I installed localcolabfold on Ubuntu running in Windows WSL.
Installation using $ bash install_colabfold_linux.sh worked well.
If I cd into colabfold and run
$ colabfold-conda/bin/python3.7 runner.py
I get the following message:
-bash: colabfold-conda/bin/python3.7: No such file or directory
If I check the folders within the colabfold folder using ls I get this:
CONTRIBUTING.md colabfold.py docker notebooks residue_constants.patch scripts
LICENSE colabfold_alphafold.patch gpurelaxation.patch pairmsa.py run_alphafold.py setup.py
README.md colabfold_alphafold.py imgs protein.patch run_alphafold_test.py
alphafold conda model.patch reformat.pl runner.py
bin config.patch modules.patch requirements.txt runner_af2advanced.py

So there is no colabfold-conda folder. Is this a typo and it should read colabfold/conda/bin... ?
I can run the script using
$ python3 runner.py
but then there's a tensorflow issue.

Hi,
Thanks for this repo it is amazing.

Can you explain the difference between runner.py and runneraf2advanced.py?
Are both run with the latest update that boosted the running time?

I run runner.py multiple times (only single compile in the beggining of course)
on gpu:
Nvidia DGX - A100 P3687 System 8 x 80GB
With varied length sequences.

Currently it run in pace of 15-20 structures in hour (each structure produced by single model with max recycke 3)
It is fast but not as thousands of models in a day.
How can I be sure that I use it correctly?

thanks

Hi, There is a recently deposited structure on the pdb with high sequence id% with respect to my query protein. However, this structure does not seem to be present on the current PDB database that the colab is using. Is there a way to us an updated version of the PDB or to provide my own set of templates to it?

Thanks.

I have installed it on my m1 mac. It's working when I run without a chain break.

But I want chain break in my protein. So as per (ColabFold AlphaFold2 notebook)- Use : to specify inter-protein chainbreaks for modeling complexes (supports homo- and hetro-oligomers) but its showing error in my system-

raise ValueError(f'Invalid character in the sequence: {aa_type}')
ValueError: Invalid character in the sequence: :

Inserted a .fasta file

First off, fantastic piece of code, thank you so much for your efforts.

We are trying to using localcolabfold to do batch predictions, but we encounter the following problem. Whenever a graphical output is being generated on the fly (i.e. a new window is opened depicting the graphic), colabfold stalls on our system until that window is closed, only to proceed to the next point where another graphics window is opened. Having set "show_images = False" which I believe suppresses the depiction of the predicted structures, still leads to stalling events for the other images that are being generated.

Can you suggest a workaround ?

Thanks in advance

Hi Yoshitaka-san,

I tried running AF2_Advanced on these two files separately:

>HHAWx4
HHAWHHAWHHAWHHAW

And

>KKAWx4
KKAWKKAWKKAWKKAW

It gave me the following error message:

running mmseqs2
  0%|          | 0/150 [elapsed: 00:00 remaining: ?]
Traceback (most recent call last):
  File "/home/ubuntu/storage1/colabfold/runner_af2advanced.py", line 163, in <module>
    hhfilter_loc="colabfold-conda/bin/hhfilter", precomputed=precomputed, TMP_DIR=output_dir)
  File "/home/ubuntu/storage1/colabfold/colabfold_alphafold.py", line 292, in prep_msa
    A3M_LINES = cf.run_mmseqs2(I["seqs"], prefix, use_filter=True, host_url=mmseqs_host_url)
  File "/home/ubuntu/storage1/colabfold/colabfold.py", line 150, in run_mmseqs2
    raise Exception(f'MMseqs2 API is giving errors. Please confirm your input is a valid protein sequence. If error persists, please try again an hour later.')
Exception: MMseqs2 API is giving errors. Please confirm your input is a valid protein sequence. If error persists, please try again an hour later.

How can I resolve the issue?

G.V.

When I run the colabfold executable on linux server (with slurm job scheduler) after installing without any glitches, I get the following output on the output file eventhough I request for nearly 12gigs of memory on the sbatch script.

2021-10-27 13:15:35.963528: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudart.so.11.0
2021-10-27 13:15:42.473577: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcuda.so.1
2021-10-27 13:15:43.476249: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1733] Found device 0 with properties:
pciBusID: 0000:05:00.0 name: Tesla K20m computeCapability: 3.5
coreClock: 0.7055GHz coreCount: 13 deviceMemorySize: 4.63GiB deviceMemoryBandwidth: 193.71GiB/s
2021-10-27 13:15:43.476936: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudart.so.11.0
2021-10-27 13:15:43.568309: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcublas.so.11
2021-10-27 13:15:43.568439: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcublasLt.so.11
2021-10-27 13:15:43.604743: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcufft.so.10
2021-10-27 13:15:43.628756: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcurand.so.10
2021-10-27 13:15:43.695103: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcusolver.so.11
2021-10-27 13:15:43.740556: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcusparse.so.11
2021-10-27 13:15:43.747635: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudnn.so.8
2021-10-27 13:15:43.758695: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1871] Adding visible gpu devices: 0
2021-10-27 13:15:54.287682: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:764] failed to alloc 19897253888 bytes unified memory; result: CUDA_ERROR_INVALID_VALUE: invalid argument
2021-10-27 13:15:54.287770: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:764] failed to alloc 17907527680 bytes unified memory; result: CUDA_ERROR_INVALID_VALUE: invalid argument
2021-10-27 13:15:54.287798: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:764] failed to alloc 16116774912 bytes unified memory; result: CUDA_ERROR_INVALID_VALUE: invalid argument
2021-10-27 13:15:54.287818: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:764] failed to alloc 14505097216 bytes unified memory; result: CUDA_ERROR_INVALID_VALUE: invalid argument
2021-10-27 13:15:54.287838: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:764] failed to alloc 13054586880 bytes unified memory; result: CUDA_ERROR_INVALID_VALUE: invalid argument
2021-10-27 13:15:55.703597: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:764] failed to alloc 11749128192 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2021-10-27 13:15:57.518297: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:764] failed to alloc 10574215168 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2021-10-27 13:15:59.337052: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:764] failed to alloc 9516793856 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2021-10-27 13:16:01.129516: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:764] failed to alloc 8565114368 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2021-10-27 13:16:02.945864: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:764] failed to alloc 7708602880 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2021-10-27 13:16:04.760092: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:764] failed to alloc 6937742336 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2021-10-27 13:16:06.572127: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:764] failed to alloc 6243968000 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2021-10-27 13:16:08.382901: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:764] failed to alloc 5619571200 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2021-10-27 13:16:10.195180: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:764] failed to alloc 5057613824 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2021-10-27 13:16:12.005831: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:764] failed to alloc 4551852544 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2021-10-27 13:16:13.820047: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:764] failed to alloc 4096667136 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2021-10-27 13:16:15.632651: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:764] failed to alloc 3687000320 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2021-10-27 13:16:17.445155: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:764] failed to alloc 3318300160 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2021-10-27 13:16:19.252805: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:764] failed to alloc 2986470144 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2021-10-27 13:16:21.063185: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:764] failed to alloc 2687823104 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2021-10-27 13:16:28.041902: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1258] Device interconnect StreamExecutor with strength 1 edge matrix:
2021-10-27 13:16:28.041959: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1264]
2021-10-27 13:16:28.140439: I tensorflow/core/platform/profile_utils/cpu_utils.cc:114] CPU Frequency: 1999895000 Hz
slurmstepd: error: Job 39188705 exceeded memory limit (1253252 > 1024000), being killed
slurmstepd: error: *** JOB 39188705 ON node101 CANCELLED AT 2021-10-27T13:16:35 ***

Hi Yoshitaka,

As colabfold_bash suggests, it recomends using a local version of the mmseq2 to generate the mas.pickle and given the recent problems with the server, do you happen to know how to te set up the search locally, including the required databases? In principle, it should be possible with the argument --host_url.

