You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
There are a few PyPI packages pixi can't install yet, where pip can.
Please paste your examples in this issue so we have a list of known packages we can track, test and benchmark with along the way.
Information we would like:
What did you run and what was the outcome?
e.g. pixi add --pypi packagex
If it doesn't recreate in an empty environment please share your pixi.toml that recreates the issue.
What error did pixi return?
e.g.
× RECORD file doesn't match wheel contents: missing hash for mediapipe/version.txt (expected sha256=-fE2KU)
Can pip install the package?
Does pip install packagex work?
What platform are you on?
e.g. linux-64
Did you find a workaround, if so please explain.
e.g. build it into a conda package, using a custom fork, etc.
Your input would greatly help us improve the pixi's experience!
Thanks in advance! ❤️
Here is the error I was seeing when trying to use pixi add:
WARN rattler_installs_packages::index::package_database: errors while processing source distributions:
× failed to resolve `pypi-dependencies`, due to underlying error
╰─▶ No metadata could be extracted for the following available artifacts:
- lxml-4.6.5.tar.gz
Error: × error while processing source distribution 'lxml-4.6.5.tar.gz':
│ could not build wheel: <string>:67: DeprecationWarning: pkg_resources is deprecated as an API. See https://
│ setuptools.pypa.io/en/latest/pkg_resources.html
│
help: Probably an error during processing of source distributions. Please check the error message above.
Posting in the discord channel got it working with this fix:
pixi add python lxml
pixi add --pypi elsie
The reason for this is:
there seems to be an issue with one of the lxml source distributions. Meaning you have to locally build the package but that is not working because of the error it is giving you. pixi reports back on errorous packages where pip probably continues trying other versions.
pixi add duckdb ❌ pixi add --pypi duckdb ❌ vs pixi run pip install duckdb ✅ on Windows
PS C:\code> pixi init duckdb-pip
✔ Initialized project in C:\code\duckdb-pip
PS C:\code> cd .\duckdb-pip\
PS C:\code\duckdb-pip> pixi add python=3.11
✔ Added python=3.11
PS C:\code\duckdb-pip> pixi add duckdb
× could not determine any available versions for duckdb on win-64. Either the package could not be found or version constraints on other
│ dependencies result in a conflict.
╰─▶ Cannot solve the request because of: No candidates were found for duckdb *.
PS C:\code\duckdb-pip> pixi add --pypi duckdb
× could not build wheel: warning: no files found matching '*.h' under directory 'duckdb'
│ warning: no files found matching '*.hpp' under directory 'duckdb'
│ warning: no files found matching '*.cpp' under directory 'duckdb'
│ warning: no files found matching '*.h' under directory 'src'
│ warning: manifest_maker: MANIFEST.in, line 6: 'recursive-include' expects <dir> <pattern1> <pattern2> ...
│
│ C:\Users\AKISLU~1\AppData\Local\Temp\.tmpkvrxrc\venv\Lib\site-packages\setuptools\command\build_py.py:207: _Warning: Package 'duckdb-
│ stubs.value' is absent from the `packages` configuration.
│ !!
│
│ ********************************************************************************
│ ############################
│ # Package would be ignored #
│ ############################
│ Python recognizes 'duckdb-stubs.value' as an importable package[^1],
│ but it is absent from setuptools' `packages` configuration.
│
│ This leads to an ambiguous overall configuration. If you want to distribute this
│ package, please make sure that 'duckdb-stubs.value' is explicitly added
│ to the `packages` configuration field.
│
│ Alternatively, you can also rely on setuptools' discovery methods
│ (for example by using `find_namespace_packages(...)`/`find_namespace:`
│ instead of `find_packages(...)`/`find:`).
│
│ You can read more about "package discovery" on setuptools documentation page:
│
│ - https://setuptools.pypa.io/en/latest/userguide/package_discovery.html
│
│ If you don't want 'duckdb-stubs.value' to be distributed and are
│ already explicitly excluding 'duckdb-stubs.value' via
│ `find_namespace_packages(...)/find_namespace` or `find_packages(...)/find`,
│ you can try to use `exclude_package_data`, or `include-package-data=False` in
│ combination with a more fine grained `package-data` configuration.
