skills.rest
Carroll Vance avatar

Carroll Vance

Community

@csvance · Houston, Texas

60Followers
|
90Public Repos
|
2Published Skills

ML Engineer @ Medical Metrics Inc.

🐙 GitHub ↗
Skills Distribution
DomainAI Models & ...AI Models & Machin.. (55%)Developer Tooling .. (30%)UI/UX & Creative M.. (15%)

Agent Skills by Carroll Vance

Showing 2 vetted skills indexed across 1 GitHub repositories.

csvancecsvance
12

timm-to-lux

Converts PyTorch timm models into numerically-equivalent Lux.jl implementations with parity-tested weight loading.

Community
Advanced
csvancecsvance
12

tachikoma-tui

Scaffolds terminal UI applications in Julia using the Tachikoma.jl framework.

Community
Intermediate

Frequently Asked Questions About Carroll Vance

FAQPage Schema
What tasks can I accomplish with Carroll Vance's skills?▼

Two tasks: porting PyTorch timm models (ResNet, ViT, EfficientNet, ConvNeXt) to numerically-equivalent Lux.jl code with HDF5 parity fixtures and isapprox tests, and bootstrapping Tachikoma.jl terminal UIs with Model/update!/view structure, layouts, widgets, and keyboard/mouse event handling.

Who should use the timm-to-lux and tachikoma-tui skills?▼

ML engineers translating PyTorch vision models into Julia's Lux ecosystem who need exact numeric parity, and Julia developers building terminal user interfaces with Tachikoma.jl who want the minimal app skeleton plus guidance on common pitfalls that cost debugging round trips.

What numeric differences does timm-to-lux help resolve?▼

It addresses cross-correlation versus convolution, padding semantics, GroupNorm/BatchNorm defaults, weight standardization, and NCHW versus WHCN layout differences between PyTorch and Lux, plus loading .safetensors weights from HuggingFace Hub within a Julia context.

What are the prerequisites for using these skills?▼

A Julia environment with Lux.jl and Tachikoma.jl respectively. The timm-to-lux skill assumes familiarity with PyTorch timm models (timm.create_model, forward_features) and access to reference weights, while tachikoma-tui assumes basic Julia knowledge for building terminal applications.