Modelpublic

emogie3D/xFormers-amd-rocm-gfx1151-cp312-torch.2.11

sourceHugging Faceapache-2.0updated 2mo agoView on Hugging Face

0likes

Model Card

xFormers / PyTorch / ROCm / Windows Diagnosis

This document was generated automatically. The diagnostic script does not modify any files or environment variables.

Diagnosis Timestamp

Property	Value
Local time	2026-08-12T19:59:42.155110+02:00
Working directory	C:\Users\xxx\Desktop
Script path	C:\Users\xxx\Desktop\xformersrocmdiag.py
Running from xFormers source tree	False

System

Property	Value
Operating System	Windows-11-10.0.26200-SP0
Windows Version	10.0.26200
Machine	AMD64
Processor	AMD64 Family 26 Model 112 Stepping 0, AuthenticAMD
Python	3.12.10 (tags/v3.12.10:0cc8128, Apr 8 2025, 12:21:36) [MSC v.1943 64 bit (AMD64)]
Python Executable	C:\Users\xxx\AppData\Local\Programs\Python\Python312\python.exe
Python Prefix	C:\Users\xxx\AppData\Local\Programs\Python\Python312
Python Base Prefix	C:\Users\xxx\AppData\Local\Programs\Python\Python312

Installed Package Versions

Property	Value
torch	2.11.0+rocm7.13.0
torchvision	0.26.0+rocm7.13.0
torchaudio	2.11.0+rocm7.13.0
xformers	0.0.32+6b467648.d20260812
triton	not installed / unavailable
ninja	1.13.0
setuptools	84.0.0
wheel	0.48.0

Relevant Environment Variables

Property	Value
ROCM_PATH	C:\AMD\ROCm\7.1\

Property	Value
ROCM_HOME	(not set)

Property	Value
HIP_PATH	C:\AMD\ROCm\7.1\

Property	Value
HIP_HOME	(not set)

Property	Value
HIP_ARCHITECTURE	(not set)

Property	Value
PYTORCHROCMARCH	gfx1151

Property	Value
XFORMERSCKFLASH_ATTN	(not set)

Property	Value
XFORMERSBUILDTYPE	(not set)

Property	Value
XFORMERSENABLEDEBUG_ASSERTIONS	(not set)

Property	Value
XFORMERSMOREDETAILS	(not set)

Property	Value
XFORMERSDISABLEDOPERATORS	ckF,ckB,ck_splitKF

Property	Value
TORCHEXTENSIONSVERBOSE	(not set)

Property	Value
MAX_JOBS	1

Property	Value
DISTUTILSUSESDK	1

Property	Value
MSSdk	(not set)

Property	Value
CUDA_PATH	(not set)

Property	Value
CUDA_HOME	(not set)

Property	Value
TORCHCUDAARCH_LIST	(not set)

Property	Value
PATH	C:\AMD\ROCm\7.1\bin;C:\AMD\ROCm\7.1\lib;C:\Users\xxx\AppData\Local\Programs\Python\Python312\Lib\site-packages\torch\include;C:\llamacpprocmgfx1151\;C:\Users\xxx\AppData\Local\Programs\Python\Python312\Lib\site-packages\torch\lib

Property	Value
LIB	(not set)

PyTorch

Property	Value
torch._version_	2.11.0+rocm7.13.0
torch.version.hip	7.13.99004
torch.version.cuda	not available
torch._file_	C:\Users\xxx\AppData\Local\Programs\Python\Python312\Lib\site-packages\torch\_init_.py
torch.cuda.is_available()	True
torch.backends.cuda.is_built()	True
GPU Name	AMD Radeon(TM) 8060S Graphics
GPU Index	0
GPU Total Memory	107.87 GiB
GPU Compute Capability	11.5
GPU gcnArchName	gfx1151
GPU Multiprocessors	20

xFormers

Property	Value
xformers._version_	0.0.32+local
xformers._file_	C:\Users\xxx\AppData\Local\Programs\Python\Python312\Lib\site-packages\xformers\_init_.py
xformers package version (metadata)	0.0.32+6b467648.d20260812

xFormers Build Metadata

json

{
  "version": {
    "cuda": null,
    "hip": "7.1.51803",
    "torch": "2.11.0",
    "python": "3.12.10",
    "flash": "0.0.0",
    "use_torch_flash": false
  },
  "env": {}
}