│
│ You can read more about "package data files" on setuptools documentation page:
│
│ - https://setuptools.pypa.io/en/latest/userguide/datafiles.html
│
│
│ [^1]: For Python, any directory (with suitable naming) can be imported,
│ even if it does not contain any `.py` files.
│ On the other hand, currently there is no concept of package data
│ directory, all directories are treated like packages.
│ ********************************************************************************
│
│ !!
│ check.warn(importable)
│ C:\Users\AKISLU~1\AppData\Local\Temp\.tmpkvrxrc\venv\Lib\site-packages\setuptools\command\build_py.py:207: _Warning: Package 'duckdb-
│ stubs.value.constant' is absent from the `packages` configuration.
│ !!
│
│ ********************************************************************************
│ ############################
│ # Package would be ignored #
│ ############################
│ Python recognizes 'duckdb-stubs.value.constant' as an importable package[^1],
│ but it is absent from setuptools' `packages` configuration.
│
│ This leads to an ambiguous overall configuration. If you want to distribute this
│ package, please make sure that 'duckdb-stubs.value.constant' is explicitly added
│ to the `packages` configuration field.
│
│ Alternatively, you can also rely on setuptools' discovery methods
│ (for example by using `find_namespace_packages(...)`/`find_namespace:`
│ instead of `find_packages(...)`/`find:`).
│
│ You can read more about "package discovery" on setuptools documentation page:
│
│ - https://setuptools.pypa.io/en/latest/userguide/package_discovery.html
│
│ If you don't want 'duckdb-stubs.value.constant' to be distributed and are
│ already explicitly excluding 'duckdb-stubs.value.constant' via
│ `find_namespace_packages(...)/find_namespace` or `find_packages(...)/find`,
│ you can try to use `exclude_package_data`, or `include-package-data=False` in
│ combination with a more fine grained `package-data` configuration.
│
│ You can read more about "package data files" on setuptools documentation page:
│
│ - https://setuptools.pypa.io/en/latest/userguide/datafiles.html
│
│
│ [^1]: For Python, any directory (with suitable naming) can be imported,
│ even if it does not contain any `.py` files.
│ On the other hand, currently there is no concept of package data
│ directory, all directories are treated like packages.
│ ********************************************************************************
│
│ !!
│ check.warn(importable)
│ C:\Users\AKISLU~1\AppData\Local\Temp\.tmpkvrxrc\venv\Lib\site-packages\setuptools\command\build_py.py:207: _Warning: Package
│ 'duckdb.experimental' is absent from the `packages` configuration.
│ !!
│
│ ********************************************************************************
│ ############################
│ # Package would be ignored #
│ ############################
│ Python recognizes 'duckdb.experimental' as an importable package[^1],
│ but it is absent from setuptools' `packages` configuration.
│
│ This leads to an ambiguous overall configuration. If you want to distribute this
│ package, please make sure that 'duckdb.experimental' is explicitly added
│ to the `packages` configuration field.
│
│ Alternatively, you can also rely on setuptools' discovery methods
│ (for example by using `find_namespace_packages(...)`/`find_namespace:`
│ instead of `find_packages(...)`/`find:`).
│
│ You can read more about "package discovery" on setuptools documentation page:
│
│ - https://setuptools.pypa.io/en/latest/userguide/package_discovery.html
│
│ If you don't want 'duckdb.experimental' to be distributed and are
│ already explicitly excluding 'duckdb.experimental' via
│ `find_namespace_packages(...)/find_namespace` or `find_packages(...)/find`,
│ you can try to use `exclude_package_data`, or `include-package-data=False` in
│ combination with a more fine grained `package-data` configuration.
│
│ You can read more about "package data files" on setuptools documentation page:
│
│ - https://setuptools.pypa.io/en/latest/userguide/datafiles.html
│
│
│ [^1]: For Python, any directory (with suitable naming) can be imported,
│ even if it does not contain any `.py` files.
