Did you know that Replicate has speech-to-text models for transcribing audio? They're actually super good, thanks to the open source community. Like not to brag but this model runs 9x faster and 10x cheaper than OpenAI's Whisper API. That's kinda crazy
@replicate
-
Replicate Hosts Raw Base Models: Llama 2 and Mistral Available
By
–
Yes actually we do host raw base models! We have Llama 2 13B: https://
replicate.com/meta/llama-2-1
3b
… And Llama 2 70B: https://
replicate.com/meta/llama-2-7
0b
… And Mistral 7B base too! https://
replicate.com/mistralai/mist
ral-7b-v0.1
… (thanks for asking) -
Replicate Partners with Vercel for Easy AI API Integration
By
–
We're partnering with @vercel to make it easy to run AI with an API. Happy hacking! https://
vercel.com/blog/ai-integr
ations
… -
Create AI App in 30 Seconds with npm
By
–
Create an AI app in 30 seconds.
— Replicate (@replicate) 8 février 2024
> npm create replicate pic.twitter.com/E1djxobYEOCreate an AI app in 30 seconds. > npm create replicate
-
Fine-tune SDXL Image Generator with Just 5 Images
By
–
We made it super easy to train your own image generator! You can finetune SDXL on any subject, scene, or style you want, with as few as 5 images. Run it on an API with Replicate as soon as it's finished training. Literally couldn't be easier
-
Replicate Sponsors Open Source Model Work by Camenduru
By
–
Incredible results with this. So glad we're able to sponsor @camenduru's open source work and get this model into people's hands! https://t.co/plNLJ7URMN pic.twitter.com/ezo337W6em
— Replicate (@replicate) 5 février 2024Incredible results with this. So glad we're able to sponsor @camenduru
's open source work and get this model into people's hands! -
AI Engineers Meetup in San Francisco on February 7th
By
–
We're having a meetup with @vercel on Feb 7th in SF! Come meet other AI Engineers and hear lightning talks from engineers at @replicate
, @vercel
, @elevenlabs
, and @brevdev
. Space is limited, RSVP here: -
Code Llama 70B Achieves 67.8 HumanEval Score
By
–
Code Llama 70B scores a 67.8 on HumanEval, making it one of the highest performing open models available today. Happy hacking!
-
Code Llama 70B Base: The Foundational Code Model
By
–
And Code Llama 70B base is the foundational code model: