Mise \u00e0 jour paquets :<\/p>\n
sudo apt update<\/span><\/code><\/li>\nsudo apt upgrade -y<\/span><\/code><\/li>\nsudo apt install -y python3-pip python3-dev build-essential<\/span><\/code><\/li>\nsudo apt install -y python3-venv<\/del>
\n<\/span><\/code><\/li>\npython3 -m venv whisper-env<\/span><\/code><\/del><\/li>\n<\/ul>\nInstallation Torch (no cache : sauf si\u00a0 m\u00e9moire > 4 GO)<\/p>\n
pip install torch --no-cache-dir<\/span><\/code><\/pre>\nInstallation FFMPEG<\/p>\n
sudo apt install -y ffmpeg<\/span><\/code><\/pre>\nCr\u00e9er l\u2019environnement virtuel \/\/\/ \/ Ne fonctionne pas avec moins de 4Go de memoire vive (3.8G \u00e0 disposition !!!) pour l’installation de Torch (l’installation de Torch hors Env Python ne fonctionne pas)<\/p>\n
source<\/span> whisper-env\/bin\/activate<\/span><\/code><\/del><\/p>\nCheck :<\/p>\n
python3 --version\r\npip3 --version<\/span><\/code><\/code> Installation Wisper :<\/pre>\n\n- \n
pip install git+https:\/\/github.com\/openai\/whisper.git<\/span><\/code><\/code><\/pre>\n<\/li>\n<\/ul>\nUtilisation de Whisper => Taille fichier audio 25Mo MAX<\/h4>\n
Sinon … perte de connexion.<\/p>\n
Heureusement on peut compresser (44Mo de MP3 deviennent 9Mo de Ogg ) … mais ogg entraine : connexion termin\u00e9e\u00a0 \ud83d\ude41<\/p>\n
ffmpeg -i audio.mp3 -vn -map_metadata -1 -ac 1 -c:a libopus -b:a 12k -application voip audio.ogg<\/code><\/p>\nLes commandes :<\/p>\n
cd \/home\/YSALMON\/audio\/<\/code><\/p>\nwhisper --model small<\/span> \"audio.mp3\" --language French --fp16 False --output_format srt \r\nwhisper --model small \"audio.mp3\" --language French --fp16 False --output_format srt --task transcribe\r\nwhisper --model small \"\/home\/YSALMON\/audio\/audio.mp3\" --language French --fp16 False --output_format srt --task transcribe\r\nwhisper --model small \"\/home\/YSALMON\/audio\/audio.ogg\" --language French --fp16 False --output_format srt --task transcribe<\/strong><\/code><\/pre>\n\n
\n# model = whisper.load_model(\"tiny.en\")
\n# model = whisper.load_model(\"base.en\")
\n# model = whisper.load_model(\"small\") # load the small model => OK
\n# model = whisper.load_model(\"medium.en\")
\n# model = whisper.load_model(\"large\") # Plantage !!!!!!!!!!<\/code><\/p>\n# model = whisper.load_model(\"medium) # Plantage !!!!!!!!!<\/code><\/p>\n<\/div>\n <\/p>\n
Fichiers de sortie dans le r\u00e9pertoire courant aux\u00a0 formats .json .tsv .txt .vtt .srt (sous-titres)<\/p>\n
Utilisable sous VLC =>Charger le fichier audio<\/p>\n
\n- menu sous-titres => Ajouter fichier sous-titre (.srt .txt .vtt)<\/li>\n
- Audio => Visualisation => Spectrom\u00e8tre (sinon marche pas !)<\/li>\n<\/ul>\n
Observations :<\/h4>\n\n- Plantage => D\u00e9connexion => d\u00e9passement m\u00e9moire (mod\u00e8les trop gros en m\u00e9moire\u00a0 : seul le small passe)<\/li>\n
- 1 minute audio = 4 minutes de temps de transcription => sur :\n
\n- Linux 5.10.0-34-amd64 on x86_64<\/span><\/li>\n
- Intel(R) Xeon(R) CPU E5-2680 v2 @ 2.80GHz, 4 cores<\/span><\/li>\n
- M\u00e9moire 3.81GiB<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n
R\u00e9f\u00e9rences :<\/h4>\n\n- https:\/\/labex.io\/fr\/tutorials\/linux-how-to-install-whisper-cli-on-linux-437909<\/a><\/li>\n
- https:\/\/www.css.cnrs.fr\/fr\/whisper-pour-retranscrire-des-entretiens\/<\/a><\/li>\n
- https:\/\/whisper-api.com\/docs\/transcription-options\/<\/a><\/li>\n
- <\/li>\n<\/ul>\n
Les commandes :<\/h4>\n
root@vxxxx:\/home\/YSALMON\/audio# whisper –help
