diff --git a/unsloth/kernels/cross_entropy_loss.py b/unsloth/kernels/cross_entropy_loss.py index 3ea4cad471..d9b9bb3a1e 100644 --- a/unsloth/kernels/cross_entropy_loss.py +++ b/unsloth/kernels/cross_entropy_loss.py @@ -217,7 +217,7 @@ class Fast_CrossEntropyLoss(torch.autograd.Function): n_rows, n_cols = logits.shape grid = lambda meta: (n_rows, triton.cdiv(n_cols, meta["BLOCK_SIZE"])) - print(logits.stride(), dlosses.stride(), dlosses.shape) + print(logits.stride(), dlosses.stride(), dlosses.shape, dlosses) _cross_entropy_backward[grid]( logits, logits.stride(0), dlosses, dlosses.stride(0),