assignment-1 run transformer-seed2-20261020-205839

assignment-1/runs/transformer-seed2-20261020-205839/checkpoint.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fbdd2bcf96156b328457e4e0e92bb660f345f0aeba7b681fc229cfa1be2cf742
+size 2175925

assignment-1/runs/transformer-seed2-20261020-205839/checksums.json

@@ -0,0 +1,10 @@
+{
+  "checkpoint.pt": "fbdd2bcf96156b328457e4e0e92bb660f345f0aeba7b681fc229cfa1be2cf742",
+  "config.yaml": "4779b7f629b3357575159fd1b6f01189208c1e76a91ecef6f1072500bcf39b79",
+  "environment.json": "0e6cb2d99270a3e06a7c1ccc1840746db19f4b623550957e3374d32e83f7ed0a",
+  "history.json": "4eba0885e3a9ccc0b780ec42f3be1b9209efb27628e22ce9f990d948eea3c247",
+  "summary.json": "6c89e33d830347ccb41ec532901f1eebda3aef2774d7d50af4a37e1ecf946a81",
+  "evaluation/val/correct.png": "650d213d661b2d1ee5715eaaa529cbbc7cc4aeb96e6e68c9209871fcc014d984",
+  "evaluation/val/incorrect.png": "67c8a939fd6e5f55c50ede9315229e0985770165ad961d729dd26c5496327a3c",
+  "evaluation/val/report.json": "dac330d03d5a325ef5281d746d8d48e34d0d70a9ee3f09a23bdb9178dd97d83d"
+}

assignment-1/runs/transformer-seed2-20261020-205839/config.yaml

@@ -0,0 +1,58 @@
+seed: 2
+dataset:
+  name: fashion-mnist
+  source: torchvision.datasets.FashionMNIST
+  root: data
+  download: true
+  val_size: 6000
+  stratified: true
+  split_file: splits/fashion-mnist-seed0.json
+  normalization:
+    mean: 0.28579843044281006
+    std: 0.3528536558151245
+hub:
+  repo_id: null
+  assignment: assignment-1
+  private: false
+input:
+  representation: columns
+  patch_size: 4
+loader:
+  batch_size: 128
+  num_workers: 4
+  augment: true
+train:
+  epochs: 30
+  optimizer: adamw
+  lr: 0.0003
+  weight_decay: 0.05
+  scheduler: cosine
+  label_smoothing: 0.0
+  early_stopping_patience: 5
+  checkpoint_metric: val_macro_f1
+  checkpoint_mode: max
+evaluation:
+  metrics:
+  - accuracy
+  - macro_f1
+  confusion_matrix: true
+  example_predictions: 8
+  output_dir: results/evaluation
+  inference:
+    batch_sizes:
+    - 1
+    - 128
+    warmup_batches: 5
+    timed_batches: 50
+model:
+  args:
+    representation: columns
+    patch_size: 4
+    d_model: 128
+    num_heads: 4
+    num_layers: 4
+    ff_dim: 256
+    dropout: 0.1
+    pooling: cls
+  name: transformer
+output_dir: results/runs

assignment-1/runs/transformer-seed2-20261020-205839/environment.json

@@ -0,0 +1,10 @@
+{
+  "commit": "42f775dbfcb780f7517ddc8ea8f33863b848da6d",
+  "python": "3.12.0",
+  "torch": "2.14.0",
+  "numpy": "2.5.3",
+  "torchvision": "0.29.0",
+  "scikit_learn": "1.9.1",
+  "hardware": "mps: Apple M4 Pro, 14 cores, 48 GB RAM",
+  "torch_threads": 10
+}

