-
Do you think you can also manage to in inherit the nn.conv2d class and replace also the mlp operations inside by KAN operations ?
This way we would have a totally KAN cnn.
-
There's a bug in attack_manager.py:
```
if self.conv_template.name == 'llama-2':
self.conv_template.messages = []
self.conv_template.append_message(self.conv_template.roles…
-
Hi, thanks for your great work!
In the run semantic_segmentation file, when trained with COD10K (camouflaged object detection dataset), no train.csv file is displayed, May I ask whether I need to ge…
-
What is the meaning of this part when extracting features from the backbone? I cannot see any mention of this part in the paper.
```
conv_current = conv[:, 1:, :, :, :]
conv_prev = conv_current - c…
-
### Describe the issue
Attempting to run [Xenova/mms-tts-eng](https://huggingface.co/Xenova/mms-tts-eng) on WebGPU produces the following error:
```
Uncaught Error: [WebGPU] Kernel "[Conv] /text_en…
-
I encountered the following error while trying to train the model `nnUNetv2_train DATASET_ID 3d_fullres 0`:
../nunetv2/utilities/plans_handling/plans_handler.py", line 116, in conv_kernel_sizes
r…
-
The following stride=2 conv now is creating extra allocations for the dequant and exceeds shared memory limit.
```mlir
func.func @main$async_dispatch_75_conv_2d_nchw_fchw_2x320x64x64x320x3x3_i32() {
…
-
1. Padding of channel (< 8) to 8 for 16 B alignment.
2. If C * window_w % 32 != 0, add support to offset by padding of 32 when writing to activation block. We already have support to pad the weights …
-
使用提供的 demo 进行略微修改后,对于这张图(`raw_img`)测试发现生成结果语言种类不稳定。
![image](https://github.com/Ucas-HaoranWei/Vary-toy/assets/108124189/31838151-57ef-4e08-9eac-20b9600a2621)
### 加载模型
```python
# Model
disable_t…
-
## 🚀 Feature
CuDNN provides flexible support for performant gemm/conv with fp8 quantization. Thunder introducing fp8 casts in its traces can benefit from cudnn fusions.
### Motivation
Today, thu…