ceos/unsloth
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Fix bug

Browse source
This commit is contained in:
Daniel Han 2024-09-03 17:30:40 -07:00
commit be310cac1e
2 changed files with 9 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

2
unsloth/models/_utils.py
View file

@ -285,6 +285,8 @@ pass
# =============================================
# Fix new Xformers versions TypeError: Multiple dispatch failed for 'torch._ops.aten.to.dtype_layout'
accelerate_old_send_to_device = None
accelerate_new_send_to_device = None
if Version(xformers_version) >= Version("0.0.27"):
import accelerate.utils.operations
if hasattr(accelerate.utils.operations, "send_to_device") and \

10
unsloth/models/llama.py
View file

@ -1371,8 +1371,10 @@ def _wrap_fast_inference(generate, device_type, dtype, model):
internal_model._flag_for_generation = True
# Must patch accelerate for Xformers
import accelerate.utils.operations
accelerate.utils.operations.send_to_device = accelerate_new_send_to_device
if accelerate_new_send_to_device is not None:
import accelerate.utils.operations
accelerate.utils.operations.send_to_device = accelerate_new_send_to_device
pass
# For newer HF
kwargs["cache_implementation"] = "dynamic"
@ -1411,7 +1413,9 @@ def _wrap_fast_inference(generate, device_type, dtype, model):
if hasattr(internal_model, "_flag_for_generation"): del internal_model._flag_for_generation
# Return accelerate back
accelerate.utils.operations.send_to_device = accelerate_old_send_to_device
if accelerate_new_send_to_device is not None:
accelerate.utils.operations.send_to_device = accelerate_old_send_to_device
pass
return output
pass