This commit is contained in:
wailovet
2024-08-13 00:10:40 +08:00
parent 81d6e720be
commit 56910da219
-2
View File
@@ -42,10 +42,8 @@ def quantize_loader(model, state_dict, bits=4, device='cuda', includes=[]):
)
q_linear.post_init()
module.cpu()
q_linear = q_linear.to(device)
set_op_by_name(model, name, q_linear)
clear_memory(module)
model.load_state_dict(state_dict, strict=False)
return model