Output of `python -m xformers.info`

Exit code: 0

text

xFormers 0.0.32+local
memory_efficient_attention.ckF:                    available
memory_efficient_attention.ckB:                    available
memory_efficient_attention.ck_splitKF:             available
memory_efficient_attention.cutlassF-pt:            unavailable
memory_efficient_attention.cutlassB-pt:            unavailable
memory_efficient_attention.fa2F@0.0.0:             unavailable
memory_efficient_attention.fa2B@0.0.0:             unavailable
memory_efficient_attention.fa3F@0.0.0:             unavailable
memory_efficient_attention.fa3B@0.0.0:             unavailable
memory_efficient_attention.fa3F_splitKV@0.0.0:     unavailable
memory_efficient_attention.triton_splitKF:         available
indexing.scaled_index_addF:                        available
indexing.scaled_index_addB:                        available
indexing.index_select:                             available
sp24.sparse24_sparsify_both_ways:                  available
sp24.sparse24_apply:                               available
sp24.sparse24_apply_dense_output:                  available
sp24._sparse24_gemm:                               available
sp24._cslt_sparse_mm_search@0.0.0:                 available
sp24._cslt_sparse_mm@0.0.0:                        available
swiglu.dual_gemm_silu:                             available
swiglu.gemm_fused_operand_sum:                     available
swiglu.fused.p.cpp:                                available
is_triton_available:                               True
pytorch.version:                                   2.11.0+rocm7.13.0
pytorch.cuda:                                      available
gpu.compute_capability:                            11.5
gpu.name:                                          AMD Radeon(TM) 8060S Graphics
dcgm_profiler:                                     unavailable
build.info:                                        available
build.cuda_version:                                None
build.hip_version:                                 7.1.51803
build.python_version:                              3.12.10
build.torch_version:                               2.11.0
source.privacy:                                    open source

Detected xFormers Operators

Property	Value
ck	ckB, ckF
ck_splitk	cksplitK1, cksplitK128, cksplitK16, cksplitK2, cksplitK32, cksplitK4, cksplitK64, cksplitK8, ck_splitKF
triton_splitk	tritonsplitK1, tritonsplitK128, tritonsplitK16, tritonsplitK2, tritonsplitK32, tritonsplitK4, tritonsplitK64, tritonsplitK8, triton_splitKF
cutlass	cutlassB-pt, cutlassF-pt
flash	fa2B@0.0.0, fa2F@0.0.0
flash3	fa3B@0.0.0, fa3F@0.0.0, fa3F_splitKV@0.0.0

Attention Function Tests

xFormers memory_efficient_attention uses the layout (Batch, Sequence, Heads, HeadDim). PyTorch SDPA uses (Batch, Heads, Sequence, HeadDim). Comparisons account for this difference.

ATTENTION: The CK kernel from ROCM Pytorch seems to has a bug when dealing with extremely short sequence lengths and very few heads. In regular Use-Cases it doesn't matter.

Seq	Heads	NAN	Good?
16	8	4096/8192	False
32	8	0/16384	True
64	8	0/32768	True
128	8	0/65536	True
256	8	0/131072	True
512	8	0/262144	True
1024	8	0/524288	True
2048	32	0/4194304	True

SDPA comparison: OK

small edge case: `(1, 4, 1, 64)`

Dtype	xFormers Result	PyTorch SDPA Result	Max Difference
`torch.float16`	`shape=(1, 4, 1, 64), dtype=torch.float16, NaN=128, Inf=0, finite=128/256, min=1.0, max=1.0`	`shape=(1, 4, 1, 64), dtype=torch.float16, NaN=0, Inf=0, finite=256/256, min=1.0, max=1.0`	`nan`
`torch.bfloat16`	`shape=(1, 4, 1, 64), dtype=torch.bfloat16, NaN=128, Inf=0, finite=128/256, min=1.0, max=1.0`	`shape=(1, 4, 1, 64), dtype=torch.bfloat16, NaN=0, Inf=0, finite=256/256, min=1.0, max=1.0`	`nan`

typical small attention: `(1, 128, 8, 64)`

Dtype	xFormers Result	PyTorch SDPA Result	Max Difference
`torch.float16`	`shape=(1, 128, 8, 64), dtype=torch.float16, NaN=0, Inf=0, finite=65536/65536, min=1.0, max=1.0`	`shape=(1, 128, 8, 64), dtype=torch.float16, NaN=0, Inf=0, finite=65536/65536, min=1.0, max=1.0`	`0.0`
`torch.bfloat16`	`shape=(1, 128, 8, 64), dtype=torch.bfloat16, NaN=0, Inf=0, finite=65536/65536, min=1.0, max=1.0`	`shape=(1, 128, 8, 64), dtype=torch.bfloat16, NaN=0, Inf=0, finite=65536/65536, min=1.0, max=1.0`	`0.0`