assignment-1/runs/transformer-seed2-20261020-205839/evaluation/val/report.json

@@ -0,0 +1,172 @@
+{
+  "run_id": "transformer-seed2-20261020-205839",
+  "model": "transformer",
+  "representation": "columns",
+  "split": "val",
+  "accuracy": 0.8951666951179504,
+  "macro_f1": 0.8948786221866699,
+  "per_class_precision": {
+    "T-shirt/top": 0.8314606741573034,
+    "Trouser": 0.9849498327759197,
+    "Pullover": 0.8207705192629816,
+    "Dress": 0.9166666666666666,
+    "Coat": 0.8003273322422259,
+    "Sandal": 0.9607201309328969,
+    "Shirt": 0.7385964912280701,
+    "Sneaker": 0.9610169491525423,
+    "Bag": 0.9747048903878583,
+    "Ankle boot": 0.9588138385502472
+  },
+  "per_class_recall": {
+    "T-shirt/top": 0.8633333333333333,
+    "Trouser": 0.9816666666666667,
+    "Pullover": 0.8166666666666667,
+    "Dress": 0.9166666666666666,
+    "Coat": 0.815,
+    "Sandal": 0.9783333333333334,
+    "Shirt": 0.7016666666666667,
+    "Sneaker": 0.945,
+    "Bag": 0.9633333333333334,
+    "Ankle boot": 0.97
+  },
+  "per_class_f1": {
+    "T-shirt/top": 0.8470973017170891,
+    "Trouser": 0.9833055091819699,
+    "Pullover": 0.8187134502923976,
+    "Dress": 0.9166666666666666,
+    "Coat": 0.8075970272502064,
+    "Sandal": 0.9694467382328654,
+    "Shirt": 0.7196581196581197,
+    "Sneaker": 0.9529411764705882,
+    "Bag": 0.9689857502095558,
+    "Ankle boot": 0.9643744821872411
+  },
+  "confusion_matrix": [
+    [
+      518,
+      1,
+      5,
+      13,
+      0,
+      0,
+      57,
+      0,
+      6,
+      0
+    ],
+    [
+      1,
+      589,
+      0,
+      5,
+      1,
+      0,
+      3,
+      0,
+      1,
+      0
+    ],
+    [
+      4,
+      0,
+      490,
+      7,
+      62,
+      0,
+      36,
+      0,
+      1,
+      0
+    ],
+    [
+      9,
+      8,
+      3,
+      550,
+      22,
+      0,
+      7,
+      0,
+      1,
+      0
+    ],
+    [
+      3,
+      0,
+      58,
+      11,
+      489,
+      0,
+      38,
+      0,
+      1,
+      0
+    ],
+    [
+      0,
+      0,
+      0,
+      0,
+      0,
+      587,
+      0,
+      10,
+      0,
+      3
+    ],
+    [
+      86,
+      0,
+      38,
+      13,
+      36,
+      0,
+      421,
+      0,
+      5,
+      1
+    ],
+    [
+      0,
+      0,
+      0,
+      0,
+      0,
+      14,
+      0,
+      567,
+      0,
+      19
+    ],
+    [
+      2,
+      0,
+      3,
+      1,
+      1,
+      5,
+      8,
+      0,
+      578,
+      2
+    ],
+    [
+      0,
+      0,
+      0,
+      0,
+      0,
+      5,
+      0,
+      13,
+      0,
+      582
+    ]
+  ],
+  "parameters": 539018,
+  "training_seconds": 551.7139692306519,
+  "inference_ms_per_image": {
+    "1": 1.0601150000002235,
+    "128": 0.018097714846589952
+  }
+}

