Update dependencies for competition models

[KartikP]

Summary

• Update PyTorch to >=2.6 for security compliance with transformers library
• Add optional competition-models dependency group for Competition 2024 models
• Fix model loading issues for CORnet-S and CVT models

pyproject.toml:

• Updated torch>=2.6 and torchvision>=0.21 in test dependencies
• Removed duplicate pytest-mock entry
• Added new competition-models optional dependency group:
  • transformers>=4.45
  • albumentations>=2.0
  • timm>=1.0
  • cornet (from git)

environment_lock.yml:

• Updated torch from 2.4.0 to 2.6.0
• Updated torchvision from 0.19.0 to 0.21.0
• Added packages required for competition models:
  • albumentations, albucore, cornet, huggingface_hub, pydantic, pydantic_core, safetensors, timm, tokenizers, transformers

Model Fixes

brainscore_vision/models/cornet_s/model.py:

• Added weights_only=False to torch.load() for PyTorch 2.6+ compatibility

brainscore_vision/models/cvt_cvt_w24_384_in22k_finetuned_in1k_4/model.py:

• Replaced AutoImageProcessor with torchvision preprocessing to resolve HuggingFace config compatibility issues

Competition 2024 Models Tested

Top 10 Models:

Model	Status
cvt_cvt-w24-384-in22k_finetuned-in1k_4 (1885)	PASS
resnext101_32x8d_wsl (646)	PASS
effnetb1_cutmixpatch_SAM_ (1112)	PASS
effnetb1_cutmixpatch_augmix_robust32_avge4e7_manylayers_324x288 (1033)	PASS
resnext101_32x32d_wsl (648)	PASS
effnetb1_272x240 (1045)	PASS
resnext101_32x48d_wsl (649)	PASS
pnasnet_large (2414)	PASS
resnet-152_v2 (2185)	PASS
focalnet_tiny_lrf_in1k (1933)	PASS

Representative Models:

Model	Status
hmax (652)	PASS
alexnet (982)	PASS
CORnet-S (849)	PASS
resnet-50-robust (623)	PASS
voneresnet-50-non_stochastic (791)	PASS
grcnn (mapped from grcnn_robust_v1)	PASS
custom_model_cv_18_dagger_408 (991)	PASS
ViT_L_32_imagenet1k (734)	PASS
mobilenet_v2_1_4_224 (2417)	PASS
pixels (768)	PASS
resnet18-local_aggregation (680)	FAIL (not in codebase)