Skip to content

do not chunk when performing beamforming #45

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
popcornell wants to merge 2 commits into
base: master
Choose a base branch
from
Open

do not chunk when performing beamforming #45

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
27d3ac1
do not chunk when performing beamforming
popcornell May 20, 2024
6daef22
apply black
popcornell May 20, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 3 additions & 1 deletion gss/bin/modes/utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -56,7 +56,9 @@ def rttm_to_supervisions_(rttm_path, out_path, channels):
def gpu_check_(num_jobs, cmd):
if cmd == "run.pl" and num_jobs > 1:
used_devices = os.environ.get("CUDA_VISIBLE_DEVICES", "0").split(",")
assert num_jobs <= len(used_devices), f"You are requesting {num_jobs} jobs but you have {len(used_devices)} GPUs available. Exiting !"
assert num_jobs <= len(
used_devices
), f"You are requesting {num_jobs} jobs but you have {len(used_devices)} GPUs available. Exiting !"
for device in used_devices:
grep_res = subprocess.check_output(("nvidia-smi", "-i", f"{device}", "-q"))
check = re.findall("Compute Mode\s+:\sDefault", str(grep_res))
Expand Down
15 changes: 3 additions & 12 deletions gss/core/enhancer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -351,18 +351,9 @@ def enhance_batch(
distortion_mask = cp.sum(masks, axis=0) - target_mask

logging.debug("Applying beamforming with computed masks")
X_hat = []
for i in range(num_chunks):
st = i * chunk_size
en = min(F, (i + 1) * chunk_size)
X_hat_chunk = self.bf_block(
Obs[:, :, st:en],
target_mask=target_mask[:, st:en],
distortion_mask=distortion_mask[:, st:en],
)
X_hat.append(X_hat_chunk)

X_hat = cp.concatenate(X_hat, axis=1) # freq axis again
X_hat = self.bf_block(
Obs, target_mask=target_mask, distortion_mask=distortion_mask
)

logging.debug("Computing inverse STFT")
x_hat = self.istft(X_hat) # returns a numpy array
Expand Down