assignment-1/runs/transformer-seed2-20261020-205839/history.json

@@ -0,0 +1,332 @@
+[
+  {
+    "epoch": 1,
+    "train_loss": 0.7806178406521126,
+    "train_accuracy": 0.7146481481481481,
+    "train_macro_f1": 0.7101357451525749,
+    "val_loss": 0.5697413370609283,
+    "val_accuracy": 0.7826666666666666,
+    "val_macro_f1": 0.7816079048142648,
+    "seconds": 18.809187173843384,
+    "learning_rate": 0.0003
+  },
+  {
+    "epoch": 2,
+    "train_loss": 0.5589737737620318,
+    "train_accuracy": 0.7915555555555556,
+    "train_macro_f1": 0.787517832826533,
+    "val_loss": 0.4944539483388265,
+    "val_accuracy": 0.8123333333333334,
+    "val_macro_f1": 0.8066616073869699,
+    "seconds": 18.325836658477783,
+    "learning_rate": 0.00029917828430524096
+  },
+  {
+    "epoch": 3,
+    "train_loss": 0.5032394073009491,
+    "train_accuracy": 0.8122222222222222,
+    "train_macro_f1": 0.8091333973202527,
+    "val_loss": 0.4620396122932434,
+    "val_accuracy": 0.825,
+    "val_macro_f1": 0.8218883757039881,
+    "seconds": 18.227479219436646,
+    "learning_rate": 0.0002967221401100708
+  },
+  {
+    "epoch": 4,
+    "train_loss": 0.469776316218906,
+    "train_accuracy": 0.8245185185185185,
+    "train_macro_f1": 0.821923838715156,
+    "val_loss": 0.4174961499373118,
+    "val_accuracy": 0.8473333333333334,
+    "val_macro_f1": 0.8424403880797138,
+    "seconds": 18.319931030273438,
+    "learning_rate": 0.00029265847744427303
+  },
+  {
+    "epoch": 5,
+    "train_loss": 0.4446907603210873,
+    "train_accuracy": 0.8337222222222223,
+    "train_macro_f1": 0.831789822056112,
+    "val_loss": 0.3929742465019226,
+    "val_accuracy": 0.854,
+    "val_macro_f1": 0.8523647003039976,
+    "seconds": 18.281349897384644,
+    "learning_rate": 0.0002870318186463901
+  },
+  {
+    "epoch": 6,
+    "train_loss": 0.4272494182145154,
+    "train_accuracy": 0.8397777777777777,
+    "train_macro_f1": 0.8379612002532209,
+    "val_loss": 0.38557294543584186,
+    "val_accuracy": 0.857,
+    "val_macro_f1": 0.8548761778149206,
+    "seconds": 18.312788009643555,
+    "learning_rate": 0.0002799038105676658
+  },
+  {
+    "epoch": 7,
+    "train_loss": 0.4126102040608724,
+    "train_accuracy": 0.8455370370370371,
+    "train_macro_f1": 0.8440143776848764,
+    "val_loss": 0.37501552097002666,
+    "val_accuracy": 0.8575,
+    "val_macro_f1": 0.858814816444092,
+    "seconds": 18.255116939544678,
+    "learning_rate": 0.0002713525491562421
+  },
+  {
+    "epoch": 8,
+    "train_loss": 0.3984466896675251,
+    "train_accuracy": 0.8516851851851852,
+    "train_macro_f1": 0.8503688875551001,
+    "val_loss": 0.34895048586527505,
+    "val_accuracy": 0.8718333333333333,
+    "val_macro_f1": 0.8702742947455822,
+    "seconds": 18.338023900985718,
+    "learning_rate": 0.00026147172382160914
+  },
+  {
+    "epoch": 9,
+    "train_loss": 0.38997761472949277,
+    "train_accuracy": 0.8549814814814815,
+    "train_macro_f1": 0.8538079963629741,
+    "val_loss": 0.3682751024564107,
+    "val_accuracy": 0.8606666666666667,
+    "val_macro_f1": 0.861118849916377,
+    "seconds": 18.29610514640808,
+    "learning_rate": 0.00025036959095382875
+  },
+  {
+    "epoch": 10,
+    "train_loss": 0.3769540901625598,
+    "train_accuracy": 0.8584444444444445,
+    "train_macro_f1": 0.8573874658336589,
+    "val_loss": 0.3431835707028707,
+    "val_accuracy": 0.8718333333333333,
+    "val_macro_f1": 0.8705019449959235,
+    "seconds": 19.08284282684326,
+    "learning_rate": 0.000238167787843871
+  },
+  {
+    "epoch": 11,
+    "train_loss": 0.368003127111329,
+    "train_accuracy": 0.8630185185185185,
+    "train_macro_f1": 0.8619319093227837,
