Survivor Logs
Weird to find errors in telegram ๐
This thing is on tiktok too. I failed to take a screenshot but the toast notification is also showing an error message in the dynamic value of it like
Kind of text
"we will remind you in {ERR: number} minutes "
Kind of text
Survivor Logs
Weird to find errors in telegram ๐
The
Was just a temporary Telegram localization bug that affects only some devices.
The issue is the app sometimes fails to load the correct text string from its Telegram's translation file and displays the raw code name instead but you can easily fix it just do :
Open Telegram
> Go to Settings
> Tap on Language
> Switch to another language like French or sth
> close the app completely
> Reopen Telegram and switch back to English.
This will force the app to re-download its language files and fixes the text immediately.
you can also simply update the app or uninstall and reinstall it
LOC_ERR: Subscribers_other
Was just a temporary Telegram localization bug that affects only some devices.
The issue is the app sometimes fails to load the correct text string from its Telegram's translation file and displays the raw code name instead but you can easily fix it just do :
Open Telegram
> Go to Settings
> Tap on Language
> Switch to another language like French or sth
> close the app completely
> Reopen Telegram and switch back to English.
This will force the app to re-download its language files and fixes the text immediately.
you can also simply update the app or uninstall and reinstall it
แณแต Illuminati แแ แตแแฝแ แ แแฐแ แฅแแฝแ แแ แญแ แแฝแ
Forwarded from Chapi Dev Talks
Hohe ASR is out. Open Amharic speech to text.
Start with the speed, because it is the part people do not expect. On four ordinary CPU cores, a five second voice note is transcribed in 0.39 seconds. A thirty second clip takes 7.75 seconds. An hour of audio goes through in under ten minutes, on a machine that costs about nine cents an hour to rent. No GPU anywhere in that sentence.
It is a CTC model, one single pass over the audio instead of generating word by word, which makes it roughly five times faster than a Whisper of similar accuracy. It also means it does not invent sentences out of silence.
Accuracy: 16.1% word error and 5.4% character error on speakers it has never heard. 880 hours of training audio, including five regional dialects and a third of it pushed through a phone line on purpose.
What it is bad at, plainly: two people talking at once, around 47% word error. Numbers come out as words. No punctuation.
Try it now, nothing to install: @dataset_ai_bot
Weights, open under CC BY 4.0: huggingface.co/snapwre/hohe-asr-amharic
Send us what it gets wrong.
Start with the speed, because it is the part people do not expect. On four ordinary CPU cores, a five second voice note is transcribed in 0.39 seconds. A thirty second clip takes 7.75 seconds. An hour of audio goes through in under ten minutes, on a machine that costs about nine cents an hour to rent. No GPU anywhere in that sentence.
It is a CTC model, one single pass over the audio instead of generating word by word, which makes it roughly five times faster than a Whisper of similar accuracy. It also means it does not invent sentences out of silence.
Accuracy: 16.1% word error and 5.4% character error on speakers it has never heard. 880 hours of training audio, including five regional dialects and a third of it pushed through a phone line on purpose.
What it is bad at, plainly: two people talking at once, around 47% word error. Numbers come out as words. No punctuation.
Try it now, nothing to install: @dataset_ai_bot
Weights, open under CC BY 4.0: huggingface.co/snapwre/hohe-asr-amharic
Send us what it gets wrong.
huggingface.co
snapwre/hohe-asr-amharic ยท Hugging Face
Weโre on a journey to advance and democratize artificial intelligence through open source and open science.