-
### 🐛 Describe the bug
Out variants of following ops are causing extra recompilations (in 3rd iteration) in torch.compile when compared to not using out variant,
torch.bmm
torch.topk
torch.chole…
-
### 🐛 Describe the bug
When using different generation configurations, such as top_k=1 or temperature=0 (while keeping other settings unchanged), why do the generated results change? They should …
-
原图：
LSM-crop裁剪出的背景块：
https://github.com/lucidrains/x-transf…
-
TopK (and sorting, being closely related) is a super important operator for streaming systems and databases as a whole, as is being able to sort data and so we should support both.
Both operations ca…
-
when I input 7350X150 image for inference, or some images with high height-width ratio ， it will get this bug, the bug detial is in below:
File "/GroundingDINO/groundingdino/models/GroundingDINO/tran…
-
if self.two_stage:
output_memory, output_proposals = self.gen_encoder_output_proposals(memory, mask_flatten, spatial_shapes)
# hack implementation for two-stage Defor…
-
It looks like regardless of the input type the indices output `B[1]` is always of type float32. I checked the c++ code and I think it's supposed to be int32/64? Is this behavior by design?
``` pyth…
-
I have played with your topk module and found something that make me confuse. I don't know whether it is bug or not. Please help to confirm. I made a test as below:
127.0.0.1:6379> TOPK.ADD testtop 5…