Thanks a lot in advance.
Best,
Ariel

Hi,
I used colabfold on my local computer successfully (Thank you guys!)
unfortunately, after I ran the update script, I have an error and I can't run it anymore:

jax._src.traceback_util.UnfilteredStackTrace: RuntimeError: INTERNAL: CustomCall failed: jaxlib/cuda_prng_kernels.cc:30: operation cudaGetLastError() failed: the provided PTX was compiled with an unsupported toolchain.

I tried using cuda 11.1 with cudnn 8.0.5
and tried cuda 11.3 with cudnn 8.2.1
and bothe does not work.

I installed jax by the instructions an I tried to uninstall and install it again with the specific cuda and cudnn versions - but it didn't help.

I tried to build the colabfold-batch from pip on new enviroment - but it acts the same too.

Any idea what went wrong?

First of all, thanks for your effort to share this nice work, it's very helpful.

I had some queries that have very low MSA coverage using mmseqs2 compared to using jackhmmer, so I am wondering if there is a new runner.py now would support jackhmmer, I noticed that the original runner.py doesn't work.
Or if is there anyway to upload a custom msa file? When I ran the py file, with custom-msa, it doesn't give me option to upload file. Thanks!

Hi there,
I was running the program when it crashed after generating the 3rd relaxed model. I ran the same protein before as a monomer (733 aa) with no issues and there was more than enough memory to continue (using a Volta V100-SXM2-32GB).

Here they mention that it's caused by OpenMM. The same OpenMM authors have a new tool that can fix this issues PDBfixer. Maybe that helps.

Cheers,
Xabi

This was the command:

colabfold --input $INPUT \
--homooligomer 2 \
--msa_method precomputed \
--precomputed $MSA \
--output_dir ${OUTDIR} \
--max_recycle 3 \
--use_ptm \
--use_turbo \
--num_relax Top5

and the error:

[...]
model rank based on pLDDT
rank_1_model_5_ptm_seed_0 pLDDT:72.23
rank_2_model_4_ptm_seed_0 pLDDT:71.97
rank_3_model_3_ptm_seed_0 pLDDT:71.74
rank_4_model_1_ptm_seed_0 pLDDT:67.31
rank_5_model_2_ptm_seed_0 pLDDT:61.61
  0%|          | 0/5 [elapsed: 00:00 remaining: ?]
Traceback (most recent call last):
  File "/g/data1a/u71/xabi/colabfold/runner_af2advanced.py", line 289, in <module>
    relaxed_pdb_lines, _, _ = amber_relaxer.process(prot=outs[key]["unrelaxed_protein"])
  File "/g/data/u71/xabi/colabfold/alphafold/relax/relax.py", line 73, in process
    min_pdb = utils.overwrite_pdb_coordinates(pdb_str, min_pos)
  File "/g/data/u71/xabi/colabfold/alphafold/relax/utils.py", line 29, in overwrite_pdb_coordinates
    openmm_app.PDBFile.writeFile(topology, pos, f)
  File "/g/data1a/u71/xabi/colabfold/colabfold-conda/lib/python3.7/site-packages/simtk/openmm/app/pdbfile.py", line 283, in writeFile
    PDBFile.writeModel(topology, positions, file, keepIds=keepIds, extraParticleIdentifier=extraParticleIdentifier)
  File "/g/data1a/u71/xabi/colabfold/colabfold-conda/lib/python3.7/site-packages/simtk/openmm/app/pdbfile.py", line 331, in writeModel
    raise ValueError('The number of positions must match the number of atoms')
ValueError: The number of positions must match the number of atoms

バグの内容

実行環境

$ nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2021 NVIDIA Corporation
Built on Sun_Feb_14_21:12:58_PST_2021
Cuda compilation tools, release 11.2, V11.2.152
Build cuda_11.2.r11.2/compiler.29618528_0

実行コマンド

colabfold-conda/bin/python3.7 runner_af2advanced.py \
--input hoge.fasta \
--output_dir hoge \
--max_recycle 6 \
--msa_method single_sequence \
--num_relax Top5

エラー内容

Traceback (most recent call last):
  File "runner_af2advanced.py", line 248, in <module>
    rank_by=rank_by, show_images=show_images)
  File "colabfold/colabfold_alphafold.py", line 775, in run_alphafold
    model_runner = prep_model_runner(opt, model_name=model_name, use_turbo=False, params_loc=params_loc)["model"]
TypeError: prep_model_runner() got an unexpected keyword argument 'use_turbo'

問題の推測

colabfold_alphafold.pyのrun_alphafoldでuse_turboがFalseの場合には、prep_model_runnerに対してopt以外の部分でuse_turboを引数として渡しているため上記のようなエラーが生じるのではないでしょうか？

https://github.com/sokrypton/ColabFold/blob/4a4552e38c89ce803ab03483a8e898943466d19a/beta/colabfold_alphafold.py#L771-L775

解決方法

use_turboがFalseの際にoptの内容を変えてみてはいかがでしょうか？
#18

Thanks a lot for providing this excellent tool for us. However, I met a problem about No GPU/TPU found. Could you please help me to fix it?

After I run
colabfold-conda/bin/python3.7 runner.py TF_CPP_MIN_LOG_LEVEL=0
It returns:

ColabFold on Mac
homooligomer: '1'
total_length: '59'
working_directory: 'prediction_test_a5e17'
running mmseqs2
0%| | 0/5 [elapsed: 00:00 remaining: ?]
WARNING:absl:No GPU/TPU found, falling back to CPU. (Set TF_CPP_MIN_LOG_LEVEL=0 and rerun for more info.)
Traceback (most recent call last):
File "runner.py", line 596, in
cf.clear_mem("gpu")
File "/Users/huacanfeng/Documents/AF2/colabfold/colabfold.py", line 55, in clear_mem
backend = jax.lib.xla_bridge.get_backend(device)
File "/Users/huacanfeng/Documents/AF2/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/lib/xla_bridge.py", line 234, in get_backend
raise RuntimeError(f"Unknown backend {platform}")
RuntimeError: Unknown backend GPU

The information for my MacBook Pro
Processor: 2.3 GHz 8-Core Intel Core i9
Memory: 16 GB 2400 MHz DDR4
Graphics: Intel UHD Graphics 630 1536 MB

what does it mean when i am installing colabfold?

Hi Yoshitaka-san,

I tried your latest colabfold_batch code. With the following command, following
your instruction in main page:

$ colabfold_batch --amber --templates --num-recycle 3 input_fasta/test.fasta af2_output

The input_fasta/test.fasta contains this:

$ cat input_fasta/test.fasta
>Gvb.fl_hhaw_x1.s1
HHAWEPEKKKPEGMKDQPSKQHHAW
>Gvb.fl_hhaw_x1.s10
HHAWSNYDERKQNGRYYIWKEHHAW

But I get the following error.
How can I resolve it?

$ colabfold_batch --amber --templates --num-recycle 3  input_fasta/test.fasta af2_output
WARNING: You are welcome to use the default MSA server, however keep in mind that it's a limited shared resource only capable of processing a few thousand MSAs per day. Please submit jobs only from a single IP address. We reserve the right to limit access to the server case-by-case when usage exceeds fair use.

