Yanbing
Yanbing
I use the wav2vec2_small.pt for fine-tuning, when I decode the fine-tuned model, I find different `--max_tokens `during inference will produce different wer results, and setting it to the same size...