JorgeAV
/

MR-JEPA

vision-language

Model card Files Files and versions

MR-JEPA / mr_jepa

166 kB

Ctrl+K

Ctrl+K

1 contributor

History: 10 commits

JorgeAV's picture

fix: target_encoder.py — respect config.jepa_loss_fn (smooth_l1/mse/cosine) instead of hardcoded MSE

ac52da3 verified about 1 month ago

configs
fix: export get_dinov2_ablation_config from configs about 1 month ago
data
Initial MR-JEPA codebase: architecture, training, evaluation, and tests about 1 month ago
evaluation
Remove __pycache__ files about 1 month ago
models
fix: target_encoder.py — respect config.jepa_loss_fn (smooth_l1/mse/cosine) instead of hardcoded MSE about 1 month ago
training
Initial MR-JEPA codebase: architecture, training, evaluation, and tests about 1 month ago
utils
fix: ablation.py — wire all disable flags, add missing experiments (dinov2, mse/cosine loss, no_sigreg, vicreg_only), add run()/load_results() methods, generate CLI commands about 1 month ago
ARCHITECTURE.md

13.7 kB
fix: ARCHITECTURE.md — complete ablation table with all 13 experiments, CLI flags, dinov2/loss_fn/sigreg/vicreg ablations, footnote on no_rollout→no_jepa about 1 month ago
__init__.py

267 Bytes
Initial MR-JEPA codebase: architecture, training, evaluation, and tests about 1 month ago