│ On the other hand, currently there is no concept of package data
│ directory, all directories are treated like packages.
│ ********************************************************************************
│
│ !!
│ check.warn(importable)
│ error: Microsoft Visual C++ 14.0 or greater is required. Get it with "Microsoft C++ Build Tools": https://visualstudio.microsoft.com/
│ visual-cpp-build-tools/
│
PS C:\code\duckdb-pip> pixi add pip
✔ Added pip
PS C:\code\duckdb-pip> pixi run pip install duckdb
Collecting duckdb
Downloading duckdb-0.9.2-cp311-cp311-win_amd64.whl.metadata (798 bytes)
Downloading duckdb-0.9.2-cp311-cp311-win_amd64.whl (10.3 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 10.3/10.3 MB 21.8 MB/s eta 0:00:00
Installing collected packages: duckdb
Successfully installed duckdb-0.9.2
work around was just to use pip to install using pixi tasks
@pablovela5620 I am also finding this necessary quite often, probably more often than @ruben-arts would like :)
In this pattern the environment definition is fragmented between pixi.toml tasks and pixi.lock. It could work, esp with pinned pip installs in tasks, but is it the intention?
In this pattern the environment definition is fragmented between pixi.toml tasks and pixi.lock. It could work, esp with pinned pip installs in tasks, but is it the intention?
@liquidcarbon This is indeed not what we want as UX but we simply need to develop more to support all PyPI packages. Its a weird bunch of requirements we have to support to be equivalent to pip. So please keep posting non working packages!
@liquidcarbon the duckdb on conda-forge has been continued on python-duckdb which should be available on windows. We'll keep the example for pypi to test more!
@tylerjw actually it also fails for the same version in pip on apple silicon, like pip we error out when we cannot build the first source distribution.
@pablovela5620 so it seems mediapipe 10.9 is a package with an invalid RECORD file, I manually checked it and it's incorrect.
This is mentioned in the PyPa
During extraction, wheel installers verify all the hashes in RECORD against the file contents. Apart from RECORD and its signatures, installation will fail if any file in the archive is not both mentioned and correctly hashed in RECORD.
Mediapipe has a version.txt that is not mentioned in the RECORD for the 10.9 release.
Which in this case triggers the error, I'm unsure why pip does no do this. But I feel its good to adhere to the standard here.
In any case, mediapipe 10.8 does seem to work, you could use that instead.
@liquidcarbon I cannot reproduce this, neither on OSX or on windows. It's strange it does not select the .whl for some reason.
@tdejager Just tried on Win10, original comment was on Win11
same thing:
PS C:\Users\a\Desktop\code\duckdb-pip> pixi add python=3.11
✔ Added python 3.11.*
PS C:\Users\a\Desktop\code\duckdb-pip> pixi add duckdb
× could not determine any available versions for duckdb on win-64. Either the package could not be found or version
│ constraints on other dependencies result in a conflict.
╰─▶ Cannot solve the request because of: No candidates were found for duckdb *.
pixi init tf-metal &&cd tf-metal
pixi add "python>=3.11""tensorflow>=2.13" pip
pixi add --pypi tensorflow-metal
> × failed to resolve `pypi-dependencies`, due to underlying error
╰─▶ The following packages are incompatible
└─ tensorflow-metal * cannot be installed because there are no viable options:
└─ tensorflow-metal 0.1.0 | 0.1.1 | 0.1.2 | 0.2.0 | 0.3.0 | 0.4.0 | 0.5.0 | 0.5.1 | 0.6.0 | 0.7.0 | 0.7.1
| 0.8.0 | 1.0.0 | 1.0.1 | 1.1.0 is excluded because none of the artifacts are compatible with the Python
interpreter or glibc version and there are no supported sdists
pixi run pip install tensorflow-metal
> Successfully installed tensorflow-metal-1.1.0
having some weirdness with rerun-sdk, feedstock isnt yet merged for latest 0.22.1 update. If I try to add via cli
(pi0-lerobot) ⚡ hocap-dataset ~/pi0-lerobot pixi add rerun-sdk==0.22.1 --pypi
Error:
× failed to solve the pypi requirements of 'default' 'linux-64'
├─▶ failed to resolve pypi dependencies
╰─▶ Because rerun-sdk==0.22.1 has no wheels with a matching platform tag (e.g., `manylinux_2_28_x86_64`) and you require rerun-sdk==0.22.1, we
can conclude that your requirements are unsatisfiable.
