Tools.
Explore ML systems in real-time with live, interactive tools
Attention Visualizer
· LIVEInspect attention patterns for any model. Drag prompts in, scrub layers and heads.
All tools & demos.
-
Attention Visualizer
Inspect attention patterns layer-by-layer for any Hugging Face model. Click any head to see its causal mask, induction behavior, and sink tokens.
-
Throughput Calculator
Estimate tokens/sec for any model, precision, batch size, and GPU combination — memory-bound roofline only, no kernel-quality wishful thinking.
-
Training Memory Calculator
Per-GPU VRAM breakdown for training and inference — params, gradients, optimizer state, activations, KV cache, with ZeRO/FSDP sharding.
-
Eval Harness Playground
Run a small set of evaluations against any inference endpoint and get back a structured scorecard — quality, latency, cost, and refusal rate side by side.
-
Model Card Generator
Generate a structured model card from a checkpoint and an evaluation log — covers intended use, training data summary, evals, limitations, and ethics.
-
Tokenizer explorer
Paste any text and see how seven popular tokenizers split it — BPE, WordPiece, SentencePiece, tiktoken, and friends, side by side.
-
Attention pattern atlas
A gallery of attention head signatures from popular open models — induction heads, name movers, sink tokens, retrieval circuits — annotated and searchable.
-
Kernel Benchmark
Side-by-side timings for Triton, CUDA, and PyTorch implementations of the same kernel — attention, layernorm, GEMM, custom — across shapes and dtypes.
Canonical references.
- Tokenization
The Tokenizer Playground
Compare how GPT-4, LLaMA, Mistral, Qwen, Gemma, and others tokenize the same text — side by side, in the browser.
Xenova · Hugging Face - Tokenization
Tiktokenizer
OpenAI-focused tokenizer playground. Visualize cl100k, o200k, and legacy encodings with per-token highlights.
dqbd - Memory & VRAM
LLM Model VRAM Calculator
Widely-referenced inference VRAM estimator for popular open-source models with quantization and context-length sliders.
NyxKrage · Hugging Face - Memory & VRAM
APXML VRAM Calculator
Inference-focused VRAM calculator covering Nvidia GPUs and Apple Silicon. Good for picking hardware for a target model.
APXML - Architecture
LLM Visualization
A 3D, animated walk through the entire forward pass of GPT-2 nano, layer by layer. The clearest mental model of how a transformer works.
Brendan Bycroft - Training & Scaling
Chinchilla Scaling Calculator
Plug in a compute budget, get the compute-optimal model and data size per Hoffmann et al. 2022. Charts the iso-loss surface too.
Nathan Godey