assignment-1 run transformer-seed1-20261020-130444

assignment-1/runs/transformer-seed1-20261020-130444/checkpoint.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ad878ba3a0d838aa36eb3aa603d937c7607c4fcbea327db113ac4b7a9c46e79
+size 2175925

assignment-1/runs/transformer-seed1-20261020-130444/checksums.json

@@ -0,0 +1,10 @@
+{
+  "checkpoint.pt": "8ad878ba3a0d838aa36eb3aa603d937c7607c4fcbea327db113ac4b7a9c46e79",
+  "config.yaml": "8f21c7cf2715ee82512adc211d90713d21877ae0ebef768acc36f691eba7946e",
+  "environment.json": "0e6cb2d99270a3e06a7c1ccc1840746db19f4b623550957e3374d32e83f7ed0a",
+  "history.json": "75994dddd7104491dce9102a7494eb62313383671310f512a503cebd8bfbbca7",
+  "summary.json": "2a2ae4063f9a254d47c1ede2d7e43d1c079db2cb51a48505162f81dad3b0ab1c",
+  "evaluation/val/correct.png": "eaa583286143883cca4db87b7e3c1bbc8671fdd22de00c281aa693df6e19a88e",
+  "evaluation/val/incorrect.png": "f1eac8fac1da4b2cb88c81ea3d5449d3324a1d883c28bd8a6fc5d6b4222357bb",
+  "evaluation/val/report.json": "189e862aec4351ba83295bab7341f62c6eb4696ed6ac006c5f3ffa340e5cc9fd"
+}

assignment-1/runs/transformer-seed1-20261020-130444/config.yaml

@@ -0,0 +1,58 @@
+seed: 1
+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-seed1-20261020-130444/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-seed1-20261020-130444/evaluation/val/report.json

@@ -0,0 +1,172 @@
+{
+  "run_id": "transformer-seed1-20261020-130444",
+  "model": "transformer",
+  "representation": "columns",
+  "split": "val",
+  "accuracy": 0.8978333473205566,
+  "macro_f1": 0.8976150453758385,
+  "per_class_precision": {
+    "T-shirt/top": 0.8411669367909238,
+    "Trouser": 0.986644407345576,
+    "Pullover": 0.8539130434782609,
+    "Dress": 0.912828947368421,
+    "Coat": 0.8022508038585209,
+    "Sandal": 0.9684908789386402,
+    "Shirt": 0.7392055267702936,
+    "Sneaker": 0.9390444810543658,
+    "Bag": 0.9715242881072027,
+    "Ankle boot": 0.9612141652613828
+  },
+  "per_class_recall": {
+    "T-shirt/top": 0.865,
+    "Trouser": 0.985,
+    "Pullover": 0.8183333333333334,
+    "Dress": 0.925,
+    "Coat": 0.8316666666666667,
+    "Sandal": 0.9733333333333334,
+    "Shirt": 0.7133333333333334,
+    "Sneaker": 0.95,
+    "Bag": 0.9666666666666667,
+    "Ankle boot": 0.95
+  },
+  "per_class_f1": {
+    "T-shirt/top": 0.8529170090386196,
+    "Trouser": 0.9858215179316097,
+    "Pullover": 0.8357446808510638,
+    "Dress": 0.9188741721854304,
+    "Coat": 0.8166939443535188,
+    "Sandal": 0.970906068162926,
+    "Shirt": 0.726039016115352,
+    "Sneaker": 0.9444904722452361,
+    "Bag": 0.9690893901420217,
+    "Ankle boot": 0.9555741827326069
+  },
+  "confusion_matrix": [
+    [
+      519,
+      1,
+      3,
+      11,
+      2,
+      0,
+      58,
+      0,
+      6,
+      0
+    ],
+    [
+      1,
+      591,
+      1,
+      4,
+      2,
+      0,
+      0,
+      0,
+      1,
+      0
+    ],
+    [
+      5,
+      0,
+      491,
+      7,
+      60,
+      0,
+      36,
+      0,
+      1,
+      0
+    ],
+    [
+      7,
+      7,
+      3,
+      555,
+      20,
+      0,
+      6,
+      0,
+      2,
+      0
+    ],
+    [
+      3,
+      0,
+      43,
+      11,
+      499,
+      0,
+      43,
+      0,
+      1,
+      0
+    ],
+    [
+      0,
+      0,
+      0,
+      1,
+      0,
+      584,
+      0,
+      12,
+      1,
+      2
+    ],
+    [
+      80,
+      0,
+      33,
+      17,
+      37,
+      0,
+      428,
+      0,
+      5,
+      0
+    ],
+    [
+      0,
+      0,
+      0,
+      0,
+      0,
+      10,
+      0,
+      570,
+      0,
+      20
+    ],
+    [
+      2,
+      0,
+      1,
+      2,
+      2,
+      4,
+      8,
+      0,
+      580,
+      1
+    ],
+    [
+      0,
+      0,
+      0,
+      0,
+      0,
+      5,
+      0,
+      25,
+      0,
+      570
+    ]
+  ],
+  "parameters": 539018,
+  "training_seconds": 546.2352831363678,
+  "inference_ms_per_image": {
+    "1": 1.035153340199031,
+    "128": 0.018120227969120606
+  }
+}

