-
Thanks for your fantastic job!
I just quantized the model with INT8, and found that the performance has dropped seriously(map from 0.4 to 0.2). I found maybe the matmul operator caused this problem. …
-
I am trying to add an attention module into resnet, and the original torch version code is like this:
```
class NonLocalModule(nn.Module):
def __init__(self, C, latent= 8):
super(Non…
-
## 🚀 Feature
Add an entropy function in PyTorch to compute entropy = -sum(p * log(p), axis=axis), which is analogous to scipy.stats.entropy.
## Motivation
It is a common thing to compute entrop…
-
In your code, the shape of gamma [batchsize,1,1,1]. I think the shape should be [1].
Besides, the attention score you get seems to be different with Han's paper. Did you calculate the attention scor…
-
### 🐛 Describe the bug
The `x.badbmm_(y, z)` method feels like something that should be fast, since it doesn't have to create a temporary tensor for `y.bmm(z)` before it adds it to `x`.
However, f…
-
### Description
Curious if it is possible to add some mathematical psychology models. Some examples being any evidence accumulation models (i.e., drift diffusion, linear ballistic accumulator, etc.),…
-
```
Torch - 0.4.1
scikit-image - 0.13.1
Python - 3.5.2
```
When running `python3 main.py --mode test_TPN`, I get the following error using the above mentioned configurations:
```
Traceback …
-
`# multiply & normalize ->(B, N, C')`
`energy = torch.bmm(proj_query, proj_key)`
`energy_new = torch.max(energy, -1, keepdim=True)[0].expand_as(energy) - energy`
I think the normalization impleme…
-
Pretty sure Arlith and I talked about this a couple of years ago. But since this was specifically mentioned and ruled on by CGL I'd post the notes:
http://bg.battletech.com/forums/index.php?topic=…
-
The operator [classic_b2b_bmm](https://github.com/facebookincubator/AITemplate/blob/main/python/aitemplate/compiler/ops/b2b_bmm/classic_b2b_bmm.py) has target specific component (`cutlass_lib`) in its…