PinnedRaymond ChangBuilding an End-to-End AI Data InfrastructureBuilding an End-to-End AI Data InfrastructureApr 11Apr 11
PinnedRaymond Chang“Feature engineering + Deep learning” model of hedge fundsNow the “feature engineering + deep learning” model of hedge funds is crushing the “multi-factor + linear combination” gameplay of…Jan 18, 2022Jan 18, 2022
PinnedRaymond ChangWhere is the soil for the commercialization of AI really?While on a conference call with clients, I quickly organized past experiences of collaborating with large companies (including PaaS/IaaS)…Mar 20Mar 20
PinnedRaymond ChangOptions on the blockchainOption is the only tool that can do long, short and neutral strategies; Buying and selling options means buying and selling volatility…Jan 8, 2022Jan 8, 2022
Raymond ChangMicrosoft and their new ZeroQuantMicrosoft and their new ZeroQuant(4+2) quantization strategy for large language models (LLMs)! This is a significant development in the…Apr 18Apr 18
Raymond ChangHow far can “Next Token Prediction” go?In recent years, Large Language Models (LLMs) have made significant breakthroughs in the field of natural language processing. One of the…Apr 15Apr 15
Raymond ChangNTP: Way to AIGC?In recent years, large language models (LLMs) have made remarkable breakthroughs in the field of natural language processing. One of the…Apr 15Apr 15
Raymond ChangMeta: 羞羞的鐵拳Meta剛剛宣布了下一代訓練和推理加速晶片 MTIAv2。 708TF/s Int8 / 353TF/s BF16 256MB SRAM,128GB 記憶體 90W TDP。每個節點 24 顆晶片,每個機架 3 個節點。 使用標準 PyTorch…Apr 15Apr 15