Can't recreate a mamba environment with pixi (getting ModuleNotFoundError: No module named 'torch')

[asmith26]

Checks

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pixi, using pixi --version.

Reproducible example

Create a pixi.toml (I'm essentially trying to create an env for florence2):

[project]
authors = ["me <me@users.noreply.github.com>"]
channels = ["nvidia", "conda-forge"]
description = "Add a short description here"
name = "florence2"
platforms = ["linux-64"]
version = "0.1.0"

[dependencies]
python = "*"
cuda-toolkit = "*"
cuda = "*"

[pypi-dependencies]
jupyterlab = "*"
transformers = "*"
datasets = "*"
matplotlib = "*"
torch = "*"
torchvision = "*"
requests = "*"
timm = "*"
flash-attn = "*"
einops = "*"
accelerate = "*"
pillow = "*"

Then running pixi install yields:

  ⠴ default:linux-64     [00:00:03] resolving flash-attn==2.6.3                                           × failed to solve the pypi requirements of 'default' 'linux-64'
  ├─▶ failed to resolve pypi dependencies
  ├─▶ Failed to download and build `flash-attn==2.6.3`
  ├─▶ Failed to build: `flash-attn==2.6.3`
  ╰─▶ Build backend failed to determine extra requires with `build_wheel()` with exit status: 1
      --- stdout:
      
      --- stderr:
      Traceback (most recent call last):
        File "<string>", line 14, in <module>
        File "~/pixi_cache/uv-cache/environments-v0/.tmp6AntAq/lib/python3.12/site-
      packages/setuptools/build_meta.py", line 332, in get_requires_for_build_wheel
          return self._get_build_requires(config_settings, requirements=[])
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
        File "~/pixi_cache/uv-cache/environments-v0/.tmp6AntAq/lib/python3.12/site-
      packages/setuptools/build_meta.py", line 302, in _get_build_requires
          self.run_setup()
        File "~/pixi_cache/uv-cache/environments-v0/.tmp6AntAq/lib/python3.12/site-
      packages/setuptools/build_meta.py", line 502, in run_setup
          super().run_setup(setup_script=setup_script)
        File "~/pixi_cache/uv-cache/environments-v0/.tmp6AntAq/lib/python3.12/site-
      packages/setuptools/build_meta.py", line 318, in run_setup
          exec(code, locals())
        File "<string>", line 21, in <module>
      ModuleNotFoundError: No module named 'torch'
      ---

Issue description

Possibly related to #1834

Expected behavior

I originally created the environment with mamba and it successfully installs/runs. Here is the exported environment.yaml

name: florence2
channels:
  - nvidia
  - conda-forge
dependencies:
  - _libgcc_mutex=0.1=conda_forge
  - _openmp_mutex=4.5=2_gnu
  - _sysroot_linux-64_curr_repodata_hack=3=h69a702a_16
  - binutils_impl_linux-64=2.40=ha1999f0_7
  - binutils_linux-64=2.40=hb3c18ed_0
  - bzip2=1.0.8=h4bc722e_7
  - ca-certificates=2024.7.4=hbcca054_0
  - cuda=12.6.0=0
  - cuda-cccl=12.6.37=0
  - cuda-cccl_linux-64=12.6.37=0
  - cuda-command-line-tools=12.6.0=0
  - cuda-compiler=12.6.0=0
  - cuda-crt-dev_linux-64=12.6.20=ha770c72_0
  - cuda-crt-tools=12.6.20=ha770c72_0
  - cuda-cudart=12.6.37=0
  - cuda-cudart-dev=12.6.37=0
  - cuda-cudart-dev_linux-64=12.6.37=0
  - cuda-cudart-static=12.6.37=0
  - cuda-cudart-static_linux-64=12.6.37=0
  - cuda-cudart_linux-64=12.6.37=0
  - cuda-cuobjdump=12.6.20=0
  - cuda-cupti=12.6.37=0
  - cuda-cupti-dev=12.6.37=0
  - cuda-cuxxfilt=12.6.20=0
  - cuda-driver-dev=12.6.37=0
  - cuda-driver-dev_linux-64=12.6.37=0
  - cuda-gdb=12.6.37=0
  - cuda-libraries=12.6.0=0
  - cuda-libraries-dev=12.6.0=0
  - cuda-nsight=12.6.20=0
  - cuda-nvcc=12.6.20=0
  - cuda-nvcc-dev_linux-64=12.6.20=he91c749_0
  - cuda-nvcc-impl=12.6.20=h85509e4_0
  - cuda-nvcc-tools=12.6.20=he02047a_0
  - cuda-nvcc_linux-64=12.6.20=0
  - cuda-nvdisasm=12.6.20=0
  - cuda-nvml-dev=12.6.37=2
  - cuda-nvprof=12.6.37=0
  - cuda-nvprune=12.6.20=0
  - cuda-nvrtc=12.6.20=0
  - cuda-nvrtc-dev=12.6.20=0
  - cuda-nvtx=12.6.37=0
  - cuda-nvvm-dev_linux-64=12.6.20=ha770c72_0
  - cuda-nvvm-impl=12.6.20=he02047a_0
  - cuda-nvvm-tools=12.6.20=he02047a_0
  - cuda-nvvp=12.6.37=0
  - cuda-opencl=12.6.37=0
  - cuda-opencl-dev=12.6.37=0
  - cuda-profiler-api=12.6.37=0
  - cuda-runtime=12.6.0=0
  - cuda-sanitizer-api=12.6.34=0
  - cuda-toolkit=12.6.0=0
  - cuda-tools=12.6.0=0
  - cuda-version=12.6=3
  - cuda-visual-tools=12.6.0=0
  - gcc_impl_linux-64=13.3.0=hfea6d02_0
  - gcc_linux-64=13.3.0=hc28eda2_0
  - gds-tools=1.11.0.15=0
  - gmp=6.3.0=hac33072_2
  - gxx_impl_linux-64=13.3.0=hdbfa832_0
  - gxx_linux-64=13.3.0=h6834431_0
  - kernel-headers_linux-64=3.10.0=h4a8ded7_16
  - ld_impl_linux-64=2.40=hf3520f5_7
  - libcublas=12.6.0.22=0
  - libcublas-dev=12.6.0.22=0
  - libcufft=11.2.6.28=0
  - libcufft-dev=11.2.6.28=0
  - libcufile=1.11.0.15=0
  - libcufile-dev=1.11.0.15=0
  - libcurand=10.3.7.37=0
  - libcurand-dev=10.3.7.37=0
  - libcusolver=11.6.4.38=0
  - libcusolver-dev=11.6.4.38=0
  - libcusparse=12.5.2.23=0
  - libcusparse-dev=12.5.2.23=0
  - libexpat=2.6.2=h59595ed_0
  - libffi=3.4.2=h7f98852_5
  - libgcc-devel_linux-64=13.3.0=h84ea5a7_100
  - libgcc-ng=14.1.0=h77fa898_0
  - libgomp=14.1.0=h77fa898_0
  - libnpp=12.3.1.23=0
  - libnpp-dev=12.3.1.23=0
  - libnsl=2.0.1=hd590300_0
  - libnvfatbin=12.6.20=0
  - libnvfatbin-dev=12.6.20=0
  - libnvjitlink=12.6.20=0
  - libnvjitlink-dev=12.6.20=0
  - libnvjpeg=12.3.3.23=0
  - libnvjpeg-dev=12.3.3.23=0
  - libsanitizer=13.3.0=heb74ff8_0
  - libsqlite=3.46.0=hde9e2c9_0
  - libstdcxx-devel_linux-64=13.3.0=h84ea5a7_100
  - libstdcxx-ng=14.1.0=hc0a3c3a_0
  - libuuid=2.38.1=h0b41bf4_0
  - libxcrypt=4.4.36=hd590300_1
  - libzlib=1.3.1=h4ab18f5_1
  - ncurses=6.5=h59595ed_0
  - nsight-compute=2024.1.1.4=0
  - openssl=3.3.1=h4bc722e_2
  - pip=24.2=pyhd8ed1ab_0
  - python=3.12.5=h2ad013b_0_cpython
  - readline=8.2=h8228510_1
  - setuptools=72.2.0=pyhd8ed1ab_0
  - sysroot_linux-64=2.17=h4a8ded7_16
  - tk=8.6.13=noxft_h4845f30_101
  - wheel=0.44.0=pyhd8ed1ab_0
  - xz=5.2.6=h166bdaf_0
  - pip:
      - accelerate==0.33.0
      - aiohappyeyeballs==2.4.0
      - aiohttp==3.10.5
      - aiosignal==1.3.1
      - attrs==24.2.0
      - certifi==2024.7.4
      - charset-normalizer==3.3.2
      - datasets==2.21.0
      - dill==0.3.8
      - einops==0.8.0
      - filelock==3.15.4
      - flash-attn==2.6.3
      - frozenlist==1.4.1
      - fsspec==2024.6.1
      - huggingface-hub==0.24.6
      - idna==3.7
      - jinja2==3.1.4
      - markupsafe==2.1.5
      - mpmath==1.3.0
      - multidict==6.0.5
      - multiprocess==0.70.16
      - networkx==3.3
      - numpy==1.26.4
      - nvidia-cublas-cu12==12.1.3.1
      - nvidia-cuda-cupti-cu12==12.1.105
      - nvidia-cuda-nvrtc-cu12==12.1.105
      - nvidia-cuda-runtime-cu12==12.1.105
      - nvidia-cudnn-cu12==9.1.0.70
      - nvidia-cufft-cu12==11.0.2.54
      - nvidia-curand-cu12==10.3.2.106
      - nvidia-cusolver-cu12==11.4.5.107
      - nvidia-cusparse-cu12==12.1.0.106
      - nvidia-nccl-cu12==2.20.5
      - nvidia-nvjitlink-cu12==12.6.20
      - nvidia-nvtx-cu12==12.1.105
      - packaging==24.1
      - pandas==2.2.2
      - pillow==10.4.0
      - psutil==6.0.0
      - pyarrow==17.0.0
      - python-dateutil==2.9.0.post0
      - pytz==2024.1
      - pyyaml==6.0.2
      - regex==2024.7.24
      - requests==2.32.3
      - safetensors==0.4.4
      - six==1.16.0
      - sympy==1.13.2
      - timm==1.0.8
      - tokenizers==0.19.1
      - torch==2.4.0
      - torchvision==0.19.0
      - tqdm==4.66.5
      - transformers==4.44.1
      - triton==3.0.0
      - typing-extensions==4.12.2
      - tzdata==2024.1
      - urllib3==2.2.2
      - xxhash==3.5.0
      - yarl==1.9.4

[tdejager]

Yes this is related to this issue, that you mentioned :)

[tdejager]

We are in the process of getting no-build-isolation in :)

[tdejager]

Hey we've now added build-isolation options in main: https://pixi.sh/dev/reference/project_configuration/#no-build-isolation. Should be released really soon, let me know if it works and feel free to re-open if it does not.