Skip to content

fix: use OrderedDict for proper LRU cache eviction in fp8_cast_bf16.py #1128

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
modimihir07 wants to merge 2 commits into
base: main
Choose a base branch
from
+11 −6
Open

fix: use OrderedDict for proper LRU cache eviction in fp8_cast_bf16.py #1128

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
2c46f06
feat: add FP8 to BF16 weight conversion and text generation scripts.
modimihir07 Mar 1, 2026
e44c87a
remove generate.py changes (moved to separate PR)
Mar 1, 2026
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
17 changes: 11 additions & 6 deletions inference/fp8_cast_bf16.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,7 @@
import os
import json
from argparse import ArgumentParser
from collections import OrderedDict
from glob import glob
from tqdm import tqdm

Expand Down Expand Up @@ -36,8 +37,8 @@ def main(fp8_path, bf16_path):
model_index = json.load(f)
weight_map = model_index["weight_map"]

# Cache for loaded safetensor files
loaded_files = {}
# Cache for loaded safetensor files (OrderedDict for LRU eviction)
loaded_files = OrderedDict()
fp8_weight_names = []

# Helper function to get tensor from the correct file
Expand All @@ -58,6 +59,9 @@ def get_tensor(tensor_name):
if file_name not in loaded_files:
file_path = os.path.join(fp8_path, file_name)
loaded_files[file_name] = load_file(file_path, device="cuda")
else:
# Mark as recently used for LRU eviction
loaded_files.move_to_end(file_name)
return loaded_files[file_name][tensor_name]

safetensor_files = list(glob(os.path.join(fp8_path, "*.safetensors")))
Expand All @@ -66,6 +70,7 @@ def get_tensor(tensor_name):
file_name = os.path.basename(safetensor_file)
current_state_dict = load_file(safetensor_file, device="cuda")
loaded_files[file_name] = current_state_dict
loaded_files.move_to_end(file_name)

new_state_dict = {}
for weight_name, weight in current_state_dict.items():
Expand All @@ -87,10 +92,10 @@ def get_tensor(tensor_name):
new_safetensor_file = os.path.join(bf16_path, file_name)
save_file(new_state_dict, new_safetensor_file)

# Memory management: keep only the 2 most recently used files
if len(loaded_files) > 2:
oldest_file = next(iter(loaded_files))
del loaded_files[oldest_file]
# Memory management: evict least recently used files, keep at most 2
while len(loaded_files) > 2:
# Evict the least recently used entry (first item in OrderedDict)
loaded_files.popitem(last=False)
torch.cuda.empty_cache()

# Update model index
Expand Down