Frank (d/acc) 🎩 💜 pfp

Frank (d/acc) 🎩 💜

@scalinglaw.eth

432 Following
253 Followers
Frank (d/acc) 🎩 💜 pfp
Frank (d/acc) 🎩 💜
@scalinglaw.eth
10 Moxie Passes available — mint yours to be eligible for upcoming airdrops, grants, Fan Tokens, and more! cc @betashop.eth @airstack
0 reply
0 recast
0 reaction
Frank (d/acc) 🎩 💜 pfp
Frank (d/acc) 🎩 💜
@scalinglaw.eth
CVPR paper: Instruct-Imagen: Image Generation with Multi-modal Instruction Innovations: - Multi-modal instruction for image generation: A new format that uses natural language to combine different modalities (text, edge, style, subject, etc.) to articulate complex generation intents in a uniform way. - Two-stage training approach for Instruct-Imagen: a) Retrieval-augmented training: Adapts a pre-trained text-to-image model to handle multi-modal inputs using retrieved similar (image, text) pairs. b) Multi-modal instruction-tuning: Fine-tunes the adapted model on diverse image generation tasks paired with multi-modal instructions. - Unified model architecture that can handle various image generation tasks - through multi-modal instructions, without task-specific designs. - Zero-shot generalization capability to unseen and more complex image generation tasks. - Adaptability to new tasks through fine-tuning on small datasets. source: https://arxiv.org/abs/2401.01952
0 reply
0 recast
1 reaction
Frank (d/acc) 🎩 💜 pfp
Frank (d/acc) 🎩 💜
@scalinglaw.eth
Prepare for my summary of next CVPR2024 paper
0 reply
0 recast
1 reaction
Frank (d/acc) 🎩 💜 pfp
Frank (d/acc) 🎩 💜
@scalinglaw.eth
CVPR 2024 Best Paper Award: Rich Human Feedback for Text-to-Image Generation - First dataset with detailed human feedback on generated images. - Rich Automatic Human Feedback (RAHF) Model (Multimodal Transformer model to predict rich human feedback) So, it enables that: - Finetuning generative models using predicted scores. - Region inpainting using predicted implausibility heatmaps. - Using aesthetic scores for classifier guidance in diffusion models. https://arxiv.org/abs/2312.10240
0 reply
0 recast
1 reaction
Frank (d/acc) 🎩 💜 pfp
Frank (d/acc) 🎩 💜
@scalinglaw.eth
I will post the gem papers I found at the CVPR 2024 conference daily here. They are either (1) innovative in new model architectures, (2) new ways to facilitate model development, or (3) fun applications with full potential.
0 reply
0 recast
0 reaction
Frank (d/acc) 🎩 💜 pfp
Frank (d/acc) 🎩 💜
@scalinglaw.eth
Minted an NFT for contributing to @jessepollak's baldness. Onchain Summer Buildathon is based. Mint yours by joining the Onchain Summer Buildathon. https://letsgetjessebald.com/token/854?ref_code=63421e6a18
0 reply
0 recast
0 reaction
Frank (d/acc) 🎩 💜 pfp
Frank (d/acc) 🎩 💜
@scalinglaw.eth
what the most innovative features on recaster?
1 reply
0 recast
0 reaction
Frank (d/acc) 🎩 💜 pfp
Frank (d/acc) 🎩 💜
@scalinglaw.eth
https://github.com/e-p-armstrong/augmentoolkit/tree/master this tool make my life much easier to generate synthetic data!
0 reply
0 recast
0 reaction
Frank (d/acc) 🎩 💜 pfp
Frank (d/acc) 🎩 💜
@scalinglaw.eth
Join me on Zora!
0 reply
0 recast
0 reaction
Frank (d/acc) 🎩 💜 pfp
Frank (d/acc) 🎩 💜
@scalinglaw.eth
https://x.com/theroaringkitty/status/1789807772542067105?s=46&t=vwylJwbnwWDZdlCwZXX_Gw “back”
0 reply
0 recast
0 reaction
Frank (d/acc) 🎩 💜 pfp
Frank (d/acc) 🎩 💜
@scalinglaw.eth
Mint Fun times ahead !!!
0 reply
0 recast
0 reaction
Frank (d/acc) 🎩 💜 pfp
Frank (d/acc) 🎩 💜
@scalinglaw.eth
https://farcaster.manifold.xyz/frame/535535856
0 reply
0 recast
0 reaction
Frank (d/acc) 🎩 💜 pfp
Frank (d/acc) 🎩 💜
@scalinglaw.eth
When AMD GPU available from runpod, I will test the vllm inference performance. Looking forward it https://rocm.blogs.amd.com/artificial-intelligence/llm-inference-optimize/README.html
0 reply
0 recast
0 reaction
Frank (d/acc) 🎩 💜 pfp
Frank (d/acc) 🎩 💜
@scalinglaw.eth
gm! generated from https://imagine.heurist.xyz/models/BluePencilRealistic-blue_pen5805
2 replies
0 recast
2 reactions
Frank (d/acc) 🎩 💜 pfp
Frank (d/acc) 🎩 💜
@scalinglaw.eth
2 replies
1 recast
2 reactions
Frank (d/acc) 🎩 💜 pfp
Frank (d/acc) 🎩 💜
@scalinglaw.eth
how's the MLX LLM models serving performance, for llama3-8b-4bit, ~ 20 tokens/s? any concurrency scheduling mechanism like continuous batching and page-attention optimization?
0 reply
0 recast
0 reaction
Frank (d/acc) 🎩 💜 pfp
Frank (d/acc) 🎩 💜
@scalinglaw.eth
Glad to see more and more systematic evaluation pipelines on LLM evaluation! https://twitter.com/lmsysorg/status/1782179997622649330
0 reply
0 recast
3 reactions
Frank (d/acc) 🎩 💜 pfp
Frank (d/acc) 🎩 💜
@scalinglaw.eth
🔵
0 reply
0 recast
0 reaction
Frank (d/acc) 🎩 💜 pfp
Frank (d/acc) 🎩 💜
@scalinglaw.eth
gm gm! have a nice day to pump!
2 replies
0 recast
1 reaction
Frank (d/acc) 🎩 💜 pfp
Frank (d/acc) 🎩 💜
@scalinglaw.eth
Shill on /lp , how bullish we are!
1 reply
0 recast
1 reaction