T5-base fine-tuned on CNN/DM Summarization dataset.

Training args:

{
            "learning_rate": 0.0001,
            "logging_steps": 5000,
            "lr_scheduler_type": "cosine",
            "num_train_epochs": 2,
            "per_device_train_batch_size": 16, # total batch size of 48
            "save_total_limit": 1,
            "weight_decay": 0.1
}

Generation kwargs:

{
            "do_sample": true,
            "max_new_tokens": 100,
            "min_length": 50,
            "temperature": 0.7,
            "top_k": 0
 },

Pre-processing: Append prompt with prefix "Summarize: " Post-processing: None

Test split metrics:

{"lexical/meteor": 0.30857827917561603, 
"lexical/rouge_rouge1": 0.41099971702474514, 
"lexical/rouge_rouge2": 0.17676173608661166, 
"lexical/rouge_rougeL": 0.2759112075051335, 
"lexical/rouge_rougeLsum": 0.34316108028094616, 
"lexical/bleu": 0.10747816852428271, 
"semantic/bert_score": 0.8760301497472277}