Ahmed Al Mahrooqi comments

Results 6 comments of


                                            Ahmed Al Mahrooqi

[diffusers] training is broken with xformers and PyTorch 2.1

Facing the same issue here!

Launching train/train.py directly without Slurm

> I was able to run train.py directly without SLURM thanks to this thread. However, now I am faced with the challenge of trying to use 2 GPUs to train...

How to generate cams along with the predictions themselves?

Hi, were you able to get the model predictions along with the generated CAMs?

[request] Additional model exports (ONNX, CoreML, ...)

Thanks @barbolo! Have you had a chance to convert a model you've trained to ONNX? I'm having issues aligning state dict keys with the HF model, as the architecture seems...

[request] Additional model exports (ONNX, CoreML, ...)

@barbolo Thanks for your prompt response, I'm now able to load my training weights directly to the torch-hub model! How did you disable the flash attention though? I'm running into...

[request] Additional model exports (ONNX, CoreML, ...)

Ah perfect, thanks a lot @barbolo. Its working fine now :)