CoolFace
Apppublic

y-agent/modular-addition-feature-learning

sourceHugging Faceupdated 6mo agoView on Hugging Face
1likes
.gitattributes212 linesDownload Raw Back to root
1*.7z filter=lfs diff=lfs merge=lfs -text2*.arrow filter=lfs diff=lfs merge=lfs -text3*.bin filter=lfs diff=lfs merge=lfs -text4*.bz2 filter=lfs diff=lfs merge=lfs -text5*.ckpt filter=lfs diff=lfs merge=lfs -text6*.ftz filter=lfs diff=lfs merge=lfs -text7*.gz filter=lfs diff=lfs merge=lfs -text8*.h5 filter=lfs diff=lfs merge=lfs -text9*.joblib filter=lfs diff=lfs merge=lfs -text10*.lfs.* filter=lfs diff=lfs merge=lfs -text11*.mlmodel filter=lfs diff=lfs merge=lfs -text12*.model filter=lfs diff=lfs merge=lfs -text13*.msgpack filter=lfs diff=lfs merge=lfs -text14*.npy filter=lfs diff=lfs merge=lfs -text15*.npz filter=lfs diff=lfs merge=lfs -text16*.onnx filter=lfs diff=lfs merge=lfs -text17*.ot filter=lfs diff=lfs merge=lfs -text18*.parquet filter=lfs diff=lfs merge=lfs -text19*.pb filter=lfs diff=lfs merge=lfs -text20*.pickle filter=lfs diff=lfs merge=lfs -text21*.pkl filter=lfs diff=lfs merge=lfs -text22*.pt filter=lfs diff=lfs merge=lfs -text23*.pth filter=lfs diff=lfs merge=lfs -text24*.rar filter=lfs diff=lfs merge=lfs -text25*.safetensors filter=lfs diff=lfs merge=lfs -text26saved_model/**/* filter=lfs diff=lfs merge=lfs -text27*.tar.* filter=lfs diff=lfs merge=lfs -text28*.tar filter=lfs diff=lfs merge=lfs -text29*.tflite filter=lfs diff=lfs merge=lfs -text30*.tgz filter=lfs diff=lfs merge=lfs -text31*.wasm filter=lfs diff=lfs merge=lfs -text32*.xz filter=lfs diff=lfs merge=lfs -text33*.zip filter=lfs diff=lfs merge=lfs -text34*.zst filter=lfs diff=lfs merge=lfs -text35*tfevents* filter=lfs diff=lfs merge=lfs -text36precomputed_results/p_015/p015_lineplot_in.png filter=lfs diff=lfs merge=lfs -text37precomputed_results/p_015/p015_lineplot_out.png filter=lfs diff=lfs merge=lfs -text38precomputed_results/p_015/p015_phase_align_approx1.png filter=lfs diff=lfs merge=lfs -text39precomputed_results/p_015/p015_phase_align_approx2.png filter=lfs diff=lfs merge=lfs -text40precomputed_results/p_015/p015_single_freq_quad.png filter=lfs diff=lfs merge=lfs -text41precomputed_results/p_015/p015_single_freq_relu.png filter=lfs diff=lfs merge=lfs -text42precomputed_results/p_023/p023_full_training_para_origin.png filter=lfs diff=lfs merge=lfs -text43precomputed_results/p_023/p023_grokk_decoded_weights_dynamic.png filter=lfs diff=lfs merge=lfs -text44precomputed_results/p_023/p023_grokk_memorization_common_to_rare.png filter=lfs diff=lfs merge=lfs -text45precomputed_results/p_023/p023_lineplot_in.png filter=lfs diff=lfs merge=lfs -text46precomputed_results/p_023/p023_lineplot_out.png filter=lfs diff=lfs merge=lfs -text47precomputed_results/p_023/p023_output_logits.png filter=lfs diff=lfs merge=lfs -text48precomputed_results/p_023/p023_overview_loss_ipr.png filter=lfs diff=lfs merge=lfs -text49precomputed_results/p_023/p023_phase_align_approx1.png filter=lfs diff=lfs merge=lfs -text50precomputed_results/p_023/p023_phase_align_approx2.png filter=lfs diff=lfs merge=lfs -text51precomputed_results/p_023/p023_single_freq_quad.png filter=lfs diff=lfs merge=lfs -text52precomputed_results/p_023/p023_single_freq_relu.png filter=lfs diff=lfs merge=lfs -text53precomputed_results/p_029/p029_full_training_para_origin.png filter=lfs diff=lfs merge=lfs -text54precomputed_results/p_029/p029_grokk_decoded_weights_dynamic.png filter=lfs diff=lfs merge=lfs -text55precomputed_results/p_029/p029_grokk_memorization_accuracy.png filter=lfs diff=lfs merge=lfs -text56precomputed_results/p_029/p029_grokk_memorization_common_to_rare.png filter=lfs diff=lfs merge=lfs -text57precomputed_results/p_029/p029_lineplot_in.png filter=lfs diff=lfs merge=lfs -text58precomputed_results/p_029/p029_lineplot_out.png filter=lfs diff=lfs merge=lfs -text59precomputed_results/p_029/p029_output_logits.png filter=lfs diff=lfs merge=lfs -text60precomputed_results/p_029/p029_overview_loss_ipr.png filter=lfs diff=lfs merge=lfs -text61precomputed_results/p_029/p029_phase_align_approx1.png filter=lfs diff=lfs merge=lfs -text62precomputed_results/p_029/p029_phase_align_approx2.png filter=lfs diff=lfs merge=lfs -text63precomputed_results/p_029/p029_single_freq_quad.png filter=lfs diff=lfs merge=lfs -text64precomputed_results/p_029/p029_single_freq_relu.png filter=lfs diff=lfs merge=lfs -text65precomputed_results/p_031/p031_full_training_para_origin.png filter=lfs diff=lfs merge=lfs -text66precomputed_results/p_031/p031_grokk_decoded_weights_dynamic.png filter=lfs diff=lfs merge=lfs -text67precomputed_results/p_031/p031_grokk_memorization_accuracy.png filter=lfs diff=lfs merge=lfs -text68precomputed_results/p_031/p031_grokk_memorization_common_to_rare.png filter=lfs diff=lfs merge=lfs -text69precomputed_results/p_031/p031_lineplot_in.png filter=lfs diff=lfs merge=lfs -text70precomputed_results/p_031/p031_lineplot_out.png filter=lfs diff=lfs merge=lfs -text71precomputed_results/p_031/p031_output_logits.png filter=lfs diff=lfs merge=lfs -text72precomputed_results/p_031/p031_overview_loss_ipr.png filter=lfs diff=lfs merge=lfs -text73precomputed_results/p_031/p031_phase_align_approx1.png filter=lfs diff=lfs merge=lfs -text74precomputed_results/p_031/p031_phase_align_approx2.png filter=lfs diff=lfs merge=lfs -text75precomputed_results/p_031/p031_single_freq_quad.png filter=lfs diff=lfs merge=lfs -text76precomputed_results/p_031/p031_single_freq_relu.png filter=lfs diff=lfs merge=lfs -text77precomputed_results/p_009/p009_lineplot_in.png filter=lfs diff=lfs merge=lfs -text78precomputed_results/p_009/p009_lineplot_out.png filter=lfs diff=lfs merge=lfs -text79precomputed_results/p_009/p009_phase_align_approx1.png filter=lfs diff=lfs merge=lfs -text80precomputed_results/p_009/p009_phase_align_approx2.png filter=lfs diff=lfs merge=lfs -text81precomputed_results/p_009/p009_phase_align_quad.png filter=lfs diff=lfs merge=lfs -text82precomputed_results/p_009/p009_single_freq_quad.png filter=lfs diff=lfs merge=lfs -text83precomputed_results/p_009/p009_single_freq_relu.png filter=lfs diff=lfs merge=lfs -text84precomputed_results/p_021/p021_grokk_decoded_weights_dynamic.png filter=lfs diff=lfs merge=lfs -text85precomputed_results/p_021/p021_grokk_memorization_accuracy.png filter=lfs diff=lfs merge=lfs -text86precomputed_results/p_021/p021_grokk_memorization_common_to_rare.png filter=lfs diff=lfs merge=lfs -text87precomputed_results/p_021/p021_lineplot_in.png filter=lfs diff=lfs merge=lfs -text88precomputed_results/p_021/p021_lineplot_out.png filter=lfs diff=lfs merge=lfs -text89precomputed_results/p_021/p021_output_logits.png filter=lfs diff=lfs merge=lfs -text90precomputed_results/p_021/p021_overview_loss_ipr.png filter=lfs diff=lfs merge=lfs -text91precomputed_results/p_021/p021_phase_align_approx1.png filter=lfs diff=lfs merge=lfs -text92precomputed_results/p_021/p021_phase_align_approx2.png filter=lfs diff=lfs merge=lfs -text93precomputed_results/p_021/p021_single_freq_quad.png filter=lfs diff=lfs merge=lfs -text94precomputed_results/p_021/p021_single_freq_relu.png filter=lfs diff=lfs merge=lfs -text95precomputed_results/p_025/p025_grokk_decoded_weights_dynamic.png