Upload folder using huggingface_hub

runs/cnn_h128-64_lr0.001/best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1950692807bf443dd5c5a6d80d20c7de1338fe659c33928040b0b9187cb1c976
+size 31377

runs/cnn_h128-64_lr0.001/latest.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a740687f35265ff75136f91233100966c485eeddd4b5369e84e5b3f713c5cdf1
+size 31453

runs/cnn_h128-64_lr0.001/metrics.jsonl

@@ -0,0 +1,10 @@
+{"epoch": 1, "train_loss": 0.6182441682815552, "train_bit_acc": 0.5669285714285714, "train_exact_acc": 0.1549999999829701, "train_opt_gap": 0.8947479520006787, "val_loss": 0.51406378698349, "val_bit_acc": 0.681875, "val_exact_acc": 0.1745, "val_opt_gap": 1.0371363951526986, "train_steps": 110, "val_steps": 32}
+{"epoch": 2, "train_loss": 0.5033679226807185, "train_bit_acc": 0.6985714286395481, "train_exact_acc": 0.2742857142857143, "train_opt_gap": 0.6968994630747729, "val_loss": 0.5010710802078248, "val_bit_acc": 0.665875, "val_exact_acc": 0.271, "val_opt_gap": 0.6363707010690486, "train_steps": 110, "val_steps": 32}
+{"epoch": 3, "train_loss": 0.4767625098909651, "train_bit_acc": 0.7360714286395482, "train_exact_acc": 0.3232857142857143, "train_opt_gap": 0.5912304033070501, "val_loss": 0.46076832938194273, "val_bit_acc": 0.754875, "val_exact_acc": 0.3335, "val_opt_gap": 0.5799466188350524, "train_steps": 110, "val_steps": 32}
+{"epoch": 4, "train_loss": 0.4325618956770216, "train_bit_acc": 0.7803214285714286, "train_exact_acc": 0.41314285717691696, "train_opt_gap": 0.4273940838703238, "val_loss": 0.4285165510177612, "val_bit_acc": 0.77475, "val_exact_acc": 0.3775, "val_opt_gap": 0.46161436827774305, "train_steps": 110, "val_steps": 32}
+{"epoch": 5, "train_loss": 0.41042464406149726, "train_bit_acc": 0.7976428572109767, "train_exact_acc": 0.4312857143197741, "train_opt_gap": 0.43528009130940765, "val_loss": 0.4149975543022156, "val_bit_acc": 0.7915, "val_exact_acc": 0.399, "val_opt_gap": 0.4746292904787352, "train_steps": 110, "val_steps": 32}
+{"epoch": 1, "train_loss": 1.9372418733324324, "train_bit_acc": 0.6768571428571428, "train_exact_acc": 0.2512857142686844, "train_mapping_acc": 0.29221428573131564, "train_mapping_exact_acc": 0.06600000000425747, "train_opt_gap": 0.8415555814988863, "val_loss": 1.8696113319396972, "val_bit_acc": 0.665375, "val_exact_acc": 0.271, "val_mapping_acc": 0.317125, "val_mapping_exact_acc": 0.0815, "val_opt_gap": 0.7234778103648245, "train_steps": 110, "val_steps": 32}
+{"epoch": 2, "train_loss": 1.8659248232160295, "train_bit_acc": 0.6813928571428571, "train_exact_acc": 0.27828571426868437, "train_mapping_acc": 0.3105, "train_mapping_exact_acc": 0.07585714285714286, "train_opt_gap": 0.7563176281128438, "val_loss": 1.8627796459197998, "val_bit_acc": 0.665375, "val_exact_acc": 0.271, "val_mapping_acc": 0.317125, "val_mapping_exact_acc": 0.0815, "val_opt_gap": 0.7234778103648245, "train_steps": 110, "val_steps": 32}
+{"epoch": 3, "train_loss": 1.8591197023391723, "train_bit_acc": 0.689, "train_exact_acc": 0.29685714289120263, "train_mapping_acc": 0.3116785714285714, "train_mapping_exact_acc": 0.07685714286140033, "train_opt_gap": 0.6922208501664494, "val_loss": 1.8541332378387452, "val_bit_acc": 0.664, "val_exact_acc": 0.271, "val_mapping_acc": 0.3165, "val_mapping_exact_acc": 0.0805, "val_opt_gap": 0.6454665685968409, "train_steps": 110, "val_steps": 32}
+{"epoch": 4, "train_loss": 1.8371378718784876, "train_bit_acc": 0.7133928572109768, "train_exact_acc": 0.32157142857142856, "train_mapping_acc": 0.31275, "train_mapping_exact_acc": 0.07928571428571428, "train_opt_gap": 0.6093321152251282, "val_loss": 1.820981255531311, "val_bit_acc": 0.721875, "val_exact_acc": 0.3235, "val_mapping_acc": 0.314, "val_mapping_exact_acc": 0.0815, "val_opt_gap": 0.5369727850583341, "train_steps": 110, "val_steps": 32}
+{"epoch": 5, "train_loss": 1.7969826451710291, "train_bit_acc": 0.7726428571428572, "train_exact_acc": 0.4362857142857143, "train_mapping_acc": 0.30535714289120264, "train_mapping_exact_acc": 0.07385714286565781, "train_opt_gap": 0.48469819825619975, "val_loss": 1.7898901481628418, "val_bit_acc": 0.740625, "val_exact_acc": 0.317, "val_mapping_acc": 0.31775, "val_mapping_exact_acc": 0.0745, "val_opt_gap": 0.5744868670974689, "train_steps": 110, "val_steps": 32}