+    "val_loss": 0.3345717782974243,
+    "val_accuracy": 0.8766666666666667,
+    "val_macro_f1": 0.8756633163896547,
+    "seconds": 19.444964170455933,
+    "learning_rate": 0.00022500000000000002
+  },
+  {
+    "epoch": 12,
+    "train_loss": 0.35966367656213266,
+    "train_accuracy": 0.8653703703703703,
+    "train_macro_f1": 0.8644443017380075,
+    "val_loss": 0.3263483492136002,
+    "val_accuracy": 0.8755,
+    "val_macro_f1": 0.8742563389464968,
+    "seconds": 19.016249656677246,
+    "learning_rate": 0.0002110104964613701
+  },
+  {
+    "epoch": 13,
+    "train_loss": 0.3515450120413745,
+    "train_accuracy": 0.8678518518518519,
+    "train_macro_f1": 0.8670250211957361,
+    "val_loss": 0.35119907355308533,
+    "val_accuracy": 0.8683333333333333,
+    "val_macro_f1": 0.8698206843999378,
+    "seconds": 18.142359972000122,
+    "learning_rate": 0.00019635254915624216
+  },
+  {
+    "epoch": 14,
+    "train_loss": 0.34299085524347095,
+    "train_accuracy": 0.8715925925925926,
+    "train_macro_f1": 0.8708377417335113,
+    "val_loss": 0.32407541648546856,
+    "val_accuracy": 0.8768333333333334,
+    "val_macro_f1": 0.8758322927862526,
+    "seconds": 18.218376874923706,
+    "learning_rate": 0.00018118675362266393
+  },
+  {
+    "epoch": 15,
+    "train_loss": 0.335953480331986,
+    "train_accuracy": 0.8730185185185185,
+    "train_macro_f1": 0.872297948957241,
+    "val_loss": 0.3176193518638611,
+    "val_accuracy": 0.8848333333333334,
+    "val_macro_f1": 0.8841871762124305,
+    "seconds": 18.292726039886475,
+    "learning_rate": 0.0001656792694901481
+  },
+  {
+    "epoch": 16,
+    "train_loss": 0.32753377961229396,
+    "train_accuracy": 0.8776851851851852,
+    "train_macro_f1": 0.8769990767750571,
+    "val_loss": 0.31427140506108603,
+    "val_accuracy": 0.8825,
+    "val_macro_f1": 0.8819530636534401,
+    "seconds": 18.40936779975891,
+    "learning_rate": 0.0001500000000000001
+  },
+  {
+    "epoch": 17,
+    "train_loss": 0.32333642591370476,
+    "train_accuracy": 0.8782962962962962,
+    "train_macro_f1": 0.8776576788203494,
+    "val_loss": 0.3118489716053009,
+    "val_accuracy": 0.883,
+    "val_macro_f1": 0.8818589983316857,
+    "seconds": 19.493489265441895,
+    "learning_rate": 0.00013432073050985205
+  },
+  {
+    "epoch": 18,
+    "train_loss": 0.31714425104635735,
+    "train_accuracy": 0.8797222222222222,
+    "train_macro_f1": 0.8791687131479206,
+    "val_loss": 0.3100802494684855,
+    "val_accuracy": 0.8836666666666667,
+    "val_macro_f1": 0.8834829191460875,
+    "seconds": 18.710447072982788,
+    "learning_rate": 0.00011881324637733612
+  },
+  {
+    "epoch": 19,
+    "train_loss": 0.3144437237757224,
+    "train_accuracy": 0.8811666666666667,
+    "train_macro_f1": 0.8805259042365343,
+    "val_loss": 0.3022348705927531,
+    "val_accuracy": 0.8861666666666667,
+    "val_macro_f1": 0.8853956744067732,
+    "seconds": 18.267027139663696,
+    "learning_rate": 0.00010364745084375793
+  },
+  {
+    "epoch": 20,
+    "train_loss": 0.3055372540862472,
+    "train_accuracy": 0.8840740740740741,
+    "train_macro_f1": 0.883507301169845,
+    "val_loss": 0.30020414996147154,
+    "val_accuracy": 0.8886666666666667,
+    "val_macro_f1": 0.8877182355397775,
+    "seconds": 18.156261920928955,
+    "learning_rate": 8.898950353863002e-05
+  },
+  {
+    "epoch": 21,
+    "train_loss": 0.30003270770885326,
+    "train_accuracy": 0.8861481481481481,
+    "train_macro_f1": 0.8856120461050034,
+    "val_loss": 0.29142689871788025,
+    "val_accuracy": 0.8908333333333334,
+    "val_macro_f1": 0.8904304372542924,
+    "seconds": 18.12278914451599,
+    "learning_rate": 7.500000000000005e-05