If you require more MSAs, please host your own API and pass it to `--host-url`
2021-12-06 15:40:37,273 Found 8 citations for tools or databases
2021-12-06 15:40:46,439 Query 1/2: Gvb.fl_hhaw_x1.s1 (length 25)
COMPLETE: 100%|████████████████████████████████████████████████████████████████████████████████████████| 150/150 [elapsed: 00:02 remaining: 00:00]
0	no_templates_found
2021-12-06 15:40:48,450 Could not get MSA/templates for Gvb.fl_hhaw_x1.s1: expected str, bytes or os.PathLike object, not NoneType
Traceback (most recent call last):
  File "/home/ubuntu/anaconda3/lib/python3.7/site-packages/colabfold/batch.py", line 603, in run
    host_url,
  File "/home/ubuntu/anaconda3/lib/python3.7/site-packages/colabfold/batch.py", line 428, in get_msa_and_templates
    query_seqs_unique[index],
  File "/home/ubuntu/anaconda3/lib/python3.7/site-packages/colabfold/batch.py", line 93, in mk_template
    obsolete_pdbs_path=None,
  File "/home/ubuntu/anaconda3/lib/python3.7/site-packages/alphafold/data/templates.py", line 836, in __init__
    if not glob.glob(os.path.join(self._mmcif_dir, '*.cif')):
  File "/home/ubuntu/anaconda3/lib/python3.7/posixpath.py", line 80, in join
    a = os.fspath(a)
TypeError: expected str, bytes or os.PathLike object, not NoneType
2021-12-06 15:40:48,451 Query 2/2: Gvb.fl_hhaw_x1.s10 (length 25)
2021-12-06 15:40:49,415 Sleeping for 9s. Reason: RATELIMIT
2021-12-06 15:40:59,395 Sleeping for 8s. Reason: RATELIMIT
2021-12-06 15:41:08,362 Sleeping for 5s. Reason: RATELIMIT
2021-12-06 15:41:14,342 Sleeping for 5s. Reason: RATELIMIT
2021-12-06 15:41:20,315 Sleeping for 10s. Reason: RATELIMIT
COMPLETE: 100%|████████████████████████████████████████████████████████████████████████████████████████| 150/150 [elapsed: 00:43 remaining: 00:00]
0	no_templates_found
2021-12-06 15:41:32,334 Could not get MSA/templates for Gvb.fl_hhaw_x1.s10: expected str, bytes or os.PathLike object, not NoneType
Traceback (most recent call last):
  File "/home/ubuntu/anaconda3/lib/python3.7/site-packages/colabfold/batch.py", line 603, in run
    host_url,
  File "/home/ubuntu/anaconda3/lib/python3.7/site-packages/colabfold/batch.py", line 428, in get_msa_and_templates
    query_seqs_unique[index],
  File "/home/ubuntu/anaconda3/lib/python3.7/site-packages/colabfold/batch.py", line 93, in mk_template
    obsolete_pdbs_path=None,
  File "/home/ubuntu/anaconda3/lib/python3.7/site-packages/alphafold/data/templates.py", line 836, in __init__
    if not glob.glob(os.path.join(self._mmcif_dir, '*.cif')):
  File "/home/ubuntu/anaconda3/lib/python3.7/posixpath.py", line 80, in join
    a = os.fspath(a)
TypeError: expected str, bytes or os.PathLike object, not NoneType
2021-12-06 15:41:32,334 Done

Hi YoshitakaMo,

Thank you very much for allowing to run colabfold on a local machine. I have a question related to pLDDT value. How we can save this value for a model structures. I have 300 model structures but it is not saving final pLDDT value of each model in a file. I need that number for quality assessment.

Currently it is modeling all the structures but not printing any scores.json file.

Thank you, S

Hi, I got the following error, any suggestion? Thanks.

2021-10-25 23:38:32.305016: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudart.so.11.0
2021-10-25 23:38:38.368438: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcuda.so.1
2021-10-25 23:38:38.420023: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-10-25 23:38:38.421306: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1733] Found device 0 with properties: 
pciBusID: 0000:41:00.0 name: RTX A6000 computeCapability: 8.6
coreClock: 1.8GHz coreCount: 84 deviceMemorySize: 47.54GiB deviceMemoryBandwidth: 715.34GiB/s
2021-10-25 23:38:38.421351: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-10-25 23:38:38.422580: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1733] Found device 1 with properties: 
pciBusID: 0000:43:00.0 name: RTX A6000 computeCapability: 8.6
coreClock: 1.8GHz coreCount: 84 deviceMemorySize: 47.54GiB deviceMemoryBandwidth: 715.34GiB/s
2021-10-25 23:38:38.422597: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudart.so.11.0
2021-10-25 23:38:38.502267: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcublas.so.11
2021-10-25 23:38:38.502329: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcublasLt.so.11
2021-10-25 23:38:38.548584: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcufft.so.10
2021-10-25 23:38:38.568160: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcurand.so.10
2021-10-25 23:38:38.662025: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcusolver.so.11
2021-10-25 23:38:38.691719: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcusparse.so.11
2021-10-25 23:38:38.696615: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudnn.so.8
2021-10-25 23:38:38.696718: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-10-25 23:38:38.698052: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-10-25 23:38:38.699305: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-10-25 23:38:38.702210: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-10-25 23:38:38.703830: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1871] Adding visible gpu devices: 0, 1
ColabFold on Linux
WARNING: For a typical Google-Colab-GPU (16G) session, the max total length is ~1400 residues. You are at 1625! Run Alphafold may crash.
homooligomer: '1'
total_length: '1625'
working_directory: 'prediction_test_37769'
running mmseqs2
  0%|          | 0/150 [elapsed: 00:00 remaining: ?]
  0%|          | 0/5 [elapsed: 00:00 remaining: ?]
2021-10-25 23:44:59.124108: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX2 FMA
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
2021-10-25 23:44:59.127144: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1258] Device interconnect StreamExecutor with strength 1 edge matrix:
2021-10-25 23:44:59.127179: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1264]      
2021-10-25 23:44:59.200029: I tensorflow/core/platform/profile_utils/cpu_utils.cc:114] CPU Frequency: 2699835000 Hz
2021-10-25 23:45:03.682089: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_asm_compiler.cc:63] cuLinkAddData fails. This is usually caused by stale driver version.
2021-10-25 23:45:03.682142: E external/org_tensorflow/tensorflow/compiler/xla/service/gpu/gpu_compiler.cc:985] The CUDA linking API did not work. Please use XLA_FLAGS=--xla_gpu_force_compilation_parallelism=1 to bypass it, but expect to get longer compilation time due to the lack of multi-threading.
Traceback (most recent call last):
  File "runner.py", line 662, in <module>
    prediction_result, (r, t) = cf.to(model_runner.predict(processed_feature_dict, random_seed=seed),"cpu")
  File "/data/colabfold/alphafold/model/model.py", line 134, in predict
    result, recycles = self.apply(self.params, jax.random.PRNGKey(random_seed), feat)
  File "/data/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/_src/random.py", line 122, in PRNGKey
    key = prng.seed_with_impl(impl, seed)
  File "/data/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/_src/prng.py", line 203, in seed_with_impl
    return PRNGKeyArray(impl, impl.seed(seed))
  File "/data/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/_src/prng.py", line 241, in threefry_seed
    k1 = convert(lax.shift_right_logical(seed_arr, lax._const(seed_arr, 32)))
  File "/data/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/_src/lax/lax.py", line 408, in shift_right_logical
    return shift_right_logical_p.bind(x, y)
  File "/data/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/core.py", line 272, in bind
    out = top_trace.process_primitive(self, tracers, params)
  File "/data/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/core.py", line 624, in process_primitive
    return primitive.impl(*tracers, **params)
  File "/data/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/interpreters/xla.py", line 312, in apply_primitive
    **params)
  File "/data/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/_src/util.py", line 187, in wrapper
    return cached(config._trace_context(), *args, **kwargs)
  File "/data/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/_src/util.py", line 180, in cached
    return f(*args, **kwargs)
  File "/data/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/interpreters/xla.py", line 335, in xla_primitive_callable
    prim.name, donated_invars, *arg_specs)
  File "/data/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/interpreters/xla.py", line 654, in _xla_callable_uncached
    *arg_specs).compile().unsafe_call
  File "/data/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/interpreters/xla.py", line 770, in compile
    self.name, self.hlo(), *self.compile_args)
  File "/data/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/interpreters/xla.py", line 798, in from_xla_computation
    compiled = compile_or_get_cached(backend, xla_computation, options)
  File "/data/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/interpreters/xla.py", line 87, in compile_or_get_cached
    return backend_compile(backend, computation, compile_options)
  File "/data/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/interpreters/xla.py", line 369, in backend_compile
    return backend.compile(built_c, compile_options=options)
RuntimeError: Unknown: no kernel image is available for execution on the device
in external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_asm_compiler.cc(66): 'status'

As issued WJluluxiu on the other issue page, there is a tensorflow error and I got the following error messages.