assignment-1/runs/transformer-seed1-20261020-130444/history.json

@@ -0,0 +1,332 @@
+[
+  {
+    "epoch": 1,
+    "train_loss": 0.7855742115091394,
+    "train_accuracy": 0.7080185185185185,
+    "train_macro_f1": 0.7028960879748465,
+    "val_loss": 0.5725893616676331,
+    "val_accuracy": 0.7795,
+    "val_macro_f1": 0.7734199368344111,
+    "seconds": 18.319252967834473,
+    "learning_rate": 0.0003
+  },
+  {
+    "epoch": 2,
+    "train_loss": 0.5643165743615892,
+    "train_accuracy": 0.7887777777777778,
+    "train_macro_f1": 0.7849194327771278,
+    "val_loss": 0.4798792770703634,
+    "val_accuracy": 0.8258333333333333,
+    "val_macro_f1": 0.8236672702992596,
+    "seconds": 18.026713132858276,
+    "learning_rate": 0.00029917828430524096
+  },
+  {
+    "epoch": 3,
+    "train_loss": 0.5082971236440871,
+    "train_accuracy": 0.8112407407407407,
+    "train_macro_f1": 0.8086276608230781,
+    "val_loss": 0.45878394691149393,
+    "val_accuracy": 0.8308333333333333,
+    "val_macro_f1": 0.8263495271144807,
+    "seconds": 18.629135847091675,
+    "learning_rate": 0.0002967221401100708
+  },
+  {
+    "epoch": 4,
+    "train_loss": 0.4765870351261563,
+    "train_accuracy": 0.8229074074074074,
+    "train_macro_f1": 0.8205341186465274,
+    "val_loss": 0.4311850510438283,
+    "val_accuracy": 0.838,
+    "val_macro_f1": 0.8359071940792123,
+    "seconds": 19.12859535217285,
+    "learning_rate": 0.00029265847744427303
+  },
+  {
+    "epoch": 5,
+    "train_loss": 0.4546221102785181,
+    "train_accuracy": 0.8307222222222223,
+    "train_macro_f1": 0.8287463788591038,
+    "val_loss": 0.4190142325560252,
+    "val_accuracy": 0.8445,
+    "val_macro_f1": 0.8432148901288541,
+    "seconds": 18.514576196670532,
+    "learning_rate": 0.0002870318186463901
+  },
+  {
+    "epoch": 6,
+    "train_loss": 0.43258711191459936,
+    "train_accuracy": 0.8376851851851852,
+    "train_macro_f1": 0.8359153049024556,
+    "val_loss": 0.3966667764186859,
+    "val_accuracy": 0.8556666666666667,
+    "val_macro_f1": 0.8547178803873212,
+    "seconds": 18.129148244857788,
+    "learning_rate": 0.0002799038105676658
+  },
+  {
+    "epoch": 7,
+    "train_loss": 0.4165567734771305,
+    "train_accuracy": 0.8434259259259259,
+    "train_macro_f1": 0.8420468427853779,
+    "val_loss": 0.3998885817527771,
+    "val_accuracy": 0.8513333333333334,
+    "val_macro_f1": 0.8501656624738894,
+    "seconds": 18.14921808242798,
+    "learning_rate": 0.0002713525491562421
+  },
+  {
+    "epoch": 8,
+    "train_loss": 0.4016880144631421,
+    "train_accuracy": 0.8497962962962963,
+    "train_macro_f1": 0.8484357348868151,
+    "val_loss": 0.35671307174364725,
+    "val_accuracy": 0.865,
+    "val_macro_f1": 0.8640694589008204,