+  },
+  {
+    "epoch": 22,
+    "train_loss": 0.2951199641801693,
+    "train_accuracy": 0.8882222222222222,
+    "train_macro_f1": 0.8876697400378186,
+    "val_loss": 0.2911049678325653,
+    "val_accuracy": 0.8908333333333334,
+    "val_macro_f1": 0.8905292332412303,
+    "seconds": 18.135189056396484,
+    "learning_rate": 6.183221215612906e-05
+  },
+  {
+    "epoch": 23,
+    "train_loss": 0.292692875429436,
+    "train_accuracy": 0.8891851851851852,
+    "train_macro_f1": 0.8886073604031923,
+    "val_loss": 0.2859543337424596,
+    "val_accuracy": 0.8911666666666667,
+    "val_macro_f1": 0.891001153051263,
+    "seconds": 18.095234155654907,
+    "learning_rate": 4.963040904617133e-05
+  },
+  {
+    "epoch": 24,
+    "train_loss": 0.2891275723245409,
+    "train_accuracy": 0.8910370370370371,
+    "train_macro_f1": 0.8905072199083325,
+    "val_loss": 0.2873632102807363,
+    "val_accuracy": 0.8911666666666667,
+    "val_macro_f1": 0.891217499250839,
+    "seconds": 18.14644718170166,
+    "learning_rate": 3.852827617839091e-05
+  },
+  {
+    "epoch": 25,
+    "train_loss": 0.28615586784150865,
+    "train_accuracy": 0.8918518518518519,
+    "train_macro_f1": 0.8914359668454302,
+    "val_loss": 0.2854286580085754,
+    "val_accuracy": 0.8946666666666667,
+    "val_macro_f1": 0.8941065670313997,
+    "seconds": 18.08373999595642,
+    "learning_rate": 2.8647450843757908e-05
+  },
+  {
+    "epoch": 26,
+    "train_loss": 0.28278572276786523,
+    "train_accuracy": 0.8933333333333333,
+    "train_macro_f1": 0.8927984560214618,
+    "val_loss": 0.2852610646486282,
+    "val_accuracy": 0.8945,
+    "val_macro_f1": 0.8942220577245017,
+    "seconds": 18.120917797088623,
+    "learning_rate": 2.00961894323342e-05
+  },
+  {
+    "epoch": 27,
+    "train_loss": 0.2804772971029635,
+    "train_accuracy": 0.8937777777777778,
+    "train_macro_f1": 0.8933842614944011,
+    "val_loss": 0.28624484809239703,
+    "val_accuracy": 0.8938333333333334,
+    "val_macro_f1": 0.8935047580418601,
+    "seconds": 18.079742908477783,
+    "learning_rate": 1.2968181353609857e-05
+  },
+  {
+    "epoch": 28,
+    "train_loss": 0.2798411499306008,
+    "train_accuracy": 0.8945925925925926,
+    "train_macro_f1": 0.8942044902065206,
+    "val_loss": 0.2820353624025981,
+    "val_accuracy": 0.8935,
+    "val_macro_f1": 0.8927859692982443,
+    "seconds": 18.093592882156372,
+    "learning_rate": 7.341522555726972e-06
+  },
+  {
+    "epoch": 29,
+    "train_loss": 0.2755305035820714,
+    "train_accuracy": 0.8954444444444445,
+    "train_macro_f1": 0.8949754523487249,
+    "val_loss": 0.2816239330768585,
+    "val_accuracy": 0.8951666666666667,
+    "val_macro_f1": 0.8948786221866699,
+    "seconds": 18.11569881439209,
+    "learning_rate": 3.2778598899291474e-06
+  },
+  {
+    "epoch": 30,
+    "train_loss": 0.27814557328047573,
+    "train_accuracy": 0.8957222222222222,
+    "train_macro_f1": 0.8953586527190678,
+    "val_loss": 0.2813730777899424,
+    "val_accuracy": 0.8946666666666667,
+    "val_macro_f1": 0.8943231055686844,
+    "seconds": 18.049081087112427,
+    "learning_rate": 8.2171569475899e-07
+  }
+]

assignment-1/runs/transformer-seed2-20261020-205839/summary.json

@@ -0,0 +1,14 @@
+{
+  "run_id": "transformer-seed2-20261020-205839",
+  "model": "transformer",
+  "representation": "columns",
+  "seed": 2,
+  "parameters": 539018,
+  "epochs_trained": 30,
+  "best_epoch": 29,
+  "checkpoint_metric": "val_macro_f1",
+  "checkpoint_value": 0.8948786221866699,
+  "training_seconds": 551.7139692306519,
+  "commit": "42f775dbfcb780f7517ddc8ea8f33863b848da6d",
+  "hardware": "mps: Apple M4 Pro, 14 cores, 48 GB RAM"
+}