CoolFace
Modelpublic

usakha/Prophetnet_GovReport_model

sourceHugging Faceupdated 3y agoView on Hugging Face
0likes23downloads
Model Card

Hyperparameters

learningrate=2e-5 perdevicetrainbatchsize=14 perdeviceevalbatchsize=14 weightdecay=0.01 savetotallimit=3 numtrainepochs=3 predictwithgenerate=True fp16=True

Training Output

globalstep=3003, trainingloss=2.5178213735600132, metrics={'trainruntime': 8703.174, 'trainsamplespersecond': 4.83, 'trainstepspersecond': 0.345, 'totalflos': 9.272950245870797e+16, 'train_loss': 2.5178213735600132, 'epoch': 3.0}

Training Results

EpochTraining LossValidation LossRouge1Rouge2RougelRougelsumBleuGen Len
12.6611002.4691110.4513000.1852000.2790000.2789000.553300141.720300
22.4341002.4036470.4569000.1928000.2845000.2845000.556800141.763100
32.3137002.3939320.4595000.1944000.2863000.2862000.559200141.571600