filter=lfs diff=lfs merge=lfs -text96precomputed_results/p_025/p025_grokk_memorization_accuracy.png filter=lfs diff=lfs merge=lfs -text97precomputed_results/p_025/p025_grokk_memorization_common_to_rare.png filter=lfs diff=lfs merge=lfs -text98precomputed_results/p_025/p025_lineplot_in.png filter=lfs diff=lfs merge=lfs -text99precomputed_results/p_025/p025_lineplot_out.png filter=lfs diff=lfs merge=lfs -text100precomputed_results/p_025/p025_output_logits.png filter=lfs diff=lfs merge=lfs -text101precomputed_results/p_025/p025_overview_loss_ipr.png filter=lfs diff=lfs merge=lfs -text102precomputed_results/p_025/p025_phase_align_approx1.png filter=lfs diff=lfs merge=lfs -text103precomputed_results/p_025/p025_phase_align_approx2.png filter=lfs diff=lfs merge=lfs -text104precomputed_results/p_025/p025_phase_align_quad.png filter=lfs diff=lfs merge=lfs -text105precomputed_results/p_025/p025_single_freq_quad.png filter=lfs diff=lfs merge=lfs -text106precomputed_results/p_025/p025_single_freq_relu.png filter=lfs diff=lfs merge=lfs -text107precomputed_results/p_027/p027_grokk_decoded_weights_dynamic.png filter=lfs diff=lfs merge=lfs -text108precomputed_results/p_027/p027_grokk_memorization_accuracy.png filter=lfs diff=lfs merge=lfs -text109precomputed_results/p_027/p027_grokk_memorization_common_to_rare.png filter=lfs diff=lfs merge=lfs -text110precomputed_results/p_027/p027_lineplot_in.png filter=lfs diff=lfs merge=lfs -text111precomputed_results/p_027/p027_lineplot_out.png filter=lfs diff=lfs merge=lfs -text112precomputed_results/p_027/p027_output_logits.png filter=lfs diff=lfs merge=lfs -text113precomputed_results/p_027/p027_overview_loss_ipr.png filter=lfs diff=lfs merge=lfs -text114precomputed_results/p_027/p027_phase_align_approx1.png filter=lfs diff=lfs merge=lfs -text115precomputed_results/p_027/p027_phase_align_approx2.png filter=lfs diff=lfs merge=lfs -text116precomputed_results/p_027/p027_single_freq_quad.png filter=lfs diff=lfs merge=lfs -text117precomputed_results/p_027/p027_single_freq_relu.png filter=lfs diff=lfs merge=lfs -text118precomputed_results/p_033/p033_grokk_decoded_weights_dynamic.png filter=lfs diff=lfs merge=lfs -text119precomputed_results/p_033/p033_grokk_memorization_accuracy.png filter=lfs diff=lfs merge=lfs -text120precomputed_results/p_033/p033_grokk_memorization_common_to_rare.png filter=lfs diff=lfs merge=lfs -text121precomputed_results/p_033/p033_lineplot_in.png filter=lfs diff=lfs merge=lfs -text122precomputed_results/p_033/p033_lineplot_out.png filter=lfs diff=lfs merge=lfs -text123precomputed_results/p_033/p033_lottery_mech_phase.png filter=lfs diff=lfs merge=lfs -text124precomputed_results/p_033/p033_output_logits.png filter=lfs diff=lfs merge=lfs -text125precomputed_results/p_033/p033_overview_loss_ipr.png filter=lfs diff=lfs merge=lfs -text126precomputed_results/p_033/p033_phase_align_approx1.png filter=lfs diff=lfs merge=lfs -text127precomputed_results/p_033/p033_phase_align_approx2.png filter=lfs diff=lfs merge=lfs -text128precomputed_results/p_033/p033_phase_align_quad.png filter=lfs diff=lfs merge=lfs -text129precomputed_results/p_033/p033_single_freq_quad.png filter=lfs diff=lfs merge=lfs -text130precomputed_results/p_033/p033_single_freq_relu.png filter=lfs diff=lfs merge=lfs -text131precomputed_results/p_037/p037_grokk_decoded_weights_dynamic.png filter=lfs diff=lfs merge=lfs -text132precomputed_results/p_037/p037_grokk_memorization_accuracy.png filter=lfs diff=lfs merge=lfs -text133precomputed_results/p_037/p037_grokk_memorization_common_to_rare.png filter=lfs diff=lfs merge=lfs -text134precomputed_results/p_037/p037_lineplot_in.png filter=lfs diff=lfs merge=lfs -text135precomputed_results/p_037/p037_lineplot_out.png filter=lfs diff=lfs merge=lfs -text136precomputed_results/p_037/p037_lottery_mech_magnitude.png filter=lfs diff=lfs merge=lfs -text137precomputed_results/p_037/p037_lottery_mech_phase.png filter=lfs diff=lfs merge=lfs -text138precomputed_results/p_037/p037_output_logits.png filter=lfs diff=lfs merge=lfs -text139precomputed_results/p_037/p037_overview_loss_ipr.png filter=lfs diff=lfs merge=lfs -text140precomputed_results/p_037/p037_phase_align_approx1.png filter=lfs diff=lfs merge=lfs -text141precomputed_results/p_037/p037_phase_align_approx2.png filter=lfs diff=lfs merge=lfs -text142precomputed_results/p_037/p037_single_freq_quad.png filter=lfs diff=lfs merge=lfs -text143precomputed_results/p_037/p037_single_freq_relu.png filter=lfs diff=lfs merge=lfs -text144precomputed_results/p_007/p007_lineplot_in.png filter=lfs diff=lfs merge=lfs -text145precomputed_results/p_007/p007_lineplot_out.png filter=lfs diff=lfs merge=lfs -text146precomputed_results/p_007/p007_phase_align_approx1.png filter=lfs diff=lfs merge=lfs -text147precomputed_results/p_007/p007_phase_align_approx2.png filter=lfs diff=lfs merge=lfs -text148precomputed_results/p_007/p007_single_freq_quad.png filter=lfs diff=lfs merge=lfs -text149precomputed_results/p_007/p007_single_freq_relu.png filter=lfs diff=lfs merge=lfs -text150precomputed_results/p_005/p005_lineplot_in.png filter=lfs diff=lfs merge=lfs -text151precomputed_results/p_005/p005_lineplot_out.png filter=lfs diff=lfs merge=lfs -text152precomputed_results/p_005/p005_phase_align_approx1.png filter=lfs diff=lfs merge=lfs -text153precomputed_results/p_005/p005_phase_align_approx2.png filter=lfs diff=lfs merge=lfs -text154precomputed_results/p_005/p005_single_freq_quad.png filter=lfs diff=lfs merge=lfs -text155precomputed_results/p_005/p005_single_freq_relu.png filter=lfs diff=lfs merge=lfs -text156precomputed_results/p_067/p067_grokk_decoded_weights_dynamic.png filter=lfs diff=lfs merge=lfs -text157precomputed_results/p_067/p067_grokk_memorization_accuracy.png filter=lfs diff=lfs merge=lfs -text158precomputed_results/p_067/p067_grokk_memorization_common_to_rare.png filter=lfs diff=lfs merge=lfs -text159precomputed_results/p_067/p067_lineplot_in.png filter=lfs diff=lfs merge=lfs -text160precomputed_results/p_067/p067_lineplot_out.png filter=lfs diff=lfs merge=lfs -text161precomputed_results/p_067/p067_lottery_mech_magnitude.png filter=lfs diff=lfs merge=lfs -text162precomputed_results/p_067/p067_lottery_mech_phase.png filter=lfs diff=lfs merge=lfs -text163precomputed_results/p_067/p067_output_logits.png filter=lfs diff=lfs merge=lfs -text164precomputed_results/p_067/p067_overview_loss_ipr.png filter=lfs diff=lfs merge=lfs -text165precomputed_results/p_067/p067_phase_align_approx1.png filter=lfs diff=lfs merge=lfs -text166precomputed_results/p_067/p067_phase_align_approx2.png filter=lfs diff=lfs merge=lfs -text167precomputed_results/p_067/p067_phase_align_quad.png filter=lfs diff=lfs merge=lfs -text168precomputed_results/p_067/p067_single_freq_quad.png filter=lfs diff=lfs merge=lfs -text169precomputed_results/p_067/p067_single_freq_relu.png filter=lfs diff=lfs merge=lfs -text170precomputed_results/p_003/p003_lineplot_in.png filter=lfs diff=lfs merge=lfs -text171precomputed_results/p_003/p003_lineplot_out.png filter=lfs diff=lfs merge=lfs -text172precomputed_results/p_003/p003_phase_align_approx1.png filter=lfs diff=lfs merge=lfs -text173precomputed_results/p_003/p003_phase_align_approx2.png