My server has 4 x 1080Ti GPUs on CentOS7.

Hi Yoshitaka-san,

As you mentioned AF2-Multimer seem to be available now.
Is there a way we can port it to command line version?
Just like your great AF2_advanced.

G.V.

Hi Yoshitaka-san,

Thank you for the great work.

Original AlphaFold2_advanced allows you to do inter protein chain-breaks in the sequence entry:

So you can do something like: AC/DE:FGH

Use / to specify intra-protein chainbreaks (for trimming regions within protein).
Use : to specify inter-protein chainbreaks (for modeling protein-protein hetero-complexes).

How can we do this with your script?
Because it only takes one file as fasta file.

Do you create a fasta file like:

 >myinput.fasta
 AC/DE:FGH

And last question, DE:FGH simply means we are docking DE to FGH, am I right?

Thanks and hope to hear from you again.

G.V.

at line 321 "pair_mode = "unpaired"" should probably be "pair_mode = args.pair_mode

Hi, I'm very interested to use other parameters of AF2 that have been metioned to provide ways to obtain alternative conformations for the models (num_ensemble, num_samples,
is_training). I tried to use them on the "ColabFold: AlphaFold2 using MMseqs2" but my protein is too large.

Can I still use these parameters in the local version? Is it doable by using this approach: "colabfold-conda/bin/python3.7 runner.py" instead of "colabfold_batch --num-models 5 --model-type AlphaFold2-multimer --templates" ?

Thanks.

Hi there, I have installed localcolabfold on the server of my school, and I can successfully run it without a problem. I am just raising a performance issue. I assigned 24, 32, 36 cores (without GPU) for my different tasks, however when I check the CPU usage, it was always <800%, meaning only 8 cores were running. Is it possible to utilize more CPUs?
Thank you!

Hello,

I have successfully installed localcolabfold on my linux system, and have been able to predict the structures of small proteins. I am now trying to run localcolabfold to predict a large protein; however, the GPU VRAM that I have is insufficient to do it. Specifically, I get the following error message:

2021-09-16 18:37:19.903673: E external/org_tensorflow/tensorflow/compiler/xla/pjrt/pjrt_stream_executor_client.cc:2040] Execution of replica 0 failed: Resource exhausted: Out of memory while trying to allocate 12026119256 bytes.

My CPU has more RAM, so I was wondering if there is a way to get localcolabfold to run on only the CPU on Linux? I realize that this will be slower, but that is not a problem for me.

Thanks in advance!

Dear experts,

I wonder if there is a way to use the wonderful power of alphafold just to model missing parts of existing pdbs, e.g. long loops?

I used to use modeller, but it would be great to use alphafold for this taks, specially for long loops.

Thanks!
Fabian

Hello,

I woould like to run localcolabfold on a cluster which compute nodes does not have internet access.

problem is one of the first step is to try to download pickled msa info from https://a3m.mmseqs.com. that of course fail
is there a way to have a local non network option ?

regards

Eric

I am trying to run colabfold on >2500 aa sequence. I read the other thread #7 regarding adjusting the following settings:

#!/bin/bash 
export TF_FORCE_UNIFIED_MEMORY=1 
export XLA_PYTHON_CLIENT_MEM_FRACTION=4.0 
colabfold-conda/bin/python3.7 runner.py

If I follow https://sbgrid.org/wiki/examples/alphafold2 recommendations:

TF_FORCE_UNIFIED_MEMORY=1
XLA_PYTHON_CLIENT_MEM_FRACTION=0.5
XLA_PYTHON_CLIENT_ALLOCATOR=platform

Either way, I have been running into the following error:

tensorflow/core/common_runtime/gpu/gpu_device.cc:1871] Adding visible gpu devices: 0, 1, 2, 3, 4, 5, 6, 7

jax._src.traceback_util.UnfilteredStackTrace: RuntimeError: Resource exhausted: Failed to allocate request for 3.25GiB (3488555008B) on device ordinal 0

Checking nvidia-smi indicated that the primary half of the memory of one GPU is being used and not the rest. I tried restarting my ssh sessions too to make sure nothing else was being unknowingly used in the background prior to running this command.

Hi, there. I have 2 RTX2080Ti(11G) GPUs with CUDA-11.2, when i input a sequence less than 1000 amino acids, it can run normally but only one of my GPU works. When i tried a ~1200 sequence or complex it will throw 'ResourceExhausted' error. So the problem is how do i let all my GPUs work and be able to calculate larger sequence or complex?

I am trying to run this on HPC for a large batch of jobs. Submitting jobs to mmseqs2 server seems to be the step slowing down for large batch. Is it easy to generate the MSA using local mmseqs2 for the same pickle format? Thanks!!

Hi, I just updated my code to the last commit and it works fine with the default num_relax = "None", but not with num_relax = "Top5":

