News for hackers. Stuff that matters.

haghiri

Atelier AI (آتلیه هوش مصنوعی) is a project by Muhammadreza Haghiri, an Iranian programmer, indie hacker and entrepreneur.
New users will have 15 credits.
Re:A 400B parameter model which is completely free on Saturday August 17, 2024 @02:34PM (#41276786)
Hi there. This is my newest project. I've been working on this project with an amazing team and we're glad to announce JabirLLM 400B, which can be considered a direct competition to famous closed-source AI models such as Claude or GPT-4o. I will be appreciating your comments and opinions. For testing the project, you only need to go to the website and provide an email address, then obtain your API key and the you can try the model. Have fun prompting.
Re: on Sunday February 11, 2024 @02:44PM (#39337951)
Well, a year ago I made Mann-E, a SD 1.5 based model which got a lot of attention from Iranian media.

After a few months, we just trained our own diffusion model using HF recipes and resources I could find here and our model has similar results to SDXL, SDXL Turbo and those "good" SD 1.5 based models. For internationalization, we translated prompts and trained model over and over.

We also added Meta's NLLB model to our backend to help with languages like Tamil which we didn't have enough data.

The website is mann-e.com. We're not supporting currencies other than Iranian rials, but if you want a premium membership, you can ping me in the chat or ask me here. I'll help you out!

Also, you'll have 5 images for 3 days. Feel free to make new accounts each 3 days

If you like it, I can make it like ideogram, each day, a few credits will be added to your wallet!

Re:Nucleus is the proof that "Small is the new Big" on Saturday January 13, 2024 @05:04AM (#38978571)
Nucleus is a small language model based on Mistral (actually, the trimmed untrained version you can find here) and trained in different steps. First, we've pretrained it on TinyStories dataset, then TinyTextBooks to make it a more specific model. This model is just a proof of concept at this point, but showed good promises in early tests. So with proper training, can be a good product over time!
Re:Maral 7B Alpha, a SOTA LLM with Support of Persian Language on Wednesday December 27, 2023 @04:16PM (#38786844)
This is another problem we may face in future. I guess we really need to launch a community where people can provide data for the future models.
Re:Maral 7B Alpha, a SOTA LLM with Support of Persian Language on Wednesday December 27, 2023 @03:40PM (#38786472)
Now as of today, I am pleased to announce that Maral is here. The 7 billion parameters bilingual model which can respond to Persian and English prompts, and can produce GPT-3.5 level of answers based on the dataset we fed to it!
This is my effort to bring an open source equivalent of Midjourney to your computers. This is version 4.1.3 and hope you enjoy prompting and art generating with this brand new model.
Re:OpenJourney: Midjourney, but Open Source on Thursday January 26, 2023 @06:55AM (#34530599)
This was my project, but since @prompthero changed their "midjourney-v4 dreambooth" model's name to openjourney, I changed my model name to "Mann-E" which is accessible here: https://huggingface.co/mann-e/mann-e_4_rev-0-1 (It's only a checkpoint and under development)
Re: on Sunday September 04, 2022 @09:53AM (#32713066)
I recently discovered BigScience's BLOOM. This was amazing, and since copilot isn't free of charge anymore, I decided to use it to create my own AI pair programmer.