SPB Git forge

spb/ai-atlas

Public
41commits 1branches 0releases
4.6 MBsize
maindefault branch
12 days agolast push
HTML 77.2% TypeScript 10.5% Python 9.6% JavaScript 2.5%
1.9 KB · 56 lines yaml
Raw Blame History
1# Hardware identities discovered on manufacturer spec pages by the `hardware_specs` connector (Apple Silicon generations published after2# the curated list, NVIDIA product variants). Only identity fields here — memory, bandwidth, TDP and compute figures are claimed by the3# connector from the cited spec page, never typed by hand. Fragment of registry/hardware.yaml.4hardware:5  - key: apple-m56    name: Apple M57    manufacturer: apple8    kind: soc9    runtimes: [mlx, llama.cpp, coreml]10    source_url: https://www.apple.com/macbook-air/specs/11  - key: apple-m5-pro12    name: Apple M5 Pro13    manufacturer: apple14    kind: soc15    runtimes: [mlx, llama.cpp, coreml]16    source_url: https://www.apple.com/macbook-pro/specs/17  - key: apple-m5-max18    name: Apple M5 Max19    manufacturer: apple20    kind: soc21    runtimes: [mlx, llama.cpp, coreml]22    source_url: https://www.apple.com/mac-studio/specs/23  - key: apple-m5-ultra24    name: Apple M5 Ultra25    manufacturer: apple26    kind: soc27    runtimes: [mlx, llama.cpp, coreml]28    source_url: https://www.apple.com/mac-studio/specs/29  - key: apple-m630    name: Apple M631    manufacturer: apple32    kind: soc33    runtimes: [mlx, llama.cpp, coreml]34    source_url: https://www.apple.com/mac-mini/specs/35  - key: nvidia-h100-nvl36    name: NVIDIA H100 NVL37    manufacturer: nvidia38    kind: gpu39    architecture: Hopper40    runtimes: [cuda, tensorrt-llm, vllm, sglang, llama.cpp]41    source_url: https://www.nvidia.com/en-us/data-center/h100/42  - key: nvidia-h200-nvl43    name: NVIDIA H200 NVL44    manufacturer: nvidia45    kind: gpu46    architecture: Hopper47    runtimes: [cuda, tensorrt-llm, vllm, sglang, llama.cpp]48    source_url: https://www.nvidia.com/en-us/data-center/h200/49  - key: nvidia-dgx-b20050    name: NVIDIA DGX B20051    manufacturer: nvidia52    kind: system53    architecture: Blackwell54    runtimes: [cuda, tensorrt-llm, vllm, sglang]55    source_url: https://www.nvidia.com/en-us/data-center/dgx-b200/56