filter=lfs diff=lfs merge=lfs -text174precomputed_results/p_003/p003_single_freq_quad.png filter=lfs diff=lfs merge=lfs -text175precomputed_results/p_003/p003_single_freq_relu.png filter=lfs diff=lfs merge=lfs -text176precomputed_results/p_019/p019_grokk_decoded_weights_dynamic.png filter=lfs diff=lfs merge=lfs -text177precomputed_results/p_019/p019_grokk_memorization_accuracy.png filter=lfs diff=lfs merge=lfs -text178precomputed_results/p_019/p019_grokk_memorization_common_to_rare.png filter=lfs diff=lfs merge=lfs -text179precomputed_results/p_019/p019_lineplot_in.png filter=lfs diff=lfs merge=lfs -text180precomputed_results/p_019/p019_lineplot_out.png filter=lfs diff=lfs merge=lfs -text181precomputed_results/p_019/p019_output_logits.png filter=lfs diff=lfs merge=lfs -text182precomputed_results/p_019/p019_overview_loss_ipr.png filter=lfs diff=lfs merge=lfs -text183precomputed_results/p_019/p019_phase_align_approx1.png filter=lfs diff=lfs merge=lfs -text184precomputed_results/p_019/p019_phase_align_approx2.png filter=lfs diff=lfs merge=lfs -text185precomputed_results/p_019/p019_phase_align_quad.png filter=lfs diff=lfs merge=lfs -text186precomputed_results/p_019/p019_single_freq_quad.png filter=lfs diff=lfs merge=lfs -text187precomputed_results/p_019/p019_single_freq_relu.png filter=lfs diff=lfs merge=lfs -text188precomputed_results/p_035/p035_grokk_decoded_weights_dynamic.png filter=lfs diff=lfs merge=lfs -text189precomputed_results/p_035/p035_grokk_memorization_accuracy.png filter=lfs diff=lfs merge=lfs -text190precomputed_results/p_035/p035_grokk_memorization_common_to_rare.png filter=lfs diff=lfs merge=lfs -text191precomputed_results/p_035/p035_lineplot_in.png filter=lfs diff=lfs merge=lfs -text192precomputed_results/p_035/p035_lineplot_out.png filter=lfs diff=lfs merge=lfs -text193precomputed_results/p_035/p035_lottery_mech_magnitude.png filter=lfs diff=lfs merge=lfs -text194precomputed_results/p_035/p035_lottery_mech_phase.png filter=lfs diff=lfs merge=lfs -text195precomputed_results/p_035/p035_output_logits.png filter=lfs diff=lfs merge=lfs -text196precomputed_results/p_035/p035_overview_loss_ipr.png filter=lfs diff=lfs merge=lfs -text197precomputed_results/p_035/p035_phase_align_approx1.png filter=lfs diff=lfs merge=lfs -text198precomputed_results/p_035/p035_phase_align_approx2.png filter=lfs diff=lfs merge=lfs -text199precomputed_results/p_035/p035_single_freq_quad.png filter=lfs diff=lfs merge=lfs -text200precomputed_results/p_035/p035_single_freq_relu.png filter=lfs diff=lfs merge=lfs -text201precomputed_results/p_045/p045_grokk_decoded_weights_dynamic.png filter=lfs diff=lfs merge=lfs -text202precomputed_results/p_045/p045_grokk_memorization_accuracy.png filter=lfs diff=lfs merge=lfs -text203precomputed_results/p_045/p045_grokk_memorization_common_to_rare.png filter=lfs diff=lfs merge=lfs -text204precomputed_results/p_045/p045_lineplot_in.png filter=lfs diff=lfs merge=lfs -text205precomputed_results/p_045/p045_lineplot_out.png filter=lfs diff=lfs merge=lfs -text206precomputed_results/p_045/p045_output_logits.png filter=lfs diff=lfs merge=lfs -text207precomputed_results/p_045/p045_overview_loss_ipr.png filter=lfs diff=lfs merge=lfs -text208precomputed_results/p_045/p045_phase_align_approx1.png filter=lfs diff=lfs merge=lfs -text209precomputed_results/p_045/p045_phase_align_approx2.png filter=lfs diff=lfs merge=lfs -text210precomputed_results/p_045/p045_single_freq_quad.png filter=lfs diff=lfs merge=lfs -text211precomputed_results/p_045/p045_single_freq_relu.png filter=lfs diff=lfs merge=lfs -text212