esahit's picture
First training run with the ul2-small-dutch model (no prompt-tuning, full model finetuning) on the increased dataset
00d5f01 verified
raw
history blame contribute delete
290 Bytes
{
"decoder_start_token_id": 0,
"diversity_penalty": 0.3,
"early_stopping": true,
"eos_token_id": 1,
"max_new_tokens": 20,
"no_repeat_ngram_size": 2,
"num_beam_groups": 5,
"num_beams": 5,
"num_return_sequences": 5,
"pad_token_id": 0,
"transformers_version": "4.44.2"
}