runs/cnn_h64-32_lr0.001/best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b9a5f83bdc408438d9d98559eb20bfb16271497993543e07de4567ef35b4ed0
+size 28113

runs/cnn_h64-32_lr0.001/latest.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:478085cb2ba29d63d9c29a39171a37e4020ad548e0abae8579c37af0ab6ef896
+size 28189

runs/cnn_h64-32_lr0.001/metrics.jsonl

@@ -0,0 +1,5 @@
+{"epoch": 1, "train_loss": 0.580137266431536, "train_bit_acc": 0.6502857143538339, "train_exact_acc": 0.177, "train_opt_gap": 0.5026852454070532, "val_loss": 0.5149517555236817, "val_bit_acc": 0.657, "val_exact_acc": 0.08, "val_opt_gap": 1.194308737310749, "train_steps": 110, "val_steps": 32}
+{"epoch": 2, "train_loss": 0.5074723009722574, "train_bit_acc": 0.6846785714285715, "train_exact_acc": 0.2487142856802259, "train_opt_gap": 0.7432729707241915, "val_loss": 0.5083248682022095, "val_bit_acc": 0.665875, "val_exact_acc": 0.271, "val_opt_gap": 0.6386823246237112, "train_steps": 110, "val_steps": 32}
+{"epoch": 3, "train_loss": 0.49414891474587574, "train_bit_acc": 0.7093214285033089, "train_exact_acc": 0.30728571425165446, "train_opt_gap": 0.6489625136015533, "val_loss": 0.48554082441329954, "val_bit_acc": 0.722875, "val_exact_acc": 0.3135, "val_opt_gap": 0.611442489767333, "train_steps": 110, "val_steps": 32}
+{"epoch": 4, "train_loss": 0.45753469807761055, "train_bit_acc": 0.746928571496691, "train_exact_acc": 0.3607142856802259, "train_opt_gap": 0.4886693076489082, "val_loss": 0.4456522121429443, "val_bit_acc": 0.751375, "val_exact_acc": 0.332, "val_opt_gap": 0.5300901962316218, "train_steps": 110, "val_steps": 32}
+{"epoch": 5, "train_loss": 0.4228967423779624, "train_bit_acc": 0.7712857142857142, "train_exact_acc": 0.34885714285714287, "train_opt_gap": 0.5479218608171623, "val_loss": 0.42140655303001406, "val_bit_acc": 0.7705, "val_exact_acc": 0.328, "val_opt_gap": 0.6148256471572304, "train_steps": 110, "val_steps": 32}