typical larger attention: `(1, 512, 8, 64)`

Dtype	xFormers Result	PyTorch SDPA Result	Max Difference
`torch.float16`	`shape=(1, 512, 8, 64), dtype=torch.float16, NaN=0, Inf=0, finite=262144/262144, min=1.0, max=1.0`	`shape=(1, 512, 8, 64), dtype=torch.float16, NaN=0, Inf=0, finite=262144/262144, min=1.0, max=1.0`	`0.0`
`torch.bfloat16`	`shape=(1, 512, 8, 64), dtype=torch.bfloat16, NaN=0, Inf=0, finite=262144/262144, min=1.0, max=1.0`	`shape=(1, 512, 8, 64), dtype=torch.bfloat16, NaN=0, Inf=0, finite=262144/262144, min=1.0, max=1.0`	`0.0`

Random Data Test

Dtype	Shape	xFormers NaN	xFormers Inf	SDPA NaN	SDPA Inf	Max Difference
`torch.float16`	`(1, 128, 8, 64)`	0	0	0	0	1.52587890625e-05
`torch.bfloat16`	`(1, 128, 8, 64)`	0	0	0	0	0.0001220703125

PyTorch Build Configuration

text

PyTorch built with:
  - C++ Version: 201703
  - clang 23.0.0
  - MSVC 194435224
  - Intel(R) MKL-DNN v3.10.2 (Git Hash f1d471933dc852f956fd05389f9313c7148783d5)
  - OpenMP 202011
  - LAPACK is enabled (usually provided by MKL)
  - CPU capability usage: AVX512
  - HIP Runtime 7.13.99004
  - MIOpen 3.5.1
  - Build settings: BLAS_INFO=open, BUILD_TYPE=Release, COMMIT_SHA=71601f273f8d82178b99832e984a059e77d970c5, CXX_COMPILER=C:/home/runner/_work/_tool/Python/3.12.10/x64/Lib/site-packages/_rocm_sdk_devel/lib/llvm/bin/clang-cl.exe, CXX_FLAGS=/DWIN32 /D_WINDOWS /EHsc /Zc:__cplusplus /bigobj /FS /utf-8 -DUSE_PTHREADPOOL -DNDEBUG -DUSE_XNNPACK -DSYMBOLICATE_MOBILE_DEBUG_HANDLE /wd4624 /wd4068 /wd4067 /wd4267 /wd4661 /wd4717 /wd4244 /wd4804 /wd4273, LAPACK_INFO=open, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, TORCH_VERSION=2.11.0, USE_CUDA=OFF, USE_CUDNN=OFF, USE_CUSPARSELT=OFF, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_GLOO=OFF, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=OFF, USE_NNPACK=OFF, USE_OPENMP=ON, USE_ROCM=ON, USE_ROCM_KERNEL_ASSERT=OFF, USE_XCCL=OFF, USE_XPU=OFF,

Relevant Files

Property	Value
c10.dll	C:\Users\xxx\AppData\Local\Programs\Python\Python312\Lib\site-packages\torch\lib\c10.dll
c10.lib	C:\Users\xxx\AppData\Local\Programs\Python\Python312\Lib\site-packages\torch\lib\c10.lib
torch_cpu.dll	C:\Users\xxx\AppData\Local\Programs\Python\Python312\Lib\site-packages\torch\lib\torch_cpu.dll
torch_cpu.lib	C:\Users\xxx\AppData\Local\Programs\Python\Python312\Lib\site-packages\torch\lib\torch_cpu.lib
torch.dll	C:\Users\xxx\AppData\Local\Programs\Python\Python312\Lib\site-packages\torch\lib\torch.dll
torch.lib	C:\Users\xxx\AppData\Local\Programs\Python\Python312\Lib\site-packages\torch\lib\torch.lib
torch_python.dll	C:\Users\xxx\AppData\Local\Programs\Python\Python312\Lib\site-packages\torch\lib\torch_python.dll
torch_python.lib	C:\Users\xxx\AppData\Local\Programs\Python\Python312\Lib\site-packages\torch\lib\torch_python.lib
amdhip64.dll	not found
amdhip64.lib	C:\AMD\ROCm\7.1\lib\amdhip64.lib
_C.pyd	C:\Users\xxx\AppData\Local\Programs\Python\Python312\Lib\site-packages\xformers\_C.pyd

Optional Export / Dependency Information

dumpbin.exe was not found or _C.pyd is not available. This section is optional.

