search

faseelmo / noc_latency_predictor

MIT License
1 stars 0 forks source link

Training Attempt #1 #22

Closed faseelmo closed 10 months ago

faseelmo commented 10 months ago

Train for 100 Epochs using MAE (L1) then use that weights to train on MSE (L2)

faseelmo commented 10 months ago

MAE and RMSE giving pretty much the same results.

faseelmo commented 10 months ago

Once MAE reaches its local minima try training using that weights and lowering the learning rate.

faseelmo commented 10 months ago

Model is overfitting if we train further than 200 Epoch