cuequivariance_ops_cu12

[dhiraj82]

Hi I tried to install boltz by git clone https://github.com/jwohlwend/boltz.git cd boltz; pip install -e . but I am getting following error. How can I fix this issue?

ERROR: Could not find a version that satisfies the requirement cuequivariance_ops_cu12>=0.5.0 (from boltz) (from versions: none) ERROR: No matching distribution found for cuequivariance_ops_cu12>=0.5.0

Thank you Dhiraj

[jwohlwend]

Are you possibly in a python version that is too old?

[dhiraj82]

My python version is 3.10. when I used pip install boltz -U, it worked fine.

[AlexWei21]

Same problem here! I'm using python version 3.11

[kimdn]

Same problem here with python3.10

(uv_boltz_2025_07_02) [kimd999@deception03 boltz]$ uv pip install -e .
Using Python 3.10.18 environment at: /qfs/people/kimd999/uv/uv_boltz_2025_07_02/.venv
  × No solution found when resolving dependencies:
  ╰─▶ Because only the following versions of cuequivariance-ops-cu12 are available:
          cuequivariance-ops-cu12<=0.5.0
          cuequivariance-ops-cu12==0.5.1
      and cuequivariance-ops-cu12>=0.5.0 has no wheels with a matching platform tag (e.g., `manylinux_2_17_x86_64`), we can conclude that
      cuequivariance-ops-cu12>=0.5.0 cannot be used.
      And because boltz==2.1.1 depends on cuequivariance-ops-cu12>=0.5.0, we can conclude that boltz==2.1.1 cannot be used.
      And because only boltz==2.1.1 is available and you require boltz, we can conclude that your requirements are unsatisfiable.

      hint: Wheels are available for `cuequivariance-ops-cu12` (v0.5.1) on the following platforms: `manylinux_2_24_x86_64`,
      `manylinux_2_28_x86_64`, `manylinux_2_39_aarch64`

[Shiyi1031]

Same problem. I'm using python version 3.11...

[S-Watanabe0123]

I was able to fix this issue by using below.

pip install boltz[cuda11]==2.2.1

gpu==RTX4070 cuda version==12.9 python==3.12

This caused another issue during the run.

line 22, in kernel_triangular_mult
    from cuequivariance_torch.primitives.triangle import triangle_multiplicative_update
ModuleNotFoundError: No module named 'cuequivariance_torch'

I used --no_kernels then it ran successfully.

[GrootLiu]

I was able to fix this issue by using below.

pip install boltz[cuda11]==2.2.1

gpu==RTX4070 cuda version==12.9 python==3.12

This caused another issue during the run.

line 22, in kernel_triangular_mult
    from cuequivariance_torch.primitives.triangle import triangle_multiplicative_update
ModuleNotFoundError: No module named 'cuequivariance_torch'

I used --no_kernels then it ran successfully.

I think when you use --no_kernels you will lose some acceleration from GPU.