colabfold-conda/bin/python3.7 runner.py
2021-09-28 07:03:03.449188: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudart.so.11.0
2021-09-28 07:03:05.429501: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcuda.so.1
2021-09-28 07:03:05.589721: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1733] Found device 0 with properties:
pciBusID: 0000:0a:00.0 name: NVIDIA GeForce GTX 1080 Ti computeCapability: 6.1
coreClock: 1.582GHz coreCount: 28 deviceMemorySize: 10.92GiB deviceMemoryBandwidth: 451.17GiB/s
2021-09-28 07:03:05.589805: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-09-28 07:03:05.592761: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1733] Found device 1 with properties:
pciBusID: 0000:42:00.0 name: NVIDIA GeForce GTX 1080 Ti computeCapability: 6.1
coreClock: 1.582GHz coreCount: 28 deviceMemorySize: 10.91GiB deviceMemoryBandwidth: 451.17GiB/s
2021-09-28 07:03:05.592788: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudart.so.11.0
2021-09-28 07:03:05.603415: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcublas.so.11
2021-09-28 07:03:05.603500: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcublasLt.so.11
2021-09-28 07:03:05.610883: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcufft.so.10
2021-09-28 07:03:05.613852: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcurand.so.10
2021-09-28 07:03:05.625022: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcusolver.so.11
2021-09-28 07:03:05.628477: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcusparse.so.11
2021-09-28 07:03:05.629806: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudnn.so.8
2021-09-28 07:03:05.633270: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-09-28 07:03:05.639817: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2021-09-28 07:03:05.642759: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1871] Adding visible gpu devices: 0, 1
2021-09-28 07:03:14.784927: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: AVX2 FMA
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
2021-09-28 07:03:14.786213: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1258] Device interconnect StreamExecutor with strength 1 edge matrix:
2021-09-28 07:03:14.786232: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1264]
2021-09-28 07:03:14.860397: I tensorflow/core/platform/profile_utils/cpu_utils.cc:114] CPU Frequency: 3393245000 Hz
ColabFold on Linux
homooligomer: '1'
total_length: '59'
working_directory: 'prediction_test_a5e17'
running mmseqs2
0%| | 0/5 [elapsed: 00:00 remaining: ?]
model_1_ptm_seed_0 recycles:3 tol:0.06 pLDDT:96.25 pTMscore:0.76
model_2_ptm_seed_0 recycles:3 tol:0.12 pLDDT:96.98 pTMscore:0.76
model_3_ptm_seed_0 recycles:3 tol:0.08 pLDDT:97.48 pTMscore:0.78
model_4_ptm_seed_0 recycles:3 tol:0.12 pLDDT:96.54 pTMscore:0.77
model_5_ptm_seed_0 recycles:3 tol:0.18 pLDDT:96.27 pTMscore:0.78
model rank based on pLDDT
rank_1_model_3_ptm_seed_0 pLDDT:97.48
rank_2_model_2_ptm_seed_0 pLDDT:96.98
rank_3_model_4_ptm_seed_0 pLDDT:96.54
rank_4_model_5_ptm_seed_0 pLDDT:96.27
rank_5_model_1_ptm_seed_0 pLDDT:96.25
0%| | 0/5 [elapsed: 00:00 remaining: ?]
Traceback (most recent call last):
File "runner.py", line 758, in
relaxed_pdb_lines, _, _ = amber_relaxer.process(prot=outs[key]["unrelaxed_protein"])
File "/usr/local/alphafold/localcolabfold/colabfold/alphafold/relax/relax.py", line 62, in process
max_outer_iterations=self._max_outer_iterations)
File "/usr/local/alphafold/localcolabfold/colabfold/alphafold/relax/amber_minimize.py", line 482, in run_pipeline
ret.update(get_violation_metrics(prot))
File "/usr/local/alphafold/localcolabfold/colabfold/alphafold/relax/amber_minimize.py", line 356, in get_violation_metrics
structural_violations, struct_metrics = find_violations(prot)
File "/usr/local/alphafold/localcolabfold/colabfold/alphafold/relax/amber_minimize.py", line 343, in find_violations
"clash_overlap_tolerance": 1.5, # Taken from model config.
File "/usr/local/alphafold/localcolabfold/colabfold/alphafold/model/folding.py", line 758, in find_structural_violations
atomtype_radius, batch['residx_atom14_to_atom37'])
File "/usr/local/alphafold/localcolabfold/colabfold/alphafold/model/utils.py", line 39, in batched_gather
return take_fn(params, indices)
File "/usr/local/alphafold/localcolabfold/colabfold/alphafold/model/utils.py", line 36, in
take_fn = lambda p, i: jnp.take(p, i, axis=axis)
File "/usr/local/alphafold/localcolabfold/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/_src/numpy/lax_numpy.py", line 5384, in take
mode)
File "/usr/local/alphafold/localcolabfold/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/_src/traceback_util.py", line 162, in reraise_with_filtered_traceback
return fun(*args, **kwargs)
File "/usr/local/alphafold/localcolabfold/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/_src/api.py", line 414, in cache_miss
donated_invars=donated_invars, inline=inline)
File "/usr/local/alphafold/localcolabfold/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/core.py", line 1618, in bind
return call_bind(self, fun, *args, **params)
File "/usr/local/alphafold/localcolabfold/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/core.py", line 1609, in call_bind
outs = primitive.process(top_trace, fun, tracers, params)
File "/usr/local/alphafold/localcolabfold/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/core.py", line 1621, in process
return trace.process_call(self, fun, tracers, params)
File "/usr/local/alphafold/localcolabfold/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/core.py", line 615, in process_call
return primitive.impl(f, *tracers, **params)
File "/usr/local/alphafold/localcolabfold/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/interpreters/xla.py", line 623, in _xla_call_impl
*unsafe_map(arg_spec, args))
File "/usr/local/alphafold/localcolabfold/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/linear_util.py", line 262, in memoized_fun
ans = call(fun, *args)
File "/usr/local/alphafold/localcolabfold/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/interpreters/xla.py", line 694, in _xla_callable
return lower_xla_callable(fun, device, backend, name, donated_invars, *arg_specs).compile().unsafe_call
File "/usr/local/alphafold/localcolabfold/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/interpreters/xla.py", line 703, in lower_xla_callable
fun, abstract_args, pe.debug_info_final(fun, "jit"))
File "/usr/local/alphafold/localcolabfold/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/interpreters/partial_eval.py", line 1522, in trace_to_jaxpr_final
jaxpr, out_avals, consts = trace_to_subjaxpr_dynamic(fun, main, in_avals)
File "/usr/local/alphafold/localcolabfold/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/interpreters/partial_eval.py", line 1500, in trace_to_subjaxpr_dynamic
ans = fun.call_wrapped(*in_tracers)
File "/usr/local/alphafold/localcolabfold/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/linear_util.py", line 166, in call_wrapped
ans = self.f(*args, **dict(self.params, **kwargs))
File "/usr/local/alphafold/localcolabfold/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/_src/numpy/lax_numpy.py", line 5390, in _take
_check_arraylike("take", a)
File "/usr/local/alphafold/localcolabfold/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/_src/numpy/lax_numpy.py", line 559, in _check_arraylike
raise TypeError(msg.format(fun_name, type(arg), pos))
jax._src.traceback_util.UnfilteredStackTrace: TypeError: take requires ndarray or scalar arguments, got <class 'list'> at position 0.

The stack trace below excludes JAX-internal frames.
The preceding is the original exception that occurred, unmodified.

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
File "runner.py", line 758, in
relaxed_pdb_lines, _, _ = amber_relaxer.process(prot=outs[key]["unrelaxed_protein"])
File "/usr/local/alphafold/localcolabfold/colabfold/alphafold/relax/relax.py", line 62, in process
max_outer_iterations=self._max_outer_iterations)
File "/usr/local/alphafold/localcolabfold/colabfold/alphafold/relax/amber_minimize.py", line 482, in run_pipeline
ret.update(get_violation_metrics(prot))
File "/usr/local/alphafold/localcolabfold/colabfold/alphafold/relax/amber_minimize.py", line 356, in get_violation_metrics
structural_violations, struct_metrics = find_violations(prot)
File "/usr/local/alphafold/localcolabfold/colabfold/alphafold/relax/amber_minimize.py", line 343, in find_violations
"clash_overlap_tolerance": 1.5, # Taken from model config.
File "/usr/local/alphafold/localcolabfold/colabfold/alphafold/model/folding.py", line 758, in find_structural_violations
atomtype_radius, batch['residx_atom14_to_atom37'])
File "/usr/local/alphafold/localcolabfold/colabfold/alphafold/model/utils.py", line 39, in batched_gather
return take_fn(params, indices)
File "/usr/local/alphafold/localcolabfold/colabfold/alphafold/model/utils.py", line 36, in
take_fn = lambda p, i: jnp.take(p, i, axis=axis)
File "/usr/local/alphafold/localcolabfold/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/_src/numpy/lax_numpy.py", line 5384, in take
mode)
File "/usr/local/alphafold/localcolabfold/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/_src/numpy/lax_numpy.py", line 5390, in _take
_check_arraylike("take", a)
File "/usr/local/alphafold/localcolabfold/colabfold/colabfold-conda/lib/python3.7/site-packages/jax/_src/numpy/lax_numpy.py", line 559, in _check_arraylike
raise TypeError(msg.format(fun_name, type(arg), pos))
TypeError: take requires ndarray or scalar arguments, got <class 'list'> at position 0.

Any idea of what could be the problem? Thank you!

Hello,

I ran the colabFold locally, but there is no .a3m file in the output folder. Could you please let me know how can I have the .a3m file?

