assignment-1 run transformer-seed1-20261020-125532

assignment-1/runs/transformer-seed1-20261020-125532/checkpoint.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:072bdb52be7ce9ce65759005a601f6fa369bfcaa7aaf60808081e74af72c58d6
+size 2175925

assignment-1/runs/transformer-seed1-20261020-125532/checksums.json

@@ -0,0 +1,10 @@
+{
+  "checkpoint.pt": "072bdb52be7ce9ce65759005a601f6fa369bfcaa7aaf60808081e74af72c58d6",
+  "config.yaml": "7c4a49aa6e8824ea73573358a583b5b2acc5b94a8531d1f5372ee4d889409fb9",
+  "environment.json": "0e6cb2d99270a3e06a7c1ccc1840746db19f4b623550957e3374d32e83f7ed0a",
+  "history.json": "6731aa72d90ad2ad26c58d3a0e9e2347da6285d656cbb13e23098a07559fbec5",
+  "summary.json": "8e8e8d5a72e9ebbdd63908827d1c9f61088b9ef0035358a97d04efbc0e177385",
+  "evaluation/val/correct.png": "bafeb0b127a58fafde060aa1635977f8198fca1854c1a77b968a72d4413a64ba",
+  "evaluation/val/incorrect.png": "59df5f5553ca8f2826aa8e2f9c289e157682ea41d27b6e43ad2e402f3a39e303",
+  "evaluation/val/report.json": "0be4830b364e817a2075001feb02d36ecafd1aa2efd8fa6ad33d366258cea721"
+}

assignment-1/runs/transformer-seed1-20261020-125532/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: rows
+  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: rows
+    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-seed1-20261020-125532/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-seed1-20261020-125532/evaluation/val/report.json

@@ -0,0 +1,172 @@
+{
+  "run_id": "transformer-seed1-20261020-125532",
+  "model": "transformer",
+  "representation": "rows",
+  "split": "val",
+  "accuracy": 0.9003333449363708,
+  "macro_f1": 0.9004399191772112,
+  "per_class_precision": {
+    "T-shirt/top": 0.8453947368421053,
+    "Trouser": 0.9866220735785953,
+    "Pullover": 0.8358974358974359,
+    "Dress": 0.9131886477462438,
+    "Coat": 0.7898089171974523,
+    "Sandal": 0.9880341880341881,
+    "Shirt": 0.7389830508474576,
+    "Sneaker": 0.9452495974235104,
+    "Bag": 0.9847972972972973,
+    "Ankle boot": 0.9797979797979798
+  },
+  "per_class_recall": {
+    "T-shirt/top": 0.8566666666666667,
+    "Trouser": 0.9833333333333333,
+    "Pullover": 0.815,
+    "Dress": 0.9116666666666666,
+    "Coat": 0.8266666666666667,
+    "Sandal": 0.9633333333333334,
+    "Shirt": 0.7266666666666667,
+    "Sneaker": 0.9783333333333334,
+    "Bag": 0.9716666666666667,
+    "Ankle boot": 0.97
+  },
+  "per_class_f1": {
+    "T-shirt/top": 0.8509933774834437,
+    "Trouser": 0.9849749582637729,
+    "Pullover": 0.8253164556962025,
+    "Dress": 0.9124270225187656,
+    "Coat": 0.8078175895765473,
+    "Sandal": 0.9755274261603376,
+    "Shirt": 0.7327731092436974,
+    "Sneaker": 0.9615069615069615,
+    "Bag": 0.9781879194630873,
+    "Ankle boot": 0.9748743718592965
+  },
+  "confusion_matrix": [
+    [
+      514,
+      1,
+      7,
+      12,
+      1,
+      0,
+      63,
+      0,
+      2,
+      0
+    ],
+    [
+      2,
+      590,
+      0,
+      6,
+      0,
+      0,
+      1,
+      0,
+      1,
+      0
+    ],
+    [
+      5,
+      0,
+      489,
+      4,
+      67,
+      0,
+      35,
+      0,
+      0,
+      0
+    ],
+    [
+      13,
+      6,
+      2,
+      547,
+      22,
+      0,
+      10,
+      0,
+      0,
+      0
+    ],
+    [
+      2,
+      1,
+      44,
+      16,
+      496,
+      0,
+      40,
+      0,
+      1,
+      0
+    ],
+    [
+      0,
+      0,
+      0,
+      1,
+      0,
+      578,
+      0,
+      18,
+      1,
+      2
+    ],
+    [
+      70,
+      0,
+      39,
+      11,
+      40,
+      0,
+      436,
+      0,
+      4,
+      0
+    ],
+    [
+      0,
+      0,
+      0,
+      0,
+      0,
+      3,
+      0,
+      587,
+      0,
+      10
+    ],
+    [
+      2,
+      0,
+      4,
+      2,
+      2,
+      1,
+      5,
+      1,
+      583,
+      0
+    ],
+    [
+      0,
+      0,
+      0,
+      0,
+      0,
+      3,
+      0,
+      15,
+      0,
+      582
+    ]
+  ],
+  "parameters": 539018,
+  "training_seconds": 549.8278179168701,
+  "inference_ms_per_image": {
+    "1": 1.0946474998490885,
+    "128": 0.018090214844050934
+  }
+}

