sanchit-gandhi
/

wav2vec2-2-rnd-regularisation

Automatic Speech Recognition Transformers PyTorch TensorBoard

speech-encoder-decoder generated_from_trainer Inference Endpoints

Model card Files Files and versions Metrics Training metrics Community

wav2vec2-2-rnd-regularisation / create_model.py

sanchit-gandhi's picture

sanchit-gandhi HF staff

up

d7a48d8 about 2 years ago

raw history blame contribute delete

No virus

1.43 kB

	from transformers import SpeechEncoderDecoderModel, AutoFeatureExtractor, AutoTokenizer, Wav2Vec2Processor, BartConfig, BartForCausalLM
	import torch

	# checkpoints to leverage
	encoder_id = "facebook/wav2vec2-large-lv60"
	decoder_id = "facebook/bart-large"

	feature_extractor = AutoFeatureExtractor.from_pretrained(encoder_id)
	feature_extractor.save_pretrained("./")
	tokenizer = AutoTokenizer.from_pretrained(decoder_id)
	tokenizer.save_pretrained("./")

	decoder_config = BartConfig.from_pretrained(decoder_id, is_decoder=True)
	decoder = BartForCausalLM(decoder_config)
	decoder.save_pretrained("decoder") # save the decoder

	model = SpeechEncoderDecoderModel.from_encoder_decoder_pretrained(encoder_id, "decoder", encoder_add_adapter=True)
	model.config.decoder_start_token_id = model.decoder.config.bos_token_id
	model.config.pad_token_id = model.decoder.config.pad_token_id
	model.config.eos_token_id = model.decoder.config.eos_token_id
	model.config.max_length = 50
	model.config.num_beams = 1
	model.config.encoder.layerdrop = 0.0
	model.config.use_cache = False
	model.config.decoder.use_cache = False
	model.config.processor_class = "Wav2Vec2Processor"

	# same enc regularisation as wav2vec2-2-bart
	model.config.encoder.feat_proj_dropout = 0.0
	model.config.encoder.final_dropout = 0.0
	model.config.encoder.mask_time_prob = 0.1

	# check if generation works
	out = model.generate(torch.ones((1, 2000)))

	model.save_pretrained("./")