バグの内容

実行環境

$ nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2021 NVIDIA Corporation
Built on Sun_Feb_14_21:12:58_PST_2021
Cuda compilation tools, release 11.2, V11.2.152
Build cuda_11.2.r11.2/compiler.29618528_0

実行コマンド

colabfold-conda/bin/python3.7 runner_af2advanced.py \
--input hoge.fasta \
--output_dir hoge \
--max_recycle 6 \
--msa_method single_sequence \
--num_relax Top5

エラー内容

Traceback (most recent call last):
  File "runner_af2advanced.py", line 161, in <module>
    pair_mode, pair_cov, pair_qid, precomputed=precomputed, TMP_DIR=TMP_DIR)
NameError: name 'precomputed' is not defined

問題の推測

msa_methodの引数をprecomputed以外にした際に、precomputedを変数として定義する部分が存在しないので161行目で変数がないというエラーになるのではないでしょうか？

解決方法

以下のようにelseを追加して、precomputedを定義する部分を作成してはいかがでしょうか？
8b2f2ff

Hi Yoshitaka,

I'm trying LocalColabFold on Windows 11 with WSL2, but it doesn't work.
Errors

2022-01-12 06:00:26,993 Running colabfold 1.2.0 (ee8e17402a3ce8aa67669d3ea22958ef99b808d9)
2022-01-12 06:00:27,001 Found 8 citations for tools or databases
2022-01-12 06:00:30.889012: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 1073741824 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:30.948473: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 966367744 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:31.001558: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 869731072 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:31.058133: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 782758144 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:31.115851: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 704482304 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:31.169997: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 634034176 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:31.222446: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 570630912 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:31.276942: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 513568000 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:31.331319: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 462211328 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:31.386146: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 415990272 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:31.442584: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 374391296 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:31.498272: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 336952320 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:31.554731: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 303257088 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:31.609158: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 272931584 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:31.671022: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 245638656 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:31.729073: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 221074944 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:31.786261: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 198967552 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:31.844108: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 179070976 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:31.895323: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 161164032 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:31.950303: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 145047808 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:32.003178: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 130543104 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:32.053485: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 117488896 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:32.105073: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 105740032 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:32.163281: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 95166208 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:32.216956: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 85649664 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:32.266061: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 77084928 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:32.500940: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 2147483648 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:32.512066: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 2147483648 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:42.538494: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 2147483648 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:42.551197: E external/org_tensorflow/tensorflow/stream_executor/cuda/cuda_driver.cc:771] failed to alloc 2147483648 bytes unified memory; result: CUDA_ERROR_OUT_OF_MEMORY: out of memory
2022-01-12 06:00:42.551242: W external/org_tensorflow/tensorflow/core/common_runtime/bfc_allocator.cc:462] Allocator (GPU_0_bfc) ran out of memory trying to allocate 12.00MiB (rounded to 12582912)requested by op
2022-01-12 06:00:42.551569: W external/org_tensorflow/tensorflow/core/common_runtime/bfc_allocator.cc:474] ****************************************************************************************************
2022-01-12 06:00:42.551647: E external/org_tensorflow/tensorflow/compiler/xla/pjrt/pjrt_stream_executor_client.cc:2085] Execution of replica 0 failed: RESOURCE_EXHAUSTED: Out of memory while trying to allocate 12582912 bytes.
BufferAssignment OOM Debugging.
BufferAssignment stats:
             parameter allocation:   12.00MiB
              constant allocation:         0B
        maybe_live_out allocation:   12.00MiB
     preallocated temp allocation:         0B
                 total allocation:   24.00MiB
              total fragmentation:         0B (0.00%)

allocation 0: 0x5576515bd320, size 12582912, output shape is |f32[48,512,128]|, maybe-live-out:
 value: <1 copy @0> (size=12582912,offset=0): f32[48,512,128]{2,1,0}
contains:<1 copy @0>
 positions:
  copy
 uses:
 from instruction:%copy = f32[48,512,128]{2,1,0} copy(f32[48,512,128]{2,1,0} %parameter.1)

allocation 1: 0x5576515bd3d0, size 12582912, parameter 0, shape |f32[48,512,128]| at ShapeIndex {}:
 value: <0 parameter.1 @0> (size=12582912,offset=0): f32[48,512,128]{2,1,0}
contains:<0 parameter.1 @0>
 positions:
  parameter.1
 uses:
  copy, operand 0
 from instruction:%parameter.1 = f32[48,512,128]{2,1,0} parameter(0)


Traceback (most recent call last):
  File "/home/yuji/colabfold_batch/colabfold-conda/bin/colabfold_batch", line 8, in <module>
    sys.exit(main())
  File "/home/yuji/colabfold_batch/colabfold-conda/lib/python3.7/site-packages/colabfold/batch.py", line 1318, in main
    zip_results=args.zip,
  File "/home/yuji/colabfold_batch/colabfold-conda/lib/python3.7/site-packages/colabfold/batch.py", line 980, in run
    rank_by=rank_by,
  File "/home/yuji/colabfold_batch/colabfold-conda/lib/python3.7/site-packages/colabfold/alphafold/models.py", line 82, in load_models_and_params
    model_name=model_name + model_suffix, data_dir=str(data_dir)
  File "/home/yuji/colabfold_batch/colabfold-conda/lib/python3.7/site-packages/alphafold/model/data.py", line 37, in get_model_haiku_params
    return utils.flat_params_to_haiku(params)
  File "/home/yuji/colabfold_batch/colabfold-conda/lib/python3.7/site-packages/alphafold/model/utils.py", line 80, in flat_params_to_haiku
    hk_params[scope][name] = jnp.array(array)
  File "/home/yuji/colabfold_batch/colabfold-conda/lib/python3.7/site-packages/jax/_src/numpy/lax_numpy.py", line 3597, in array
    out = lax._convert_element_type(out, dtype, weak_type=weak_type)
  File "/home/yuji/colabfold_batch/colabfold-conda/lib/python3.7/site-packages/jax/_src/lax/lax.py", line 481, in _convert_element_type
    weak_type=new_weak_type)
  File "/home/yuji/colabfold_batch/colabfold-conda/lib/python3.7/site-packages/jax/core.py", line 272, in bind
    out = top_trace.process_primitive(self, tracers, params)
  File "/home/yuji/colabfold_batch/colabfold-conda/lib/python3.7/site-packages/jax/core.py", line 624, in process_primitive
    return primitive.impl(*tracers, **params)
  File "/home/yuji/colabfold_batch/colabfold-conda/lib/python3.7/site-packages/jax/interpreters/xla.py", line 418, in apply_primitive
    return compiled_fun(*args)
  File "/home/yuji/colabfold_batch/colabfold-conda/lib/python3.7/site-packages/jax/interpreters/xla.py", line 442, in <lambda>
    return lambda *args, **kw: compiled(*args, **kw)[0]
  File "/home/yuji/colabfold_batch/colabfold-conda/lib/python3.7/site-packages/jax/interpreters/xla.py", line 1100, in _execute_compiled
    out_bufs = compiled.execute(input_bufs)
RuntimeError: RESOURCE_EXHAUSTED: Out of memory while trying to allocate 12582912 bytes.
BufferAssignment OOM Debugging.
BufferAssignment stats:
             parameter allocation:   12.00MiB
              constant allocation:         0B
        maybe_live_out allocation:   12.00MiB
     preallocated temp allocation:         0B
                 total allocation:   24.00MiB
              total fragmentation:         0B (0.00%)

allocation 0: 0x5576515bd320, size 12582912, output shape is |f32[48,512,128]|, maybe-live-out:
 value: <1 copy @0> (size=12582912,offset=0): f32[48,512,128]{2,1,0}
contains:<1 copy @0>
 positions:
  copy
 uses:
 from instruction:%copy = f32[48,512,128]{2,1,0} copy(f32[48,512,128]{2,1,0} %parameter.1)

allocation 1: 0x5576515bd3d0, size 12582912, parameter 0, shape |f32[48,512,128]| at ShapeIndex {}:
 value: <0 parameter.1 @0> (size=12582912,offset=0): f32[48,512,128]{2,1,0}
contains:<0 parameter.1 @0>
 positions:
  parameter.1
 uses:
  copy, operand 0
 from instruction:%parameter.1 = f32[48,512,128]{2,1,0} parameter(0)

nvcc --version and nvidia-smi information are below.

