Files
RFdiffusion/tests/test_diffusion.py
woodsh17 dd7643d640 Revert "Remove try block with except FileExistsError that isn't needed"
After running the test, I found that this was needed
This reverts commit ead721f326.
2025-11-12 13:45:07 -06:00

416 lines
14 KiB
Python

import unittest
import subprocess
import glob
import datetime
import os
import torch
from shutil import copyfile
from rfdiffusion.inference import utils as iu
from rfdiffusion.util import calc_rmsd
import sys, json
script_dir = os.path.dirname(os.path.abspath(__file__))
class TestSubmissionCommands(unittest.TestCase):
"""
Test harness for checking that commands in the examples folder,
when run in deterministic mode, produce the same output as the
reference outputs.
Requirements:
- example command must be written on a single line
- outputs must be written to example_outputs folder
- needs to be run on the same hardware as the reference outputs (A100 GPU)
For speed, we only run the first 2 steps of diffusion, and set inference.num_designs=1
This means that outputs DO NOT look like proteins, but we can still check that the
outputs are the same as the reference outputs.
"""
failed_tests = []
# number of chunks to split examples into
total_chunks = 1
# which chunk to run
chunk_index = 1
out_f = None
results = {}
exec_status = {}
@classmethod
def setUpClass(cls):
"""
Class-level setup: Grabs files from the examples folder, discover & rewrite example commands, then execute them once.
"""
submissions = glob.glob(f"{script_dir}/../examples/*.sh")
# get datetime for output folder, in YYYY_MM_DD_HH_MM_SS format
chunks = cls.total_chunks
idx = cls.chunk_index
if chunks < 1:
raise ValueError("total_chunks must be at least 1")
if idx < 1 or idx > chunks:
raise ValueError(
"chunk_index must be between 1 and total_chunks (inclusive)"
)
if chunks > 1:
submissions = [
submissions[i]
for i in range(len(submissions))
if i % chunks == (idx - 1)
]
print(
f"Running chunk {idx}/{chunks}, {len(submissions)} submissions to run"
)
if not submissions:
raise ValueError("No submissions selected for chunk {idx} of {chunks}")
now = datetime.datetime.now()
now = now.strftime("%Y_%m_%d_%H_%M_%S")
cls.out_f = f"{script_dir}/tests_{now}_{idx}"
os.mkdir(cls.out_f)
# Make sure we have access to all the relevant files
exclude_dirs = ["outputs", "example_outputs"]
for filename in os.listdir(f"{script_dir}/../examples"):
if (
filename not in exclude_dirs
and not os.path.exists(os.path.join(script_dir, filename))
and os.path.isdir(os.path.join(f"{script_dir}/../examples", filename))
):
try:
os.symlink(
os.path.join(f"{script_dir}/../examples", filename),
os.path.join(script_dir, filename),
)
except FileExistsError:
pass
for submission in submissions:
cls._write_command(submission, cls.out_f)
print(
f"Running commands in {cls.out_f}, two steps of diffusion, deterministic=True"
)
cls.results = {}
cls.exec_status = {}
for bash_file in sorted(glob.glob(f"{cls.out_f}/*.sh"), reverse=False):
test_name = os.path.basename(bash_file)[: -len(".sh")]
res, output = execute(
f"Running {test_name}",
f"bash {bash_file}",
return_="tuple",
add_message_and_command_line_to_output=True,
)
cls.exec_status[test_name] = (res, output)
cls.results[test_name] = dict(
state="failed" if res else "passed",
log=output,
)
# subprocess.run(["bash", bash_file], stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
# subprocess.run(["bash", bash_file])
def test_examples_run_without_errors(self):
for name, (exit_code, output) in sorted(self.__class__.exec_status.items()):
with self.subTest(example=name):
if exit_code != 0:
self.__class__.failed_tests.append(f"{name}")
self.assertEqual(
exit_code,
0,
msg=f"Example '{name}' exited with {exit_code}\n{output}",
)
sys.stderr.write("\n==== EXAMPLE FAILURE SUMMARY ====\n")
for line in self.__class__.failed_tests:
sys.stderr.write(f" - {line}\n")
sys.stderr.write("=========================\n\n")
sys.stderr.flush()
def test_commands(self):
"""
Runs all the commands in the test_f folder
"""
reference = f"{script_dir}/reference_outputs"
os.makedirs(reference, exist_ok=True)
test_files = glob.glob(f"{self.__class__.out_f}/example_outputs/*pdb")
print(f"{self.__class__.out_f=} {test_files=}")
# first check that we have the right number of outputs
# self.assertEqual(len(test_files), len(glob.glob(f"{self.out_f}/*.sh"))), "One or more of the example commands didn't produce an output (check the example command is formatted correctly)"
result = self.defaultTestResult()
for test_file in test_files:
with self.subTest(test_file=test_file):
test_pdb = iu.parse_pdb(test_file)
if not os.path.exists(f"{reference}/{os.path.basename(test_file)}"):
copyfile(test_file, f"{reference}/{os.path.basename(test_file)}")
print(
f"Created reference file {reference}/{os.path.basename(test_file)}"
)
else:
ref_pdb = iu.parse_pdb(f"{reference}/{os.path.basename(test_file)}")
rmsd = calc_rmsd(
test_pdb["xyz"][:, :3].reshape(-1, 3),
ref_pdb["xyz"][:, :3].reshape(-1, 3),
)[0]
try:
self.assertAlmostEqual(rmsd, 0, 2)
result.addSuccess(self)
print(f"Subtest {test_file} passed")
state = "passed"
log = f"Subtest {test_file} passed"
except AssertionError as e:
result.addFailure(self, e)
print(f"Subtest {test_file} failed")
state = "failed"
log = f"Subtest {test_file} failed:\n{e!r}"
self.results["pdb-diff." + test_file.rpartition("/")[-1]] = dict(
state=state, log=log
)
with open(".results.json", "w") as f:
json.dump(self.results, f, sort_keys=True, indent=2)
self.assertTrue(result.wasSuccessful(), "One or more subtests failed")
@classmethod
def _write_command(cls, bash_file, test_f) -> None:
"""
Takes a bash file from the examples folder, and writes
a version of it to the test_f folder.
It appends to the python command the following arguments:
inference.deterministic=True
if partial_T is in the command, it grabs partial T and sets:
inference.final_step=partial_T-2
else:
inference.final_step=48
"""
out_lines = []
command_lines = []
in_command = False
with open(bash_file, "r") as f:
for line in f:
stripped = line.strip()
if stripped.startswith("python") or stripped.startswith("../"):
in_command = True
if in_command:
# Remove trailing line continuation slashes
if stripped.endswith("\\"):
command_lines.append(stripped[:-1].strip())
else:
command_lines.append(stripped)
in_command = False # End of command
else:
out_lines.append(line)
if not command_lines:
raise ValueError(f"No valid python command found in {bash_file}")
command = " ".join(command_lines)
# get the partial_T
if "partial_T" in command:
final_step = int(command.split("partial_T=")[1].split(" ")[0]) - 2
else:
final_step = 48
output_command = (
f"{command} inference.deterministic=True inference.final_step={final_step}"
)
# replace inference.num_designs with 1
if "inference.num_designs=" in output_command:
output_command = f'{output_command.split("inference.num_designs=")[0]}inference.num_designs=1 {" ".join(output_command.split("inference.num_designs=")[1].split(" ")[1:])}'
else:
output_command = f"{output_command} inference.num_designs=1"
# replace 'example_outputs' with f'{self.out_f}/example_outputs'
output_command = f'{output_command.split("example_outputs")[0]}{cls.out_f}/example_outputs{output_command.split("example_outputs")[1]}'
# write the new command
with open(f"{test_f}/{os.path.basename(bash_file)}", "w") as f:
for line in out_lines:
f.write(line)
f.write(output_command)
def execute_through_pty(command_line):
import pty, select
if sys.platform == "darwin":
master, slave = pty.openpty()
p = subprocess.Popen(
command_line,
shell=True,
stdout=slave,
stdin=slave,
stderr=subprocess.STDOUT,
close_fds=True,
)
buffer = []
while True:
try:
if select.select([master], [], [], 0.2)[0]: # has something to read
data = os.read(master, 1 << 22)
if data:
buffer.append(data)
elif (p.poll() is not None) and (
not select.select([master], [], [], 0.2)[0]
):
break # process is finished and output buffer if fully read
except OSError:
break # OSError will be raised when child process close PTY descriptior
output = b"".join(buffer).decode(encoding="utf-8", errors="backslashreplace")
os.close(master)
os.close(slave)
p.wait()
exit_code = p.returncode
"""
buffer = []
while True:
if select.select([master], [], [], 0.2)[0]: # has something to read
data = os.read(master, 1 << 22)
if data: buffer.append(data)
# else: break # # EOF - well, technically process _should_ be finished here...
# elif time.sleep(1) or (p.poll() is not None): # process is finished (sleep here is intentional to trigger race condition, see solution for this on the next few lines)
# assert not select.select([master], [], [], 0.2)[0] # should be nothing left to read...
# break
elif (p.poll() is not None) and (not select.select([master], [], [], 0.2)[0] ): break # process is finished and output buffer if fully read
assert not select.select([master], [], [], 0.2)[0] # should be nothing left to read...
os.close(slave)
os.close(master)
output = b''.join(buffer).decode(encoding='utf-8', errors='backslashreplace')
exit_code = p.returncode
"""
else:
master, slave = pty.openpty()
p = subprocess.Popen(
command_line,
shell=True,
stdout=slave,
stdin=slave,
stderr=subprocess.STDOUT,
close_fds=True,
)
os.close(slave)
buffer = []
while True:
try:
data = os.read(master, 1 << 22)
if data:
buffer.append(data)
except OSError:
break # OSError will be raised when child process close PTY descriptior
output = b"".join(buffer).decode(encoding="utf-8", errors="backslashreplace")
os.close(master)
p.wait()
exit_code = p.returncode
return exit_code, output
def execute(
message,
command_line,
return_="status",
until_successes=False,
terminate_on_failure=True,
silent=False,
silence_output=False,
silence_output_on_errors=False,
add_message_and_command_line_to_output=False,
):
if not silent:
print(message)
print(command_line)
sys.stdout.flush()
while True:
# exit_code, output = execute_through_subprocess(command_line)
# exit_code, output = execute_through_pexpect(command_line)
exit_code, output = execute_through_pty(command_line)
if (exit_code and not silence_output_on_errors) or not (
silent or silence_output
):
print(output)
sys.stdout.flush()
if exit_code and until_successes:
pass # Thats right - redability COUNT!
else:
break
print("Error while executing {}: {}\n".format(message, output))
print("Sleeping 60s... then I will retry...")
sys.stdout.flush()
time.sleep(60)
if add_message_and_command_line_to_output:
output = message + "\nCommand line: " + command_line + "\n" + output
if return_ == "tuple" or return_ == tuple:
return (exit_code, output)
if exit_code and terminate_on_failure:
print("\nEncounter error while executing: " + command_line)
if return_ == True:
return True
else:
print("\nEncounter error while executing: " + command_line + "\n" + output)
raise BenchmarkError(
"\nEncounter error while executing: " + command_line + "\n" + output
)
if return_ == "output":
return output
else:
return exit_code
if __name__ == "__main__":
import argparse
parser = argparse.ArgumentParser()
parser.add_argument(
"--total_chunks",
type=int,
default=1,
help="total number of chunks to split the examples into (default: 1)",
)
parser.add_argument(
"--chunk_index",
type=int,
default=1,
help="Which chunk to run (1-based index, default:1)",
)
args, remaining = parser.parse_known_args()
TestSubmissionCommands.total_chunks = args.total_chunks
TestSubmissionCommands.chunk_index = args.chunk_index
unittest.main(argv=[sys.argv[0]] + remaining)