You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with model.to('cuda')
I am running on a newly setup H100 and all seems to be working as the miner downloads the shards, but I am seeing this error in the logs.
I am getting loss of 7-8 .
The text was updated successfully, but these errors were encountered:
error:
You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with
model.to('cuda')
I am running on a newly setup H100 and all seems to be working as the miner downloads the shards, but I am seeing this error in the logs.
I am getting loss of 7-8 .
The text was updated successfully, but these errors were encountered: