Running RFdiffusion on Intel macbook pro without Nvidia GPU

[shanghongsim]

Hi all,

May I check if it is possible to run the code in this repo on a Intel Macbook without Nvidia GPU? I installed Pytorch but this error keeps coming up:

I installed Pytorch like this:

I searched Nvidia's website for CUDA tookit 11.1 but it seesm like there isn't an option for Mac.

If it is possible, may I know how I can install the missing packages?

Greatly appreciate any help! Thank you!

[roccomoretti]

I think that cuda versions are only needed if you want to use NVIDIA GPUs to speed things up. I think (but have not tested) that you can simply adjust the installation scripts such that they use the CPU versions of the packages. (That is, edit SE3nv.yml to remove the cudatoolkit line, and change the dgl-cuda11.1 line to just dgl.)

[jadolfbr]

So, if you remove the CUDA-specific code, add code for the inference script to run on mps instead of Cuda/CPU, the script progresses. However, this is as far as I've got:

    with nvtx_range('spherical harmonics'):
  File "/Users/jadolfbr/miniforge3/envs/pytorch_m1/lib/python3.10/contextlib.py", line 135, in __enter__
    return next(self.gen)
  File "/Users/jadolfbr/miniforge3/envs/pytorch_m1/lib/python3.10/site-packages/torch/cuda/nvtx.py", line 89, in range
    range_push(msg.format(*args, **kwargs))
  File "/Users/jadolfbr/miniforge3/envs/pytorch_m1/lib/python3.10/site-packages/torch/cuda/nvtx.py", line 31, in range_push
    return _nvtx.rangePushA(msg)
  File "/Users/jadolfbr/miniforge3/envs/pytorch_m1/lib/python3.10/site-packages/torch/cuda/nvtx.py", line 10, in _fail
    raise RuntimeError(
RuntimeError: NVTX functions not installed. Are you sure you have a CUDA build?

Can't seem to figure my way past this one. I have the pytorch nightly builds with MPS enabled comp. I can run many ML models, including ProteinMPNN. However, this SE3transformer seems to want NVTX model, which does not work outside of CUDA. Pretty strange. So not sure how to move forward here.

[jadolfbr]

Removing all the nvtx_range code within the SE3transformer code and then rebuilding, (which is in a 'with' statement, and seems to be primarily for profiling GPU calls), it made it farther and now I get another error. Gonna leave it here for anyone else trying to get this working.

    return self._call_impl(*args, **kwargs)
  File "/Users/jadolfbr/miniforge3/envs/pytorch_m1/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1528, in _call_impl
    return forward_call(*args, **kwargs)
  File "/Users/jadolfbr/miniforge3/envs/pytorch_m1/lib/python3.10/site-packages/se3_transformer-1.0.0-py3.10.egg/se3_transformer/model/layers/attention.py", line 158, in forward
    z_concat = aggregate_residual(node_features, z, 'cat')
  File "/Users/jadolfbr/miniforge3/envs/pytorch_m1/lib/python3.10/site-packages/se3_transformer-1.0.0-py3.10.egg/se3_transformer/runtime/utils.py", line 43, in aggregate_residual
    return {k: torch.cat([v, feats1[k]], dim=1) if k in feats1 else v for k, v in feats2.items()}
AttributeError: 'NoneType' object has no attribute 'items'

[danielguion]

Thanks for sharing. I'm working on a similar problem trying to make RFdiffusion GPU-independent to run on my intel macbook pro. Getting this to work on CPUs only is a challenge. Any progress?

[sokrypton]

SE3transformer was rewritten by nivida... last thing they'll want to make sure is that it works on CPU :P

reverting to the original SE3transformer might help:
https://github.com/FabianFuchsML/se3-transformer-public

[jadolfbr]

Just saw the 'intel' part. You may have a lot more luck here than my for my M1!!

[ardhe-qb]

Any updates on this since Sep 13? Trying to run RFdiffusion on M1 Max chip

[YaoYinYing]

@jadolfbr Check this fork. looks super good in testing of design_enzyme.sh

[danielguion]

Any updates on trying to get a CPU-only version to work? I made it as far as installing the software on my intel macbook pro and editing the cuda and nvtx-dependent code throughout the codebase to rely on CPUs. However I still get the following error:

(SE3nv) % ./scripts/run_inference.py 'contigmap.contigs=[150-150]' inference.output_prefix=test_outputs/test inference.num_designs=10

