# Installation \> Compute Platform: ROCm 6.1 **URL:** \ **Category:** Uncategorized\ **Created:** [October 17, 2024, 2:55am UTC](https://discuss.pytorch.org/t/installation-compute-platform-rocm-6-1/211694 "2024-10-17T02:55:03Z")\ **Posts on this page:** 13\ **Page:** 1
**Author:** ![dcpetit](https://discuss.pytorch.org/letter_avatar_proxy/v4/letter/d/e68b1a/32.png) [@dcpetit](https://discuss.pytorch.org/u/dcpetit)\ **Post date:** [October 17, 2024, 2:55am UTC](https://discuss.pytorch.org/t/installation-compute-platform-rocm-6-1/211694/1 "2024-10-17T02:55:04Z")
Hello! I am trying to install PyTorch. How do I upgrade my ROCm to 6.1 (to install PyTorch)? I have an Ubuntu linux with the following specifications: - **Hardware Model:** Lenovo ThinkPad X13 Gen 4 … - **Processor:** AMD Ryzen™ 7 PRO 7840U w/ Radeon™ 780M Graphics × 16 - **Graphics:** AMD Radeon™ Graphics - **Firmware Version:** R29ET56W (1.30 ) … - **OS Name:** Ubuntu 24.04.1 LTS - **OS Build:** (null) - **OS Type:** 64-bit - **GNOME Version:** 46 - **Windowing System:** Wayland - **Kernel Version:** Linux 6.8.0-47-generic I seem to have an older version(s) of ROCm. When I use “apt show rocm…” in the command line I get: 1.) Package: rocminfo Version: 5.7.1-3build1 Priority: optional Section: universe/devel Origin: Ubuntu Installed-Size: 95.2 kB Depends: libc6 (\>= 2.34), libgcc-s1 (\>= 3.3.1), libhsa-runtime64-1 (\>= 5.7.1~), libstdc++6 (\>= 11), python3, pciutils, kmod Download-Size: 25.6 kB 2.) Package: rocm-cmake Version: 6.0.0-1 Priority: optional Section: universe/devel Origin: Ubuntu Installed-Size: 138 kB Download-Size: 25.6 kB 3.) Package: rocm-device-libs-17 Version: 6.0+git20231212.5a852ed-2 Priority: optional Section: universe/libs Source: rocm-device-libs Origin: Ubuntu Installed-Size: 3,324 kB Download-Size: 549 kB 4.) Package: rocm-smi Version: 5.7.0-1 Priority: optional Section: universe/utils Source: rocm-smi-lib Origin: Ubuntu Installed-Size: 227 kB Depends: python3:any, librocm-smi64-1 (= 5.7.0-1) Download-Size: 52.9 kB None of the version numbers at 6.1 or higher, so I think I should upgrade before installing, yes? Anyone familiar with upgrading ROCm that can help me here 🙂 ---
**Author:** ![Maksym\_Oberemchenko](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/maksym_oberemchenko/32/74402_2.png) [@Maksym\_Oberemchenko](https://discuss.pytorch.org/u/Maksym_Oberemchenko)\ **Post date:** [December 7, 2024, 1:13pm UTC](https://discuss.pytorch.org/t/installation-compute-platform-rocm-6-1/211694/2 "2024-12-07T13:13:03Z")
> [@dcpetit](#): > > 780M Graphics 780M Graphics is not supported! See the ROCm docs. You may want to try 7600 it looks like it’s on the list. ---
**Author:** ![dcpetit](https://discuss.pytorch.org/letter_avatar_proxy/v4/letter/d/e68b1a/32.png) [@dcpetit](https://discuss.pytorch.org/u/dcpetit)\ **Post date:** [January 6, 2025, 5:25am UTC](https://discuss.pytorch.org/t/installation-compute-platform-rocm-6-1/211694/3 "2025-01-06T05:25:28Z")
Hello Maksym\_Oberemchenko, Thank you for your reply! I am looking at the ROCm docs and not finding relevant information on what hardware is supported; is it in these (AMD) GPUs [here](https://rocm.docs.amd.com/projects/install-on-linux/en/latest/reference/system-requirements.html#system-requirements)? How can I try 7600? I don’t see it on any of the lists now, nor know how to use it for installing ROCm and then PyTorch 🙂 ---
**Author:** ![fluidnumerics\_joe](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/fluidnumerics_joe/32/75754_2.png) [@fluidnumerics\_joe](https://discuss.pytorch.org/u/fluidnumerics_joe)\ **Post date:** [March 3, 2025, 9:02pm UTC](https://discuss.pytorch.org/t/installation-compute-platform-rocm-6-1/211694/4 "2025-03-03T21:02:06Z")
Hey @dcpetit , There’s a good deal of information on installing ROCm at [Quick start installation guide — ROCm installation (Linux)](https://rocm.docs.amd.com/projects/install-on-linux/en/latest/install/quick-start.html) Definitely check out the system requirements at [System requirements (Linux) — ROCm installation (Linux)](https://rocm.docs.amd.com/projects/install-on-linux/en/latest/reference/system-requirements.html#supported-distributions) We’ve had varied success with GPUs that are not in the “Supported GPUs” list; we’ve found that making sure your OS and the Linux kernel version are supported is definitely critical. From what you’ve shared, Ubuntu 24.04 with Linux kernel 6.8.0-47-generic is supported. I’d recommend uninstalling uninstalling the existing rocm and amdgpu-dkms packages you have currently installed. From there, use the amdgpu-install helper script to install the latest ROCm 6.1 release (6.1.4) ```auto wget https://repo.radeon.com/amdgpu-install/6.1.4/ubuntu/jammy/amdgpu-install_6.1.60104-1_all.deb sudo apt install ./amdgpu-install_6.1.60104-1_all.deb sudo apt update sudo apt install amdgpu-dkms rocm ``` Once installed, reboot your system. Check to see if your GPUs are picked up by `rocm-smi` and `rocminfo` Hope this helps ---
**Author:** ![dcpetit](https://discuss.pytorch.org/letter_avatar_proxy/v4/letter/d/e68b1a/32.png) [@dcpetit](https://discuss.pytorch.org/u/dcpetit)\ **Post date:** [March 4, 2025, 4:19am UTC](https://discuss.pytorch.org/t/installation-compute-platform-rocm-6-1/211694/5 "2025-03-04T04:19:12Z")
Hi Joe! Thanks for your response. I’ve taken your advice and gotten some results, I think they are good. My original post is quite old, so ROCm’s latest version seems to be 6.3 now, and I edited your recommended code to reflect that. After a long installation I get this: ```auto d...@...:~$: rocm-smi =========================================== ROCm System Management Interface =========================================== ===================================================== Concise Info ===================================================== Device Node IDs Temp Power Partitions SCLK MCLK Fan Perf PwrCap VRAM% GPU% (DID, GUID) (Edge) (Socket) (Mem, Compute, ID) ======================================================================================================================== 0 1 0x15bf, 26266 43.0°C 12.073W N/A, N/A, 0 None 800Mhz 0% auto Unsupported 73% 1% ======================================================================================================================== ================================================= End of ROCm SMI Log ================================================== (base) d...@...:~$ ``` and ```auto :~$ rocm-smi -i ============================ ROCm System Management Interface ============================ =========================================== ID =========================================== GPU[0] : Device Name: Phoenix1 GPU[0] : Device ID: 0x15bf GPU[0] : Device Rev: 0xdd GPU[0] : Subsystem ID: 0x50d0 GPU[0] : GUID: 26266 ========================================================================================== ================================== End of ROCm SMI Log =================================== ``` and ```auto (base) d...@...:~$:~$ rocminfo ROCk module version 6.10.5 is loaded ===================== HSA System Attributes ===================== Runtime Version: 1.14 Runtime Ext Version: 1.6 System Timestamp Freq.: 1000.000000MHz Sig. Max Wait Duration: 18446744073709551615 (0xFFFFFFFFFFFFFFFF) (timestamp count) Machine Model: LARGE System Endianness: LITTLE Mwaitx: DISABLED DMAbuf Support: YES ========== HSA Agents ========== ******* Agent 1 ******* Name: AMD Ryzen 7 PRO 7840U w/ Radeon 780M Graphics Uuid: CPU-XX Marketing Name: AMD Ryzen 7 PRO 7840U w/ Radeon 780M Graphics Vendor Name: CPU Feature: None specified Profile: FULL_PROFILE Float Round Mode: NEAR Max Queue Number: 0(0x0) Queue Min Size: 0(0x0) Queue Max Size: 0(0x0) Queue Type: MULTI Node: 0 Device Type: CPU Cache Info: L1: 32768(0x8000) KB Chip ID: 0(0x0) ASIC Revision: 0(0x0) Cacheline Size: 64(0x40) Max Clock Freq. (MHz): 5132 BDFID: 0 Internal Node ID: 0 Compute Unit: 16 SIMDs per CU: 0 Shader Engines: 0 Shader Arrs. per Eng.: 0 WatchPts on Addr. Ranges:1 Memory Properties: Features: None Pool Info: Pool 1 Segment: GLOBAL; FLAGS: FINE GRAINED Size: 31452120(0x1dfebd8) KB Allocatable: TRUE Alloc Granule: 4KB Alloc Recommended Granule:4KB Alloc Alignment: 4KB Accessible by all: TRUE Pool 2 Segment: GLOBAL; FLAGS: EXTENDED FINE GRAINED Size: 31452120(0x1dfebd8) KB Allocatable: TRUE Alloc Granule: 4KB Alloc Recommended Granule:4KB Alloc Alignment: 4KB Accessible by all: TRUE Pool 3 Segment: GLOBAL; FLAGS: KERNARG, FINE GRAINED Size: 31452120(0x1dfebd8) KB Allocatable: TRUE Alloc Granule: 4KB Alloc Recommended Granule:4KB Alloc Alignment: 4KB Accessible by all: TRUE Pool 4 Segment: GLOBAL; FLAGS: COARSE GRAINED Size: 31452120(0x1dfebd8) KB Allocatable: TRUE Alloc Granule: 4KB Alloc Recommended Granule:4KB Alloc Alignment: 4KB Accessible by all: TRUE ISA Info: ******* Agent 2 ******* Name: gfx1103 Uuid: GPU-XX Marketing Name: AMD Radeon Graphics Vendor Name: AMD Feature: KERNEL_DISPATCH Profile: BASE_PROFILE Float Round Mode: NEAR Max Queue Number: 128(0x80) Queue Min Size: 64(0x40) Queue Max Size: 131072(0x20000) Queue Type: MULTI Node: 1 Device Type: GPU Cache Info: L1: 32(0x20) KB L2: 2048(0x800) KB Chip ID: 5567(0x15bf) ASIC Revision: 9(0x9) Cacheline Size: 64(0x40) Max Clock Freq. (MHz): 2700 BDFID: 49920 Internal Node ID: 1 Compute Unit: 12 SIMDs per CU: 2 Shader Engines: 1 Shader Arrs. per Eng.: 2 WatchPts on Addr. Ranges:4 Coherent Host Access: FALSE Memory Properties: APU Features: KERNEL_DISPATCH Fast F16 Operation: TRUE Wavefront 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 4294967295(0xffffffff) y 4294967295(0xffffffff) z 4294967295(0xffffffff) Max fbarriers/Workgrp: 32 Packet Processor uCode:: 40 SDMA engine uCode:: 21 IOMMU Support:: None Pool Info: Pool 1 Segment: GLOBAL; FLAGS: COARSE GRAINED Size: 15726060(0xeff5ec) KB Allocatable: TRUE Alloc Granule: 4KB Alloc Recommended Granule:2048KB Alloc Alignment: 4KB Accessible by all: FALSE Pool 2 Segment: GLOBAL; FLAGS: EXTENDED FINE GRAINED Size: 15726060(0xeff5ec) KB Allocatable: TRUE Alloc Granule: 4KB Alloc Recommended Granule:2048KB Alloc Alignment: 4KB Accessible by all: FALSE Pool 3 Segment: GROUP Size: 64(0x40) KB Allocatable: FALSE Alloc Granule: 0KB Alloc Recommended Granule:0KB Alloc Alignment: 0KB Accessible by all: FALSE ISA Info: ISA 1 Name: amdgcn-amd-amdhsa--gfx1103 Machine Models: HSA_MACHINE_MODEL_LARGE Profiles: HSA_PROFILE_BASE Default Rounding Mode: NEAR Default Rounding Mode: NEAR Fast f16: TRUE Workgroup 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 4294967295(0xffffffff) y 4294967295(0xffffffff) z 4294967295(0xffffffff) FBarrier Max Size: 32 ***Done*** (base) d...@...:~$ ``` Looks like it works. Is the best place to learn how to install PyTorch [here](https://pytorch.org/get-started/locally/)? It is suggesting that I use `pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm6.2.4`, however I did just install 6.3 intead of 6.2(.4). Does it make a difference? ---
**Author:** ![fluidnumerics\_joe](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/fluidnumerics_joe/32/75754_2.png) [@fluidnumerics\_joe](https://discuss.pytorch.org/u/fluidnumerics_joe)\ **Post date:** [March 4, 2025, 4:04pm UTC](https://discuss.pytorch.org/t/installation-compute-platform-rocm-6-1/211694/6 "2025-03-04T16:04:13Z")
@dcpetit , This is looking good. Yes, [Start Locally | PyTorch](https://pytorch.org/get-started/locally/) is the easiest way to get started. Definitely match the major.minor version of your ROCm installation with the appropriate build of pytorch. Since you’re working with ROCm 6.3, you can use the nightly installs of pytorch ```auto pip3 install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/rocm6.3 ``` Note that your GPU is not on AMD’s supported GPU list. What this means is that AMD does not regularly test ROCm against that GPU. In my experience, this simply means your mileage will vary. For gfx1103 GPUs, like yours, you can try setting the environment variable `HSA_OVERRIDE_GFX_VERSION` to `11.0.0`, e.g. ```auto export HSA_OVERRIDE_GFX_VERSION=11.0.0 ``` before running pytorch applications ( you can even add this to your `~/.bashrc` so you don’t have to remember this every time ). This will force pytorch to view your GPU as a `gfx1100` card. It’d be interesting to see what this does with some minimal pytorch examples. Let me know if you run into any more trouble. 🙂 ---
**Author:** ![dcpetit](https://discuss.pytorch.org/letter_avatar_proxy/v4/letter/d/e68b1a/32.png) [@dcpetit](https://discuss.pytorch.org/u/dcpetit)\ **Post date:** [March 5, 2025, 12:46am UTC](https://discuss.pytorch.org/t/installation-compute-platform-rocm-6-1/211694/7 "2025-03-05T00:46:04Z")
It’s looking rather good now Joe, thanks for your help. I’ve gotten PyTorch up and running, but I think I’m running into some trouble though with ROCm. I’ve made a very simple script to test this; it has: ```auto import torch if torch.version.hip is not None: device = torch.device("hip") a = torch.randn(3, 3, device=device) print(a) else: print("ROCm/HIP not available.") ``` I was hoping it would print ‘a’ without any issues, but I get a very long error message when I do this. Do you know what `not run aten::empty.memory_format' with arguments from the 'HIP' backend.` means by chance? The full error message is pasted below. ```auto --------------------------------------------------------------------------- NotImplementedError Traceback (most recent call last) Cell In[11], line 7 5 if torch.version.hip is not None: 6 device = torch.device("hip") ----> 7 a = torch.randn(3, 3, device=device) 8 print(a) 9 else: NotImplementedError: Could not run 'aten::empty.memory_format' with arguments from the 'HIP' backend. This could be because the operator doesn't exist for this backend, or was omitted during the selective/custom build process (if using custom build). If you are a Facebook employee using PyTorch on mobile, please visit https://fburl.com/ptmfixes for possible resolutions. 'aten::empty.memory_format' is only available for these backends: [CPU, CUDA, Meta, QuantizedCPU, QuantizedCUDA, QuantizedMeta, MkldnnCPU, SparseCPU, SparseCUDA, SparseMeta, SparseCsrCPU, SparseCsrCUDA, SparseCsrMeta, BackendSelect, Python, FuncTorchDynamicLayerBackMode, Functionalize, Named, Conjugate, Negative, ZeroTensor, ADInplaceOrView, AutogradOther, AutogradCPU, AutogradCUDA, AutogradHIP, AutogradXLA, AutogradMPS, AutogradIPU, AutogradXPU, AutogradHPU, AutogradVE, AutogradLazy, AutogradMTIA, AutogradPrivateUse1, AutogradPrivateUse2, AutogradPrivateUse3, AutogradMeta, AutogradNestedTensor, Tracer, AutocastCPU, AutocastMTIA, AutocastXPU, AutocastMPS, AutocastCUDA, FuncTorchBatched, BatchedNestedTensor, FuncTorchVmapMode, Batched, VmapMode, FuncTorchGradWrapper, PythonTLSSnapshot, FuncTorchDynamicLayerFrontMode, PreDispatch, PythonDispatcher]. CPU: registered at /pytorch/build/aten/src/ATen/RegisterCPU_1.cpp:2517 [kernel] CUDA: registered at /pytorch/build/aten/src/ATen/RegisterCUDA_0.cpp:9209 [kernel] Meta: registered at /pytorch/build/aten/src/ATen/RegisterMeta_0.cpp:5433 [kernel] QuantizedCPU: registered at /pytorch/build/aten/src/ATen/RegisterQuantizedCPU_0.cpp:311 [kernel] QuantizedCUDA: registered at /pytorch/build/aten/src/ATen/RegisterQuantizedCUDA_0.cpp:194 [kernel] QuantizedMeta: registered at /pytorch/build/aten/src/ATen/RegisterQuantizedMeta_0.cpp:116 [kernel] MkldnnCPU: registered at /pytorch/build/aten/src/ATen/RegisterMkldnnCPU_0.cpp:230 [kernel] SparseCPU: registered at /pytorch/build/aten/src/ATen/RegisterSparseCPU_0.cpp:843 [kernel] SparseCUDA: registered at /pytorch/build/aten/src/ATen/RegisterSparseCUDA_0.cpp:881 [kernel] SparseMeta: registered at /pytorch/build/aten/src/ATen/RegisterSparseMeta_0.cpp:187 [kernel] SparseCsrCPU: registered at /pytorch/build/aten/src/ATen/RegisterSparseCsrCPU_0.cpp:737 [kernel] SparseCsrCUDA: registered at /pytorch/build/aten/src/ATen/RegisterSparseCsrCUDA_0.cpp:835 [kernel] SparseCsrMeta: registered at /pytorch/build/aten/src/ATen/RegisterSparseCsrMeta_0.cpp:702 [kernel] BackendSelect: registered at /pytorch/build/aten/src/ATen/RegisterBackendSelect.cpp:792 [kernel] Python: registered at /pytorch/aten/src/ATen/core/PythonFallbackKernel.cpp:194 [backend fallback] FuncTorchDynamicLayerBackMode: registered at /pytorch/aten/src/ATen/functorch/DynamicLayer.cpp:479 [backend fallback] Functionalize: registered at /pytorch/aten/src/ATen/FunctionalizeFallbackKernel.cpp:349 [backend fallback] Named: registered at /pytorch/aten/src/ATen/core/NamedRegistrations.cpp:7 [backend fallback] Conjugate: fallthrough registered at /pytorch/aten/src/ATen/ConjugateFallback.cpp:21 [kernel] Negative: fallthrough registered at /pytorch/aten/src/ATen/native/NegateFallback.cpp:22 [kernel] ZeroTensor: fallthrough registered at /pytorch/aten/src/ATen/ZeroTensorFallback.cpp:90 [kernel] ADInplaceOrView: fallthrough registered at /pytorch/aten/src/ATen/core/VariableFallbackKernel.cpp:100 [backend fallback] AutogradOther: registered at /pytorch/torch/csrc/autograd/generated/VariableType_2.cpp:20142 [autograd kernel] AutogradCPU: registered at /pytorch/torch/csrc/autograd/generated/VariableType_2.cpp:20142 [autograd kernel] AutogradCUDA: registered at /pytorch/torch/csrc/autograd/generated/VariableType_2.cpp:20142 [autograd kernel] AutogradHIP: registered at /pytorch/torch/csrc/autograd/generated/VariableType_2.cpp:20142 [autograd kernel] AutogradXLA: registered at /pytorch/torch/csrc/autograd/generated/VariableType_2.cpp:20142 [autograd kernel] AutogradMPS: registered at /pytorch/torch/csrc/autograd/generated/VariableType_2.cpp:20142 [autograd kernel] AutogradIPU: registered at /pytorch/torch/csrc/autograd/generated/VariableType_2.cpp:20142 [autograd kernel] AutogradXPU: registered at /pytorch/torch/csrc/autograd/generated/VariableType_2.cpp:20142 [autograd kernel] AutogradHPU: registered at /pytorch/torch/csrc/autograd/generated/VariableType_2.cpp:20142 [autograd kernel] AutogradVE: registered at /pytorch/torch/csrc/autograd/generated/VariableType_2.cpp:20142 [autograd