CoolFace
Modelpublic

pcam-interpretability/resnet50-val09050-resnet-tuned

sourceHugging Faceupdated 1y agoView on Hugging Face
0likes
Model Card

resnet50

Best Validation Accuracy: 0.9050

Metadata

  • Model Name: resnet50
  • Optimizer: adamw
  • Scheduler: cosine
  • Weight Decay: 0.0005
  • Warmup Epochs: 3
  • Patience: 7
  • Amp: True
  • Seed: 42
  • Batch Size: 32
  • Initial Lr: 0.0005
  • Total Epochs Ran: 14
  • Early Stopped: True
  • Training Time Seconds: 33306.75762796402
  • Num Parameters: 22067009
  • Device: NVIDIA A100-SXM4-40GB
  • Run Id: resnet-tuned

Training Configuration

  • Epochs: 14
  • Batch size: 32
  • Learning rate (initial): 0.0005

Training Logs (Per Epoch)

EpochTrain LossTrain AccVal LossVal AccLR
10.26620.89210.29350.87860.000167
20.22450.91250.33810.87630.000333
30.20900.91960.43550.81740.000500
40.18410.93040.34260.87510.000500
50.16780.93760.37540.86110.000499
60.15540.94300.45090.84150.000495
70.14470.94680.29050.90500.000488
80.13730.95020.45050.84440.000478
90.12970.95390.47260.84730.000467
100.12240.95690.33710.88670.000452
110.11670.95830.37820.87410.000436
120.11150.96040.47380.86450.000417
130.10520.96260.34770.88960.000397
140.09980.96470.40670.88450.000375