assignment-1/runs/transformer-seed1-20261020-125532/history.json

@@ -0,0 +1,332 @@
+[
+  {
+    "epoch": 1,
+    "train_loss": 0.8477474351459079,
+    "train_accuracy": 0.6822037037037038,
+    "train_macro_f1": 0.6771563776665724,
+    "val_loss": 0.6515940836270651,
+    "val_accuracy": 0.7555,
+    "val_macro_f1": 0.7492436993166336,
+    "seconds": 18.27364492416382,
+    "learning_rate": 0.0003
+  },
+  {
+    "epoch": 2,
+    "train_loss": 0.5942427578325625,
+    "train_accuracy": 0.7768148148148148,
+    "train_macro_f1": 0.773640386669675,
+    "val_loss": 0.4822105747858683,
+    "val_accuracy": 0.8178333333333333,
+    "val_macro_f1": 0.8165323582475528,
+    "seconds": 18.04147505760193,
+    "learning_rate": 0.00029917828430524096
+  },
+  {
+    "epoch": 3,
+    "train_loss": 0.5241721116878368,
+    "train_accuracy": 0.806,
+    "train_macro_f1": 0.8034334252569117,
+    "val_loss": 0.4577854286034902,
+    "val_accuracy": 0.8258333333333333,
+    "val_macro_f1": 0.8222828109977229,
+    "seconds": 18.081893920898438,
+    "learning_rate": 0.0002967221401100708
+  },
+  {
+    "epoch": 4,
+    "train_loss": 0.4868325994986075,
+    "train_accuracy": 0.8188518518518518,
+    "train_macro_f1": 0.8165625679656792,
+    "val_loss": 0.4455214460690816,
+    "val_accuracy": 0.8321666666666667,
+    "val_macro_f1": 0.83092069028235,
+    "seconds": 18.066375255584717,
+    "learning_rate": 0.00029265847744427303
+  },
+  {
+    "epoch": 5,
+    "train_loss": 0.46030119357285676,
+    "train_accuracy": 0.827962962962963,
+    "train_macro_f1": 0.826002512217395,
+    "val_loss": 0.41558539724349974,
+    "val_accuracy": 0.8433333333333334,
+    "val_macro_f1": 0.8408446886157979,
+    "seconds": 18.156234979629517,
+    "learning_rate": 0.0002870318186463901
+  },
+  {
+    "epoch": 6,
+    "train_loss": 0.4375398625885999,
+    "train_accuracy": 0.8359444444444445,
+    "train_macro_f1": 0.8341862326567837,
+    "val_loss": 0.3675247021516164,
+    "val_accuracy": 0.8621666666666666,
+    "val_macro_f1": 0.8621007550297856,
+    "seconds": 18.117130994796753,
+    "learning_rate": 0.0002799038105676658
+  },
+  {
+    "epoch": 7,
+    "train_loss": 0.4249583039107146,
+    "train_accuracy": 0.842,
+    "train_macro_f1": 0.8404289869985376,
+    "val_loss": 0.3741024050712585,
+    "val_accuracy": 0.8613333333333333,
+    "val_macro_f1": 0.8604320044659186,
+    "seconds": 18.07611083984375,
+    "learning_rate": 0.0002713525491562421
+  },
+  {
+    "epoch": 8,
+    "train_loss": 0.4092225739161173,
+    "train_accuracy": 0.8464074074074074,
+    "train_macro_f1": 0.84501777507071,
+    "val_loss": 0.36287187751134237,
+    "val_accuracy": 0.8646666666666667,
+    "val_macro_f1": 0.8628324289887314,
+    "seconds": 18.110090970993042,
+    "learning_rate": 0.00026147172382160914
+  },
+  {
+    "epoch": 9,
+    "train_loss": 0.3990704091743187,
+    "train_accuracy": 0.8505,
+    "train_macro_f1": 0.8493060867800366,
+    "val_loss": 0.35981582657496136,
+    "val_accuracy": 0.865,
+    "val_macro_f1": 0.863891282462722,
+    "seconds": 18.107638835906982,
+    "learning_rate": 0.00025036959095382875
+  },
+  {
+    "epoch": 10,
+    "train_loss": 0.38687067361231203,
+    "train_accuracy": 0.8551296296296297,
+    "train_macro_f1": 0.8539314516865295,
+    "val_loss": 0.3549997150103251,
+    "val_accuracy": 0.8636666666666667,
+    "val_macro_f1": 0.8610561005431894,
+    "seconds": 18.114627838134766,
+    "learning_rate": 0.000238167787843871
+  },
+  {
+    "epoch": 11,
+    "train_loss": 0.3783643187240318,
+    "train_accuracy": 0.8577592592592592,
+    "train_macro_f1": 0.8567511825248841,
+    "val_loss": 0.33545671923955284,
+    "val_accuracy": 0.8733333333333333,
+    "val_macro_f1": 0.8717027893589411,
+    "seconds": 18.134950160980225,
+    "learning_rate": 0.00022500000000000002
+  },
+  {
+    "epoch": 12,
+    "train_loss": 0.36875384525899535,
+    "train_accuracy": 0.8622777777777778,
+    "train_macro_f1": 0.8612942386995188,
+    "val_loss": 0.3295309015115102,
+    "val_accuracy": 0.8746666666666667,
+    "val_macro_f1": 0.8738068671834329,
+    "seconds": 18.411638021469116,
+    "learning_rate": 0.0002110104964613701
+  },
+  {
+    "epoch": 13,
+    "train_loss": 0.36504547277203314,
+    "train_accuracy": 0.8629074074074075,
+    "train_macro_f1": 0.8618998569962801,
+    "val_loss": 0.333461507320404,
+    "val_accuracy": 0.876,
+    "val_macro_f1": 0.8770821933983255,
+    "seconds": 18.99585199356079,
+    "learning_rate": 0.00019635254915624216
+  },
+  {
+    "epoch": 14,
+    "train_loss": 0.3555319798699132,
+    "train_accuracy": 0.8663518518518518,
+    "train_macro_f1": 0.8653675681747348,
+    "val_loss": 0.32645242555936177,
+    "val_accuracy": 0.8766666666666667,
+    "val_macro_f1": 0.8763226422325163,
+    "seconds": 19.066322803497314,
+    "learning_rate": 0.00018118675362266393
+  },
+  {
+    "epoch": 15,
+    "train_loss": 0.34858067391536857,
+    "train_accuracy": 0.8690740740740741,
+    "train_macro_f1": 0.8681582739867284,
+    "val_loss": 0.3091099932193756,
+    "val_accuracy": 0.8836666666666667,
+    "val_macro_f1": 0.8834165054087284,
+    "seconds": 19.657222270965576,
+    "learning_rate": 0.0001656792694901481
+  },
+  {
+    "epoch": 16,
+    "train_loss": 0.34016810108997203,
+    "train_accuracy": 0.8723888888888889,
+    "train_macro_f1": 0.8715154994139331,
+    "val_loss": 0.3082869652907054,
+    "val_accuracy": 0.8846666666666667,
+    "val_macro_f1": 0.884371191013668,
+    "seconds": 18.39118790626526,
+    "learning_rate": 0.0001500000000000001
+  },
+  {
+    "epoch": 17,
+    "train_loss": 0.33520478791660735,
+    "train_accuracy": 0.8740370370370371,
+    "train_macro_f1": 0.8732856266925302,
+    "val_loss": 0.3143378590742747,
+    "val_accuracy": 0.8811666666666667,
+    "val_macro_f1": 0.8804653882974585,
+    "seconds": 18.297115802764893,
+    "learning_rate": 0.00013432073050985205
+  },
+  {
+    "epoch": 18,
+    "train_loss": 0.3300457675457001,
+    "train_accuracy": 0.8759259259259259,
+    "train_macro_f1": 0.8751402894818654,
+    "val_loss": 0.2968063197930654,
+    "val_accuracy": 0.8876666666666667,
+    "val_macro_f1": 0.8865160626891038,
+    "seconds": 19.070133686065674,
+    "learning_rate": 0.00011881324637733612
+  },
+  {
+    "epoch": 19,
+    "train_loss": 0.32542695063131827,
+    "train_accuracy": 0.8774444444444445,
+    "train_macro_f1": 0.8767313299098654,
+    "val_loss": 0.3111685937643051,
+    "val_accuracy": 0.8795,
+    "val_macro_f1": 0.8786597386361249,
+    "seconds": 18.5068199634552,
+    "learning_rate": 0.00010364745084375793
+  },
+  {
+    "epoch": 20,
+    "train_loss": 0.32060488950323174,
+    "train_accuracy": 0.8792037037037037,
+    "train_macro_f1": 0.8785814986027575,
+    "val_loss": 0.29465187430381773,
+    "val_accuracy": 0.8883333333333333,
+    "val_macro_f1": 0.8875407471391806,
+    "seconds": 18.202396869659424,
+    "learning_rate": 8.898950353863002e-05
+  },
+  {
+    "epoch": 21,
+    "train_loss": 0.31707376568847234,
+    "train_accuracy": 0.8807407407407407,
+    "train_macro_f1": 0.8799765732537151,
+    "val_loss": 0.28822426482041674,
+    "val_accuracy": 0.8928333333333334,
+    "val_macro_f1": 0.8929718111748752,
+    "seconds": 18.107588052749634,
+    "learning_rate": 7.500000000000005e-05
+  },
+  {
+    "epoch": 22,
+    "train_loss": 0.31266511322833873,
+    "train_accuracy": 0.8821481481481481,
+    "train_macro_f1": 0.8815029004888913,
+    "val_loss": 0.28188495604197183,
+    "val_accuracy": 0.8948333333333334,
+    "val_macro_f1": 0.8942267101376549,
+    "seconds": 18.078409910202026,
+    "learning_rate": 6.183221215612906e-05
+  },
+  {
+    "epoch": 23,
+    "train_loss": 0.3056478698871754,
+    "train_accuracy": 0.8862962962962962,
+    "train_macro_f1": 0.885649515449159,
+    "val_loss": 0.29011006212234497,
+    "val_accuracy": 0.8928333333333334,
+    "val_macro_f1": 0.8925547654078256,
+    "seconds": 18.098785161972046,
+    "learning_rate": 4.963040904617133e-05
+  },
+  {
+    "epoch": 24,
+    "train_loss": 0.307222580048773,
+    "train_accuracy": 0.8845555555555555,
+    "train_macro_f1": 0.8839531275076504,
+    "val_loss": 0.2853318722248077,
