Developer Tools●Mid
Black Forest Labs CLI – let coding agents paint
API wrapper with agent-friendly flags—but FLUX, Replicate, and Fal already have CLIs.
Ship It
mackenzie_bowes
213mo ago
Back to browse
GitHub Repository
GPU ML architecture exploration tool
3 starsGo
Mixlab, an ML arch lab in Go. JSON config, Metal and CUDA, 1.6s builds
by mrothroc·Apr 22, 2026·2 points·1 comment
AI Analysis
●●●BangerBig BrainNiche GemShip It
JSON-defined ML architectures in Go beat Python boilerplate for rapid iteration.
Strengths
- •JSON configs eliminate Python boilerplate for architecture search experiments.
- •Metal and CUDA parity requires zero code changes between platforms.
- •Go compilation speed enables faster CLI iteration than Python scripts.
Weaknesses
- •Niche audience limits adoption to Go-loving ML researchers specifically.
- •Depends entirely on MLX ecosystem maturity and backend support.
Category
Target Audience
ML researchers and engineers prototyping architectures
Similar To
MLX · PyTorch Lightning · Hydra
Post Description
I built a tool for quickly testing different ML architectures. Define a model in JSON, train on your Mac (Metal) or ship the same config to a cloud GPU (CUDA). No code changes between platforms.
Why: I wanted to compare attention vs Mamba vs GQA at different parameter budgets without writing PyTorch for each experiment. Edit a JSON config, hit enter, see loss numbers. It will race different configs for you. The number one goal is iteration speed.
JSON config lets you chain together common ML blocks (attention, GQA, mamba, RetNet, and several more) and optimizers (muon, adamw) and compiles them to MLX IR, which can either run on Metal or CUDA backends.
Why Go: 1.6s builds, built-in profiling (mixlab -cpuprofile gives you a flame graph), import-based extensibility for custom blocks. No C++ extensions, no custom build systems. And personally I prefer strongly-typed, compiled languages.
On a Shakespeare benchmark matching nanoGPT (6L, 6H, d=384, 10.8M params): val loss 1.5527 on M1 Max, 1.5588 on A40. PyTorch numerical parity confirmed to 8 decimal places.
brew install mrothroc/tap/mixlab
Similar Projects
Design●●Solid
Pu-erh Lab, a CUDA-accelerated RAW photo editor
CUDA pipeline hits 60 FPS on 45MP RAW files, competing with Darktable.
WizardryBold Bet
yurunzi
113mo ago
Developer Tools●Mid
ChronosBench – Terminal CPU/Metal GPU stress tester with live load bars
Metal GPU stress testing in terminal, but is the workload realistic for benchmarking?
Niche GemShip It
nish068
103mo ago
Developer Tools●●Solid
JSON-up – Like database migrations, but for JSON
Type-safe JSON migrations inspired by database patterns—clean API but crowded space.
Solve My ProblemShip It
mrspence
103mo ago
Developer Tools●●Solid
Ucpify – JSON config to UCP-compliant commerce server
Drop a merchant-config.json into the CLI and you get 13 ready-made UCP endpoints — catalog, checkout lifecycle, orders, plus OAuth metadata — with Stripe/PayPal and SQLite wired up. The built-in OAuth identity-linking for agent workflows is the most interesting move; it's a focused, practical tool for prototyping UCP-compliant stores but not a substitute when you need bespoke business logic or complex fulfillment.
Solve My ProblemNiche Gem
init0
203mo ago
Developer Tools●●●Banger
jsoon, a streaming JSON parser and query engine in C
GPU-accelerated JSON parsing with 91ns selective queries versus simdjson's 24ms.
WizardryBig Brain
enduku
302mo ago