weights from AIStudio (supports resumable downloads)
5+
- 1. During FastDeploy deployment, specify the ``model`` parameter as the model name in the table below to automatically download model weights from AIStudio (supports resumable downloads)
66
- 2. Download Paddle-version ERNIE models from [HuggingFace/baidu/models](https://huggingface.co/baidu/models), e.g., `baidu/ERNIE-4.5-0.3B-Paddle`
77
- 3. Search for corresponding Paddle-version ERNIE models on [ModelScope/PaddlePaddle](https://www.modelscope.cn/models?name=PaddlePaddle&page=1&tabKey=task), e.g., `ERNIE-4.5-0.3B-Paddle`
88

9-
For the first method (auto-download), the default download path is ```~/``` (user home directory). Users can modify this path by setting the ```FD_MODEL_CACHE``` environment variable, e.g.:
9+
For the first method (auto-download), the default download path is ``~/`` (user home directory). Users can modify this path by setting the ``FD_MODEL_CACHE`` environment variable, e.g.:
10+
1011
```bash
1112
export FD_MODEL_CACHE=/ssd1/download_models
1213
```
1314

14-
| Model Name | Context Length | Quantization | Minimum Deployment Resources | Notes |
15-
| :--------- | :------------- | :----------- | :-------------------------- | :---- |
16-
| baidu/ERNIE-4.5-VL-424B-A47B-Paddle | 32K/128K | WINT2 | 1*141G GPU VRAM/1T RAM | Chunked Prefill required for 128K |
17-
| baidu/ERNIE-4.5-VL-424B-A47B-Paddle | 32K/128K | WINT4 | 4*80G GPU VRAM/1T RAM | Chunked Prefill required for 128K |
18-
| baidu/ERNIE-4.5-VL-424B-A47B-Paddle | 32K/128K | WINT8 | 8*80G GPU VRAM/1T RAM | Chunked Prefill required for 128K |
19-
| baidu/ERNIE-4.5-300B-A47B-Paddle | 32K/128K | WINT4 | 4*64G GPU VRAM/600G RAM | Chunked Prefill required for 128K |
20-
| baidu/ERNIE-4.5-300B-A47B-Paddle | 32K/128K | WINT8 | 8*64G GPU VRAM/600G RAM | Chunked Prefill required for 128K |
21-
| baidu/ERNIE-4.5-300B-A47B-W4A8C8-TP4-Paddle | 32K/128K | W4A8C8 | 4*64G GPU VRAM/160G RAM | Fixed 4-GPU setup, Chunked Prefill recommended |
22-
| baidu/ERNIE-4.5-300B-A47B-FP8-Paddle | 32K/128K | FP8 | 8*64G GPU VRAM/600G RAM | Chunked Prefill recommended, only supports PD Disaggragated Deployment with EP parallelism |
23-
| baidu/ERNIE-4.5-300B-A47B-Base-Paddle | 32K/128K | WINT4 | 4*64G GPU VRAM/600G RAM | Chunked Prefill recommended |
24-
| baidu/ERNIE-4.5-300B-A47B-Base-Paddle | 32K/128K | WINT8 | 8*64G GPU VRAM/600G RAM | Chunked Prefill recommended |
25-
| baidu/ERNIE-4.5-VL-28B-A3B-Paddle | 32K | WINT4 | 1*24G GPU VRAM/128G RAM | Chunked Prefill required |
26-
| baidu/ERNIE-4.5-VL-28B-A3B-Paddle | 128K | WINT4 | 1*48G GPU VRAM/128G RAM | Chunked Prefill required |
27-
| baidu/ERNIE-4.5-VL-28B-A3B-Paddle | 32K/128K | WINT8 | 1*48G GPU VRAM/128G RAM | Chunked Prefill required |
28-
| baidu/ERNIE-4.5-21B-A3B-Paddle | 32K/128K | WINT4 | 1*24G GPU VRAM/128G RAM | Chunked Prefill required for 128K |
29-
| baidu/ERNIE-4.5-21B-A3B-Paddle | 32K/128K | WINT8 | 1*48G GPU VRAM/128G RAM | Chunked Prefill required for 128K |
30-