runs/mlp_h128-64_lr0.001/best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0ace26fff2992f9148b4121be13df3831fe451f3f07d4e2ae57b2ec84a24ce74
+size 362817

runs/mlp_h128-64_lr0.001/latest.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b0d01c8ba1de136b44c49b87db1425186b1e22ac9dc3c3b9a1e66bb4b8def6d3
+size 362877

runs/mlp_h128-64_lr0.001/metrics.jsonl

@@ -0,0 +1,10 @@
+{"epoch": 1, "train_loss": 0.3645717891795295, "train_bit_acc": 0.8385000000681196, "train_exact_acc": 0.4900000000681196, "train_opt_gap": 0.4510521061875763, "val_loss": 0.20793694293498993, "val_bit_acc": 0.927875, "val_exact_acc": 0.7135, "val_opt_gap": 0.18489978510798985, "train_steps": 110, "val_steps": 32}
+{"epoch": 2, "train_loss": 0.13140959232194083, "train_bit_acc": 0.9572857142175947, "train_exact_acc": 0.8324285713604518, "train_opt_gap": 0.10856030916905282, "val_loss": 0.09674916106462479, "val_bit_acc": 0.9665, "val_exact_acc": 0.869, "val_opt_gap": 0.053046944697105865, "train_steps": 110, "val_steps": 32}
+{"epoch": 3, "train_loss": 0.07681818102938788, "train_bit_acc": 0.9774642857824053, "train_exact_acc": 0.9125714286395482, "train_opt_gap": 0.03140815307417969, "val_loss": 0.06602364593744278, "val_bit_acc": 0.981625, "val_exact_acc": 0.9285, "val_opt_gap": 0.020702274490847366, "train_steps": 110, "val_steps": 32}
+{"epoch": 4, "train_loss": 0.05211190024656909, "train_bit_acc": 0.98625, "train_exact_acc": 0.9462857142857143, "train_opt_gap": 0.01022859023778252, "val_loss": 0.04416496294736862, "val_bit_acc": 0.98675, "val_exact_acc": 0.948, "val_opt_gap": 0.006393083893364323, "train_steps": 110, "val_steps": 32}
+{"epoch": 5, "train_loss": 0.03320242746387209, "train_bit_acc": 0.992821428503309, "train_exact_acc": 0.9721428570747376, "train_opt_gap": 0.002376981958738094, "val_loss": 0.028211043529212474, "val_bit_acc": 0.992375, "val_exact_acc": 0.9705, "val_opt_gap": 0.0, "train_steps": 110, "val_steps": 32}
+{"epoch": 1, "train_loss": 1.7748839757101877, "train_bit_acc": 0.8134285714285714, "train_exact_acc": 0.421571428503309, "train_mapping_acc": 0.30678571425165446, "train_mapping_exact_acc": 0.07428571428571429, "train_opt_gap": 0.6493494041131348, "val_loss": 1.6149305572509767, "val_bit_acc": 0.91275, "val_exact_acc": 0.6865, "val_mapping_acc": 0.324875, "val_mapping_exact_acc": 0.087, "val_opt_gap": 0.39023795096807307, "train_steps": 110, "val_steps": 32}
+{"epoch": 2, "train_loss": 1.5291621843065535, "train_bit_acc": 0.9366428570747376, "train_exact_acc": 0.7565714285033089, "train_mapping_acc": 0.31657142853736875, "train_mapping_exact_acc": 0.08385714284011296, "train_opt_gap": 0.2659146966517049, "val_loss": 1.4789328384399414, "val_bit_acc": 0.956625, "val_exact_acc": 0.8295, "val_mapping_acc": 0.322125, "val_mapping_exact_acc": 0.0845, "val_opt_gap": 0.12581568844264457, "train_steps": 110, "val_steps": 32}
+{"epoch": 3, "train_loss": 1.4458946569987705, "train_bit_acc": 0.9662857143538339, "train_exact_acc": 0.8698571429252625, "train_mapping_acc": 0.32289285714285715, "train_mapping_exact_acc": 0.0840000000170299, "train_opt_gap": 0.12463749809460628, "val_loss": 1.4437950382232665, "val_bit_acc": 0.969875, "val_exact_acc": 0.88, "val_mapping_acc": 0.32225, "val_mapping_exact_acc": 0.0855, "val_opt_gap": 0.10015244655319303, "train_steps": 110, "val_steps": 32}
+{"epoch": 4, "train_loss": 1.4106814974376134, "train_bit_acc": 0.976428571496691, "train_exact_acc": 0.9100000000681195, "train_mapping_acc": 0.3383928571769169, "train_mapping_exact_acc": 0.09514285714711462, "train_opt_gap": 0.09075825861856833, "val_loss": 1.4061835584640503, "val_bit_acc": 0.974875, "val_exact_acc": 0.902, "val_mapping_acc": 0.337875, "val_mapping_exact_acc": 0.1, "val_opt_gap": 0.09120948045398729, "train_steps": 110, "val_steps": 32}
+{"epoch": 5, "train_loss": 1.3765283599581037, "train_bit_acc": 0.9826428570747375, "train_exact_acc": 0.9337142856461661, "train_mapping_acc": 0.36017857139451165, "train_mapping_exact_acc": 0.11314285715988705, "train_opt_gap": 0.07302094313579098, "val_loss": 1.3741267080307007, "val_bit_acc": 0.984375, "val_exact_acc": 0.9385, "val_mapping_acc": 0.371, "val_mapping_exact_acc": 0.123, "val_opt_gap": 0.062400219656476, "train_steps": 110, "val_steps": 32}

