Instructions to use SZLHOLDINGS/YARQA-ATTN with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Kernels
How to use SZLHOLDINGS/YARQA-ATTN with Kernels:
# !pip install kernels from kernels import get_kernel kernel = get_kernel("SZLHOLDINGS/YARQA-ATTN") - Notebooks
- Google Colab
- Kaggle
File size: 795 Bytes
621477d | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 | # SPDX-FileCopyrightText: 2026 SZL Holdings
# SPDX-License-Identifier: Apache-2.0
"""Source-tree example of yarqa-attn. Not a benchmark. Shapes only.
Hub get_kernel is MEASURE after KERNEL bind. This script is not import-LIVE.
"""
from __future__ import annotations
import sys
from pathlib import Path
sys.path.insert(0, str(Path(__file__).resolve().parent / "torch-ext"))
import torch
from yarqa_attn import ReceiptChain, canal_bounds, selfcheck, yarqa_attn
q = torch.randn(1, 2, 16, 32)
k = torch.randn(1, 2, 16, 32)
v = torch.randn(1, 2, 16, 32)
chain = ReceiptChain()
out = yarqa_attn(q, k, v, 4, chain=chain)
ok, depth, first_break = chain.verify()
print("bounds", canal_bounds(16, 4))
print("out", tuple(out.shape), "verify", ok, depth, first_break)
print("selfcheck", selfcheck())
|