transcription stops before end of audio file

[pantau000]

transcription of a 1h34 audio file stops after 1h13 minutes.

last message:

INFO: Finished transcription for Transcription of Caxias 09.mp3 in 32723 seconds
'HTTPSConnectionPool(host='huggingface.co', port=443): Max retries exceeded with url: /facebook/bart-large-mnli/resolve/main/config.json (Caused by ConnectTimeoutError(<urllib3.connection.HTTPSConnection object at 0x0000021AFA45DFF0>, 'Connection to huggingface.co timed out. (connect timeout=10)'))' thrown while requesting HEAD https://huggingface.co/facebook/bart-large-mnli/resolve/main/config.json

[octimot]

I think there must have been an error with accessing huggingface.co when the tool tried to download the bart-large-mnli model (maybe for question labeling?).

If you haven't tried this again already, I would give it another try. It should work!

Feel free to re-open this and continue the conversation if it doesn't.

Cheers

[pantau000]

just checked again, it continues to stop transcription before the end

[octimot]

Could you check if you can access https://huggingface.co/facebook/bart-large-mnli/resolve/main/config.json directly from your browser? If so, there might be an issue with some ssl certificates on your machine. Are you using the standalone or the git version of the tool?

[pantau000]

I would rather suggest that this has maybe to do with the metadata in the audio file?

Anyway, I checked, and the page opens without problem:

{
"_num_labels": 3,
"activation_dropout": 0.0,
"activation_function": "gelu",
"add_final_layer_norm": false,
"architectures": [
"BartForSequenceClassification"
],
"attention_dropout": 0.0,
"bos_token_id": 0,
"classif_dropout": 0.0,
"classifier_dropout": 0.0,
"d_model": 1024,
"decoder_attention_heads": 16,
"decoder_ffn_dim": 4096,
"decoder_layerdrop": 0.0,
"decoder_layers": 12,
"decoder_start_token_id": 2,
"dropout": 0.1,
"encoder_attention_heads": 16,
"encoder_ffn_dim": 4096,
"encoder_layerdrop": 0.0,
"encoder_layers": 12,
"eos_token_id": 2,
"forced_eos_token_id": 2,
"gradient_checkpointing": false,
"id2label": {
"0": "contradiction",
"1": "neutral",
"2": "entailment"
},
"init_std": 0.02,
"is_encoder_decoder": true,
"label2id": {
"contradiction": 0,
"entailment": 2,
"neutral": 1
},
"max_position_embeddings": 1024,
"model_type": "bart",
"normalize_before": false,
"num_hidden_layers": 12,
"output_past": false,
"pad_token_id": 1,
"scale_embedding": false,
"transformers_version": "4.7.0.dev0",
"use_cache": true,
"vocab_size": 50265
}

[octimot]

I would rather suggest that this has maybe to do with the metadata in the audio file?

Not according to the error though.

Are you using the standalone or the git version?

[pantau000]

git

[pantau000]

just checked again, this happens independently of the hugging face error (which i didn't get again), with different audio files. maybe a problem of mp3 files?

[octimot]

Did you find a workaround for this or is it still an issue?

Cheers!

[pantau000]

haven't tried again, will do so with the newest version

[pantau000]

just checked, wiht another audio file, unfortunately the problem continues. audio is 1:28:15 and transcription stops at 01:00:33,660.

[pantau000]

bump