Update README.md
Browse files
README.md
CHANGED
|
@@ -4,12 +4,11 @@ language:
|
|
| 4 |
- en
|
| 5 |
tags:
|
| 6 |
- LLM
|
| 7 |
-
-
|
| 8 |
-
- Belle
|
| 9 |
---
|
| 10 |
-
## Model Card for
|
| 11 |
|
| 12 |
-
lyraBelle is currently the **fastest BELLE model** available. To the best of our knowledge, it is the **first accelerated version of
|
| 13 |
|
| 14 |
The inference speed of lyraBelle has achieved **10x** acceleration upon the ealry original version. We are still working hard to further improve the performance.
|
| 15 |
|
|
@@ -17,7 +16,6 @@ Among its main features are:
|
|
| 17 |
|
| 18 |
- weights: original BELLE-7B-2M weights released by BelleGroup.
|
| 19 |
- device: Nvidia Ampere architechture or newer (e.g A100)
|
| 20 |
-
- batch_size: compiled with dynamic batch size, max batch_size = 8
|
| 21 |
|
| 22 |
Note that:
|
| 23 |
**Some interface/code were set for future uses(see demo below).**
|
|
@@ -30,7 +28,15 @@ Note that:
|
|
| 30 |
### test environment
|
| 31 |
|
| 32 |
- device: Nvidia A100 40G
|
| 33 |
-
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 34 |
|
| 35 |
|
| 36 |
|
|
@@ -77,12 +83,12 @@ print(output_texts)
|
|
| 77 |
``` bibtex
|
| 78 |
@Misc{lyraBelle2023,
|
| 79 |
author = {Kangjian Wu, Zhengtao Wang, Bin Wu},
|
| 80 |
-
title = {
|
| 81 |
-
howpublished = {\url{https://huggingface.co/TMElyralab/
|
| 82 |
year = {2023}
|
| 83 |
}
|
| 84 |
```
|
| 85 |
|
| 86 |
## Report bug
|
| 87 |
-
- start a discussion to report any bugs!--> https://huggingface.co/TMElyralab/
|
| 88 |
- report bug with a `[bug]` mark in the title.
|
|
|
|
| 4 |
- en
|
| 5 |
tags:
|
| 6 |
- LLM
|
| 7 |
+
- BELLE
|
|
|
|
| 8 |
---
|
| 9 |
+
## Model Card for lyraBELLE
|
| 10 |
|
| 11 |
+
lyraBelle is currently the **fastest BELLE model** available. To the best of our knowledge, it is the **first accelerated version of BELLE**.
|
| 12 |
|
| 13 |
The inference speed of lyraBelle has achieved **10x** acceleration upon the ealry original version. We are still working hard to further improve the performance.
|
| 14 |
|
|
|
|
| 16 |
|
| 17 |
- weights: original BELLE-7B-2M weights released by BelleGroup.
|
| 18 |
- device: Nvidia Ampere architechture or newer (e.g A100)
|
|
|
|
| 19 |
|
| 20 |
Note that:
|
| 21 |
**Some interface/code were set for future uses(see demo below).**
|
|
|
|
| 28 |
### test environment
|
| 29 |
|
| 30 |
- device: Nvidia A100 40G
|
| 31 |
+
- warmup: 10 rounds
|
| 32 |
+
- percision：fp16
|
| 33 |
+
- batch size for our version: 64 (maximum under A100 40G)
|
| 34 |
+
- batch size for original: xx (maximum under A100 40G)
|
| 35 |
+
|
| 36 |
+
|version|batch size|speed|
|
| 37 |
+
|:-:|:-:|
|
| 38 |
+
|original|xxx|
|
| 39 |
+
|lyraBELLE|80|3030.36/sec|
|
| 40 |
|
| 41 |
|
| 42 |
|
|
|
|
| 83 |
``` bibtex
|
| 84 |
@Misc{lyraBelle2023,
|
| 85 |
author = {Kangjian Wu, Zhengtao Wang, Bin Wu},
|
| 86 |
+
title = {lyraBELLE: Accelerating BELLE by 10x+},
|
| 87 |
+
howpublished = {\url{https://huggingface.co/TMElyralab/lyraBELLE},
|
| 88 |
year = {2023}
|
| 89 |
}
|
| 90 |
```
|
| 91 |
|
| 92 |
## Report bug
|
| 93 |
+
- start a discussion to report any bugs!--> https://huggingface.co/TMElyralab/lyraBELLE/discussions
|
| 94 |
- report bug with a `[bug]` mark in the title.
|