fix(model): resolve NFS flock error when loading pretrained timm weights

[gomezzz]

Summary

• Use cache_dir for pretrained weight downloads: Pass timm's native cache_dir parameter pointing to anomaly_match/pretrained_cache/, so HuggingFace Hub downloads and file locks happen on a local filesystem instead of NFS (which doesn't support fcntl.flock).
• Bundle default efficientnet-lite0 weights via Git LFS: Ship the default tf_efficientnet_lite0.in1k pretrained weights (~18 MB) with the repo, eliminating the need for network access when using the default model.
• Skip redundant pretrained download for eval_model: The eval_model's pretrained weights were immediately overwritten by copying from train_model, making the download wasteful. Now eval_model is created with pretrained=False.

Fixes OSError: [Errno 37] No locks available when running on NFS filesystems.

Context

Release 1.3.0 switched from efficientnet_lite_pytorch to timm, which uses huggingface_hub for pretrained weight downloads. huggingface_hub uses fcntl.flock() for cache file locking, which fails on NFS filesystems that don't support POSIX file locks.

Test plan

• All 267 unit tests pass
• Verified timm loads from local pretrained_cache/ without network access
• Verified LFS tracking for the model blob
• Test on NFS filesystem

[github-actions]

Coverage Report

File	Stmts	Miss	Cover	Missing
__init__.py	8	0	100%
data_io
SessionIOHandler.py	354	47	86%	112, 129–130, 174–175, 185–186, 253–254, 319–320, 323, 363–365, 407, 425–427, 449, 451–452, 456–458, 523, 598, 619, 633–634, 636–637, 639–641, 785–790, 804, 808–809, 813–815
find_images_in_folder.py	21	0	100%
load_images.py	91	28	69%	51–53, 58–60, 81, 93, 110, 140–147, 150, 156–157, 189, 248, 256, 277, 285, 292–294
metadata_handler.py	80	8	90%	70–71, 117, 122–123, 170–172
save_config.py	33	3	90%	92–94
datasets
AnomalyDetectionDataset.py	248	39	84%	126–127, 356, 362–363, 365, 369, 371–372, 374, 395–396, 398, 404–405, 409, 422–423, 484–485, 491–496, 498–499, 503–506, 508–510, 512–514, 525
BasicDataset.py	52	4	92%	59, 61, 97, 103
Label.py	5	0	100%
SSL_Dataset.py	68	3	95%	136, 139, 209
__init__.py	0	0	100%
data_utils.py	56	2	96%	80, 199
datasets/augmentation
randaugment.py	92	11	88%	222, 225–226, 245, 328–330, 332–335
randaugment_multispectral.py	77	23	70%	62–63, 77–79, 120–121, 148–149, 166, 170, 174, 248, 253, 270–272, 274–277, 280–281
image_processing
transforms.py	57	3	94%	46–47, 76
models
FixMatch.py	214	29	86%	108, 201–202, 217, 222, 254, 281–282, 285, 288–290, 293, 301, 306–307, 366, 404, 430–434, 495, 497–498, 502, 519–520
pipeline
SessionTracker.py	122	3	97%	127, 219–220
session.py	535	85	84%	131–134, 384–387, 432, 445, 597, 654, 656–657, 665, 668, 679, 687, 717, 721, 725–726, 730, 735–736, 743, 747–748, 760, 770–771, 773, 777, 791–792, 794–796, 798, 801–802, 860–861, 863–866, 868–872, 878, 884, 889–893, 895, 903, 906–907, 923, 964–965, 971–972, 976–978, 980, 1005, 1031, 1065–1066, 1074, 1085, 1093, 1097, 1101, 1106, 1109, 1153, 1156
utils
accuracy.py	13	0	100%
consistency_loss.py	18	0	100%
create_model_string.py	4	0	100%
cross_entropy_loss.py	9	0	100%
cutana_stream_utils.py	70	9	87%	64–65, 88, 108, 118–122
get_cosine_schedule_with_warmup.py	11	1	90%	39
get_default_cfg.py	61	0	100%
get_net_builder.py	49	7	85%	81–82, 89, 92, 146–147, 151
get_optimizer.py	21	2	90%	54, 58
print_cfg.py	45	4	91%	89, 95, 117, 119
set_log_level.py	15	0	100%
set_seeds.py	13	2	84%	25–26
validate_config.py	131	11	91%	180, 199, 214, 228, 244, 248, 264, 290, 379–381
TOTAL	2573	324	87%

Tests	Skipped	Failures	Errors	Time
365	0 💤	0 ❌	0 🔥	58.877s ⏱️