hint: Wheels are available for `rerun-sdk` (v0.22.1) on the following platforms: `manylinux_2_31_aarch64`, `manylinux_2_31_x86_64`,
`macosx_10_12_x86_64`, `macosx_11_0_arm64`, `win_amd64`
Error:
× failed to solve the pypi requirements of 'default' 'linux-64'
├─▶ failed to resolve pypi dependencies
╰─▶ Because only the following versions of rerun-sdk are available:
rerun-sdk<=0.21.0
rerun-sdk==0.22.0
rerun-sdk==0.22.1
and rerun-sdk>=0.21.0 has no wheels with a matching platform tag (e.g., `manylinux_2_28_x86_64`), we can conclude that rerun-sdk>=0.21.0
cannot be used.
And because lerobot==0.1.0 depends on rerun-sdk>=0.21.0, we can conclude that lerobot==0.1.0 cannot be used.
And because only lerobot[pusht]==0.1.0 is available and you require lerobot[pusht], we can conclude that your requirements are unsatisfiable.
hint: Pre-releases are available for `rerun-sdk` in the requested range (e.g., 0.22.1rc1), but pre-releases weren't enabled (try:
`--prerelease=allow`)
hint: Wheels are available for `rerun-sdk` (v0.22.1) on the following platforms: `manylinux_2_31_aarch64`, `manylinux_2_31_x86_64`,
`macosx_10_12_x86_64`, `macosx_11_0_arm64`, `win_amd64`
[workspace]
channels = ["conda-forge"]
name = "pixi-example"platforms = ["win-64"]
version = "0.1.0"
[tasks]
[dependencies]
python = ">=3.11,<3.13"pip = ">=25.0.1,<26"setuptools = ">=75.8.2,<76"
[pypi-dependencies]
torch = { version = "<2.6", index = "https://download.pytorch.org/whl/cu124" }
torchvision = { version = "<0.21", index = "https://download.pytorch.org/whl/cu124" }
torchaudio = { version = "<2.6", index = "https://download.pytorch.org/whl/cu124" }
packaging = ">=24.2, <25"einops = ">=0.8.1, <0.9"
What error did pixi return?
Error: × Failed to update PyPI packages for environment 'default'
├─▶ Failed to prepare distributions
├─▶ Failed to build `flash-attn==2.7.4.post1`
├─▶ The build backend returned an error
╰─▶ Call to `setuptools.build_meta:__legacy__.build_wheel` failed (exit code: 1)
[stderr]
Traceback (most recent call last):
File "<string>", line 14, in <module>
File "C:\Users\***\AppData\Local\rattler\cache\uv-cache\builds-v0\.tmp36pAkT\Lib\site-packages\setuptools\build_meta.py", line 334, in get_requires_for_build_wheel
return self._get_build_requires(config_settings, requirements=[])
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\***\AppData\Local\rattler\cache\uv-cache\builds-v0\.tmp36pAkT\Lib\site-packages\setuptools\build_meta.py", line 304, in _get_build_requires
self.run_setup()
File "C:\Users\***\AppData\Local\rattler\cache\uv-cache\builds-v0\.tmp36pAkT\Lib\site-packages\setuptools\build_meta.py", line 522, in run_setup
super().run_setup(setup_script=setup_script)
File "C:\Users\***\AppData\Local\rattler\cache\uv-cache\builds-v0\.tmp36pAkT\Lib\site-packages\setuptools\build_meta.py", line 320, in run_setup
exec(code, locals())
File "<string>", line 22, in <module>
ModuleNotFoundError: No module named 'torch'
hint: This error likely indicates that `[email protected]` depends on `torch`, but doesn't declare it as a build dependency. If `flash-attn` is a first-party package, consider adding `torch` to its `build-system.requires`. Otherwise, `uv
pip install torch` into the environment and re-run with `--no-build-isolation`.
import torch works of course using pixi run python
Can pip install the package?
