-
**Describe the bug**
The behavior is a bit random. **When the text generation input size < batch size from the previous step** and replica > 1. The final output could missing some samples. This does …
-
To my understanding, I need to match the image size with the attention weight size as (h, w) and (h*w, h*w) each. But in your code, I cannot find where you change the cam size. It seems like they just…
-
Hi, Coming from HTMX.I feel that EHTML has a different focus and inclination. However, it does offer some additional features from a practical standpoint. I'll try to use both.
The main issue is th…
-
I've tried rev.ng on quite a few vmprotect samples, and it errors out on almost all of them. Of a subset of 31 samples, only 4 could be lifted; the others could not be lifted with the same error. Exam…
-
Is there anywhere in the code, particularly around hybrid_decoder_layer, where the state size is made explicit? I am looking for `N'`, and am trying to derive it from hidden size and state size
-
In spite of the RawBinaryCompressed mode, my vtu files are still way too large to transfer over the internet. I would like to find a good way to compress them even more.
My code does use single preci…
-
I observed that the script functions correctly with larger input sizes such as 224x224. Could you please advise if there are specific parameter adjustments or configurations recommended for smaller da…
-
Hi there,
I’m encountering issues with the model crashing frequently. Could you please provide guidance on how to adjust the model parameters to reduce memory usage?
Also, could you share the am…
-
Is there currently a way to dynamically change the size of uniform/storage buffers at runtime?
It doesn't seem like I can change it after calling `.add_uniform`, `.add_staging` or `.add[_rw]_storage`…
-
### Checklist
- [x] I checked, but didn't find any duplicates (open OR closed) of this issue in the repo.
- [x] This issue contains only one feature request. I will open one issue for e…