-
Hello,
Going via the training.
Some small ideas for improvements.
#######################
Transformers, what can they do?
https://huggingface.co/learn/nlp-course/en/chapter1/3
A)
Curren…
-
As in the title.. I spent a bit of time debugging it but haven't figured out the cause yet. E.g. running
```
tune run --nproc_per_node 2 full_finetune_distributed --config llama2/7B_full fsdp_cpu_…
-
### 🐛 Describe the bug
As can be seen from the following example, the output of `Conv1d` is different when it process the same token in a sequence. There is a small difference, although the result sh…
-
PyTorch/XLA currently has no means to clear cached memory, i.e. something similar to `torch.cuda.empty_cache()`. This is relevant for benchmarking a model on both PyTorch CUDA and XLA:CUDA.
For com…
-
### Why it needs to get done
The DSS project allows users to select the image of the Notebook server to be used, some of which are built and distributed by this team.
### What needs to get done
1. …
-
**Description**
bug when deploying Macbert
**Triton Information**
I use the official image: nvcr.io/nvidia/tritonserver:21.09-py3
```
NVIDIA Release 21.09 (build 27443074)
Copyright (c) 20…
-
### 🐛 Describe the bug
For some really weird reason I can't explain importing inductor in an eager mode complex test causes inaccuracies in complex numbers arithmetic, after a lot of time trying to c…
-
So I been trying so hard with this cuda backend problem, I intalled cuda 12.1, python 3.10, pytorch 2.4.0 and installed all the requirement without error, but when ever i run run.py this error popup
…
-
## Device
OS: CentOS Linux 7
GPU: gtx 1080
### Issue
Hi! I get the following error running any of the examples scripts
```
RuntimeError: NVTX functions not installed. Are you sure you have a C…
-
One piece of compiled autograd debugging that I think might aid in debugging would be that:
* in the "non-compiled autograd flow", when you inspect a given backward graph and I see some input that is…