kernel] AutogradLazy: registered at /pytorch/torch/csrc/autograd/generated/VariableType_2.cpp:20142 [autograd kernel] AutogradMTIA: registered at /pytorch/torch/csrc/autograd/generated/VariableType_2.cpp:20142 [autograd kernel] AutogradPrivateUse1: registered at /pytorch/torch/csrc/autograd/generated/VariableType_2.cpp:20142 [autograd kernel] AutogradPrivateUse2: registered at /pytorch/torch/csrc/autograd/generated/VariableType_2.cpp:20142 [autograd kernel] AutogradPrivateUse3: registered at /pytorch/torch/csrc/autograd/generated/VariableType_2.cpp:20142 [autograd kernel] AutogradMeta: registered at /pytorch/torch/csrc/autograd/generated/VariableType_2.cpp:20142 [autograd kernel] AutogradNestedTensor: registered at /pytorch/torch/csrc/autograd/generated/VariableType_2.cpp:20142 [autograd kernel] Tracer: registered at /pytorch/torch/csrc/autograd/generated/TraceType_2.cpp:17801 [kernel] AutocastCPU: fallthrough registered at /pytorch/aten/src/ATen/autocast_mode.cpp:322 [backend fallback] AutocastMTIA: fallthrough registered at /pytorch/aten/src/ATen/autocast_mode.cpp:466 [backend fallback] AutocastXPU: fallthrough registered at /pytorch/aten/src/ATen/autocast_mode.cpp:504 [backend fallback] AutocastMPS: fallthrough registered at /pytorch/aten/src/ATen/autocast_mode.cpp:209 [backend fallback] AutocastCUDA: fallthrough registered at /pytorch/aten/src/ATen/autocast_mode.cpp:165 [backend fallback] FuncTorchBatched: registered at /pytorch/aten/src/ATen/functorch/LegacyBatchingRegistrations.cpp:731 [backend fallback] BatchedNestedTensor: registered at /pytorch/aten/src/ATen/functorch/LegacyBatchingRegistrations.cpp:758 [backend fallback] FuncTorchVmapMode: fallthrough registered at /pytorch/aten/src/ATen/functorch/VmapModeRegistrations.cpp:27 [backend fallback] Batched: registered at /pytorch/aten/src/ATen/LegacyBatchingRegistrations.cpp:1075 [backend fallback] VmapMode: fallthrough registered at /pytorch/aten/src/ATen/VmapModeRegistrations.cpp:33 [backend fallback] FuncTorchGradWrapper: registered at /pytorch/aten/src/ATen/functorch/TensorWrapper.cpp:208 [backend fallback] PythonTLSSnapshot: registered at /pytorch/aten/src/ATen/core/PythonFallbackKernel.cpp:202 [backend fallback] FuncTorchDynamicLayerFrontMode: registered at /pytorch/aten/src/ATen/functorch/DynamicLayer.cpp:475 [backend fallback] PreDispatch: registered at /pytorch/aten/src/ATen/core/PythonFallbackKernel.cpp:206 [backend fallback] PythonDispatcher: registered at /pytorch/aten/src/ATen/core/PythonFallbackKernel.cpp:198 [backend fallback] ``` Have a great day! ---
**Author:** ![fluidnumerics\_joe](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/fluidnumerics_joe/32/75754_2.png) [@fluidnumerics\_joe](https://discuss.pytorch.org/u/fluidnumerics_joe)\ **Post date:** [March 5, 2025, 1:51am UTC](https://discuss.pytorch.org/t/installation-compute-platform-rocm-6-1/211694/8 "2025-03-05T01:51:31Z")
It may seem odd, but the device needs to be set to `cuda` . The pytorch folks opted to stick with the CUDA naming conventions to mean running on a GPU backend so that developers didn’t need to change existing pytorch codes to run on AMD GPUs. Here’s a modified version of your example code that should work ```python import torch if torch.version.hip is not None: device = torch.device("cuda") a = torch.randn(3, 3, device=device) print(a) else: print("ROCm/HIP not available.") ``` ---
**Author:** ![dcpetit](https://discuss.pytorch.org/letter_avatar_proxy/v4/letter/d/e68b1a/32.png) [@dcpetit](https://discuss.pytorch.org/u/dcpetit)\ **Post date:** [March 5, 2025, 3:59am UTC](https://discuss.pytorch.org/t/installation-compute-platform-rocm-6-1/211694/9 "2025-03-05T03:59:28Z")
Never mind, I had a problem (below), but solved it by simply restarting the program and kernel. _### ### ### ### ### ### ### ### ### ###_ That changes things, but unfortunately it does result in a different error at the same step: ```auto --------------------------------------------------------------------------- RuntimeError Traceback (most recent call last) Cell In[2], line 7 5 if torch.version.hip is not None: 6 device = torch.device("cuda") ----> 7 a = torch.randn(3, 3, device=device) 8 print(a) 9 else: RuntimeError: HIP error: invalid device function HIP kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect. For debugging consider passing AMD_SERIALIZE_KERNEL=3 Compile with `TORCH_USE_HIP_DSA` to enable device-side assertions. ``` Do I have to change the hip to cuda in the `if torch.version.hip is not None:` line as well? ---
**Author:** ![fluidnumerics\_joe](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/fluidnumerics_joe/32/75754_2.png) [@fluidnumerics\_joe](https://discuss.pytorch.org/u/fluidnumerics_joe)\ **Post date:** [March 5, 2025, 6:53pm UTC](https://discuss.pytorch.org/t/installation-compute-platform-rocm-6-1/211694/10 "2025-03-05T18:53:21Z")
I’m glad it got resolved. Oddly enough, `torch.version.hip` is the API call to check for the HIP version used by the “cuda” backend. I know… odd decisions were made. ---
**Author:** ![dcpetit](https://discuss.pytorch.org/letter_avatar_proxy/v4/letter/d/e68b1a/32.png) [@dcpetit](https://discuss.pytorch.org/u/dcpetit)\ **Post date:** [March 6, 2025, 12:17am UTC](https://discuss.pytorch.org/t/installation-compute-platform-rocm-6-1/211694/11 "2025-03-06T00:17:04Z")
I hope this helps anyone reading this thread that also has gfx1103 GPUs, like mine. This is so strange Joe, I got the error in my last post, but then restarted the kernel and it worked great. Then later I shut down my computer. Today, after turning my computer back on, I ran the same code and got the same error that I originally got in the last post. Confused, I tried a few things and eventually remembered I had forced pytorch to view my GPU as `export HSA_OVERRIDE_GFX_VERSION=11.0.0`. After setting that in the terminal, my code would run fine again… I hope that the error message can point towards this solution in the future, or better yet the 11.0.3 version gets supported and this workaround no longer becomes necessary. ---
**Author:** ![fluidnumerics\_joe](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/fluidnumerics_joe/32/75754_2.png) [@fluidnumerics\_joe](https://discuss.pytorch.org/u/fluidnumerics_joe)\ **Post date:** [March 6, 2025, 3:37pm UTC](https://discuss.pytorch.org/t/installation-compute-platform-rocm-6-1/211694/12 "2025-03-06T15:37:52Z")
@dcpetit , In your home directory, there is a file `~/.bashrc` . When ever you start a shell session, this file gets sourced and can be used to configure your environment. If you run the following from your terminal ```auto cat "export HSA_OVERRIDE_GFX_VERSION=11.0.0" >> ~/.bashrc ``` this will ensure this environment variable is set each time you start a shell session. ---
**Author:** ![fedorgoncharov](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/fedorgoncharov/32/76365_2.png) [@fedorgoncharov](https://discuss.pytorch.org/u/fedorgoncharov)\ **Post date:** [March 27, 2025, 9:49am UTC](https://discuss.pytorch.org/t/installation-compute-platform-rocm-6-1/211694/13 "2025-03-27T09:49:37Z")
Thanks, I followed this thread and all worked! I am so glad that I can press out maximum of my poor personal laptop 🙂