yuji@DESKTOP-B10GKOG:~$ nvidia-smi
Wed Jan 12 05:57:10 2022
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 511.04.01    Driver Version: 511.09       CUDA Version: 11.6     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA GeForce ...  On   | 00000000:01:00.0  On |                  N/A |
|  0%   37C    P8    10W / 170W |    521MiB / 12288MiB |      3%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+
yuji@DESKTOP-B10GKOG:~$ nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2021 NVIDIA Corporation
Built on Thu_Nov_18_09:45:30_PST_2021
Cuda compilation tools, release 11.5, V11.5.119
Build cuda_11.5.r11.5/compiler.30672275_0

お忙しいところ恐縮ですが、ご教授お願いします。

Dear all,

I am running the following command:

export NVIDIA_VISIBLE_DEVICES='all'
export TF_FORCE_UNIFIED_MEMORY='1'
export XLA_PYTHON_CLIENT_MEM_FRACTION='4.0'
colabfold-conda/bin/python3.7 runner_af2advanced.py --input RctB.fasta --output_dir RctB_monomer --max_recycle 3 --use_ptm --use_turbo --num_relax Top1

The sequence has 658 residues and it has been running for more than 12 hours on a 4xGPUs machine (RTX 2080Ti with 12 GB).

ColabFold on Linux
2021-10-09 13:03:38.900883: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudart.so.11.0
2021-10-09 13:03:39.713378: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcuda.so.1
2021-10-09 13:03:39.734361: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1733] Found device 0 with properties:
pciBusID: 0000:19:00.0 name: NVIDIA GeForce RTX 2080 Ti computeCapability: 7.5
coreClock: 1.545GHz coreCount: 68 deviceMemorySize: 10.76GiB deviceMemoryBandwidth: 573.69GiB/s
2021-10-09 13:03:39.735057: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1733] Found device 1 with properties:
pciBusID: 0000:1a:00.0 name: NVIDIA GeForce RTX 2080 Ti computeCapability: 7.5
coreClock: 1.545GHz coreCount: 68 deviceMemorySize: 10.76GiB deviceMemoryBandwidth: 573.69GiB/s
2021-10-09 13:03:39.735707: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1733] Found device 2 with properties:
pciBusID: 0000:67:00.0 name: NVIDIA GeForce RTX 2080 Ti computeCapability: 7.5
coreClock: 1.545GHz coreCount: 68 deviceMemorySize: 10.76GiB deviceMemoryBandwidth: 573.69GiB/s
2021-10-09 13:03:39.736356: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1733] Found device 3 with properties:
pciBusID: 0000:68:00.0 name: NVIDIA GeForce RTX 2080 Ti computeCapability: 7.5
coreClock: 1.545GHz coreCount: 68 deviceMemorySize: 10.76GiB deviceMemoryBandwidth: 573.69GiB/s
2021-10-09 13:03:39.736390: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudart.so.11.0
2021-10-09 13:03:39.738905: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcublas.so.11
2021-10-09 13:03:39.738953: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcublasLt.so.11
2021-10-09 13:03:39.739855: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcufft.so.10
2021-10-09 13:03:39.740062: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcurand.so.10
2021-10-09 13:03:39.742704: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcusolver.so.11
2021-10-09 13:03:39.743337: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcusparse.so.11
2021-10-09 13:03:39.743445: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudnn.so.8
2021-10-09 13:03:39.749167: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1871] Adding visible gpu devices: 0, 1, 2, 3
Input ID: RctB_V.Cholereae
Input Sequence:
homooligomer: 1
total_length: 658
output_dir: RctB_monomer
running mmseqs2
0%| | 0/5 [elapsed: 00:00 remaining: ?]
2021-10-09 13:03:44.719056: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: AVX2 AVX512F FMA
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
2021-10-09 13:03:44.721174: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1258] Device interconnect StreamExecutor with strength 1 edge matrix:
2021-10-09 13:03:44.721200: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1264]
2021-10-09 13:03:44.762135: I tensorflow/core/platform/profile_utils/cpu_utils.cc:114] CPU Frequency: 3100000000 Hz
model_1_ptm_seed_0 recycles:3 tol:6.20 pLDDT:54.26 pTMscore:0.30

I get figures 1 and 2. From figure 2 it seems that it already got a decent model but I do not have any pdb file. Moreover, after a command "top" I cannot see any process related to the run. and after "nvidia-smi" I get this:

I can see that colabfold-conda/bin.python3.7 is running in all four GPUS, but only the memory of GPU 0 is currently been relatively used, but at the same time the GPU-Utilization is 0 %. Is this the normal behaviour or I am doing something wrong?

Thanks a lot in advance.
Best regards
Ariel

hello, I meet the problem when I use mmseqs2. I don't know how to solve it

Thank you

It seems that command line argments are only available in runner_af2advanced.py according to README.md.
However, when runner_af2advanced.py was added, runner.py was also changed to require the command line argment "num_relax".
Even though num_relax passed to runner.py, it gives an error because args is not defined in runner.py.

Hi Yoshitaka-san,

I was trying my_af2_input.fasta.txt file. Notice that there, I use : for modeling protein-protein hetero-complexes.

I used this command line:

/home/ubuntu/storage1/colabfold/bin/colabfold --input my_af2_input.fasta.txt \
    --output my_output_dir \
    --use_ptm \
    -p paired \
    -r 10 \
    --num_relax Top5

Basically I wanted to optimize the result with pairing option.

I get the following error

COLABFOLD_PATH is set to /home/ubuntu/storage1/colabfold
Input ID: lrp1/hmgb1.144
Input Sequence: CSNLCLLSPGGGHKCACPTNFYLGSDGRTCVSNCTASQFVCKNDKCIPFWWKCDTEDDCGDHSDEPPDCPEFKCRPGQFQCSTGICTNPAFICDGDNDCQDNSDEANCDIHVCLPSQFKCTNTNRCIPGIFRCNGQDNCGDGEDERDCPEVTCAPNQFQCSITKRCIPRVWVCDRDNDCVDGSDEPANCTQMTCGVDEFRCKDSGRCIPARWKCDGEDDCGDGSDEPKEECDERTCEPYQFRCKNNRCVPGRWQCDYDNDCGDNSDEESCTPRPCSESEFSCANGRCIAGRWKCDGDHDCADGSDEKDCTPRCDMDQFQCKSGHCIPLRWRCDADADCMDGSDEEACGTGVRTCPLDEFQCNNTLCKPLAWKCDGEDDCGDNSDENPEECARFVCPPNRPFRCKNDRVCLWIGRQCDGTDNCGDGTDEEDCEPPTAHTTHCKDKKEFLCRNQRCLSSSLRCNMFDDCGDGSDEEDCSIDPKLTSCATNASICGDEARCVR:MGKGDPKKPRGKMSSYAFFVQTCREEHKKKHPDASVNFSEFSKK
homooligomer: 1:1
total_length: 544
output_dir: /home/ubuntu/storage1/make_or_dock_peptide_with_alphafold2/af2_output/my_output_dir
attempting to pair some sequences...
running mmseqs2_noenv_nofilter on all seqs
prepping seq_0
prepping seq_1
attempting pairwise stitch for 0 1
  0%|          | 0/13025 [elapsed: 00:00 remaining: ?]
sh: 1: colabfold-conda/bin/hhfilter: not found
Traceback (most recent call last):
  File "/home/ubuntu/storage1/colabfold/runner_af2advanced.py", line 163, in <module>
    hhfilter_loc="colabfold-conda/bin/hhfilter", precomputed=precomputed, TMP_DIR=output_dir)
  File "/home/ubuntu/storage1/colabfold/colabfold_alphafold.py", line 368, in prep_msa
    for line in open(f"{TMP_DIR}/tmp.id90.fas","r"):
FileNotFoundError: [Errno 2] No such file or directory: '/home/ubuntu/storage1/make_or_dock_peptide_with_alphafold2/af2_output/hmgb1.144_dock_lrp1_yamada.paired.r10/tmp.id90.fas'

It has no problem when I use -p option.
How can I resolve it?

