Abusing AI graphical & nlp tools to feed creative addiction and hyperactive curiosity. creator of https://teapawt.org/t.co/sFMboQ8QRJ & Imaginarium comic
Cryptoverse
Joined November 2021
- Tweets360
- Following241
- Followers167
- Likes2K
Pinned Tweet
WIP - Experimental AI LLM gradio UI chatbot for GGUF with a twist, enabling anyone to chain prompts piloted by a csv and more... github.com/dbddv01/GPT-Seque… @huggingface @Gradio @ggerganov @abetlen @TheBlokeAI @takyon236 @GrantSlatton
DB ddv01 retweeted
BioMistral is a new 7B foundation model for medical domains, based on Mistral and further trained PubMed Central.
- top open-source medical Large Language Model (LLM) in its weight class
- Apache License
- includes base models, fine tunes, and quantized versions.
ALT Screenshot of the model card on the Hub.
DB ddv01 retweeted
Does 2 bit quantization excite you?
Well it excites me. Imagine the knowledge of the world encoded in 00, 01, 10, 11 sequences. My own model, codeCherryPop, is perfectly coherent and reasons at 2bit as well.
"QuIP: 2-Bit Quantization of Large Language Models With Guarantees" is one such amazing paper discussing how 2 bit quantization performance is spectacular for models
DB ddv01 retweeted
This is huge: Llama-v2 is open source, with a license that authorizes commercial use!
This is going to change the landscape of the LLM market.
Llama-v2 is available on Microsoft Azure and will be available on AWS, Hugging Face and other providers
Pretrained and fine-tuned models are available with 7B, 13B and 70B parameters.
Llama-2 website: ai.meta.com/llama/
Llama-2 paper: ai.meta.com/research/publica…
A number of personalities from industry and academia have endorsed our open source approach: about.fb.com/news/2023/07/ll…
DB ddv01 retweeted
OpenLLaMA 13B Released
model: huggingface.co/openlm-resear…
present a permissively licensed open source reproduction of Meta AI's LLaMA large language model. We are releasing 3B, 7B and 13B models trained on 1T tokens. We provide PyTorch and JAX weights of pre-trained OpenLLaMA models, as well as evaluation results and comparison against the original LLaMA models.
DB ddv01 retweeted
Nous-Hermes-13b fp16 weights have been released.
Totallly uncensored, over 300,000 GPT-4 outputs, on base LLaMA.
We’re proud to compete with modern models with this release as we use it as our current basis for tool+agent-based finetuned models.
huggingface.co/NousResearch/…
DB ddv01 retweeted
Added arbitrary context free grammar constraints to llama.cpp
Can now plug in any llama.cpp compatible model and give an exact grammar spec: JSON, etc
Excited to use with more powerful local models as they are released
Thanks @ggerganov & friends for such a wonderful project.
#AIart is too serious ? Luckily, #FoolDiffusion comic is here. pls RT if you 😃. Drawed with #stablediffusion #comics #webcomics #comicstrip #digitalart #cartoon #aiartcommunity #bandedessinee #ComicArt #funnytwitter #twitterhumor #aiia #chatgpt
DB ddv01 retweeted
📢 Introducing MPT: a new family of open-source commercially usable LLMs from @MosaicML. Trained on 1T tokens of text+code, MPT models match and - in many ways - surpass LLaMa-7B. This release includes 4 models: MPT-Base, Instruct, Chat, & StoryWriter (🧵)
mosaicml.com/blog/mpt-7b
ALT Introducing MPT-7B, the latest entry in our MosaicML Foundation Series. MPT-7B is a transformer trained from scratch on 1T tokens of text and code.
DB ddv01 retweeted
The first RedPajama models are here! The 3B and 7B models are now available under Apache 2.0 license, including instruction-tuned and chat versions!
This project demonstrates the power of the open-source AI community with many contributors ... 🧵 together.xyz/blog/redpajama-…
Run local Vicuna-v1-7b model, at 16 token/sec output speed, on my win pc 6gb vram (gx1060). Model in 3bit and <3gb size. Thanks to github.com/mlc-ai/mlc-llm ! Quality of int3 still to be assessed but already worth to keep an eye on... easy install!
DB ddv01 retweeted
.@Replit just announced their own LLaMa style code LLM at their developer day!
replit-code-v1-3b
- 2.7b params
- 20 languages
- 525B tokens (“20x Chinchilla?”)
- beats all open source code models on HumanEval benchmark
- trained in 10 days with @NaveenGRao @MosaicML
DB ddv01 retweeted
EVERYONE NEEDS AN AGENT
Inspired by talk with @connerruhl, everyone of us and businesses need a bot which play a role of interactive assistant. Personal bot needs to be local running LLM with Lora that know your habits and self evoluted by interaction with u and other agents.
[1]
DB ddv01 retweeted
GeorgiaTechResearchInstitute/galpaca-30b
Please somebody 4int this!
huggingface.co/GeorgiaTechRe…
DB ddv01 retweeted
it may be useful to establish a "proof of humanity" word, which your trusted contacts can ask you for, in case they get a strange and urgent voice or video call from you
this can help assure them they are actually speaking with you, and not a deepfaked/deepcloned version of you
DB ddv01 retweeted
#RWKV is One Dev's Journey to Dethrone GPT Transformers. The largest RNN ever (up to 14B). Parallelizable. Faster inference & training. Supports INT8/4. No KV cache. 3 years of hard work. DEMO: huggingface.co/spaces/BlinkD… Computation sponsored by @StabilityAI @AiEleuther @EMostaque
DB ddv01 retweeted
Not sure how this was compiled, but I expect significantly better performance if it is built correctly
See for example how we compile whisper.cpp for android. Make sure your are compiling ggml.c with the correct -mfpu flag to enable SIMD
github.com/ggerganov/whisper…
