Skip to content

Add to overload for GGMLTensor, so calling to on the model moves the quantized data. #7949

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 13 commits into
base: main
Choose a base branch
from
Open

Add to overload for GGMLTensor, so calling to on the model moves the quantized data. #7949

Changes from all commits
Commits
Show all changes
13 commits
Select commit Hold shift + click to select a range
372f313
Add to overload for GGMLTensor, so calling to on the model moves the …
Vargol Apr 22, 2025
645930b
raise exected exception when attempting to change dtype
Vargol Apr 22, 2025
8ebc955
fix missing bracket
Vargol Apr 22, 2025
a159fb3
fix picky ruff issue
Vargol Apr 22, 2025
8628f54
Add to overload for GGMLTensor, so calling to on the model moves the …
Vargol Apr 22, 2025
6d7f35b
raise exected exception when attempting to change dtype
Vargol Apr 22, 2025
89ebdac
fix missing bracket
Vargol Apr 22, 2025
fcc452b
fix picky ruff issue
Vargol Apr 22, 2025
7ef8bdf
revert to overload due to failing tests, use Torch futures instead
Vargol Apr 29, 2025
1cf14ab
fix offload device
Vargol Apr 29, 2025
288c0cc
add check for state_dict, required to load TI's
Vargol Apr 29, 2025
1cceef8
out of sync with upstream
Vargol Apr 29, 2025
f77f019
fix import ordering, remove code I reverted that the resync added back
Vargol Apr 29, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
22 changes: 20 additions & 2 deletions invokeai/backend/model_manager/load/model_cache/cached_model/cached_model_only_full_load.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,6 +2,8 @@

import torch

from invokeai.backend.quantization.gguf.ggml_tensor import GGMLTensor


class CachedModelOnlyFullLoad:
"""A wrapper around a PyTorch model to handle full loads and unloads between the CPU and the compute device.
Expand Down Expand Up @@ -76,7 +78,15 @@ def full_load_to_vram(self) -> int:
for k, v in self._cpu_state_dict.items():
new_state_dict[k] = v.to(self._compute_device, copy=True)
self._model.load_state_dict(new_state_dict, assign=True)
self._model.to(self._compute_device)

check_for_gguf = hasattr(self._model, "state_dict") and self._model.state_dict().get("img_in.weight")
if isinstance(check_for_gguf, GGMLTensor):
old_value = torch.__future__.get_overwrite_module_params_on_conversion()
torch.__future__.set_overwrite_module_params_on_conversion(True)
self._model.to(self._compute_device)
torch.__future__.set_overwrite_module_params_on_conversion(old_value)
else:
self._model.to(self._compute_device)

self._is_in_vram = True
return self._total_bytes
Expand All @@ -92,7 +102,15 @@ def full_unload_from_vram(self) -> int:

if self._cpu_state_dict is not None:
self._model.load_state_dict(self._cpu_state_dict, assign=True)
self._model.to(self._offload_device)

check_for_gguf = hasattr(self._model, "state_dict") and self._model.state_dict().get("img_in.weight")
if isinstance(check_for_gguf, GGMLTensor):
old_value = torch.__future__.get_overwrite_module_params_on_conversion()
torch.__future__.set_overwrite_module_params_on_conversion(True)
self._model.to(self._offload_device)
torch.__future__.set_overwrite_module_params_on_conversion(old_value)
else:
self._model.to(self._offload_device)

self._is_in_vram = False
return self._total_bytes
Loading