该方法不依赖与特定的Linux内核，例如我的Wsl内核如下
Linux DESKTOP-H25OAU7 6.6.87.2-microsoft-standard-WSL2 #1 SMP PREEMPT_DYNAMIC Thu Jun 5 18:30:46 UTC 2025 x86_64 x86_64 x86_64 GNU/Linux
可以看到，远低于Rocm的要求。为此AMD专门为了wsl开发了特殊的环境，安装方法如下。注意，在WSL中不要换源！！！！否则Rocm会提示找不到包
该方法来自于https://github.com/ROCm/ROCm/issues/4952
How to Set up ROCm in WSL on Strix Halo Systems
First, wsl --install Ubuntu-24.04
In Ubuntu 24.04:
Install deps:
sudo apt update
sudo apt install -y ca-certificates wget gpg
Add AMD repo signing key:
sudo mkdir --parents --mode=0755 /etc/apt/keyrings
wget https://repo.radeon.com/rocm/rocm.gpg.key -O - | \gpg --dearmor | sudo tee /etc/apt/keyrings/rocm.gpg > /dev/null
Add AMD apt sources:
sudo tee /etc/apt/sources.list.d/rocm.list << EOF
deb [arch=amd64 signed-by=/etc/apt/keyrings/rocm.gpg] https://repo.radeon.com/rocm/apt/latest noble main
deb [arch=amd64 signed-by=/etc/apt/keyrings/rocm.gpg] https://repo.radeon.com/graphics/latest/ubuntu noble main
EOF
Set higher priority for AMD repo:
sudo tee /etc/apt/preferences.d/rocm-pin-600 << EOF
Package: *
Pin: release o=repo.radeon.com
Pin-Priority: 600
EOF
Install WSL-specific runtime:
sudo apt install hsa-runtime-rocr4wsl-amdgpu
Then get latest deb from https://github.com/ROCm/librocdxg/releases and install it.
Then you can install rocm meta package:
sudo apt install rocm
Finally, add these lines to your .bashrc:
export HSA_ENABLE_DXG_DETECTION=1
export PATH=/opt/rocm/bin:$PATH
Verification
sudo apt install rocminfo
rocminfo
WSL environment detected.
Load librocdxg.so successully!
Load all DTIF APIs OK!
=====================
HSA System Attributes
=====================
Runtime Version: 1.1
Runtime Ext Version: 1.15
System Timestamp Freq.: 1000.000000MHz
Sig. Max Wait Duration: 18446744073709551615 (0xFFFFFFFFFFFFFFFF) (timestamp count)
Machine Model: LARGE
System Endianness: LITTLE
Mwaitx: DISABLED
XNACK enabled: NO
DMAbuf Support: YES
VMM Support: YES==========
HSA Agents
==========
*******
Agent 1
*******Name: AMD RYZEN AI MAX+ 395 w/ Radeon 8060SUuid: CPU-XXMarketing Name: AMD RYZEN AI MAX+ 395 w/ Radeon 8060SVendor Name: CPUFeature: None specifiedProfile: FULL_PROFILEFloat Round Mode: NEARMax Queue Number: 0(0x0)Queue Min Size: 0(0x0)Queue Max Size: 0(0x0)Queue Type: MULTINode: 0Device Type: CPUCache Info:L1: 49152(0xc000) KBChip ID: 0(0x0)Cacheline Size: 64(0x40)BDFID: 0Internal Node ID: 0Compute Unit: 32SIMDs per CU: 0Shader Engines: 0Shader Arrs. per Eng.: 0Memory Properties:Features: NonePool Info:Pool 1Segment: GLOBAL; FLAGS: FINE GRAINEDSize: 32419956(0x1eeb074) KBAllocatable: TRUEAlloc Granule: 4KBAlloc Recommended Granule:4KBAlloc Alignment: 4KBAccessible by all: TRUEPool 2Segment: GLOBAL; FLAGS: EXTENDED FINE GRAINEDSize: 32419956(0x1eeb074) KBAllocatable: TRUEAlloc Granule: 4KBAlloc Recommended Granule:4KBAlloc Alignment: 4KBAccessible by all: TRUEPool 3Segment: GLOBAL; FLAGS: KERNARG, FINE GRAINEDSize: 32419956(0x1eeb074) KBAllocatable: TRUEAlloc Granule: 4KBAlloc Recommended Granule:4KBAlloc Alignment: 4KBAccessible by all: TRUEPool 4Segment: GLOBAL; FLAGS: COARSE GRAINEDSize: 32419956(0x1eeb074) KBAllocatable: TRUEAlloc Granule: 4KBAlloc Recommended Granule:4KBAlloc Alignment: 4KBAccessible by all: TRUEISA Info:
*******
Agent 2
*******Name: gfx1151Uuid: GPU-ffffffffffffffffMarketing Name: AMD Radeon(TM) 8060S GraphicsVendor Name: AMDFeature: KERNEL_DISPATCHProfile: BASE_PROFILEFloat Round Mode: NEARMax Queue Number: 128(0x80)Queue Min Size: 64(0x40)Queue Max Size: 131072(0x20000)Queue Type: MULTINode: 1Device Type: GPUCache Info:L1: 32(0x20) KBL2: 2048(0x800) KBL3: 32768(0x8000) KBChip ID: 5510(0x1586)Cacheline Size: 64(0x40)Max Clock Freq. (MHz): 2900BDFID: 50176Internal Node ID: 1Compute Unit: 40SIMDs per CU: 2Shader Engines: 2Shader Arrs. per Eng.: 2Coherent Host Access: FALSEMemory Properties:Features: KERNEL_DISPATCHFast F16 Operation: TRUEWavefront Size: 32(0x20)Workgroup Max Size: 1024(0x400)Workgroup Max Size per Dimension:x 1024(0x400)y 1024(0x400)z 1024(0x400)Max Waves Per CU: 32(0x20)Max Work-item Per CU: 1024(0x400)Grid Max Size: 4294967295(0xffffffff)Grid Max Size per Dimension:x 2147483647(0x7fffffff)y 65535(0xffff)z 65535(0xffff)Max fbarriers/Workgrp: 32Packet Processor uCode:: 29SDMA engine uCode:: 14IOMMU Support:: NonePool Info:Pool 1Segment: GLOBAL; FLAGS: COARSE GRAINEDSize: 33107874(0x1f92fa2) KBAllocatable: TRUEAlloc Granule: 4KBAlloc Recommended Granule:2048KBAlloc Alignment: 4KBAccessible by all: FALSEPool 2Segment: GLOBAL; FLAGS: EXTENDED FINE GRAINEDSize: 33107874(0x1f92fa2) KBAllocatable: TRUEAlloc Granule: 4KBAlloc Recommended Granule:2048KBAlloc Alignment: 4KBAccessible by all: FALSEPool 3Segment: GROUPSize: 64(0x40) KBAllocatable: FALSEAlloc Granule: 0KBAlloc Recommended Granule:0KBAlloc Alignment: 0KBAccessible by all: FALSEISA Info:ISA 1Name: amdgcn-amd-amdhsa--gfx1151Machine Models: HSA_MACHINE_MODEL_LARGEProfiles: HSA_PROFILE_BASEDefault Rounding Mode: NEARDefault Rounding Mode: NEARFast f16: TRUEWorkgroup Max Size: 1024(0x400)Workgroup Max Size per Dimension:x 1024(0x400)y 1024(0x400)z 1024(0x400)Grid Max Size: 4294967295(0xffffffff)Grid Max Size per Dimension:x 2147483647(0x7fffffff)y 65535(0xffff)z 65535(0xffff)FBarrier Max Size: 32ISA 2Name: amdgcn-amd-amdhsa--gfx11-genericMachine Models: HSA_MACHINE_MODEL_LARGEProfiles: HSA_PROFILE_BASEDefault Rounding Mode: NEARDefault Rounding Mode: NEARFast f16: TRUEWorkgroup Max Size: 1024(0x400)Workgroup Max Size per Dimension:x 1024(0x400)y 1024(0x400)z 1024(0x400)Grid Max Size: 4294967295(0xffffffff)Grid Max Size per Dimension:x 2147483647(0x7fffffff)y 65535(0xffff)z 65535(0xffff)FBarrier Max Size: 32
*** Done ***
Unload librocdxg.so successully!
PyTorch
pip install --index-url https://repo.amd.com/rocm/whl/gfx1151/ -U torch torchvision torchaudio