cannot install xFormers from Source anymore since installing latest Automatic1111 version

[Marcophono2]

Hello!
After a longer while (may be 8 months) I reinstalled 1111 from the scratch but I am not able anylonger to install xFormers from the source. Sure, this has nothing to do with the 1111 ver. 1.7.0 but probably with a conflict(?) between the new package versions coming with 1111 1.7.0. May be torch, diffusers and xFormers. I hope someone can help me.

Ubuntu 23.04
RTX 4090

nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2022 NVIDIA Corporation
Built on Wed_Sep_21_10:33:58_PDT_2022
Cuda compilation tools, release 11.8, V11.8.89
Build cuda_11.8.r11.8/compiler.31833905_0

Python 3.10.11

diffusers 0.25.0
torch 2.1.2+cu118
torchaudio 2.1.2+cu118
torchdiffeq 0.2.3
torchmetrics 0.11.4
torchsde 0.2.6
torchvision 0.16.2+cu118
nvidia-cublas-cu11 11.10.3.66
nvidia-cublas-cu12 12.1.3.1
nvidia-cuda-cupti-cu11 11.7.101
nvidia-cuda-cupti-cu12 12.1.105
nvidia-cuda-nvrtc-cu11 11.7.99
nvidia-cuda-nvrtc-cu12 12.1.105
nvidia-cuda-runtime-cu11 11.7.99
nvidia-cuda-runtime-cu12 12.1.105
nvidia-cudnn-cu11 8.9.4.25
nvidia-cudnn-cu12 8.9.2.26
nvidia-cufft-cu11 10.9.0.58
nvidia-cufft-cu12 11.0.2.54
nvidia-curand-cu11 10.2.10.91
nvidia-curand-cu12 10.3.2.106
nvidia-cusolver-cu11 11.4.0.1
nvidia-cusolver-cu12 11.4.5.107
nvidia-cusparse-cu11 11.7.4.91
nvidia-cusparse-cu12 12.1.0.106
nvidia-ml-py 12.535.108
nvidia-ml-py3 7.352.0
nvidia-nccl-cu11 2.14.3
nvidia-nccl-cu12 2.18.1
nvidia-nvjitlink-cu12 12.3.101
nvidia-nvtx-cu11 11.7.91
nvidia-nvtx-cu12 12.1.105

When I try

git clone https://github.com/facebookresearch/xformers.git
cd xformers
git submodule update --init --recursive
pip install -r requirements.txt
pip install -e .

the command shell process stopps after about 2 minute by closing itsself. xFormers isn't installed.
Has anyone an idea what the reason can be? I read anywhere that meanwhile xFormers isn't needed for performance reasons but in my case it makes abig difference. Although when I chose manually one of the optimization settings for layer attention in the best case "only" 60% of performance are missing.

Another point: Also the image quality is worse now. Do I have to change some settings manually to get the same results when using the same model, same sampler and else?

Best regards
Marc

[Marcophono2]

Okay, it seems that the building process takes too much ram though I have 128 GB from which > 100 GB were available. Setting max jobs to 1 seems to have solved the issue. Still running but the remaining RAM never is below 58 GB.

[Marcophono2]

xFormer was correctly installed but there is still no increase of performance. Do I have to change any of the default settings in the GUI? There are much more options now to change.

[Marcophono2]

May be it is helpful if I post the output of xformers.info here. Therea re no entries for the built environment. Could this point to the reason behind my performance issue?

python3 -m xformers.info
Unable to find python bindings at /usr/local/dcgm/bindings/python3. No data will be captured.
xFormers 0.0.24+042abc8.d20240104
memory_efficient_attention.cutlassF:               available
memory_efficient_attention.cutlassB:               available
memory_efficient_attention.decoderF:               available
[email protected]:        available
[email protected]:        available
memory_efficient_attention.smallkF:                available
memory_efficient_attention.smallkB:                available
memory_efficient_attention.tritonflashattF:        unavailable
memory_efficient_attention.tritonflashattB:        unavailable
memory_efficient_attention.triton_splitKF:         available
indexing.scaled_index_addF:                        available
indexing.scaled_index_addB:                        available
indexing.index_select:                             available
swiglu.dual_gemm_silu:                             available
swiglu.gemm_fused_operand_sum:                     available
swiglu.fused.p.cpp:                                available
is_triton_available:                               True
pytorch.version:                                   2.1.2+cu118
pytorch.cuda:                                      available
gpu.compute_capability:                            8.9
gpu.name:                                          NVIDIA GeForce RTX 4090
dcgm_profiler:                                     unavailable
build.info:                                        available
build.cuda_version:                                1108
build.python_version:                              3.10.11
build.torch_version:                               2.1.2+cu118
build.env.TORCH_CUDA_ARCH_LIST:                    None
build.env.XFORMERS_BUILD_TYPE:                     None
build.env.XFORMERS_ENABLE_DEBUG_ASSERTIONS:        None
build.env.NVCC_FLAGS:                              None
build.env.XFORMERS_PACKAGE_FROM:                   None
build.nvcc_version:                                11.8.89
source.privacy:                                    open source