[2023-12-21 01:25:31,029][__main__][INFO] - Setting device to CPU.
Reading models from 
/RFdiffusion/rfdiffusion/inference/../../models
[2023-12-21 01:25:31,030][rfdiffusion.inference.model_runners][INFO] - Reading checkpoint from /RFdiffusion/rfdiffusion/inference/../../models/Base_ckpt.pt
This is inf_conf.ckpt_path
/RFdiffusion/rfdiffusion/inference/../../models/Base_ckpt.pt
Assembling -model, -diffuser and -preprocess configs from checkpoint
USING MODEL CONFIG: self._conf[model][n_extra_block] = 4
USING MODEL CONFIG: self._conf[model][n_main_block] = 32
USING MODEL CONFIG: self._conf[model][n_ref_block] = 4
USING MODEL CONFIG: self._conf[model][d_msa] = 256
USING MODEL CONFIG: self._conf[model][d_msa_full] = 64
USING MODEL CONFIG: self._conf[model][d_pair] = 128
USING MODEL CONFIG: self._conf[model][d_templ] = 64
USING MODEL CONFIG: self._conf[model][n_head_msa] = 8
USING MODEL CONFIG: self._conf[model][n_head_pair] = 4
USING MODEL CONFIG: self._conf[model][n_head_templ] = 4
USING MODEL CONFIG: self._conf[model][d_hidden] = 32
USING MODEL CONFIG: self._conf[model][d_hidden_templ] = 32
USING MODEL CONFIG: self._conf[model][p_drop] = 0.15
USING MODEL CONFIG: self._conf[model][SE3_param_full] = {'num_layers': 1, 'num_channels': 32, 'num_degrees': 2, 'n_heads': 4, 'div': 4, 'l0_in_features': 8, 'l0_out_features': 8, 'l1_in_features': 3, 'l1_out_features': 2, 'num_edge_features': 32}
USING MODEL CONFIG: self._conf[model][SE3_param_topk] = {'num_layers': 1, 'num_channels': 32, 'num_degrees': 2, 'n_heads': 4, 'div': 4, 'l0_in_features': 64, 'l0_out_features': 64, 'l1_in_features': 3, 'l1_out_features': 2, 'num_edge_features': 64}
USING MODEL CONFIG: self._conf[model][freeze_track_motif] = False
USING MODEL CONFIG: self._conf[model][use_motif_timestep] = True
USING MODEL CONFIG: self._conf[diffuser][T] = 50
USING MODEL CONFIG: self._conf[diffuser][b_0] = 0.01
USING MODEL CONFIG: self._conf[diffuser][b_T] = 0.07
USING MODEL CONFIG: self._conf[diffuser][schedule_type] = linear
USING MODEL CONFIG: self._conf[diffuser][so3_type] = igso3
USING MODEL CONFIG: self._conf[diffuser][crd_scale] = 0.25
USING MODEL CONFIG: self._conf[diffuser][so3_schedule_type] = linear
USING MODEL CONFIG: self._conf[diffuser][min_b] = 1.5
USING MODEL CONFIG: self._conf[diffuser][max_b] = 2.5
USING MODEL CONFIG: self._conf[diffuser][min_sigma] = 0.02
USING MODEL CONFIG: self._conf[diffuser][max_sigma] = 1.5
USING MODEL CONFIG: self._conf[preprocess][sidechain_input] = False
USING MODEL CONFIG: self._conf[preprocess][motif_sidechain_input] = True
USING MODEL CONFIG: self._conf[preprocess][d_t1d] = 22
USING MODEL CONFIG: self._conf[preprocess][d_t2d] = 44
USING MODEL CONFIG: self._conf[preprocess][prob_self_cond] = 0.5
USING MODEL CONFIG: self._conf[preprocess][str_self_cond] = True
USING MODEL CONFIG: self._conf[preprocess][predict_previous] = False
[2023-12-21 01:25:36,369][rfdiffusion.inference.model_runners][INFO] - Loading checkpoint.
[2023-12-21 01:25:36,805][rfdiffusion.diffusion][INFO] - Using cached IGSO3.
Successful diffuser __init__
[2023-12-21 01:25:36,827][__main__][INFO] - Making design test_outputs/test_0
[2023-12-21 01:25:36,833][rfdiffusion.inference.model_runners][INFO] - Using contig: ['150-150']
With this beta schedule (linear schedule, beta_0 = 0.04, beta_T = 0.28), alpha_bar_T = 0.00013696050154976547
[2023-12-21 01:25:36,892][rfdiffusion.inference.model_runners][INFO] - Sequence init: ------------------------------------------------------------------------------------------------------------------------------------------------------
Error executing job with overrides: ['contigmap.contigs=[150-150]', 'inference.output_prefix=test_outputs/test', 'inference.num_designs=10']
Traceback (most recent call last):
  File "/RFdiffusion/./scripts/run_inference.py", line 87, in main
    px0, x_t, seq_t, plddt = sampler.sample_step(
  File "/RFdiffusion/rfdiffusion/inference/model_runners.py", line 664, in sample_step
    msa_prev, pair_prev, px0, state_prev, alpha, logits, plddt = self.model(msa_masked,
  File "/Users/Daniel/opt/anaconda3/envs/SE3nv/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1518, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
  File "/Users/Daniel/opt/anaconda3/envs/SE3nv/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1527, in _call_impl
    return forward_call(*args, **kwargs)
  File "/RFdiffusion/rfdiffusion/RoseTTAFoldModel.py", line 102, in forward
    msa, pair, R, T, alpha_s, state = self.simulator(seq, msa_latent, msa_full, pair, xyz[:,:,:3],
  File "/Users/Daniel/opt/anaconda3/envs/SE3nv/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1518, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
  File "/Users/Daniel/opt/anaconda3/envs/SE3nv/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1527, in _call_impl
    return forward_call(*args, **kwargs)
  File "/RFdiffusion/rfdiffusion/Track_module.py", line 420, in forward
    msa_full, pair, R_in, T_in, state, alpha = self.extra_block[i_m](msa_full,
  File "/Users/Daniel/opt/anaconda3/envs/SE3nv/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1518, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
  File "/Users/Daniel/opt/anaconda3/envs/SE3nv/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1527, in _call_impl
    return forward_call(*args, **kwargs)
  File "RFdiffusion/rfdiffusion/Track_module.py", line 332, in forward
    R, T, state, alpha = self.str2str(msa, pair, R_in, T_in, xyz, state, idx, motif_mask=motif_mask, top_k=0)
  File "/Users/Daniel/opt/anaconda3/envs/SE3nv/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1518, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
  File "/Users/Daniel/opt/anaconda3/envs/SE3nv/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1527, in _call_impl
    return forward_call(*args, **kwargs)
  File "/RFdiffusion/rfdiffusion/Track_module.py", line 266, in forward
    shift = self.se3(G, node.reshape(B*L, -1, 1), l1_feats, edge_feats)
  File "/Users/Daniel/opt/anaconda3/envs/SE3nv/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1518, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
  File "/Users/Daniel/opt/anaconda3/envs/SE3nv/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1527, in _call_impl
    return forward_call(*args, **kwargs)
  File "/RFdiffusion/rfdiffusion/SE3_network.py", line 83, in forward
    return self.se3(G, node_features, edge_features)
  File "/Users/Daniel/opt/anaconda3/envs/SE3nv/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1518, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
  File "/Users/Daniel/opt/anaconda3/envs/SE3nv/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1527, in _call_impl
    return forward_call(*args, **kwargs)
  File "/Users/Daniel/opt/anaconda3/envs/SE3nv/lib/python3.9/site-packages/se3_transformer-1.0.0-py3.9.egg/se3_transformer/model/transformer.py", line 140, in forward
    basis = basis or get_basis(graph.edata['rel_pos'], max_degree=self.max_degree, compute_gradients=False,
  File "/Users/Daniel/opt/anaconda3/envs/SE3nv/lib/python3.9/site-packages/se3_transformer-1.0.0-py3.9.egg/se3_transformer/model/basis.py", line 166, in get_basis
    with nvtx_range('spherical harmonics'):
  File "/Users/Daniel/opt/anaconda3/envs/SE3nv/lib/python3.9/contextlib.py", line 119, in __enter__
    return next(self.gen)
  File "/Users/Daniel/opt/anaconda3/envs/SE3nv/lib/python3.9/site-packages/torch/cuda/nvtx.py", line 89, in range
    range_push(msg.format(*args, **kwargs))
  File "/Users/Daniel/opt/anaconda3/envs/SE3nv/lib/python3.9/site-packages/torch/cuda/nvtx.py", line 31, in range_push
    return _nvtx.rangePushA(msg)
  File "/Users/Daniel/opt/anaconda3/envs/SE3nv/lib/python3.9/site-packages/torch/cuda/nvtx.py", line 10, in _fail
    raise RuntimeError(
RuntimeError: NVTX functions not installed. Are you sure you have a CUDA build?

Set the environment variable HYDRA_FULL_ERROR=1 for a complete stack trace.
(SE3nv) Daniel@Daniels-MacBook-Pro RFdiffusion % 

[ardhe-qb]

I got it working on my silicon MBP using similar stuff to what is mentioned in the fork mentioned by @YaoYinYing in this comment

[danielguion]

That worked. It is on my 2015 macbook pro (no apple silicon) too. Thank you @YaoYinYing & @ardhe-qb.

[mehedihasandesu]

Hi @YaoYinYing & @ardhe-qb, I could install RFdiffusion on macs M1 and M2 by following the ways you suggested in this fork (https://github.com/YaoYinYing/RFdiffusion#conda-install-se3-transformer). But while executing the example commands, I found the following error message: FileNotFoundError: Cannot find DGL C++ graphbolt library at /Users/mehedi/miniconda3/envs/SE3nv/lib/python3.9/site-packages/dgl/graphbolt/libgraphbolt_pytorch_2.3.1.dylib. Do you have any suggestions? Thank you.

[YaoYinYing]

@mehedihasandesu Hi, sorry for my late reply. It seems like you are using Pytorch 2.3.1 + latest DGL(2.2.1?). I checked DGL's release log and found that current release of DGL only support Pytorch 2.3.0. I have tested this again and update the README. Thanks for your response!

[mehedihasandesu]

It is working now. Thank you so much @YaoYinYing.

[jadolfbr]

Awesome @YaoYinYing ! Will try this out!

[jnielsen20]

Out of curiosity. Does anyone know how the performance of the CPU only version of RFdiffusion compares with the GPU enabled version?