Non-cuda installation error

[kunleng]

Hi there,

Thanks for making a great tool for folks to easily implement SAM! I am using a MacBook Pro (Intel processor) and ran into the error below on installation. Would appreciate any assistance!

Best,
Kun
UCSF

(samapi) kunleng@Kuns-MacBook-Pro ~ % python -m pip install git+https://github.com/ksugar/samapi.git
Collecting git+https://github.com/ksugar/samapi.git
Cloning https://github.com/ksugar/samapi.git to /private/var/folders/1v/tq470sc561vbjlwxjph3kbmm0000gn/T/pip-req-build-lmosy9s0
Running command git clone --filter=blob:none --quiet https://github.com/ksugar/samapi.git /private/var/folders/1v/tq470sc561vbjlwxjph3kbmm0000gn/T/pip-req-build-lmosy9s0
Resolved https://github.com/ksugar/samapi.git to commit 914703a
Installing build dependencies ... done
Getting requirements to build wheel ... done
Preparing metadata (pyproject.toml) ... done
Collecting mobile_sam@ git+https://github.com/ksugar/MobileSAM.git (from samapi==0.6.1)
Cloning https://github.com/ksugar/MobileSAM.git to /private/var/folders/1v/tq470sc561vbjlwxjph3kbmm0000gn/T/pip-install-ybt1e0uw/mobile-sam_3ae06ccdfc4042fbb1a09b3827480bb0
Running command git clone --filter=blob:none --quiet https://github.com/ksugar/MobileSAM.git /private/var/folders/1v/tq470sc561vbjlwxjph3kbmm0000gn/T/pip-install-ybt1e0uw/mobile-sam_3ae06ccdfc4042fbb1a09b3827480bb0
Resolved https://github.com/ksugar/MobileSAM.git to commit 41ac534a20a4779fab4f8c84e158ca7c7b40ff26
Preparing metadata (setup.py) ... done
Collecting sam2@ git+https://github.com/ksugar/segment-anything-2@support-non-cuda (from samapi==0.6.1)
Cloning https://github.com/ksugar/segment-anything-2 (to revision support-non-cuda) to /private/var/folders/1v/tq470sc561vbjlwxjph3kbmm0000gn/T/pip-install-ybt1e0uw/sam2_86192826f71646b68cc6ccfc01581ad2
Running command git clone --filter=blob:none --quiet https://github.com/ksugar/segment-anything-2 /private/var/folders/1v/tq470sc561vbjlwxjph3kbmm0000gn/T/pip-install-ybt1e0uw/sam2_86192826f71646b68cc6ccfc01581ad2
Running command git checkout -b support-non-cuda --track origin/support-non-cuda
Switched to a new branch 'support-non-cuda'
branch 'support-non-cuda' set up to track 'origin/support-non-cuda'.
Resolved https://github.com/ksugar/segment-anything-2 to commit 00ff21e508132e3b351d8c2c4b713970b03df001
Installing build dependencies ... error
error: subprocess-exited-with-error

× pip subprocess to install build dependencies did not run successfully.
│ exit code: 1
╰─> [4 lines of output]
Collecting setuptools>=61.0
Downloading setuptools-75.1.0-py3-none-any.whl.metadata (6.9 kB)
ERROR: Could not find a version that satisfies the requirement torch>=2.3.1 (from versions: 1.11.0, 1.12.0, 1.12.1, 1.13.0, 1.13.1, 2.0.0, 2.0.1, 2.1.0, 2.1.1, 2.1.2, 2.2.0, 2.2.1, 2.2.2)
ERROR: No matching distribution found for torch>=2.3.1
[end of output]

note: This error originates from a subprocess, and is likely not a problem with pip.
error: subprocess-exited-with-error

× pip subprocess to install build dependencies did not run successfully.
│ exit code: 1
╰─> See above for output.

note: This error originates from a subprocess, and is likely not a problem with pip.

[bertrandchauveau]

Hi kunleng,

Did you solve this issue? I have the exact same problem with my Apple Silicon.
In the meantime, pointing the version 0.4.1 during installation does work:

python -m pip install git+https://github.com/ksugar/samapi.git@v0.4.1

But it is not an answer to our problem.

@ksugar: could you take a look at this please? Seems to be a dependency conflict.

[ksugar]

Hi @kunleng and @bertrandchauveau, thank you for reporting the issue, and apologies for the delayed response.

I tested the installation on my Apple Silicon Mac (M2) using the exact commands provided in the README, and it completed successfully without any errors:

conda create -n samapi -y python=3.10
conda activate samapi
python -m pip install git+https://github.com/ksugar/samapi.git

Could you please share your system specifications by running the following command?

sw_vers

For reference, here’s the output on my system:

ProductName:		macOS
ProductVersion:		14.3.1
BuildVersion:		23D60

I hope this helps me gather the necessary information to resolve the issue quickly.

[bertrandchauveau]

Hi ksugar,

Thanks for your reply.
Here is the output:

ProductName:		macOS
ProductVersion:		15.1.1
BuildVersion:		24B2091

To complete my previous message, the installation runs successfully for version 0.4.1, but the issue occurs starting from v0.5

[bertrandchauveau]

Sorry for the double post, but I found out how to resolve the issue (at least mine).
As I can understand, it is due to a "confusion" between an arm64 and x86_64 conda environment, so for Mac users M chips versus Intel chips.

For some reasons, x86_64 seems to only support up to the 2.2.2 pytorch version.

This behavior of my installation of miniconda is probably due to the data transfer between my Intel Mac to my recent M chip Mac.
By forcing the creation of an arm64 conda environment, like this, I can confirm that the installation runs flawlessly:

CONDA_SUBDIR=osx-arm64 conda create -n samapi -y python=3.10

[kunleng]

Thanks for figuring that out! Unfortunately I have an Intel Mac and like you said it seems like PyTorch supports up to the 2.2.2 version :( Get Outlook for iOS<https://aka.ms/o0ukef>

…

________________________________ From: Bertrand Chauveau ***@***.***> Sent: Sunday, December 22, 2024 9:42:02 AM To: ksugar/samapi ***@***.***> Cc: Leng, Kun ***@***.***>; Mention ***@***.***> Subject: Re: [ksugar/samapi] Non-cuda installation error (Issue #29) Sorry for the double post, but I found out how to resolve the issue (at least mine). As I can understand, it is due to a "confusion" between an arm64 and x86_64 conda environment, so for Mac users M chips versus Intel chips. For some reasons, ZjQcmQRYFpfptBannerStart This Message Is From an External Sender This message came from outside your organization. ZjQcmQRYFpfptBannerEnd Sorry for the double post, but I found out how to resolve the issue (at least mine). As I can understand, it is due to a "confusion" between an arm64 and x86_64 conda environment, so for Mac users M chips versus Intel chips. For some reasons, x86_64 seems to only support up to the 2.2.2 pytorch version. This behavior of my installation of miniconda is probably due to the data transfer between my Intel Mac to my recent M chip Mac. By forcing the creation of an arm64 conda environment, like this, I can confirm that the installation runs flawlessly: CONDA_SUBDIR=osx-arm64 conda create -n samapi -y python=3.10 — Reply to this email directly, view it on GitHub<https://urldefense.com/v3/__https://github.com/ksugar/samapi/issues/29*issuecomment-2558533164__;Iw!!LQC6Cpwp!u25Acwc_QP44IRtx1biR6H-_nJAVYrI9WbUoe2-5X9Rpzj4kDdv8GnluTiiC8CfuNKcuLW1T1IWhoWa883rT9ld52Q$>, or unsubscribe<https://urldefense.com/v3/__https://github.com/notifications/unsubscribe-auth/AKDFHZPQ46D6XEJPSVXXHMT2G32WVAVCNFSM6AAAAABP4BZUH2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDKNJYGUZTGMJWGQ__;!!LQC6Cpwp!u25Acwc_QP44IRtx1biR6H-_nJAVYrI9WbUoe2-5X9Rpzj4kDdv8GnluTiiC8CfuNKcuLW1T1IWhoWa883qaKYdrMg$>. You are receiving this because you were mentioned.Message ID: ***@***.***>

[ksugar]

Hi @bertrandchauveau and @kunleng, thank you for identifying the cause and sharing the workaround.
The dependency on torch>=2.3.1 is inherited from segment-anything-2, and we need to consider whether it will work with version 2.2.2.

If anyone knows how to install SAM2 on an Intel Mac with dependencies, please share your insights.