pip Package Information

`pip show torch`

Exit code: 0

text

Name: torch
Version: 2.11.0+rocm7.13.0
Summary: Tensors and Dynamic neural networks in Python with strong GPU acceleration
Home-page: https://pytorch.org
Author: 
Author-email: PyTorch Team <packages@pytorch.org>
License: BSD-3-Clause
Location: C:\Users\xxx\AppData\Local\Programs\Python\Python312\Lib\site-packages
Requires: filelock, fsspec, jinja2, networkx, rocm, setuptools, sympy, typing-extensions
Required-by: accelerate, auto-round, bitsandbytes, clean-fid, clip-anytorch, compressed-tensors, controlnet_aux, cut-cross-entropy, descript-audio-codec, descript-audiotools, docling-ibm-models, easyocr, fairscale, flair, invisible-watermark, julius, kornia, lightning, lpips, lycoris_lora, open_clip_torch, optimum-quanto, peft, pytorch-fid, pytorch-lightning, pytorch-wavelets, pytorch_revgrad, SAM-2, sentence-transformers, silero-vad, spandrel, timm, torch-stoi, torchdiffeq, torchmetrics, torchsde, torchvision, transformer-smaller-training-vocab, turboquant, ultralytics, ultralytics-thop, unsloth, unsloth_zoo, xformers

`pip show torchvision`

Exit code: 0

text

Name: torchvision
Version: 0.26.0+rocm7.13.0
Summary: image and video datasets and models for torch deep learning
Home-page: https://github.com/pytorch/vision
Author: PyTorch Core Team
Author-email: soumith@pytorch.org
License: BSD
Location: C:\Users\xxx\AppData\Local\Programs\Python\Python312\Lib\site-packages
Requires: numpy, pillow, torch
Required-by: clean-fid, clip-anytorch, controlnet_aux, docling-ibm-models, easyocr, lpips, open_clip_torch, pytorch-fid, SAM-2, spandrel, timm, ultralytics, unsloth

`pip show torchaudio`

Exit code: 0

text

Name: torchaudio
Version: 2.11.0+rocm7.13.0
Summary: An audio package for PyTorch
Home-page: https://github.com/pytorch/audio
Author: Soumith Chintala, David Pollack, Sean Naren, Peter Goldsborough, Moto Hira, Caroline Chen, Jeff Hwang, Zhaoheng Ni, Xiaohui Zhang
Author-email: soumith@pytorch.org
License: 
Location: C:\Users\xxx\AppData\Local\Programs\Python\Python312\Lib\site-packages
Requires: 
Required-by: descript-audio-codec, descript-audiotools, silero-vad, torch-stoi

`pip show xformers`

Exit code: 0

text

Name: xformers
Version: 0.0.32+6b467648.d20260812
Summary: XFormers: A collection of composable Transformer building blocks.
Home-page: https://facebookresearch.github.io/xformers/
Author: Facebook AI Research
Author-email: oncall+xformers@xmail.facebook.com
License: 
Location: C:\Users\xxx\AppData\Local\Programs\Python\Python312\Lib\site-packages
Requires: numpy, torch
Required-by: unsloth

`pip show triton`

Exit code: 1

text

WARNING: Package(s) not found: triton

Automatic Summary

—✅ The script was not started from a recognized xFormers source tree.
—✅ PyTorch reports HIP/ROCm: 7.13.99004
—✅ PyTorch detects at least one GPU.
—✅ xFormers could be imported.

xFormers / PyTorch / ROCm / Windows Diagnosis

Diagnosis Timestamp

System

Installed Package Versions

Relevant Environment Variables

PyTorch

xFormers

xFormers Build Metadata

Output of python -m xformers.info

Detected xFormers Operators

Attention Function Tests

ATTENTION: The CK kernel from ROCM Pytorch seems to has a bug when dealing with extremely short sequence lengths and very few heads. In regular Use-Cases it doesn't matter.

small edge case: (1, 4, 1, 64)

typical small attention: (1, 128, 8, 64)

typical larger attention: (1, 512, 8, 64)

Random Data Test

PyTorch Build Configuration

Relevant Files

Optional Export / Dependency Information

pip Package Information

pip show torch

pip show torchvision

pip show torchaudio

pip show xformers

pip show triton

Automatic Summary

Output of `python -m xformers.info`

small edge case: `(1, 4, 1, 64)`

typical small attention: `(1, 128, 8, 64)`

typical larger attention: `(1, 512, 8, 64)`

`pip show torch`

`pip show torchvision`

`pip show torchaudio`

`pip show xformers`

`pip show triton`