runs/mlp_h64-32_lr0.001/best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e9c3ff80c2ccac5c72efee607f1a576b5e2c6c1433e317c64b71cca386d588d3
+size 154113

runs/mlp_h64-32_lr0.001/latest.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5a663579db617f6081fe04d796e36d1bb001be963e50d8cdb52559b1a26770e7
+size 154173

runs/mlp_h64-32_lr0.001/metrics.jsonl

@@ -0,0 +1,5 @@
+{"epoch": 1, "train_loss": 0.44679641819000243, "train_bit_acc": 0.8003571427890233, "train_exact_acc": 0.4117142857142857, "train_opt_gap": 0.5156520087839899, "val_loss": 0.2899297285079956, "val_bit_acc": 0.90225, "val_exact_acc": 0.645, "val_opt_gap": 0.35159553472301014, "train_steps": 110, "val_steps": 32}
+{"epoch": 2, "train_loss": 0.2030170758962631, "train_bit_acc": 0.9297500000681196, "train_exact_acc": 0.728428571496691, "train_opt_gap": 0.23026969726325242, "val_loss": 0.14096497011184692, "val_bit_acc": 0.956125, "val_exact_acc": 0.8255, "val_opt_gap": 0.10383042466360437, "train_steps": 110, "val_steps": 32}
+{"epoch": 3, "train_loss": 0.11499322422913143, "train_bit_acc": 0.9644642857142857, "train_exact_acc": 0.8587142857142858, "train_opt_gap": 0.07512260565253527, "val_loss": 0.10989983016252518, "val_bit_acc": 0.962625, "val_exact_acc": 0.851, "val_opt_gap": 0.06235965729883336, "train_steps": 110, "val_steps": 32}
+{"epoch": 4, "train_loss": 0.08548988983886582, "train_bit_acc": 0.9764642856461662, "train_exact_acc": 0.9069999999318804, "train_opt_gap": 0.03449031781373011, "val_loss": 0.08120208898186684, "val_bit_acc": 0.978, "val_exact_acc": 0.9125, "val_opt_gap": 0.03125507681200336, "train_steps": 110, "val_steps": 32}
+{"epoch": 5, "train_loss": 0.06947661690626826, "train_bit_acc": 0.9818928571428571, "train_exact_acc": 0.9292857142857143, "train_opt_gap": 0.018181676038943845, "val_loss": 0.06612214753031731, "val_bit_acc": 0.98175, "val_exact_acc": 0.9275, "val_opt_gap": 0.016602754593125057, "train_steps": 110, "val_steps": 32}

