Serverless GPU computing with Modal (for custom models)
Using Modal Serverless GPU computing from Python for ML research. Custom models, custom code, high speeds.
Early check on Byte Latent Transformation (Dec 2024)
Byte Latent Transformation as a Deep Learning based alternative to Byte-Pair Encoding. Pros and Cons, code and measurements, model analysis. Traditional Machine Learning approaches behind the scenes.
Experience report: AI, LLM and data
How to build your own LLM environment: with LangChain, FAISS, Mistral and other free OpenSource components. All self-hosted, GPU and CPU.