update todolist
Browse files
README.md
CHANGED
|
@@ -98,7 +98,11 @@ print(res)
|
|
| 98 |
|
| 99 |
1. 训练深度神经网络需要大量的计算资源,特别是在训练深度神经网络时,需要更多的计算资源,因此需要更快的训练速度。
|
| 100 |
|
|
|
|
| 101 |
|
|
|
|
|
|
|
|
|
|
| 102 |
|
| 103 |
## Citation
|
| 104 |
``` bibtex
|
|
|
|
| 98 |
|
| 99 |
1. 训练深度神经网络需要大量的计算资源,特别是在训练深度神经网络时,需要更多的计算资源,因此需要更快的训练速度。
|
| 100 |
|
| 101 |
+
### TODO：
|
| 102 |
|
| 103 |
+
We have implemented some special operators in ChatGLM, such as 2D rotary embedding, alpha residual, etcs.
|
| 104 |
+
|
| 105 |
+
We plan to add these operators on top of FasterTransformer to release a faster version.
|
| 106 |
|
| 107 |
## Citation
|
| 108 |
``` bibtex
|