runs/vector_h128-64_lr0.001/best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c42c28082533e9aa245b0868db40abfb2da1126800605d2460060b8f6104908
+size 852593

runs/vector_h128-64_lr0.001/latest.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1f925b6237699c5605b381fe78e098e9c9129ded5b03f07a17d131bd0f3c8d96
+size 852701

runs/vector_h128-64_lr0.001/metrics.jsonl

@@ -0,0 +1,10 @@
+{"epoch": 1, "train_loss": 0.13961991440993735, "train_bit_acc": 0.9322142857142857, "train_exact_acc": 0.8117142857142857, "train_opt_gap": 0.14961570667634536, "val_loss": 0.004755560530349612, "val_bit_acc": 0.999625, "val_exact_acc": 0.9995, "val_opt_gap": 0.0, "train_steps": 110, "val_steps": 32}
+{"epoch": 2, "train_loss": 0.004610078685484561, "train_bit_acc": 0.9997142857142857, "train_exact_acc": 0.9992857142857143, "train_opt_gap": 8.7717858101044e-05, "val_loss": 0.014474161362275481, "val_bit_acc": 0.995625, "val_exact_acc": 0.9835, "val_opt_gap": 0.004978418436734743, "train_steps": 110, "val_steps": 32}
+{"epoch": 3, "train_loss": 0.004731700062286109, "train_bit_acc": 0.9996071428571428, "train_exact_acc": 0.9988571428571429, "train_opt_gap": 0.0003457115583982322, "val_loss": 0.027983925372362137, "val_bit_acc": 0.992375, "val_exact_acc": 0.9705, "val_opt_gap": 0.010528722909128251, "train_steps": 110, "val_steps": 32}
+{"epoch": 4, "train_loss": 0.0037781222663013196, "train_bit_acc": 0.9996071428571428, "train_exact_acc": 0.9988571428571429, "train_opt_gap": 0.0008599790009906273, "val_loss": 0.005080595665611327, "val_bit_acc": 0.999625, "val_exact_acc": 0.9995, "val_opt_gap": 0.0, "train_steps": 110, "val_steps": 32}
+{"epoch": 5, "train_loss": 0.002524018256259816, "train_bit_acc": 0.9997857142857143, "train_exact_acc": 0.9995714285714286, "train_opt_gap": 1.7199580019812547e-06, "val_loss": 0.003144834449281916, "val_bit_acc": 0.999625, "val_exact_acc": 0.9995, "val_opt_gap": 0.0, "train_steps": 110, "val_steps": 32}
+{"epoch": 1, "train_loss": 1.565076292037964, "train_bit_acc": 0.9202142857142858, "train_exact_acc": 0.7404285714285714, "train_mapping_acc": 0.31307142857142856, "train_mapping_exact_acc": 0.08228571428997176, "train_opt_gap": 0.2829805510698709, "val_loss": 1.3179095268249512, "val_bit_acc": 0.991375, "val_exact_acc": 0.9665, "val_mapping_acc": 0.3705, "val_mapping_exact_acc": 0.1085, "val_opt_gap": 0.06046372594613642, "train_steps": 110, "val_steps": 32}
+{"epoch": 2, "train_loss": 1.1731938363483974, "train_bit_acc": 0.9966785714285714, "train_exact_acc": 0.9885714285714285, "train_mapping_acc": 0.4326785714626312, "train_mapping_exact_acc": 0.17785714285714285, "train_opt_gap": 0.038658251693232795, "val_loss": 1.0859097728729248, "val_bit_acc": 0.9995, "val_exact_acc": 0.999, "val_mapping_acc": 0.48325, "val_mapping_exact_acc": 0.221, "val_opt_gap": 0.030068921754722488, "train_steps": 110, "val_steps": 32}
+{"epoch": 3, "train_loss": 1.0434970055988857, "train_bit_acc": 0.9989642857142857, "train_exact_acc": 0.9977142857142857, "train_mapping_acc": 0.4909285713945116, "train_mapping_exact_acc": 0.23985714284011295, "train_opt_gap": 0.029825405208623368, "val_loss": 1.0224776043891908, "val_bit_acc": 0.999625, "val_exact_acc": 0.9995, "val_mapping_acc": 0.4735, "val_mapping_exact_acc": 0.21, "val_opt_gap": 0.02957383112974894, "train_steps": 110, "val_steps": 32}
+{"epoch": 4, "train_loss": 0.9976327650887625, "train_bit_acc": 0.9991071428571429, "train_exact_acc": 0.9982857142857143, "train_mapping_acc": 0.5084642857483455, "train_mapping_exact_acc": 0.2549999999659402, "train_opt_gap": 0.028982050111471334, "val_loss": 0.9903940615653992, "val_bit_acc": 0.9995, "val_exact_acc": 0.999, "val_mapping_acc": 0.52025, "val_mapping_exact_acc": 0.276, "val_opt_gap": 0.027342789832399794, "train_steps": 110, "val_steps": 32}
+{"epoch": 5, "train_loss": 0.9588834537778582, "train_bit_acc": 0.9991428571428571, "train_exact_acc": 0.9984285714285714, "train_mapping_acc": 0.5425357143538339, "train_mapping_exact_acc": 0.3090000000340598, "train_opt_gap": 0.024876289521726894, "val_loss": 0.9603699526786804, "val_bit_acc": 0.9995, "val_exact_acc": 0.999, "val_mapping_acc": 0.5455, "val_mapping_exact_acc": 0.3225, "val_opt_gap": 0.023425933622166044, "train_steps": 110, "val_steps": 32}