+    "val_accuracy": 0.8931666666666667,
+    "val_macro_f1": 0.893128058932229,
+    "seconds": 18.169212102890015,
+    "learning_rate": 3.852827617839091e-05
+  },
+  {
+    "epoch": 25,
+    "train_loss": 0.3010576022730933,
+    "train_accuracy": 0.8859074074074074,
+    "train_macro_f1": 0.8852287541756363,
+    "val_loss": 0.27492775921026863,
+    "val_accuracy": 0.8983333333333333,
+    "val_macro_f1": 0.8975407198026571,
+    "seconds": 18.178268909454346,
+    "learning_rate": 2.8647450843757908e-05
+  },
+  {
+    "epoch": 26,
+    "train_loss": 0.2998311327475089,
+    "train_accuracy": 0.887537037037037,
+    "train_macro_f1": 0.8869532966817468,
+    "val_loss": 0.274651545882225,
+    "val_accuracy": 0.9003333333333333,
+    "val_macro_f1": 0.9004399191772112,
+    "seconds": 18.09980010986328,
+    "learning_rate": 2.00961894323342e-05
+  },
+  {
+    "epoch": 27,
+    "train_loss": 0.29762842460473377,
+    "train_accuracy": 0.8888518518518519,
+    "train_macro_f1": 0.8883522974479632,
+    "val_loss": 0.2772646809418996,
+    "val_accuracy": 0.8958333333333334,
+    "val_macro_f1": 0.895731980595478,
+    "seconds": 18.10190200805664,
+    "learning_rate": 1.2968181353609857e-05
+  },
+  {
+    "epoch": 28,
+    "train_loss": 0.29677929993028995,
+    "train_accuracy": 0.8888518518518519,
+    "train_macro_f1": 0.8882789787485106,
+    "val_loss": 0.27571588663260144,
+    "val_accuracy": 0.8971666666666667,
+    "val_macro_f1": 0.8970122652050889,
+    "seconds": 18.073351860046387,
+    "learning_rate": 7.341522555726972e-06
+  },
+  {
+    "epoch": 29,
+    "train_loss": 0.29406067896772314,
+    "train_accuracy": 0.8895185185185185,
+    "train_macro_f1": 0.8890023449165094,
+    "val_loss": 0.2735966603755951,
+    "val_accuracy": 0.8986666666666666,
+    "val_macro_f1": 0.8984582388417737,
+    "seconds": 18.41309905052185,
+    "learning_rate": 3.2778598899291474e-06
+  },
+  {
+    "epoch": 30,
+    "train_loss": 0.2939305702580346,
+    "train_accuracy": 0.8903518518518518,
+    "train_macro_f1": 0.8897660239232319,
+    "val_loss": 0.2736836162805557,
+    "val_accuracy": 0.899,
+    "val_macro_f1": 0.898815424122921,
+    "seconds": 18.372551918029785,
+    "learning_rate": 8.2171569475899e-07
+  }
+]

assignment-1/runs/transformer-seed1-20261020-125532/summary.json

@@ -0,0 +1,14 @@
+{
+  "run_id": "transformer-seed1-20261020-125532",
+  "model": "transformer",
+  "representation": "rows",
+  "seed": 1,
+  "parameters": 539018,
+  "epochs_trained": 30,
+  "best_epoch": 26,
+  "checkpoint_metric": "val_macro_f1",
+  "checkpoint_value": 0.9004399191772112,
+  "training_seconds": 549.8278179168701,
+  "commit": "42f775dbfcb780f7517ddc8ea8f33863b848da6d",
+  "hardware": "mps: Apple M4 Pro, 14 cores, 48 GB RAM"
+}