search

facebookresearch / DiT

Official PyTorch Implementation of "Scalable Diffusion Models with Transformers"
Other
6.37k stars 570 forks source link

Fixed the sampling issue in the dataset with varying categories #37

Open Littleor opened 1 year ago

Littleor commented 1 year ago

This pull request aims to solve the below issue when sampling in a different classification datasets, and this issue occurs because the y_null and class_labels are hardcoded in the sample.py file.

RuntimeError: CUDA error: CUBLAS_STATUS_EXECUTION_FAILED when calling `cublasSgemmStridedBatched( handle, opa, opb, m, n, k, &alpha, a, lda, stridea, b, ldb, strideb, &beta, c, ldc, stridec, num_batches)`