+    "seconds": 18.145450830459595,
+    "learning_rate": 0.00026147172382160914
+  },
+  {
+    "epoch": 9,
+    "train_loss": 0.39201156942932697,
+    "train_accuracy": 0.8546111111111111,
+    "train_macro_f1": 0.8533707863424697,
+    "val_loss": 0.3660530693531036,
+    "val_accuracy": 0.8661666666666666,
+    "val_macro_f1": 0.866723979543561,
+    "seconds": 18.116605043411255,
+    "learning_rate": 0.00025036959095382875
+  },
+  {
+    "epoch": 10,
+    "train_loss": 0.3770344776135904,
+    "train_accuracy": 0.8592407407407407,
+    "train_macro_f1": 0.8579744089619472,
+    "val_loss": 0.34514867758750917,
+    "val_accuracy": 0.8708333333333333,
+    "val_macro_f1": 0.8693746354976506,
+    "seconds": 18.109079837799072,
+    "learning_rate": 0.000238167787843871
+  },
+  {
+    "epoch": 11,
+    "train_loss": 0.37299344863714995,
+    "train_accuracy": 0.8610925925925926,
+    "train_macro_f1": 0.8601310016894994,
+    "val_loss": 0.33409516859054567,
+    "val_accuracy": 0.8745,
+    "val_macro_f1": 0.8737784240034807,
+    "seconds": 18.103153944015503,
+    "learning_rate": 0.00022500000000000002
+  },
+  {
+    "epoch": 12,
+    "train_loss": 0.3633330173315825,
+    "train_accuracy": 0.8642407407407408,
+    "train_macro_f1": 0.8632466305847248,
+    "val_loss": 0.3444760547478994,
+    "val_accuracy": 0.8715,
+    "val_macro_f1": 0.8707349905131994,
+    "seconds": 18.293263912200928,
+    "learning_rate": 0.0002110104964613701
+  },
+  {
+    "epoch": 13,
+    "train_loss": 0.35729905661830197,
+    "train_accuracy": 0.8652407407407408,
+    "train_macro_f1": 0.864345352158159,
+    "val_loss": 0.33548590588569643,
+    "val_accuracy": 0.8721666666666666,
+    "val_macro_f1": 0.8720244109544888,
+    "seconds": 18.126598834991455,
+    "learning_rate": 0.00019635254915624216
+  },
+  {
+    "epoch": 14,
+    "train_loss": 0.34843457345609313,
+    "train_accuracy": 0.8689814814814815,
+    "train_macro_f1": 0.8681020867016069,
+    "val_loss": 0.3385181994438171,
+    "val_accuracy": 0.874,
+    "val_macro_f1": 0.8723406213384033,
+    "seconds": 18.64314579963684,
+    "learning_rate": 0.00018118675362266393
+  },
+  {
+    "epoch": 15,
+    "train_loss": 0.34048544459872776,
+    "train_accuracy": 0.8708333333333333,
+    "train_macro_f1": 0.8700100118833719,
+    "val_loss": 0.3283892033100128,
+    "val_accuracy": 0.8753333333333333,
+    "val_macro_f1": 0.8745928711873976,
+    "seconds": 18.51517105102539,
+    "learning_rate": 0.0001656792694901481
+  },
+  {
+    "epoch": 16,
+    "train_loss": 0.33723165916513514,
+    "train_accuracy": 0.8740925925925926,
+    "train_macro_f1": 0.8731989897128993,
+    "val_loss": 0.30849884247779846,
+    "val_accuracy": 0.8843333333333333,
+    "val_macro_f1": 0.8835580729863967,
