AI/ML●●Solid
WebGPU LLM inference comprehensive benchmark
Sequential-dispatch methodology corrects 20x overestimation in prior WebGPU benchmarks.
Big BrainNiche Gem
yu3zhou4
222mo ago
Back to browse
GitHub Repository
A suite to benchmark CPU/GPU Python performance in training ML models and running local LLMs
19 starsPython
AI/ML benchmark for local LLM inference and XGBoost training on GPU/CPU
by albedan·May 16, 2026·2 points·0 comments
AI Analysis
●●SolidSolve My ProblemNiche Gem
One-command benchmark suite comparing Ollama and XGBoost performance with a shared Streamlit dashboard.
Strengths
- •Combines LLM token throughput and XGBoost training metrics in a single unified runner.
- •Encrypted result upload creates a crowdsourced hardware performance database automatically.
- •Uses uv for fast, reproducible environment setup without manual dependency hell.
Weaknesses
- •Hardware benchmarking is a crowded space with established tools like Phoronix.
- •Reference data relies on user submissions which may lack statistical rigor or verification.
Category
Target Audience
Data scientists and ML engineers comparing local hardware for model training
Similar To
Phoronix Test Suite · MLPerf · UserBenchmark
Similar Projects
AI/ML●●●Banger
Llama CPU Benchmarks
Proves speculative decoding slows down 4B models on 4-core CPUs despite marketing claims.
Big BrainDark Horse
muthuishere
2029d ago
Infrastructure●●●Banger
Physics-based simulator for distributed LLM training and inference
Estimates LLM training MFU, memory, timeline across 70 models and parallelism strategies—genuinely useful before GPUs commit.
WizardrySolve My ProblemBig Brain
zhebrak
113mo ago
AI/ML●●Solid
Doppler.js – WebGPU inference, faster/simpler than transformer.js
Explicit kernel control over TVM-style black boxes, but benchmarks show mixed wins vs Transformers.js.
Big BrainWizardry
clocksmith
303mo ago
AI/ML○Pass
Progressive Cognitive Architecture – Training LLMs in 4 Phases
Blog post masquerading as a product; no code, no reproducible implementation.
dexmac221
103mo ago
Infrastructure●●Solid
Go LLM inference with a Vulkan GPU back end that beats Ollama's CUDA
28% faster Vulkan-to-CUDA on Qwen, but llm.c and llama.cpp already own inference.
WizardryBig BrainNiche Gem
computerex
103mo ago