G.V.

Dear all,

Question:
I am test colabfold_batch on my linux machine, and while it works like a charm, but it seems that when I run it on the background with "&” and log out, it start running but it dies during the first model prediction.

Solved:
I simply used nohup to run on the background and it worked perfectly.

Thanks a lot in any case,
Fabian Glaser
Technion, Israel

Hi, thank you for a great effort. I have a problem after installing on Linux CentOS with message related to the jaxlib version:

ValueError: jaxlib is version 0.1.72, but this version of jax requires version 0.1.74.

And if I try to run update script it returns error as well:

bash-4.2$ pwd

/hpc/shared/colabfold_batch

bash-4.2$ ls

bin  colabfold-conda  conda  stereo_chemical_props.txt  update_linux.sh

bash-4.2$ ./update_linux.sh

Error! colabfold-conda directory is not present in .

Hi,

I'm trying the most recent version of localcolabfold, with this sequence:

AVLKIIQGALDTRELLKAYQEEACAKNFGAFCVFVGIVRKEDNIQGLSFDIYEA

My command line is this:

colabfold/bin/colabfold --input $INPUT_FASTA \
    --output $OUTPUT_DIR \
    --use_turbo \
    --use_ptm \
    --num_relax Top5 \
    --homooligomer 1

The error I get is this:

model rank based on pLDDT
rank_1_model_3_ptm_seed_0 pLDDT:88.87
rank_2_model_4_ptm_seed_0 pLDDT:88.20
rank_3_model_2_ptm_seed_0 pLDDT:86.50
rank_4_model_5_ptm_seed_0 pLDDT:85.70
rank_5_model_1_ptm_seed_0 pLDDT:84.93
  0%|          | 0/5 [elapsed: 00:00 remaining: ?]
Traceback (most recent call last):
  File "/home/ubuntu/storage1/colabfold/runner_af2advanced.py", line 289, in <module>
    relaxed_pdb_lines, _, _ = amber_relaxer.process(prot=outs[key]["unrelaxed_protein"])
  File "/home/ubuntu/storage1/colabfold/alphafold/relax/relax.py", line 62, in process
    max_outer_iterations=self._max_outer_iterations)
  File "/home/ubuntu/storage1/colabfold/alphafold/relax/amber_minimize.py", line 476, in run_pipeline
    max_attempts=max_attempts)
  File "/home/ubuntu/storage1/colabfold/alphafold/relax/amber_minimize.py", line 415, in _run_one_iteration
    raise ValueError(f"Minimization failed after {max_attempts} attempts.")

But when I replace it with :

--num_relax None

It has no problem at all.

Before I never encounter such issue, even with the same sequence.
Is there a way we can deal with this?

The relaxed version is the preferred one, am I right?

Thanks and hope to hear from you again.

G.V.

Running a very long CPU only job (3 days so far) and model 3 finally finished, model 4 just started. Can I access model 3 pdb at this stage, it is not in the output folder?

Thank you again, Yoshitaka for this amazing tool!

Everything works well with installation, however when I try to run colab

colabfold_batch --amber --templates --num-recycle 3 --num-models 1 /Users/jihadsnobre1/Desktop/df.csv /Users/jihadsnobre1/Applications/output_files/

I get this error:

Running colabfold 1.2.0 (f5d0cec9e4045666cb615ec9ad89b3bdc1ecda62)
 Found 8 citations for tools or databases
Query 1/1: rpoB (length 1172)
Sequence 0 found templates: [b'6dv9_C' b'6edt_C' b'5zx3_C' b'5uaj_C' b'4yfk_I' b'6pst_I' b'6j9e_C'
 b'2be5_M' b'2o5i_C' b'6woy_C' b'6j9f_C' b'6kf9_B' b'6rfl_B' b'4ayb_B'
 b'5iy9_B' b'6gmh_B' b'5fja_B' b'6rie_B' b'4qiw_B' b'6tut_B']
Running model_3
E external/org_tensorflow/tensorflow/compiler/xla/service/slow_operation_alarm.cc:55] 
********************************
Very slow compile?  If you want to file a bug, run with envvar XLA_FLAGS=--xla_dump_to=/tmp/foo and attach the results.
Compiling module jit_apply_fn.109930
********************************

I tried to reduce --num-models to 1 and to run and also -- cpu mode, I keep getting the same error.
Any idea of how to fix that?

Many thanks

At line 46, it should be:

parser.add_argument("--num_samples", default=1, type=int, help="number of random_seeds to try. Default is 1.")

Apologies for not doing this correctly but my local advanced.py has other differences. For recycle tolerance to work, this must be added:

parser.add_argument("--tol", default=0, type=float, help="tolerance for deciding when to stop (CA-RMS between recycles)")

Attached is a patch (fix_tol.txt) to make this change and also two others in my local advanced script that are useful for me:

1. Reuse existing pickle.msa or specify a directory where one is; should add a check that the sequence matches
2. Change --num_models behavior to take a list of models to run as a string e.g. "--num_models 35" runs models 3 and 5 only.

Thank you for making this available -- very nice to get around the colab size limit and increasing policing of interactive use; this installed with no trouble on google cloud program deep learning VMs (except for the same array/list issue for relaxation) using the default K40 and A100 configurations.

Thank you very much for this useful tool!

Installation worked fine, but now I don't seem to be able to run colabfold.

When I run:
colabfold_batch --amber --templates --num-recycle 3 /Users/jihadsnobre1/Applications/input_files/seq.csv /Users/jihadsnobre1/Applications/output_files/

I get this error:
Traceback (most recent call last): File "/Users/jihadsnobre1/Applications/colabfold_batch/colabfold-conda/bin/colabfold_batch", line 8, in <module> sys.exit(main()) File "/Users/jihadsnobre1/Applications/colabfold_batch/colabfold-conda/lib/python3.7/site-packages/colabfold/batch.py", line 1288, in main queries, is_complex = get_queries(args.input, args.sort_queries_by) File "/Users/jihadsnobre1/Applications/colabfold_batch/colabfold-conda/lib/python3.7/site-packages/colabfold/batch.py", line 385, in get_queries assert "id" in df.columns and "sequence" in df.columns AssertionError

Any idea of how to fix this? I wrote ta single protein sequence in a csv file and used that as input.

Perhaps a redundant question but I'm asking it anyway as I lack the expertise to verify it entirely on my own, but is it theoretically possible (and if so, how) to run colabfold locally with an AMD GPU (specifically a radeon rx 580) through another API such as Vulkan or openCL instead of Cuda?

If amino acids joined with ':' are passed as sequence in runner.py, it will crash because pair_mode is not defiend.

https://github.com/YoshitakaMo/localcolabfold/blob/main/runner.py#L303

Hi, the last commit seems to have an issue by hanging here:

(...)
Successfully installed jax-0.2.21 jaxlib-0.1.69+cuda111
Applying OpenMM patch...
can't find file to patch at input line 5
Perhaps you used the wrong -p or --strip option?
The text leading up to this was:

|Index: simtk/openmm/app/topology.py
|===================================================================
|--- simtk.orig/openmm/app/topology.py
|+++ simtk/openmm/app/topology.py

File to patch: