2024-09-23

Models

• Llama 3.2: Revolutionizing edge AI and vision with open, customizable models #vlm
• molmo.allenai.org/blog molmo multimodal models #vlm
• stepfun-ai/GOT-OCR2_0 · Hugging Face #ocr
• GitHub - ByungKwanLee/Phantom: [Under Review] Official PyTorch implementation code for realizing the technical part of Phantom of Latent representing equipped with enlarged hidden dimension to build super frontier vision language models. #vlm

Papers

• [2409.13523] EMMeTT: Efficient Multimodal Machine Translation Training
• [2409.14683] Reducing the Footprint of Multi-Vector Retrieval with Minimal Performance Impact via Token Pooling
• [2409.15278] PixWizard: Versatile Image-to-Image Visual Assistant with Open-Language Instructions

Code

• RWKV-LM/RWKV-v7 at main · BlinkDL/RWKV-LM · GitHub
• GitHub - willccbb/mlx_parallm: Fast parallel LLM inference for MLX #mlx #inference

Articles

• The Practitioner’s Guide to the Maximal Update Parameterization | EleutherAI Blog #scaling
• Understanding how LLM inference works with llama.cpp llama.cpp
• Techniques for KV Cache Optimization in Large Language Models #kvcache #llm #inference
• The basic idea behind FlashAttention #flash-attention #softmax
  • From Online Softmax to FlashAttention
• FlexAttention: The Flexibility of PyTorch with the Performance of FlashAttention | PyTorch
• Tune Llama3 405B on AMD MI300x (our journey) - Felafax Blog - Obsidian Publish #amd #jax
• Exploring Parallel Strategies with Jax | AstraBlog #distributed #jax
• Power of Diffusion Models | AstraBlog #diffusion
• GenAI Handbook

Videos

• Boris Hanin | Scaling Limits of Neural Networks - YouTube
• MLBBQ: Flash Atttention by Mike Doan - YouTube #flash-attention

Other

• [ ]