Epidemiology & Technology

Blog Archive

Qwen-3 ASR MLX Runtime for Realtime Transcription on Mac

Realtime transcription is not only an ASR model problem. A usable transcription system needs audio capture, speech boundary detection, batching, decoding, latency monitoring, memory control, and a stable interface that an application can call repeatedly. On a Mac, it also needs to use Apple Silicon efficiently without making the application

LLM based Text extraction from PDFs

I received CVs today for a project job interview. I tried to make LLMs help me. all had to run locally for obvious reasons. Here are the results Vision Model Benchmark — Document Extraction Vision Model Benchmark — Document Extraction Task: Evaluate local LLMs for extracting structured data from scanned

Benchmarking MLX Models on Laptop using oMLX

Link: https://omlx.ai/my/92e9df628cd488f548c7885cb732afb0f7870d4d7f40ae6024f8d4f85724ca20 gemma-4-26B-A4B-it-QAT-MLX-4bit With TurboQuant Qwen3.6-35B-A3B-NSC-ACE-SABER-8bit-MTPLX-Optimized-Speed With Turboquant Failed for pp131072 - Prefill context too large for available memory (pre-chunk guard at 83968 tokens, kv_len=83968): predicted peak would exceed prefill safety cap 46.7GB (90% of effective ceiling 51.8GB) gemma-4-31b-it-8bit Note: Fans went wild after pp4096 LFM2-24B-A2B-MLX-4bit medgemma-27b-text-it-MLX-4bit Note: fans went

glm_env for Switching GLM / Anthropic Claude Opus-Sonnet models easily in Claude Code

binary code, binary, binary system, byte, bits, computer, digital, software, code, developer, software development, programming, binary code, binary, binary, binary, binary, computer, digital, digital, digital, digital, software, software, software, code, code, code, code, code, developer, programming, programming, programming
Add following to your ~/.bashrc or ~/.zshrc terminal profile file Close and reopen terminal or source the profile and set and unset the GLM env variables easily

Resoucres for working with OCT images

Meta / reviews of datasets Datasets Tools Codes