-
## 🐛 Bug
PyTorch uses `tensor.numel()` while Thunder's TensorProxy and in the codebase it's used as a property `tensor.numel`. The property attribute makes more sense, but it conflicts with setting…
-
## Code
```py
N = Tensor.zeros((3, 3)).contiguous()
for i in range(3):
for j in range(3):
N[i][j] += 1
```
## Expectation
`N` to be a 3x3 matrix of 1's.
## Reality
tinygrad…
-
When we use burn with a Wgpu backend, the underlying storage of the Tensors are wgpu::Buffers. It would be really helpful for a lot of use cases if we could directly have access to the wgpu::Buffer. …
-
如题,在使用toolkit2.1版本,目标平台3588上转模型出现dataconvert type -1 is unsupport in current 这个问题
I rknn-toolkit2 version: 2.1.0+708089d1
I Loading : 100%|███████████████████████████████████████████████| 209/209 …
-
### 🐛 Describe the bug
`delta` argument of [HuberLoss()](https://pytorch.org/docs/stable/generated/torch.nn.HuberLoss.html) with a `complex` value gets the error message `delta` must be `float` as sh…
-
Traceback (most recent call last):
File "main.py", line 174, in
main()
File "main.py", line 170, in main
learner.train()
File "/home/research/lanier.m/Desktop/SDVT-master/metalearn…
-
I got a training result as below:
targets: tensor([[ 80.],
[ 71.],
[ 65.],
[149.]], device='cuda:0'),
pred: tensor([[33.4143],
[ nan],
[32.8215],
…
-
Hi,
code:
`
def joint_embed(
self,
input_ids: torch.Tensor = None,
frames: torch.Tensor = None,
):
if frames is None:
return se…
-
**Describe the bug**
Untilization (`tensor.to(tt::tt_metal::Layour::ROW_MAJOR)`) crashes if the tesnor is in the TILE layout but not padded. This can happen if the tensor is reshaped from a padded…
-
Trying to compile the program with flash attention enabled errors out
apparently, the ggml_flash_attn() function was replaced by ggml_flash_attn_ext() in ggml in this pr: https://github.com/ggerganov…