2024-09-30

Models

• SAM 2.1 with training code
  • x.com
• Liquid Foundation Models: Our First Series of Generative AI Models
• GitHub - THUDM/CogView3: text to image to generation: CogView3-Plus and CogView3(ECCV 2024)
• Accelerating Leaderboard-Topping ASR Models 10x with NVIDIA NeMo | NVIDIA Technical Blog
• vidore/colqwen2-v0.1 · Hugging Face
• Announcing FLUX1.1 [pro] and the BFL API - Black Forest Labs

Papers

• [2409.18869v1] Emu3: Next-Token Prediction is All You Need
• [2409.17692] MIO: A Foundation Model on Multimodal Tokens
• [2405.03882] Trio-ViT: Post-Training Quantization and Acceleration for Softmax-Free Efficient Vision Transformer
• [2409.16280] MonoFormer: One Transformer for Both Diffusion and Autoregression
• Site Unreachable
• [2409.20370] The Perfect Blend: Redefining RLHF with Mixture of Judges
• [2408.05088] UNIC: Universal Classification Models via Multi-teacher Distillation
• [2407.09111] Inference Optimization of Foundation Models on AI Accelerators
• [2402.10376] Interpreting CLIP with Sparse Linear Concept Embeddings (SpLiCE)
  • x.com
• [2410.01806v1] Samba: Synchronized Set-of-Sequences Modeling for Multiple Object Tracking
• [2410.01201] Were RNNs All We Needed? #rnn
• [2410.01679] VinePPO: Unlocking RL Potential For LLM Reasoning Through Refined Credit Assignment
• https://arxiv.org/abs/2409.20370
• Movie Gen: A Cast of Media Foundation Models
• [2410.02746] Contrastive Localized Language-Image Pre-Training

Code

• GitHub - THUDM/SwissArmyTransformer: SwissArmyTransformer is a flexible and powerful library to develop your own Transformer variants.
• x.com
  • Self contained example of how pipeline parallel works (AFAB and 1F1B) in 200 LOC · GitHub
• GitHub - evanatyourservice/kron_torch: An implementation of PSGD Kron second-order optimizer for PyTorch

Articles

• Recreating PyTorch from Scratch (with GPU Support and Automatic Differentiation) | by Lucas de Lima Nogueira | Towards Data Science
• Accelerating Leaderboard-Topping ASR Models 10x with NVIDIA NeMo | NVIDIA Technical Blog
• Cheng Luo - MINI-SEQUENCE TRANSFORMER (MST)
• Distributed Training Of Deep Learning Models : Part ~ 1 #distributed
• How to train a model on 10k H100 GPUs? #distributed
• Self contained example of how pipeline parallel works (AFAB and 1F1B) in 200 LOC · GitHub
• Deploy SkyPilot on existing machines — SkyPilot documentation #skypilot
• Transformers Inference Optimization Toolset | AstraBlog

Videos

• PyTorch Conference 2024 - YouTube
• 2024 RL Conference #rl
• Modern GPU Architecture

Other

• ECCV
  • ILR2024 instance level recognition
• litellm/model_prices_and_context_window.json at main · BerriAI/litellm · GitHub LLM specs
• Llm Pricing - a Hugging Face Space by philschmid