Arm backend: Fix bmm quantization bug (#19798) #24

	name: Test RISC-V Backend

	on:
	push:
	branches:
	- main
	- release/*
	tags:
	- ciflow/trunk/*
	pull_request:
	paths:
	- .github/workflows/riscv64.yml
	- .ci/scripts/test_riscv_qemu.sh
	- tools/cmake/preset/riscv64_linux.cmake
	- examples/riscv/**
	workflow_dispatch:
	schedule:
	- cron: '0 10 * * *' # Runs daily at 2 AM PST

	concurrency:
	group: ${{ github.workflow }}-${{ github.event.pull_request.number \|\| github.sha }}-${{ github.event_name == 'workflow_dispatch' }}-${{ github.event_name == 'schedule' }}
	cancel-in-progress: true

	jobs:
	test-riscv:
	name: test-riscv
	uses: ./.github/workflows/_test_riscv.yml
	strategy:
	fail-fast: false
	matrix:
	model:
	- add
	- mv2
	- mobilebert
	- llama2
	- resnet18
	- yolo26
	xnnpack: [true, false]
	quantize: [true, false]
	exclude:
	# We only enable quantization with XNNPACK
	- xnnpack: false
	quantize: true
	# We don't test quantization for Yolo26
	- model: yolo26
	quantize: true
	permissions:
	id-token: write
	contents: read
	with:
	model: ${{ matrix.model }}
	xnnpack: ${{ matrix.xnnpack }}
	quantize: ${{ matrix.quantize }}
	# If XNNPACK, test with multiple RVV length, disabled otherwise
	qemu-cpu: >-
	${{
	case(
	matrix.xnnpack, '[
	"rv64,zba=true,zbb=true,zbs=true,v=true,vlen=128,elen=64,vext_spec=v1.0",
	"rv64,zba=true,zbb=true,zbs=true,v=true,vlen=256,elen=64,vext_spec=v1.0",
	"rv64,zba=true,zbb=true,zbs=true,v=true,vlen=512,elen=64,vext_spec=v1.0"
	]',
	'[
	"rv64,zba=true,zbb=true,zbs=true,v=false"
	]'
	)
	}}

Provide feedback