| baidu/ERNIE-4.5-21B-A3B-Base-Paddle | 32K/128K | WINT4 | 1*24G GPU VRAM/128G RAM | Chunked Prefill required for 128K |
31-
| baidu/ERNIE-4.5-21B-A3B-Base-Paddle | 32K/128K | WINT8 | 1*48G GPU VRAM/128G RAM | Chunked Prefill required for 128K |
32-
| baidu/ERNIE-4.5-0.3B-Paddle | 32K/128K | BF16 | 1*16G GPU VRAM/2G RAM | |
33-
| baidu/ERNIE-4.5-0.3B-Base-Paddle | 32K/128K | BF16 | 1*16G GPU VRAM/2G RAM | |
15+
| Model Name | Context Length | Quantization | Minimum Deployment Resources | Notes |
16+
| :------------------------------------------ | :------------- | :----------- | :--------------------------- | :----------------------------------------------------------------------------------------- |
17+
| baidu/ERNIE-4.5-VL-424B-A47B-Paddle | 32K/128K | WINT2 | 1*141G GPU VRAM/1T RAM | Chunked Prefill required for 128K |
18+
| baidu/ERNIE-4.5-VL-424B-A47B-Paddle | 32K/128K | WINT4 | 4*80G GPU VRAM/1T RAM | Chunked Prefill required for 128K |
19+
| baidu/ERNIE-4.5-VL-424B-A47B-Paddle | 32K/128K | WINT8 | 8*80G GPU VRAM/1T RAM | Chunked Prefill required for 128K |
20+
| baidu/ERNIE-4.5-300B-A47B-Paddle | 32K/128K | WINT4 | 4*64G GPU VRAM/600G RAM | Chunked Prefill required for 128K |
21+
| baidu/ERNIE-4.5-300B-A47B-Paddle | 32K/128K | WINT8 | 8*64G GPU VRAM/600G RAM | Chunked Prefill required for 128K |
22+
| baidu/ERNIE-4.5-300B-A47B-W4A8C8-TP4-Paddle | 32K/128K | W4A8C8 | 4*64G GPU VRAM/160G RAM | Fixed 4-GPU setup, Chunked Prefill recommended |
23+
| baidu/ERNIE-4.5-300B-A47B-FP8-Paddle | 32K/128K | FP8 | 8*64G GPU VRAM/600G RAM | Chunked Prefill recommended, only supports PD Disaggragated Deployment with EP parallelism |
24+
| baidu/ERNIE-4.5-300B-A47B-Base-Paddle | 32K/128K | WINT4 | 4*64G GPU VRAM/600G RAM | Chunked Prefill recommended |
25+
| baidu/ERNIE-4.5-300B-A47B-Base-Paddle | 32K/128K | WINT8 | 8*64G GPU VRAM/600G RAM | Chunked Prefill recommended |
26+
| baidu/ERNIE-4.5-VL-28B-A3B-Paddle | 32K | WINT4 | 1*24G GPU VRAM/128G RAM | Chunked Prefill required |
27+
| baidu/ERNIE-4.5-VL-28B-A3B-Paddle | 128K | WINT4 | 1*48G GPU VRAM/128G RAM | Chunked Prefill required |
28+
| baidu/ERNIE-4.5-VL-28B-A3B-Paddle | 32K/128K | WINT8 | 1*48G GPU VRAM/128G RAM | Chunked Prefill required |
29+
| baidu/ERNIE-4.5-21B-A3B-Paddle | 32K/128K | WINT4 | 1*24G GPU VRAM/128G RAM | Chunked Prefill required for 128K |
30+
| baidu/ERNIE-4.5-21B-A3B-Paddle | 32K/128K | WINT8 | 1*48G GPU VRAM/128G RAM | Chunked Prefill required for 128K |
31+
| baidu/ERNIE-4.5-21B-A3B-Base-Paddle | 32K/128K | WINT4 | 1*24G GPU VRAM/128G RAM | Chunked Prefill required for 128K |
32+
| baidu/ERNIE-4.5-21B-A3B-Base-Paddle | 32K/128K | WINT8 | 1*48G GPU VRAM/128G RAM | Chunked Prefill required for 128K |
33+
| baidu/ERNIE-4.5-0.3B-Paddle | 32K/128K | BF16 | 1*6G/12G GPU VRAM/2G RAM | |
34+
| baidu/ERNIE-4.5-0.3B-Base-Paddle | 32K/128K | BF16 | 1*6G/12G GPU VRAM/2G RAM | |
3435

3536
More models are being supported. You can submit requests for new model support via [Github Issues](https://github.com/PaddlePaddle/FastDeploy/issues).

