Codelab issue Fine-Tune Whisper For Multilingual ASR with 🤗 Transformers about blog HOT 2 CLOSED

Found solution to enforce single language as below in the model config:

MODEL_NAME = "openai/whisper-medium"
LANGUAGE = "Vietnamese"
TASK = "transcribe"
MAX_LENGTH = 224

model.config.forced_decoder_ids = processor.tokenizer.get_decoder_prompt_ids(
language=LANGUAGE, task=TASK
)
model.config.suppress_tokens = []
model.generation_config.forced_decoder_ids = processor.tokenizer.get_decoder_prompt_ids(
language=LANGUAGE, task=TASK
)
model.generation_config.suppress_tokens = []

from blog.

Hi, I tried implementing your solution but I'm still stuck. I've set my transformer to have logging log verbosely, and this is what I see.

It just stops and shows nothing when running evaluation. Any idea on why this is?

from blog.