runs/vector_h64-32_lr0.001/best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:89ca0c267ae3cb02f5b717d7bf22965bbec4b869822a01fc1bdf07e78d138dc9
+size 386417

runs/vector_h64-32_lr0.001/latest.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c8d18f81fd351a861c3eab12122211bf5c1eda75588195b39164a60c52abcd8
+size 386525

runs/vector_h64-32_lr0.001/metrics.jsonl

@@ -0,0 +1,5 @@
+{"epoch": 1, "train_loss": 0.1859106548407248, "train_bit_acc": 0.9181428571428571, "train_exact_acc": 0.7505714285714286, "train_opt_gap": 0.19270065462597566, "val_loss": 0.0316567013412714, "val_bit_acc": 0.988875, "val_exact_acc": 0.9565, "val_opt_gap": 0.0031182838575920147, "train_steps": 110, "val_steps": 32}
+{"epoch": 2, "train_loss": 0.00878694873043735, "train_bit_acc": 0.9988928571428571, "train_exact_acc": 0.996, "train_opt_gap": 0.000830739714956946, "val_loss": 0.004599129845853895, "val_bit_acc": 0.999625, "val_exact_acc": 0.9995, "val_opt_gap": 0.0, "train_steps": 110, "val_steps": 32}
+{"epoch": 3, "train_loss": 0.004165871840219812, "train_bit_acc": 0.9997857142857143, "train_exact_acc": 0.9995714285714286, "train_opt_gap": 1.7199580019812547e-06, "val_loss": 0.004101993801537901, "val_bit_acc": 0.999625, "val_exact_acc": 0.9995, "val_opt_gap": 0.0, "train_steps": 110, "val_steps": 32}
+{"epoch": 4, "train_loss": 0.005737838335069163, "train_bit_acc": 0.9992142857142857, "train_exact_acc": 0.9972857142857143, "train_opt_gap": 0.0008599790009906275, "val_loss": 0.027336503296159206, "val_bit_acc": 0.9915, "val_exact_acc": 0.967, "val_opt_gap": 0.014790778838037798, "train_steps": 110, "val_steps": 32}
+{"epoch": 5, "train_loss": 0.004519833138412131, "train_bit_acc": 0.9994642857142857, "train_exact_acc": 0.9982857142857143, "train_opt_gap": 0.0008049403449272273, "val_loss": 0.003976050034747459, "val_bit_acc": 0.999625, "val_exact_acc": 0.9995, "val_opt_gap": 0.0, "train_steps": 110, "val_steps": 32}