gpu-mode / discord-cluster-manager

Hook up GPUs to your Discord channel and start running jobs via DMs!
5 stars 3 forks source link

Upload train.py in train_workflow.yml #22

Closed b9r5 closed 1 week ago

b9r5 commented 1 week ago

This change adds train.py to the artifacts that are uploaded to GitHub actions. Previously just training.log was uploaded.

b9r5 commented 1 week ago

Testing on my branch: https://github.com/gpu-mode/discord-cluster-manager/actions/runs/11849246328

train.py is in training-artifacts.zip:

$ unzip training-artifacts.zip 
Archive:  training-artifacts.zip
  inflating: train.py                
  inflating: training.log            
$ cat train.py 
import torch

a = torch.Tensor([1, 2, 3, 4, 5]).cuda()
b= torch.Tensor([1, 2, 3, 4, 5]).cuda()

print(a)
print(b)
print(a + b)%
$ cat training.log 
/__w/_tool/Python/3.10.15/x64/lib/python3.10/site-packages/torch/_subclasses/functional_tensor.py:295: UserWarning: Failed to initialize NumPy: No module named 'numpy' (Triggered internally at ../torch/csrc/utils/tensor_numpy.cpp:84.)
  cpu = _conversion_method_template(device=torch.device("cpu"))
tensor([1., 2., 3., 4., 5.], device='cuda:0')
tensor([1., 2., 3., 4., 5.], device='cuda:0')
tensor([ 2.,  4.,  6.,  8., 10.], device='cuda:0')