\nusage: whisper<\/p>\n
\n- [-h]<\/li>\n
- [–model MODEL]<\/li>\n
- [–model_dir MODEL_DIR]<\/li>\n
- [–device DEVICE]<\/li>\n
- [–output_dir OUTPUT_DIR]<\/li>\n
- [–output_format {txt,vtt,srt,tsv,json,all}]<\/li>\n
- [–verbose VERBOSE]<\/li>\n
- [–task {transcribe,translate}]<\/li>\n
- [–language {af,am,ar,as,az,ba,be,bg,bn,bo,br,bs,ca,cs,cy,da,de,el,en,es,et,eu,fa,fi,fo,fr,gl,gu,ha,haw,he,hi,hr,ht,hu,hy,id,is,it,ja,jw,ka,kk,km,kn,ko,la,lb,ln,lo,lt,lv,mg,mi,mk,ml,mn,mr,ms,mt,my,ne,nl,nn,no,oc,pa,pl,ps,pt,ro,ru,sa,sd,si,sk,sl,sn,so,sq,sr,su,sv,sw,ta,te,tg,th,tk,tl,tr,tt,uk,ur,uz,vi,yi,yo,yue,zh,Afrikaans,Albanian,Amharic,Arabic,Armenian,Assamese,Azerbaijani,Bashkir,Basque,Belarusian,Bengali,Bosnian,Breton,Bulgarian,Burmese,Cantonese,Castilian,Catalan,Chinese,Croatian,Czech,Danish,Dutch,English,Estonian,Faroese,Finnish,Flemish,French,Galician,Georgian,German,Greek,Gujarati,Haitian,Haitian Creole,Hausa,Hawaiian,Hebrew,Hindi,Hungarian,Icelandic,Indonesian,Italian,Japanese,Javanese,Kannada,Kazakh,Khmer,Korean,Lao,Latin,Latvian,Letzeburgesch,Lingala,Lithuanian,Luxembourgish,Macedonian,Malagasy,Malay,Malayalam,Maltese,Mandarin,Maori,Marathi,Moldavian,Moldovan,Mongolian,Myanmar,Nepali,Norwegian,Nynorsk,Occitan,Panjabi,Pashto,Persian,Polish,Portuguese,Punjabi,Pushto,Romanian,Russian,Sanskrit,Serbian,Shona,Sindhi,Sinhala,Sinhalese,Slovak,Slovenian,Somali,Spanish,Sundanese,Swahili,Swedish,Tagalog,Tajik,Tamil,Tatar,Telugu,Thai,Tibetan,Turkish,Turkmen,Ukrainian,Urdu,Uzbek,Valencian,Vietnamese,Welsh,Yiddish,Yoruba}]
\n[–temperature TEMPERATURE]<\/li>\n - [–best_of BEST_OF]<\/li>\n
- [–beam_size BEAM_SIZE]\u00a0 5 par defaut = nombre d’hypoth\u00e9ses faites => l’augmenter am\u00e9liore la qualit\u00e9 et la m\u00e9moire utilsi\u00e9e + temps de calcul.<\/li>\n
- [–patience PATIENCE] : 1.0 par ddefaut : temps accord\u00e9 pour trouver une meilleure hypoth\u00e9se.
\n[–length_penalty LENGTH_PENALTY] : 1.0 par defaut => augmenter = allonger la taille de la sortie.<\/li>\n - [–suppress_tokens SUPPRESS_TOKENS]
\n[–initial_prompt INITIAL_PROMPT] : le context : ex: \u00ab\u00a0initial_prompt=Meeting transcript:<\/code><\/li>\n- [–carry_initial_prompt CARRY_INITIAL_PROMPT]<\/li>\n
- [–condition_on_previous_text CONDITION_ON_PREVIOUS_TEXT]<\/li>\n
- [–fp16 FP16]<\/li>\n
- [–temperature_increment_on_fallback TEMPERATURE_INCREMENT_ON_FALLBACK]<\/li>\n
- [–compression_ratio_threshold COMPRESSION_RATIO_THRESHOLD]<\/li>\n
- [–logprob_threshold LOGPROB_THRESHOLD]<\/li>\n
- [–no_speech_threshold NO_SPEECH_THRESHOLD]<\/li>\n
- [–word_timestamps WORD_TIMESTAMPS]<\/li>\n
- [–prepend_punctuations PREPEND_PUNCTUATIONS]<\/li>\n
- [–append_punctuations APPEND_PUNCTUATIONS]<\/li>\n
- [–highlight_words HIGHLIGHT_WORDS]<\/li>\n
- [–max_line_width MAX_LINE_WIDTH]<\/li>\n
- [–max_line_count MAX_LINE_COUNT]<\/li>\n
- [–max_words_per_line MAX_WORDS_PER_LINE]<\/li>\n
- [–threads THREADS]<\/li>\n
- [–clip_timestamps CLIP_TIMESTAMPS]<\/li>\n
- [–hallucination_silence_threshold HALLUCINATION_SILENCE_THRESHOLD]<\/li>\n
- audio<\/li>\n<\/ul>\n
positional arguments:
\naudio audio file(s) to transcribe<\/p>\n
optional arguments:
\n-h, –help show this help message and exit
\n–model MODEL name of the Whisper model to use (default: turbo)
\n–model_dir MODEL_DIR
\nthe path to save model files; uses ~\/.cache\/whisper by default (default: None)
\n–device DEVICE device to use for PyTorch inference (default: cpu)
\n–output_dir OUTPUT_DIR, -o OUTPUT_DIR
\ndirectory to save the outputs (default: .)