It works in pure conda env, but does not work with pip inside the pixi (pixi run pip install ...) resulting with ImportError: cannot import name 'setup' from 'setuptools' (unknown location).
What platform are you on?
win-64
Did you find a workaround, if so please explain.
I guess one could install torch using conda packages (pytorch-gpu, etc) but there is no torchaudio for win-64 and I couldn't manage to install torchaudio separately using --pypi either. So no workaround so far.
hint: This error likely indicates that `[email protected]` depends on `torch`, but doesn't declare it as a build dependency. If `flash-attn` is a first-party package, consider adding `torch` to its `build-system.requires`. Otherwise, `uv
pip install torch` into the environment and re-run with `--no-build-isolation`.
Probably you can disable build-isolation for flash-attn
Thank you, it helped. I overlooked the thing about build isolation, got confused after seeing uv mentioned, sorry. It would be great though if there would be a message from pixi with the link you provided, but not sure it worth investing.
When trying to install a pyproject.toml project where a dependency depends on pywin32, pixi will throw a failed to resolve pypi dependencies error, even if the dependency was marked with sys_platform == 'win32'
❯ pixi install --all
Error: × failed to solve the pypi requirements of 'default''linux-64'
├─▶ failed to resolve pypi dependencies
╰─▶ Because only the following versions of pywin32 are available:
pywin32<=223
pywin32==224
pywin32==225
pywin32==226
pywin32==227
pywin32==228
pywin32==300
pywin32==301
pywin32==302
pywin32==303
pywin32==304
pywin32==305
pywin32==306
pywin32==307
pywin32==308
pywin32==309
pywin32==310
and pywin32>=223,<=305 has no wheels with a matching Python ABI tag (e.g., `cp312`), we can conclude that pywin32>=223,<=305 cannot
be used.
And because pywin32>=306 has no wheels with a matching platform tag (e.g., `manylinux_2_28_x86_64`) and httpx-negotiate-sspi==0.28.1
depends on pywin32>=223, we can conclude that httpx-negotiate-sspi==0.28.1 cannot be used.
And because only httpx-negotiate-sspi<=0.28.1 is available and you require httpx-negotiate-sspi>=0.28.1, we can conclude that your
requirements are unsatisfiable.
hint: You require CPython 3.12 (`cp312`), but we only found wheels for`pywin32` (v305) with the following Python ABI tags: `cp36m`,
`cp37m`, `cp38`, `cp39`, `cp310`, `cp311`
hint: Wheels are available for`pywin32` (v310) on the following platforms: `win32`, `win_amd64`, `win_arm64`
Note
This is when trying to install the project from a linux box.
I have to have things working on both Windows and Linux so this one is definitely biting me.
It's not a show-stopper though - I just make those deps [tool.pixi.target.win-64.dependencies] which works fine for pixi envs. Things are a little broken for pip users, but I'm trying to discourage the use of pip anyway! 🤣
I have run into the same issue related to a "pywin32; sys_platform == 'win32'" dependency, but I need to continue supporting pip users and producing pip packages. There isn't really a workaround as far as I can tell other than potentially moving away from using the pyproject.toml support and having a separate pixi.toml with relevant information duplicated across them.
