Skip to content

dispatch model with real max memory #1022

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

Merged
merged 1 commit into from
Nov 12, 2025
Merged

dispatch model with real max memory #1022

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
8 changes: 2 additions & 6 deletions auto_round/compressors/base.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,7 +25,7 @@
import accelerate
import torch
from accelerate.big_modeling import dispatch_model, infer_auto_device_map
from accelerate.utils import get_balanced_memory
from accelerate.utils import get_max_memory
from torch import autocast
from tqdm import tqdm
from transformers import set_seed
Expand Down Expand Up @@ -1992,11 +1992,7 @@ def try_cache_inter_data_gpucpu(self, block_names, nsamples, layer_names=None, l
if str(self.model.device) == "cpu" and (not self.device.startswith("hpu")):
no_split_modules = getattr(self.model, "_no_split_modules", [])
devices = parse_available_devices(self.device_map)
max_memory = get_balanced_memory(
self.model,
max_memory=None,
no_split_module_classes=no_split_modules,
)
max_memory = get_max_memory()
new_max_memory = {}
for device in devices:
if ":" in device:
Expand Down