Skip to content

Commit b2079ff

Browse files
Revise README for inference script usage and outputs
Updated instructions for running inference scripts and clarified expected outputs.
1 parent 2524ce7 commit b2079ff

File tree

1 file changed

+30
-6
lines changed

1 file changed

+30
-6
lines changed

README.md

Lines changed: 30 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -86,9 +86,10 @@ Equivalent python environments can be set up using conda, or an environment mana
8686
All dependencies and packages are specified in the files found in `rf_diffusion/environment` (see `environment.yml` and `SE3nv.spec`)
8787

8888
## Testing that everything works (the *DEMO*):
89-
Change directory to wherever you want to perform design, and run example inference script:
89+
Change directory to run from within the RFDpoly directory useful for filepath searches if you are not providing an input pdb), and run example inference script.
90+
Be sure to specify your desired output directory with `$DESIGN_DIR`.
9091
```Bash
91-
cd $DESIGN_DIR
92+
cd $DESIGN_DIR/
9293

9394
$APPTAINER_PATH $RFDPOLY_DIR/polydiff/rf_diffusion/run_inference.py --config-name=multi_polymer \
9495
diffuser.T=50 \
@@ -101,7 +102,26 @@ inference.output_prefix=$DESIGN_DIR/test_outputs/basic_uncond_test01
101102
The initial run will take a little while to precompute the IGSO3 cache, but subsequent runs will be more direct and quick.
102103

103104

104-
If the example command above works, proceed to exploration of the full [design tutorial](https://github.com/andrewfavor95/polydiff/blob/main/RFDpoly_tutorial.pdf).
105+
if that throws errors, then try:
106+
```
107+
$APPTAINER_PATH $RFDPOLY_DIR/polydiff/rf_diffusion/run_inference.py --config-name=multi_polymer \
108+
diffuser.T=50 \
109+
inference.input_pdb=$RFDPOLY_DIR/polydiff/rf_diffusion/test_data/DBP035.pdb \
110+
inference.num_designs=3 \
111+
contigmap.contigs=[\'33\ 33\ 75\'] \
112+
contigmap.polymer_chains=[\'dna\',\'rna\',\'protein\'] \
113+
inference.output_prefix=$DESIGN_DIR/test_outputs/basic_uncond_test01
114+
115+
116+
```
117+
explanation: model initialization searches for an input pdb filepath, even if you aren't performing motif scaffolding. Providing a (real) dummy filepath will fix this is the default search paths are unsuccessful.
118+
119+
**Expected output:** a three chain .pdb file, containing a complex of DNA (chain A), RNA (chain B), and protein (chain C).
120+
121+
122+
## If the example (demo) command above works, proceed to exploration of the full [design tutorial](https://github.com/andrewfavor95/polydiff/blob/main/RFDpoly_tutorial.pdf).
123+
The full design tutorial contains many inference commands for the types of designs reported in the RFDpoly paper.
124+
The design tutorial also is intented to provide documentation, and explain arguments in the context of their use-cases.
105125

106126

107127

@@ -123,12 +143,16 @@ If the example command above works, proceed to exploration of the full [design t
123143
- Minimum: 16 GB RAM and 40 GB free disk space for the repository, checkpoints, and cache.
124144
- Recommended: NVIDIA GPU with ≥ 16 GB VRAM for practical throughput. CPU-only execution is supported but 5–10× slower. No additional specialized hardware is required.
125145

126-
## Times to install provided files and run inference (tested on Linux):
146+
## Times to install provided files and run inference (tested on Linux operating system, Ubuntu 24.04.2):
127147
- Downloading apptainer .sif file:
128148
- Total: 1 minute, 15 seconds
129149
- Downloading model weight .ckpt file:
130150
- Total: 7 seconds
151+
- Cloning the repository:
152+
- Total: 52 seconds
131153
- Demo inference run:
132-
- first run, including generation of IGSO3 cache: Total: 2 minutes, 13 seconds
133-
- subsequent runs (per trajectory):
154+
- first run, including generation of IGSO3 cache: Total: 2 minutes, 26 seconds
155+
- subsequent runs (per trajectory): 50 seconds.
134156

157+
## Testing:
158+
- This software has only been tested on Linux (Ubuntu 24.04.3 LTS).

0 commit comments

Comments
 (0)