Update utils.py

This commit is contained in:
Daniel Han-Chen 2024-01-22 22:21:29 +11:00
commit e1cbc9e423

View file

@ -180,6 +180,7 @@ pass
def fast_linear_forward(proj, X, temp_lora = None, out = None):
W, W_quant, lora_A, lora_B, lora_S = get_lora_parameters(proj)
out = fast_gemv(X, W, W_quant, out = out)
print(out.shape, X.shape)
if lora_A is not None:
dtype = X.dtype
temp_lora = torch.matmul(X, lora_A.to(dtype).t(), out = temp_lora)