Activity
pablovela5620 commentedon Feb 2, 2024
Mediapipe
× RECORD file doesn't match wheel contents: missing hash for mediapipe/version.txt (expected sha256=-fE2KU)python -m pip install mediapipedoes workosx-arm64tylerjw commentedon Feb 4, 2024
I wanted to use this library that is packaged in pypi: https://github.com/spirali/elsie
Here is the error I was seeing when trying to use pixi add:
Posting in the discord channel got it working with this fix:
The reason for this is:
liquidcarbon commentedon Feb 12, 2024
pixi add duckdb❌pixi add --pypi duckdb❌ vspixi run pip install duckdb✅ on Windowsliquidcarbon commentedon Feb 12, 2024
@pablovela5620 I am also finding this necessary quite often, probably more often than @ruben-arts would like :)
In this pattern the environment definition is fragmented between
pixi.tomltasks andpixi.lock. It could work, esp with pinned pip installs in tasks, but is it the intention?ruben-arts commentedon Feb 12, 2024
@liquidcarbon This is indeed not what we want as UX but we simply need to develop more to support all PyPI packages. Its a weird bunch of requirements we have to support to be equivalent to
pip. So please keep posting non working packages!ruben-arts commentedon Feb 12, 2024
@liquidcarbon the
duckdbonconda-forgehas been continued onpython-duckdbwhich should be available on windows. We'll keep the example for pypi to test more!tdejager commentedon Feb 15, 2024
@liquidcarbon I cannot reproduce this, neither on OSX or on windows. It's strange it does not select the
.whlfor some reason.tdejager commentedon Feb 15, 2024
@tylerjw actually it also fails for the same version in pip on apple silicon, like pip we error out when we cannot build the first source distribution.
tdejager commentedon Feb 15, 2024
@pablovela5620 so it seems mediapipe 10.9 is a package with an invalid RECORD file, I manually checked it and it's incorrect.
This is mentioned in the PyPa
Mediapipe has a
version.txtthat is not mentioned in the RECORD for the 10.9 release.Which in this case triggers the error, I'm unsure why pip does no do this. But I feel its good to adhere to the standard here.
In any case, mediapipe 10.8 does seem to work, you could use that instead.
Also see: google-ai-edge/mediapipe#5025
fix: issue with zip and artifact:from_file (#221)
liquidcarbon commentedon Feb 15, 2024
@tdejager Just tried on Win10, original comment was on Win11
same thing:
PS C:\Users\a\Desktop\code\duckdb-pip> pixi add python=3.11
✔ Added python 3.11.*
PS C:\Users\a\Desktop\code\duckdb-pip> pixi add duckdb
× could not determine any available versions for duckdb on win-64. Either the package could not be found or version
│ constraints on other dependencies result in a conflict.
╰─▶ Cannot solve the request because of: No candidates were found for duckdb *.
I was on Pixi 0.9.1, upped to 0.13.0 - same thing
But
pixi add --pypi duckdbworks. 🤷♂️ruben-arts commentedon Feb 15, 2024
The conda package you add with
pixi add duckdbshould bepixi add python-duckdbhttps://prefix.dev/channels/conda-forge/packages/python-duckdb
liquidcarbon commentedon Feb 15, 2024
@ruben-arts noted -- I just wasn't sure which part @tdejager was trying to reproduce
awray3 commentedon Feb 25, 2024
Tensorflow metal on Apple silicon MacOS 14.2.1:
baszalmstra commentedon Feb 25, 2024
Most likely you are missing a system requirement: https://pixi.sh/latest/configuration/#the-system-requirements-table
Most likely macos=12.0
58 remaining items
ruben-arts commentedon Feb 18, 2025
@Goooyi
pixi globaldoesn't supportpypidependencies: #2261pablovela5620 commentedon Feb 25, 2025
having some weirdness with rerun-sdk, feedstock isnt yet merged for latest 0.22.1 update. If I try to add via cli
but if I directly link to the url
it works just fine
traversaro commentedon Feb 25, 2025
Kind OT, but anyhow: I pinged in conda-forge/rerun-sdk-feedstock#61, feel free to ping yourself in the future if something like this happens, thanks!
pablovela5620 commentedon Feb 25, 2025
Yes, I should have started with that!
dlyz commentedon Apr 3, 2025
What did you run and what was the outcome?
Command:
pixi add --pypi flash-attnpixi.toml
What error did pixi return?
Error: × Failed to update PyPI packages for environment 'default' ├─▶ Failed to prepare distributions ├─▶ Failed to build `flash-attn==2.7.4.post1` ├─▶ The build backend returned an error ╰─▶ Call to `setuptools.build_meta:__legacy__.build_wheel` failed (exit code: 1) [stderr] Traceback (most recent call last): File "<string>", line 14, in <module> File "C:\Users\***\AppData\Local\rattler\cache\uv-cache\builds-v0\.tmp36pAkT\Lib\site-packages\setuptools\build_meta.py", line 334, in get_requires_for_build_wheel return self._get_build_requires(config_settings, requirements=[]) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\***\AppData\Local\rattler\cache\uv-cache\builds-v0\.tmp36pAkT\Lib\site-packages\setuptools\build_meta.py", line 304, in _get_build_requires self.run_setup() File "C:\Users\***\AppData\Local\rattler\cache\uv-cache\builds-v0\.tmp36pAkT\Lib\site-packages\setuptools\build_meta.py", line 522, in run_setup super().run_setup(setup_script=setup_script) File "C:\Users\***\AppData\Local\rattler\cache\uv-cache\builds-v0\.tmp36pAkT\Lib\site-packages\setuptools\build_meta.py", line 320, in run_setup exec(code, locals()) File "<string>", line 22, in <module> ModuleNotFoundError: No module named 'torch' hint: This error likely indicates that `[email protected]` depends on `torch`, but doesn't declare it as a build dependency. If `flash-attn` is a first-party package, consider adding `torch` to its `build-system.requires`. Otherwise, `uv pip install torch` into the environment and re-run with `--no-build-isolation`.import torchworks of course usingpixi run pythonCan pip install the package?
It works in pure conda env, but does not work with pip inside the pixi (
pixi run pip install ...) resulting withImportError: cannot import name 'setup' from 'setuptools' (unknown location).What platform are you on?
win-64
Did you find a workaround, if so please explain.
I guess one could install torch using conda packages (
pytorch-gpu, etc) but there is no torchaudio for win-64 and I couldn't manage to install torchaudio separately using --pypi either. So no workaround so far.traversaro commentedon Apr 3, 2025
The hint is relevant:
Probably you can disable build-isolation for
flash-attn, see https://pixi.sh/latest/reference/pixi_manifest/#no-build-isolation ? Note that in that case probably you also nee to install other build dependencies to the environment explicitly.dlyz commentedon Apr 3, 2025
Thank you, it helped. I overlooked the thing about build isolation, got confused after seeing uv mentioned, sorry. It would be great though if there would be a message from pixi with the link you provided, but not sure it worth investing.
dhirschfeld commentedon Apr 9, 2025
When trying to install a
pyproject.tomlproject where a dependency depends onpywin32,pixiwill throw afailed to resolve pypi dependencieserror, even if the dependency was marked withsys_platform == 'win32'Example
pyproject.tomlNote
This is when trying to install the project from a linux box.
tdejager commentedon Apr 10, 2025
Yes, you are right this is kind of a big issue we should pass the environment markers correctly to uv. Which we currently do not seem to do.
dhirschfeld commentedon Apr 10, 2025
I have to have things working on both Windows and Linux so this one is definitely biting me.
It's not a show-stopper though - I just make those deps
[tool.pixi.target.win-64.dependencies]which works fine forpixienvs. Things are a little broken forpipusers, but I'm trying to discourage the use ofpipanyway! 🤣tdejager commentedon Apr 10, 2025
Aha okay! Good to know 😄
atonderski commentedon Apr 26, 2025
Also ran into this issue, and the proposed workaround breaks compatibility with for example direct
uv pip install.Any estimate on if/when this will be fixed?
kkraus14 commentedon Jul 25, 2025
I have run into the same issue related to a
"pywin32; sys_platform == 'win32'"dependency, but I need to continue supporting pip users and producing pip packages. There isn't really a workaround as far as I can tell other than potentially moving away from using thepyproject.tomlsupport and having a separatepixi.tomlwith relevant information duplicated across them.