‎docs/zh/supported_models.md

Lines changed: 20 additions & 20 deletions
Original file line numberDiff line numberDiff line change
@@ -12,25 +12,25 @@ FastDeploy目前支持模型列表如下，以下模型提供如下3种下载方
1212
export FD_MODEL_CACHE=/ssd1/download_models
1313
```
1414

15-
| 模型名 | 上下文长度 | 量化方式 | 最小部署资源 | 说明 |
16-
| :----- | :-------------- | :----------- |:----------- |:----------- |
17-
| baidu/ERNIE-4.5-VL-424B-A47B-Paddle | 32K/128K | WINT2 | 1卡*141G显存/1T内存 | 128K需要开启Chunked Prefill |
18-
| baidu/ERNIE-4.5-VL-424B-A47B-Paddle | 32K/128K | WINT4 | 4卡*80G显存/1T内存 | 128K需要开启Chunked Prefill |
19-
| baidu/ERNIE-4.5-VL-424B-A47B-Paddle | 32K/128K | WINT8 | 8卡*80G显存/1T内存 | 128K需要开启Chunked Prefill |
20-
| baidu/ERNIE-4.5-300B-A47B-Paddle | 32K/128K | WINT4 | 4卡*64G显存/600G内存 | 128K需要开启Chunked Prefill |
21-
| baidu/ERNIE-4.5-300B-A47B-Paddle | 32K/128K | WINT8 | 8卡*64G显存/600G内存 | 128K需要开启Chunked Prefill |
22-
| baidu/ERNIE-4.5-300B-A47B-W4A8C8-TP4-Paddle | 32K/128K | W4A8C8 | 4卡*64G显存/160G内存 | 限定4卡，建议开启Chunked Prefill |
23-
| baidu/ERNIE-4.5-300B-A47B-FP8-Paddle| 32K/128K | FP8 | 8卡*64G显存/600G内存 | 建议开启Chunked Prefill，仅在PD分离EP并行下支持 |
24-
| baidu/ERNIE-4.5-300B-A47B-Base-Paddle | 32K/128K | WINT4 | 4卡*64G显存/600G内存 | 建议开启Chunked Prefill |
25-
| baidu/ERNIE-4.5-300B-A47B-Base-Paddle | 32K/128K | WINT8 | 8卡*64G显存/600G内存 | 建议开启Chunked Prefill |
26-
| baidu/ERNIE-4.5-VL-28B-A3B-Paddle | 32K | WINT4 | 1卡*24G/128G内存 | 需要开启Chunked Prefill |
27-
| baidu/ERNIE-4.5-VL-28B-A3B-Paddle | 128K | WINT4 | 1卡*48G/128G内存 | 需要开启Chunked Prefill |
28-
| baidu/ERNIE-4.5-VL-28B-A3B-Paddle | 32K/128K | WINT8 | 1卡*48G/128G内存 | 需要开启Chunked Prefill |
29-
| baidu/ERNIE-4.5-21B-A3B-Paddle | 32K/128K | WINT4 | 1卡*24G/128G内存 | 128K需要开启Chunked Prefill |
30-
| baidu/ERNIE-4.5-21B-A3B-Paddle | 32K/128K | WINT8 | 1卡*48G/128G内存 | 128K需要开启Chunked Prefill |
31-
| baidu/ERNIE-4.5-21B-A3B-Base-Paddle | 32K/128K | WINT4 | 1卡*24G/128G内存 | 128K需要开启Chunked Prefill |
32-
| baidu/ERNIE-4.5-21B-A3B-Base-Paddle | 32K/128K | WINT8 | 1卡*48G/128G内存 | 128K需要开启Chunked Prefill |
33-
| baidu/ERNIE-4.5-0.3B-Paddle | 32K/128K | BF16 | 1卡*16G显存/2G内存 | |
34-
| baidu/ERNIE-4.5-0.3B-Base-Paddle | 32K/128K | BF16 | 1卡*16G显存/2G内存 | |
