i keep getting this error of RuntimeError: CUDA error: CUBLAS_STATUS_EXECUTION_FAILED when calling cublasLtMatmul with transpose_mat1 1 transpose_mat2 0 m 512 n 16 k 864 mat1_ld 864 mat2_ld 864 result_ld 512 abcType 0 computeType 68 scaleType 0 whenever i try to train my pytorch model. i will give details of my dataloader and model below.

I’m trying to use a weighted loss function to handle class imbalance in my data. My problem is a multi-class and multi-output problem. For example (my data has five output/target columns (output_1, output_2, output_3) and I have three classes (class_0, class_1, and class_2) in each target column. I am currently using pytorch’s cross entropy loss function https://pytorch.org/docs/stable/generated/torch.nn.CrossEntropyLoss.html and I see that it has a weight parameter but my understanding is that this the same weight would be applied uniformly to each output/target, but I want to apply separate weights for each class in each output/target.

I have the following code using AdamW optimizer from Pytorch:

I have trained a RegNet model on a custom dataset for an image classification task. That was in August 2023. Now I want to train exactly the same model again, using the same dataset. I would expect this new model to achieve about the same performance as the previous one from August 2023, since nothing has changed: