Repository navigation
What changed
CPU support. The model now runs on a machine without a GPU: same weights, same 18 labels, same threshold of 0.1041.
intraspannet._gdn_cpuis a plain PyTorch port of the Gated DeltaNet block with identical parameter names, so the released checkpoints load unchanged. It follows the published gated delta rule and fla's reference implementation.pip install intraspannetis now torch-only.pip install "intraspannet[gpu]"adds flash-linear-attention and the Triton kernel, which is roughly three orders of magnitude faster.- Measured on labelled lab captures: 9 of 10 flows correct with the CPU path, about two seconds per flow per arm.
scripts/verify_cpu_backend.pyreproduces that on your own captures. - CLI gains
--armsand--gatefor a quick single-arm run, and no longer aborts on Windows consoles when printing the Chinese tier names.
The weights did not change. They are on Hugging Face and in the v0.1.0 weights asset.
Checksums
acb731488017a0198e00bb214b853c9ca5ef0071d0ea93b9fc6172037d99a33b intraspannet-0.2.0-source.zip