jens/pyRBM
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Optimizer.train() shall ignore param.do_batch_sample

Browse Source
This commit is contained in:
jens
2025-12-19 17:29:32 +01:00
parent d3c9fe4681
commit 5f5c7c6d77
1 changed files with 4 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/rbm/train.py
+4 -6
View File
@@ -145,17 +145,13 @@ class Optimizer:
d_progress = 100.0 / self.params.num_epochs d_progress = 100.0 / self.params.num_epochs
progress = 0 progress = 0
v_states = batch
if self.params.do_batch_sample:
v_states = sample(batch)
forward = None
for epochs in range(self.params.num_epochs): for epochs in range(self.params.num_epochs):
forward = self.epoch(v_states) self.epoch(batch)
# check if status update is needed # check if status update is needed
if status.want_report(round(progress)): if status.want_report(round(progress)):
# Calculate error # Calculate error
forward = self.entity.forward(batch)
result = self.entity.reconstruct(forward) result = self.entity.reconstruct(forward)
err_rms = rms_error_accu(batch - result) err_rms = rms_error_accu(batch - result)
if not status.on_change({"progress": {"value": round(progress), "unit": "%"}, if not status.on_change({"progress": {"value": round(progress), "unit": "%"},
@@ -164,6 +160,7 @@ class Optimizer:
progress += d_progress progress += d_progress
forward = self.entity.forward(batch)
err_rms = rms_error_accu(batch - self.entity.reconstruct(forward)) err_rms = rms_error_accu(batch - self.entity.reconstruct(forward))
status.on_change({"progress": {"value": round(progress), "unit": "%"}, status.on_change({"progress": {"value": round(progress), "unit": "%"},
"err_rms": {"value": err_rms, "unit": ""}}) "err_rms": {"value": err_rms, "unit": ""}})
@@ -174,6 +171,7 @@ class Optimizer:
forward = None forward = None
training_remain = batch.shape[0] training_remain = batch.shape[0]
batch_row_index = 0 batch_row_index = 0
while training_remain > 0: while training_remain > 0:
batch_size = min(self.params.mini_batch_size, training_remain) batch_size = min(self.params.mini_batch_size, training_remain)
if batch_size == 0: if batch_size == 0: