AlphaFold3-Sapelo2 - Revision history

Shtsai at 14:02, 30 March 2026

2026-03-30T14:02:23Z

← Older revision		Revision as of 10:02, 30 March 2026
Line 1:		Line 1:
	[[Category:Sapelo2]][[Category:Software]][[Category:Bioinformatics]]		[[Category:Sapelo2]][[Category:Software]][[Category:Bioinformatics]]

	=== Category ===		== Category ==
	Bioinformatics		Bioinformatics

	=== Program On ===		== Program On ==
	Sapelo2		Sapelo2

	=== Version ===		== Version ==
	3.0.0, 3.0.1		3.0.0, 3.0.1

	=== Author / Distributor ===		== Author / Distributor ==
	Please see https://github.com/google-deepmind/alphafold3		Please see https://github.com/google-deepmind/alphafold3

	=== Description ===		== Description ==
	From https://github.com/google-deepmind/alphafold3: "This package provides an implementation of the inference pipeline of AlphaFold 3. "		From https://github.com/google-deepmind/alphafold3: "This package provides an implementation of the inference pipeline of AlphaFold 3. "

	=== Running Program ===		== Running Program ==
	Also refer to [[Running Jobs on Sapelo2]]		Also refer to [[Running Jobs on Sapelo2]]

Line 110:		Line 110:
	Note: the above homodimer protein runs on both the H100 and A100 nodes in under an hour with 32 CPUs, so if you find your job is sitting in the queue for long periods of time, you could consider requesting significantly less than 120 hours (especially if your protein is either small or is otherwise well characterized and thus likely to align to the databases relatively quickly).		Note: the above homodimer protein runs on both the H100 and A100 nodes in under an hour with 32 CPUs, so if you find your job is sitting in the queue for long periods of time, you could consider requesting significantly less than 120 hours (especially if your protein is either small or is otherwise well characterized and thus likely to align to the databases relatively quickly).

	=== Documentation ===		== Documentation ==

	Details and references are at: https://github.com/google-deepmind/alphafold3		Details and references are at: https://github.com/google-deepmind/alphafold3

Jordan: /* Running Program */

2025-12-05T15:05:13Z

Running Program

← Older revision		Revision as of 11:05, 5 December 2025
Line 81:		Line 81:
	--bind /Path/to/model_parameters:/root/models \		--bind /Path/to/model_parameters:/root/models \
	--bind /db/AlphaFold3/20241114:/root/public_databases \		--bind /db/AlphaFold3/20241114:/root/public_databases \
	/apps/singularity-images/alphafold-3.0.0.sif \		/apps/singularity-images/alphafold-3.0.1-JSON-4.sif \
	python /app/alphafold/run_alphafold.py \		python /app/alphafold/run_alphafold.py \
	--json_path=/root/af_input/NAME_OF_INPUT_FILE.json \		--json_path=/root/af_input/NAME_OF_INPUT_FILE.json \
Line 90:		Line 90:
	</pre>		</pre>

	~~Example~~ input JSON ~~file~~:		You can find extensive information on formatting input JSON files <u>[https://github.com/google-deepmind/alphafold3/blob/main/docs/input.md here]</u>, but the following is a simple example:
	<pre class="gcommand">		<pre class="gcommand">
	{		{

Jordan: /* Running Program */

2025-09-25T20:27:14Z

Running Program

← Older revision		Revision as of 16:27, 25 September 2025
Line 53:		Line 53:
	</pre>		</pre>

	Please also note that like many AI tools, AlphaFold is prone to "hallucinations" when dealing with inputs that are too large; this mostly manifests as results with polypeptides predicted to occupy the same 3D space, which obviously makes these results useless. While tests with a complex of up to ~9,700 tokens (the urease complex from <i>Helicobacter pylori</i>) completed successfully, to date, the largest complex observed to run <b>correctly</b> was ~7,~~500~~ tokens (the ornithine decarboxylase, specifically <i>SpeF</i>, complex from <i>Salmonella</i>).		Please also note that like many AI tools, AlphaFold is prone to "hallucinations" when dealing with inputs that are too large; this mostly manifests as results with polypeptides predicted to occupy the same 3D space, which obviously makes these results useless. While tests with a complex of up to ~9,700 tokens (the urease complex from <i>Helicobacter pylori</i>) completed successfully, to date, the largest complex observed to run <b>correctly</b> was ~7,300 tokens (the ornithine decarboxylase, specifically <i>SpeF</i>, complex from <i>Salmonella</i>).
	</br>		</br>
	</br>		</br>

Jordan: /* Running Program */

2025-09-24T13:55:49Z

Running Program

← Older revision		Revision as of 09:55, 24 September 2025
Line 46:		Line 46:
	singularity exec /apps/singularity-images/alphafold-3.0.1.sif python /app/alphafold/run_alphafold.py --helpfull		singularity exec /apps/singularity-images/alphafold-3.0.1.sif python /app/alphafold/run_alphafold.py --helpfull
	</pre>		</pre>
	Note that AlphaFold 3 is GPU dependent while also being CPU and memory intensive, meaning that it will only run on the A100 and H100 nodes and needs numerous CPUs to run in a timely/efficient manner (see the example job submission script below). <b>Furthemore, the "unifiedmem" and "JSON-4" versions have unified memory enabled, which allows the GPU device to "borrow" host memory from the node for proteins and other complexes that are large enough to exceed the standard maximum of 5,120 tokens; not only does this substantially increase run times, it will also <i>only</i> work on the A100 nodes due to architectural conflicts with the H100 nodes. To use either of these versions with a large protein or other complex, please use the following Slurm header in your submission script in place of the corresponding "--gres" header seen in the sample job below (and note that you ~~don't need~~ the "constraint" header for this):</b>		Note that AlphaFold 3 is GPU dependent while also being CPU and memory intensive, meaning that it will only run on the A100 and H100 nodes and needs numerous CPUs to run in a timely/efficient manner (see the example job submission script below). <b>Furthemore, the "unifiedmem" and "JSON-4" versions have unified memory enabled, which allows the GPU device to "borrow" host memory from the node for proteins and other complexes that are large enough to exceed the standard maximum of 5,120 tokens; not only does this substantially increase run times, it will also <i>only</i> work on the A100 nodes due to architectural conflicts with the H100 nodes. To use either of these versions with a large protein or other complex, please use the following Slurm header in your submission script in place of the corresponding "--gres" header seen in the sample job below (and note that you can remove the "#SBATCH --constraint" header for this):</b>
	<pre class="gscript">		<pre class="gscript">
	...		...
Line 53:		Line 53:
	</pre>		</pre>

			Please also note that like many AI tools, AlphaFold is prone to "hallucinations" when dealing with inputs that are too large; this mostly manifests as results with polypeptides predicted to occupy the same 3D space, which obviously makes these results useless. While tests with a complex of up to ~9,700 tokens (the urease complex from <i>Helicobacter pylori</i>) completed successfully, to date, the largest complex observed to run <b>correctly</b> was ~7,500 tokens (the ornithine decarboxylase, specifically <i>SpeF</i>, complex from <i>Salmonella</i>).
			</br>
			</br>
	AlphaFold 3 also depends on a set of database files, the current versions of which can be found at /db/AlphaFold3/20241114, and a file containing the model parameters; currently this is only available directly from Google, and can be obtained by completing the form found here: https://docs.google.com/forms/d/e/1FAIpQLSfWZAgo1aYk0O4MuAXZj8xRQ8DafeFJnldNOnh_13qAx2ceZw/viewform. Once these are obtained and placed somewhere in your storage space on Sapelo2 (/home or /work is recommended), they must be included with the '-B' or "--bind" option at runtime, as must the location of the input file and output directory (see example below).		AlphaFold 3 also depends on a set of database files, the current versions of which can be found at /db/AlphaFold3/20241114, and a file containing the model parameters; currently this is only available directly from Google, and can be obtained by completing the form found here: https://docs.google.com/forms/d/e/1FAIpQLSfWZAgo1aYk0O4MuAXZj8xRQ8DafeFJnldNOnh_13qAx2ceZw/viewform. Once these are obtained and placed somewhere in your storage space on Sapelo2 (/home or /work is recommended), they must be included with the '-B' or "--bind" option at runtime, as must the location of the input file and output directory (see example below).

Jordan: /* Running Program */

2025-09-12T19:21:28Z

Running Program

← Older revision		Revision as of 15:21, 12 September 2025
Line 46:		Line 46:
	singularity exec /apps/singularity-images/alphafold-3.0.1.sif python /app/alphafold/run_alphafold.py --helpfull		singularity exec /apps/singularity-images/alphafold-3.0.1.sif python /app/alphafold/run_alphafold.py --helpfull
	</pre>		</pre>
	Note that AlphaFold 3 is GPU dependent while also being CPU and memory intensive, meaning that it will only run on the A100 and H100 nodes and needs numerous CPUs to run in a timely/efficient manner (see the example job submission script below). <b>Furthemore, the "unifiedmem" and "JSON-4" versions have unified memory enabled, which allows the GPU device to "borrow" host memory from the node for proteins and other complexes that are large enough to exceed the standard maximum of 5,120 tokens; not only does this substantially increase run times, it will also <i>only</i> work on the A100 nodes due to architectural conflicts with the H100 nodes~~; to~~ use ~~this version~~, please use the following Slurm header in your submission script in place of the corresponding "--gres" header seen in the sample job below (and note that you don't need the "constraint" header for this):</b>		Note that AlphaFold 3 is GPU dependent while also being CPU and memory intensive, meaning that it will only run on the A100 and H100 nodes and needs numerous CPUs to run in a timely/efficient manner (see the example job submission script below). <b>Furthemore, the "unifiedmem" and "JSON-4" versions have unified memory enabled, which allows the GPU device to "borrow" host memory from the node for proteins and other complexes that are large enough to exceed the standard maximum of 5,120 tokens; not only does this substantially increase run times, it will also <i>only</i> work on the A100 nodes due to architectural conflicts with the H100 nodes. To use either of these versions with a large protein or other complex, please use the following Slurm header in your submission script in place of the corresponding "--gres" header seen in the sample job below (and note that you don't need the "constraint" header for this):</b>
	<pre class="gscript">		<pre class="gscript">
	...		...

Jordan: /* Running Program */

2025-09-12T19:20:11Z

Running Program

← Older revision		Revision as of 15:20, 12 September 2025
Line 39:		Line 39:

	/apps/singularity-images/alphafold-3.0.1-unifiedmem.sif		/apps/singularity-images/alphafold-3.0.1-unifiedmem.sif

			/apps/singularity-images/alphafold-3.0.1-JSON-4.sif
	</pre>		</pre>
	You can view the documentation for this version of AlphaFold with the following command, in an interactive job (i.e., NOT a login/submit node, where Singularity does not work):		You can view the documentation for this version of AlphaFold with the following command, in an interactive job (i.e., NOT a login/submit node, where Singularity does not work):
Line 44:		Line 46:
	singularity exec /apps/singularity-images/alphafold-3.0.1.sif python /app/alphafold/run_alphafold.py --helpfull		singularity exec /apps/singularity-images/alphafold-3.0.1.sif python /app/alphafold/run_alphafold.py --helpfull
	</pre>		</pre>
	Note that AlphaFold 3 is GPU dependent while also being CPU and memory intensive, meaning that it will only run on the A100 and H100 nodes and needs numerous CPUs to run in a timely/efficient manner (see the example job submission script below). <b>~~More specifically~~, the "unifiedmem" ~~version is~~ for proteins and other complexes that are large enough to exceed the standard maximum of 5,120 tokens, ~~and~~ <i>~~this~~</i> ~~version will only~~ work on the A100 nodes due to architectural conflicts with the H100 nodes; to use this version, please use the following Slurm header in your submission script in place of the corresponding header seen in the sample job below (and note that you don't need the "constraint" header for this):</b>		Note that AlphaFold 3 is GPU dependent while also being CPU and memory intensive, meaning that it will only run on the A100 and H100 nodes and needs numerous CPUs to run in a timely/efficient manner (see the example job submission script below). <b>Furthemore, the "unifiedmem" and "JSON-4" versions have unified memory enabled, which allows the GPU device to "borrow" host memory from the node for proteins and other complexes that are large enough to exceed the standard maximum of 5,120 tokens; not only does this substantially increase run times, it will also <i>only</i> work on the A100 nodes due to architectural conflicts with the H100 nodes; to use this version, please use the following Slurm header in your submission script in place of the corresponding "--gres" header seen in the sample job below (and note that you don't need the "constraint" header for this):</b>
	<pre class="gscript">		<pre class="gscript">
	...		...

Jordan: /* Running Program */

2025-03-25T14:39:58Z

Running Program

← Older revision		Revision as of 10:39, 25 March 2025
Line 51:		Line 51:
	</pre>		</pre>

	AlphaFold 3 also depends on a set of database files, the current versions of which can be found at /db/AlphaFold3/20241114, and a file containing the model parameters; currently this is only available directly from Google, and can be obtained by completing the form found here: https://docs.google.com/forms/d/e/1FAIpQLSfWZAgo1aYk0O4MuAXZj8xRQ8DafeFJnldNOnh_13qAx2ceZw/viewform. Once these are obtained and placed somewhere in your storage space on Sapelo2 (/home or /work is recommended), they must be included with the '-B' or "--bind" option at runtime, as must the location of the input file and output directory (see example below). Also note that although AlphaFold 3 was created for CUDA v12.6 and we currently only have CUDA v12.2, AlphaFold 3 will still work (it prints a message noting the fact that it's running in an old version of CUDA, but then proceeds to run anyway).		AlphaFold 3 also depends on a set of database files, the current versions of which can be found at /db/AlphaFold3/20241114, and a file containing the model parameters; currently this is only available directly from Google, and can be obtained by completing the form found here: https://docs.google.com/forms/d/e/1FAIpQLSfWZAgo1aYk0O4MuAXZj8xRQ8DafeFJnldNOnh_13qAx2ceZw/viewform. Once these are obtained and placed somewhere in your storage space on Sapelo2 (/home or /work is recommended), they must be included with the '-B' or "--bind" option at runtime, as must the location of the input file and output directory (see example below).

	Sample job submission script to run the singularity container for v3.0.0 on a GPU:		Sample job submission script to run the singularity container for v3.0.0 on a GPU:

Jordan: /* Running Program */

2025-03-25T14:16:36Z

Running Program

← Older revision		Revision as of 10:16, 25 March 2025
Line 44:		Line 44:
	singularity exec /apps/singularity-images/alphafold-3.0.1.sif python /app/alphafold/run_alphafold.py --helpfull		singularity exec /apps/singularity-images/alphafold-3.0.1.sif python /app/alphafold/run_alphafold.py --helpfull
	</pre>		</pre>
	Note that AlphaFold 3 is GPU dependent while also being CPU and memory intensive, meaning that it will only run on the A100 and H100 nodes and needs numerous CPUs to run in a timely/efficient manner (see the example job submission script below). More specifically, the "unifiedmem" version is for proteins and other complexes that are large enough to exceed the standard maximum of 5,120 tokens, and <i>this</i> version will only work on the A100 nodes due to architectural conflicts with the H100 nodes; to use this version, please use the following Slurm header in your submission script in place of the corresponding header seen in the sample job below (and note that you don't need the "constraint" header for this):		Note that AlphaFold 3 is GPU dependent while also being CPU and memory intensive, meaning that it will only run on the A100 and H100 nodes and needs numerous CPUs to run in a timely/efficient manner (see the example job submission script below). <b>More specifically, the "unifiedmem" version is for proteins and other complexes that are large enough to exceed the standard maximum of 5,120 tokens, and <i>this</i> version will only work on the A100 nodes due to architectural conflicts with the H100 nodes; to use this version, please use the following Slurm header in your submission script in place of the corresponding header seen in the sample job below (and note that you don't need the "constraint" header for this):</b>
	<pre class="gscript">		<pre class="gscript">
	...		...

Jordan: /* Running Program */

2025-03-25T14:14:09Z

Running Program

← Older revision		Revision as of 10:14, 25 March 2025
Line 27:		Line 27:
	/apps/singularity-images/alphafold-3.0.0-CCDpatched.sif		/apps/singularity-images/alphafold-3.0.0-CCDpatched.sif
	</pre>		</pre>
	Both of these containers have worked for multiple users, but the second one ("CCDpatched") seems to work better for users working on proteins with one or more ligands. You can view the documentation for this version of AlphaFold with the following command, on an interactive (i.e., NOT a login/submit) ~~node~~:		Both of these containers have worked for multiple users, but the second one ("CCDpatched") seems to work better for users working on proteins with one or more ligands. You can view the documentation for this version of AlphaFold with the following command, in an interactive job (i.e., NOT a login/submit node, where Singularity does not work):
	<pre class="gcommand">		<pre class="gcommand">
	singularity exec /apps/singularity-images/alphafold-3.0.0.sif python /app/alphafold/run_alphafold.py --helpfull		singularity exec /apps/singularity-images/alphafold-3.0.0.sif python /app/alphafold/run_alphafold.py --helpfull
Line 37:		Line 37:
	<pre class="gscript">		<pre class="gscript">
	/apps/singularity-images/alphafold-3.0.1.sif		/apps/singularity-images/alphafold-3.0.1.sif

			/apps/singularity-images/alphafold-3.0.1-unifiedmem.sif
	</pre>		</pre>
	You can view the documentation for this version of AlphaFold with the following command, on an interactive (i.e., NOT a login/submit) ~~node~~:		You can view the documentation for this version of AlphaFold with the following command, in an interactive job (i.e., NOT a login/submit node, where Singularity does not work):
	<pre class="gcommand">		<pre class="gcommand">
	singularity exec /apps/singularity-images/alphafold-3.0.1.sif python /app/alphafold/run_alphafold.py --helpfull		singularity exec /apps/singularity-images/alphafold-3.0.1.sif python /app/alphafold/run_alphafold.py --helpfull
	</pre>		</pre>
	Note that AlphaFold 3 is GPU dependent while also being CPU and memory intensive, meaning that it will only run on the A100 and H100 nodes and needs numerous CPUs to run in a timely/efficient manner (see the example job submission script below). ~~AlphaFold 3 also depends on a set of database files~~, the ~~current versions~~ of ~~which can be found at /db/AlphaFold3/20241114~~, and ~~a file containing~~ the ~~model parameters~~; ~~currently~~ this ~~is only available directly from Google~~, ~~and can be obtained by completing~~ the ~~form found here: https://docs.google.com/forms/d/e/1FAIpQLSfWZAgo1aYk0O4MuAXZj8xRQ8DafeFJnldNOnh_13qAx2ceZw/viewform. Once these are obtained and placed somewhere~~ in your ~~storage space on Sapelo2~~ (~~/home or /work is recommended), they must be included with~~ the ~~'-B' or~~ "~~--bind~~" ~~option at runtime, as must the location of the input file and output directory (see example below~~). ~~Also note that although AlphaFold 3 was created for CUDA v12~~.~~6 and we currently only have CUDA v12~~.~~2, AlphaFold 3 will still work (it prints a message noting the fact that it's running in an old version of CUDA, but then proceeds to run anyway)~~.		Note that AlphaFold 3 is GPU dependent while also being CPU and memory intensive, meaning that it will only run on the A100 and H100 nodes and needs numerous CPUs to run in a timely/efficient manner (see the example job submission script below). More specifically, the "unifiedmem" version is for proteins and other complexes that are large enough to exceed the standard maximum of 5,120 tokens, and <i>this</i> version will only work on the A100 nodes due to architectural conflicts with the H100 nodes; to use this version, please use the following Slurm header in your submission script in place of the corresponding header seen in the sample job below (and note that you don't need the "constraint" header for this):
			<pre class="gscript">
			...
			#SBATCH --gres=gpu:A100:1
			...
			</pre>

			AlphaFold 3 also depends on a set of database files, the current versions of which can be found at /db/AlphaFold3/20241114, and a file containing the model parameters; currently this is only available directly from Google, and can be obtained by completing the form found here: https://docs.google.com/forms/d/e/1FAIpQLSfWZAgo1aYk0O4MuAXZj8xRQ8DafeFJnldNOnh_13qAx2ceZw/viewform. Once these are obtained and placed somewhere in your storage space on Sapelo2 (/home or /work is recommended), they must be included with the '-B' or "--bind" option at runtime, as must the location of the input file and output directory (see example below). Also note that although AlphaFold 3 was created for CUDA v12.6 and we currently only have CUDA v12.2, AlphaFold 3 will still work (it prints a message noting the fact that it's running in an old version of CUDA, but then proceeds to run anyway).

	Sample job submission script to run the singularity container for v3.0.0 on a GPU:		Sample job submission script to run the singularity container for v3.0.0 on a GPU:

Jordan: /* Version */

2025-03-11T13:06:52Z

Version

← Older revision		Revision as of 09:06, 11 March 2025
Line 8:		Line 8:

	=== Version ===		=== Version ===
	3.0.0		3.0.0, 3.0.1

	=== Author / Distributor ===		=== Author / Distributor ===