15+
| 模型名 | 上下文长度 | 量化方式 | 最小部署资源 | 说明 |
16+
| :------------------------------------------ | :--------- | :------- | :-------------------- | :---------------------------------------------- |
17+
| baidu/ERNIE-4.5-VL-424B-A47B-Paddle | 32K/128K | WINT2 | 1卡*141G显存/1T内存 | 128K需要开启Chunked Prefill |
18+
| baidu/ERNIE-4.5-VL-424B-A47B-Paddle | 32K/128K | WINT4 | 4卡*80G显存/1T内存 | 128K需要开启Chunked Prefill |
19+
| baidu/ERNIE-4.5-VL-424B-A47B-Paddle | 32K/128K | WINT8 | 8卡*80G显存/1T内存 | 128K需要开启Chunked Prefill |
20+
| baidu/ERNIE-4.5-300B-A47B-Paddle | 32K/128K | WINT4 | 4卡*64G显存/600G内存 | 128K需要开启Chunked Prefill |
21+
| baidu/ERNIE-4.5-300B-A47B-Paddle | 32K/128K | WINT8 | 8卡*64G显存/600G内存 | 128K需要开启Chunked Prefill |
22+
| baidu/ERNIE-4.5-300B-A47B-W4A8C8-TP4-Paddle | 32K/128K | W4A8C8 | 4卡*64G显存/160G内存 | 限定4卡，建议开启Chunked Prefill |
23+
| baidu/ERNIE-4.5-300B-A47B-FP8-Paddle | 32K/128K | FP8 | 8卡*64G显存/600G内存 | 建议开启Chunked Prefill，仅在PD分离EP并行下支持 |
24+
| baidu/ERNIE-4.5-300B-A47B-Base-Paddle | 32K/128K | WINT4 | 4卡*64G显存/600G内存 | 建议开启Chunked Prefill |
25+
| baidu/ERNIE-4.5-300B-A47B-Base-Paddle | 32K/128K | WINT8 | 8卡*64G显存/600G内存 | 建议开启Chunked Prefill |
26+
| baidu/ERNIE-4.5-VL-28B-A3B-Paddle | 32K | WINT4 | 1卡*24G/128G内存 | 需要开启Chunked Prefill |
27+
| baidu/ERNIE-4.5-VL-28B-A3B-Paddle | 128K | WINT4 | 1卡*48G/128G内存 | 需要开启Chunked Prefill |
28+
| baidu/ERNIE-4.5-VL-28B-A3B-Paddle | 32K/128K | WINT8 | 1卡*48G/128G内存 | 需要开启Chunked Prefill |
29+
| baidu/ERNIE-4.5-21B-A3B-Paddle | 32K/128K | WINT4 | 1卡*24G/128G内存 | 128K需要开启Chunked Prefill |
30+
| baidu/ERNIE-4.5-21B-A3B-Paddle | 32K/128K | WINT8 | 1卡*48G/128G内存 | 128K需要开启Chunked Prefill |
31+
| baidu/ERNIE-4.5-21B-A3B-Base-Paddle | 32K/128K | WINT4 | 1卡*24G/128G内存 | 128K需要开启Chunked Prefill |
32+
| baidu/ERNIE-4.5-21B-A3B-Base-Paddle | 32K/128K | WINT8 | 1卡*48G/128G内存 | 128K需要开启Chunked Prefill |
33+
| baidu/ERNIE-4.5-0.3B-Paddle | 32K/128K | BF16 | 1卡*6G/12G显存/2G内存 | |
34+
| baidu/ERNIE-4.5-0.3B-Base-Paddle | 32K/128K | BF16 | 1卡*6G/12G显存/2G内存 | |
3535

3636
更多模型同步支持中，你可以通过[Github Issues](https://github.com/PaddlePaddle/FastDeploy/issues)向我们提交新模型的支持需求。

0 commit comments

Comments
 (0)