\n–output_format {txt,vtt,srt,tsv,json,all}, -f {txt,vtt,srt,tsv,json,all}
\nformat of the output file; if not specified, all available formats will be produced
\n(default: all)
\n–verbose VERBOSE whether to print out the progress and debug messages (default: True)
\n–task {transcribe,translate}
\nwhether to perform X->X speech recognition (‘transcribe’) or X->English translation
\n(‘translate’) (default: transcribe)
\n–language {af,am,ar,as,az,ba,be,bg,bn,bo,br,bs,ca,cs,cy,da,de,el,en,es,et,eu,fa,fi,fo,fr,gl,gu,ha,haw,he,hi,hr,ht,hu,hy,id,is,it,ja,jw,ka,kk,km,kn,ko,la,lb,ln,lo,lt,lv,mg,mi,mk,ml,mn,mr,ms,mt,my,ne,nl,nn,no,oc,pa,pl,ps,pt,ro,ru,sa,sd,si,sk,sl,sn,so,sq,sr,su,sv,sw,ta,te,tg,th,tk,tl,tr,tt,uk,ur,uz,vi,yi,yo,yue,zh,Afrikaans,Albanian,Amharic,Arabic,Armenian,Assamese,Azerbaijani,Bashkir,Basque,Belarusian,Bengali,Bosnian,Breton,Bulgarian,Burmese,Cantonese,Castilian,Catalan,Chinese,Croatian,Czech,Danish,Dutch,English,Estonian,Faroese,Finnish,Flemish,French,Galician,Georgian,German,Greek,Gujarati,Haitian,Haitian Creole,Hausa,Hawaiian,Hebrew,Hindi,Hungarian,Icelandic,Indonesian,Italian,Japanese,Javanese,Kannada,Kazakh,Khmer,Korean,Lao,Latin,Latvian,Letzeburgesch,Lingala,Lithuanian,Luxembourgish,Macedonian,Malagasy,Malay,Malayalam,Maltese,Mandarin,Maori,Marathi,Moldavian,Moldovan,Mongolian,Myanmar,Nepali,Norwegian,Nynorsk,Occitan,Panjabi,Pashto,Persian,Polish,Portuguese,Punjabi,Pushto,Romanian,Russian,Sanskrit,Serbian,Shona,Sindhi,Sinhala,Sinhalese,Slovak,Slovenian,Somali,Spanish,Sundanese,Swahili,Swedish,Tagalog,Tajik,Tamil,Tatar,Telugu,Thai,Tibetan,Turkish,Turkmen,Ukrainian,Urdu,Uzbek,Valencian,Vietnamese,Welsh,Yiddish,Yoruba}
\nlanguage spoken in the audio, specify None to perform language detection (default: None)
\n–temperature TEMPERATURE
\ntemperature to use for sampling (default: 0)
\n–best_of BEST_OF number of candidates when sampling with non-zero temperature (default: 5)
\n–beam_size BEAM_SIZE
\nnumber of beams in beam search, only applicable when temperature is zero (default: 5)
\n–patience PATIENCE optional patience value to use in beam decoding, as in https:\/\/arxiv.org\/abs\/2204.05424,
\nthe default (1.0) is equivalent to conventional beam search (default: None)
\n–length_penalty LENGTH_PENALTY
\noptional token length penalty coefficient (alpha) as in
\nhttps:\/\/arxiv.org\/abs\/1609.08144, uses simple length normalization by default (default:
\nNone)
\n–suppress_tokens SUPPRESS_TOKENS
\ncomma-separated list of token ids to suppress during sampling; ‘-1’ will suppress most
\nspecial characters except common punctuations (default: -1)
\n–initial_prompt INITIAL_PROMPT
\noptional text to provide as a prompt for the first window. (default: None)
\n–carry_initial_prompt CARRY_INITIAL_PROMPT
\nif True, prepend initial_prompt to every internal decode() call. May reduce the
\neffectiveness of condition_on_previous_text (default: False)
\n–condition_on_previous_text CONDITION_ON_PREVIOUS_TEXT
\nif True, provide the previous output of the model as a prompt for the next window;
\ndisabling may make the text inconsistent across windows, but the model becomes less
\nprone to getting stuck in a failure loop (default: True)
\n–fp16 FP16 whether to perform inference in fp16; True by default (default: True)
\n–temperature_increment_on_fallback TEMPERATURE_INCREMENT_ON_FALLBACK
\ntemperature to increase when falling back when the decoding fails to meet either of the
\nthresholds below (default: 0.2)
\n–compression_ratio_threshold COMPRESSION_RATIO_THRESHOLD
\nif the gzip compression ratio is higher than this value, treat the decoding as failed
\n(default: 2.4)
\n–logprob_threshold LOGPROB_THRESHOLD
\nif the average log probability is lower than this value, treat the decoding as failed
\n(default: -1.0)
\n–no_speech_threshold NO_SPEECH_THRESHOLD
\nif the probability of the <|nospeech|> token is higher than this value AND the decoding
\nhas failed due to `logprob_threshold`, consider the segment as silence (default: 0.6)
\n–word_timestamps WORD_TIMESTAMPS
\n(experimental) extract word-level timestamps and refine the results based on them
\n(default: False)
\n–prepend_punctuations PREPEND_PUNCTUATIONS
\nif word_timestamps is True, merge these punctuation symbols with the next word (default:
\n\u00ab\u00a0‘\u201c\u00bf([{-)
\n–append_punctuations APPEND_PUNCTUATIONS
\nif word_timestamps is True, merge these punctuation symbols with the previous word
\n(default: \u00ab\u00a0‘.\u3002,\uff0c!\uff01?\uff1f:\uff1a\u201d)]}\u3001)
\n–highlight_words HIGHLIGHT_WORDS
\n(requires –word_timestamps True) underline each word as it is spoken in srt and vtt
\n(default: False)
\n–max_line_width MAX_LINE_WIDTH
\n(requires –word_timestamps True) the maximum number of characters in a line before
\nbreaking the line (default: None)
\n–max_line_count MAX_LINE_COUNT
\n(requires –word_timestamps True) the maximum number of lines in a segment (default:
\nNone)
\n–max_words_per_line MAX_WORDS_PER_LINE
\n(requires –word_timestamps True, no effect with –max_line_width) the maximum number of
\nwords in a segment (default: None)
\n–threads THREADS number of threads used by torch for CPU inference; supercedes
\nMKL_NUM_THREADS\/OMP_NUM_THREADS (default: 0)
\n–clip_timestamps CLIP_TIMESTAMPS
\ncomma-separated list start,end,start,end,… timestamps (in seconds) of clips to
\nprocess, where the last end timestamp defaults to the end of the file (default: 0)
\n–hallucination_silence_threshold HALLUCINATION_SILENCE_THRESHOLD
\n(requires –word_timestamps True) skip silent periods longer than this threshold (in
\nseconds) when a possible hallucination is detected (default: None)<\/p>\n
<\/p>\n","protected":false},"excerpt":{"rendered":"
Installation : Mise \u00e0 jour paquets : sudo apt update sudo apt upgrade -y sudo apt install -y python3-pip python3-dev build-essential sudo […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"class_list":["post-5446","post","type-post","status-publish","format-standard","hentry","category-non-classe"],"_links":{"self":[{"href":"http:\/\/www.blue-bears.com\/blog\/index.php?rest_route=\/wp\/v2\/posts\/5446","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/www.blue-bears.com\/blog\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/www.blue-bears.com\/blog\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/www.blue-bears.com\/blog\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/www.blue-bears.com\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=5446"}],"version-history":[{"count":11,"href":"http:\/\/www.blue-bears.com\/blog\/index.php?rest_route=\/wp\/v2\/posts\/5446\/revisions"}],"predecessor-version":[{"id":5494,"href":"http:\/\/www.blue-bears.com\/blog\/index.php?rest_route=\/wp\/v2\/posts\/5446\/revisions\/5494"}],"wp:attachment":[{"href":"http:\/\/www.blue-bears.com\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=5446"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/www.blue-bears.com\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=5446"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/www.blue-bears.com\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=5446"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}