usakha/Prophetnet_multiNews_model

Hyperparameters

learning_rate=2e-5
per_device_train_batch_size=14
per_device_eval_batch_size=14
weight_decay=0.01
save_total_limit=3
num_train_epochs=3
predict_with_generate=True
fp16=True

Training Output

global_step=7710,
training_loss=2.8554159399445727,
metrics={'train_runtime': 21924.7566,
'train_samples_per_second': 4.923,
'train_steps_per_second': 0.352,
'total_flos': 2.3807388210639667e+17,
'train_loss': 2.8554159399445727,
'epoch': 3.0}

Training Results

Epoch	Training Loss	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Bleu	Gen Len
1	2.981200	2.831641	0.414500	0.147000	0.230700	0.230600	0.512800	140.734900
2	2.800900	2.789402	0.417300	0.148400	0.231800	0.231700	0.516000	141.158200
3	2.680300	2.780862	0.418300	0.148400	0.232200	0.232100	0.516800	140.872300