create_model.py

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

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

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

model = SpeechEncoderDecoderModel.from_encoder_decoder_pretrained(encoder_id, decoder_id, encoder_add_adapter=False)
model.config.encoder.feat_proj_dropout = 0.0
model.config.encoder.final_dropout = 0.0
model.config.encoder.mask_time_prob = 0.1
model.config.decoder_start_token_id = tokenizer.cls_token_id
model.config.pad_token_id = tokenizer.pad_token_id
model.config.eos_token_id = tokenizer.sep_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"

# freeze entire encoder
for param in model.encoder.parameters():
    param.requires_grad = False

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

model.save_pretrained("./")