After a long wait, Ikari and me finally made a new release of our last model on NeverSleep repo: Lumimaid-v0.2.
This model can be used in different size, from the small Llama-3.1-8B to the gigantic Mistral-Large-123B!
huggingface.co/collections/N…
Hope you will enjoy them!
Undi retweeted
🚨 Model Merging competition @NeurIPSConf!🚀
Can you revolutionize model selection and merging?Let's create the best LLMs!🧠✨
💻Come for science
💰Stay for $8K
💬Discord: discord.gg/dPBHEVnV
🔗Sign up: llm-merging.github.io/
Sponsors: @huggingface @SakanaAILabs @arcee_ai
Undi retweeted
Noromaid Mixtral 8x7B Instruct
An awesome new Mixtral fine-tune from IkariDev and @undi95, the creators of Remm Slerp and Noromaid, suitable for roleplaying: openrouter.ai/models/neversl…
Undi retweeted
Frankensteining models is a phenomenon that only exists thanks to a bunch of scrappy GPU poor open sorcerers to cut costs on crafting new larger and smaller models from existing pretrains and finetunes, rather than training from scratch. Truly a form of digital alchemy. @nisten @Undi95 @chargoddard
Undi retweeted
🚀 Other top users:
@Lvminzhang 800 followers 🖼️
@erhartford 800 followers 🐬
@Undi95 600 followers 🤖
@julien_c 600 followers 🤗
@multimodalart 500 followers 🎨
@radamar 500 followers🤗
Thank you all!
Undi retweeted
🎉 Introducing Followers Leaderboard!
🚀 Tracking the follower counts!
- @TheBlokeAI 12k followers 🤯
- @mervenoyann 2.2k followers 🤗
- @fffiloni 2.2k followers 🖼️
- @_akhaliq 1k followers 🤗
- @teknium 1k followers 🤖
hf.co/spaces/Weyaxi/follower…
Undi retweeted
Launching three new models tonight:
MythoMist 7B, from the maker of MythoMax: openrouter.ai/models/gryphe/…
Noromaid 20B: a collab by @Undi95 and IkariDev, getting rave reviews and strong when paired with specific instruct settings in ST: openrouter.ai/models/neversl…
Intel's Neural Chat 7B: a new big corp is entering the arena with a top-ranking 7b model: openrouter.ai/models/intel/n…
Undi retweeted
I've become a big fan of "frankenmerge" LLMs that combine weights of open-source models and (inexplicably) result in something better than the sum of the parts. Feels like a very underexplored area of ML.
Examples: Toppy by @Undi95 is now top of the fastest growing models on @openrouter, above GPT-4 Turbo, and MythoMax (red bar) has outpaced everyone. Goliath 120b is also new this weekend.
Here's a search to track them all: openrouter.ai/models?q=%23me…
Undi retweeted
🔍 Searching for @iclr_conf 2024 paper on Model Merging/Fusion & related topic:
Dive into the latest advancements in model merging, fusion, and weight interpolations from #ICLR2024 through the list below!
Let me know if I missed some relevant papers.
[Thread ⬇️]
Undi retweeted
Open Orca is a really good dataset
ive got a better one ive been sitting on for months constantly tooling and tweaking
problem is that its 2x the size of orca currently (and likely ends a little smaller as i continue to filter it out)
Anyone feel like pitching in compute in exchange for being credited and largely responsible for the next SOTA model?
We've not released a lackluster model yet, and we're aiming to really blow the doors off of this whole thing (again) next chance we get.