+    "seconds": 18.09110689163208,
+    "learning_rate": 0.0001500000000000001
+  },
+  {
+    "epoch": 17,
+    "train_loss": 0.3310878303050995,
+    "train_accuracy": 0.8750370370370371,
+    "train_macro_f1": 0.8742838532618865,
+    "val_loss": 0.3179968660672506,
+    "val_accuracy": 0.8843333333333333,
+    "val_macro_f1": 0.8826581372144577,
+    "seconds": 18.11937117576599,
+    "learning_rate": 0.00013432073050985205
+  },
+  {
+    "epoch": 18,
+    "train_loss": 0.32238500963317024,
+    "train_accuracy": 0.8769444444444444,
+    "train_macro_f1": 0.8762114783998181,
+    "val_loss": 0.30773649231592815,
+    "val_accuracy": 0.8873333333333333,
+    "val_macro_f1": 0.8858031251808042,
+    "seconds": 18.056949138641357,
+    "learning_rate": 0.00011881324637733612
+  },
+  {
+    "epoch": 19,
+    "train_loss": 0.32011558523884526,
+    "train_accuracy": 0.8787962962962963,
+    "train_macro_f1": 0.8781481873429486,
+    "val_loss": 0.3036822547117869,
+    "val_accuracy": 0.8895,
+    "val_macro_f1": 0.8891466710216724,
+    "seconds": 18.03196692466736,
+    "learning_rate": 0.00010364745084375793
+  },
+  {
+    "epoch": 20,
+    "train_loss": 0.3107262215172803,
+    "train_accuracy": 0.8813518518518518,
+    "train_macro_f1": 0.8806316044164954,
+    "val_loss": 0.3029062040646871,
+    "val_accuracy": 0.887,
+    "val_macro_f1": 0.8868803952922599,
+    "seconds": 17.954805850982666,
+    "learning_rate": 8.898950353863002e-05
+  },
+  {
+    "epoch": 21,
+    "train_loss": 0.3042101956164395,
+    "train_accuracy": 0.8860925925925925,
+    "train_macro_f1": 0.885526407721634,
+    "val_loss": 0.296605263710022,
+    "val_accuracy": 0.8908333333333334,
+    "val_macro_f1": 0.8906397889550979,
+    "seconds": 17.881686210632324,
+    "learning_rate": 7.500000000000005e-05
+  },
+  {
+    "epoch": 22,
+    "train_loss": 0.30081665700894816,
+    "train_accuracy": 0.8864074074074074,
+    "train_macro_f1": 0.8858509948520294,
+    "val_loss": 0.29177012006441755,
+    "val_accuracy": 0.8921666666666667,
+    "val_macro_f1": 0.8916198436637061,
+    "seconds": 18.152956247329712,
+    "learning_rate": 6.183221215612906e-05
+  },
+  {
+    "epoch": 23,
+    "train_loss": 0.2975855053177586,
+    "train_accuracy": 0.8876481481481482,
+    "train_macro_f1": 0.8870135552626083,
+    "val_loss": 0.29523315517107646,
+    "val_accuracy": 0.8933333333333333,
+    "val_macro_f1": 0.8926013962500601,
+    "seconds": 18.063379049301147,
+    "learning_rate": 4.963040904617133e-05
+  },
+  {
+    "epoch": 24,
+    "train_loss": 0.29114541902807023,
+    "train_accuracy": 0.8893518518518518,
+    "train_macro_f1": 0.8887786523664254,
+    "val_loss": 0.2899388891061147,
+    "val_accuracy": 0.8933333333333333,
+    "val_macro_f1": 0.8932757092522504,
+    "seconds": 18.191206693649292,
+    "learning_rate": 3.852827617839091e-05
+  },
+  {
+    "epoch": 25,
+    "train_loss": 0.28601285768438267,
+    "train_accuracy": 0.8917037037037037,
+    "train_macro_f1": 0.8912389085838923,
+    "val_loss": 0.2896527975400289,
+    "val_accuracy": 0.8938333333333334,
+    "val_macro_f1": 0.8934155369601562,
+    "seconds": 18.06678295135498,
+    "learning_rate": 2.8647450843757908e-05
+  },
+  {
+    "epoch": 26,
+    "train_loss": 0.2852685272340421,
+    "train_accuracy": 0.8920925925925925,
+    "train_macro_f1": 0.8916670414514097,
+    "val_loss": 0.2850552248160044,
+    "val_accuracy": 0.8978333333333334,
+    "val_macro_f1": 0.8976150453758385,
+    "seconds": 18.03009009361267,
+    "learning_rate": 2.00961894323342e-05
+  },
+  {
+    "epoch": 27,
+    "train_loss": 0.2820568954679701,
+    "train_accuracy": 0.8932037037037037,
+    "train_macro_f1": 0.8927950555041703,
+    "val_loss": 0.2850688910484314,
+    "val_accuracy": 0.8965,
+    "val_macro_f1": 0.8962809487219344,
+    "seconds": 18.089646100997925,
+    "learning_rate": 1.2968181353609857e-05
+  },
+  {
+    "epoch": 28,
+    "train_loss": 0.2823549322110635,
+    "train_accuracy": 0.893462962962963,
+    "train_macro_f1": 0.8929643928265879,
+    "val_loss": 0.28619178891181946,
+    "val_accuracy": 0.8966666666666666,
+    "val_macro_f1": 0.8963559501087637,
+    "seconds": 18.071300983428955,
+    "learning_rate": 7.341522555726972e-06
+  },
+  {
+    "epoch": 29,
+    "train_loss": 0.28011440164954576,
+    "train_accuracy": 0.8948518518518519,
+    "train_macro_f1": 0.8943991525366254,
+    "val_loss": 0.2841674795945485,
+    "val_accuracy": 0.8978333333333334,
+    "val_macro_f1": 0.8974150119116713,
+    "seconds": 18.102773904800415,
+    "learning_rate": 3.2778598899291474e-06
+  },
+  {
+    "epoch": 30,
+    "train_loss": 0.2788687227478734,
+    "train_accuracy": 0.8936296296296297,
+    "train_macro_f1": 0.8930715819404806,
+    "val_loss": 0.2845633130868276,
+    "val_accuracy": 0.8973333333333333,
+    "val_macro_f1": 0.8970030126511048,
+    "seconds": 18.077155828475952,
+    "learning_rate": 8.2171569475899e-07
+  }
+]

assignment-1/runs/transformer-seed1-20261020-130444/summary.json

@@ -0,0 +1,14 @@
+{
+  "run_id": "transformer-seed1-20261020-130444",
+  "model": "transformer",
+  "representation": "columns",
+  "seed": 1,
+  "parameters": 539018,
+  "epochs_trained": 30,
+  "best_epoch": 26,
+  "checkpoint_metric": "val_macro_f1",
+  "checkpoint_value": 0.8976150453758385,
+  "training_seconds": 546.2352831363678,
+  "commit": "42f775dbfcb780f7517ddc8ea8f33863b848da6d",
+  "hardware": "mps: Apple M4 Pro, 14 cores, 48 GB RAM"
+}