fix no prox bug (#13)

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "hotaru"
-version = "3.2.4"
+version = "3.3.0"
 description = "High performance Optimizer to extract spike Timing And cell location from calcium imaging data via lineaR impUlse"
 license = "GPL-3.0-only"
 authors = ["TAKEKAWA Takashi <[email protected]>"]
@@ -10,7 +10,7 @@ keywords = ["Calcium Imaging", "Spike Detection", "Cell Extraction"]
 
 [tool.poetry.dependencies]
 python = "^3.7"
-tensorflow = "^2.2.1"
+tensorflow = "^2.3.1"
 tifffile = "^2020.5.11"
 matplotlib = "^3.2.1"
 cleo = "^0.8.1"

src/hotaru/console/clean.py

@@ -19,7 +19,7 @@ class CleanCommand(Command):
 '''
 
     options = [
-        _option('job-dir', 'j', ''),
+        _option('job-dir', 'j', 'target directory'),
         option('force', 'f', ''),
     ]
 

src/hotaru/console/data.py

@@ -22,7 +22,7 @@ class DataCommand(Command):
 '''
 
     options = [
-        _option('job-dir'),
+        _option('job-dir', 'j', 'target directory'),
         option('force', 'f'),
     ]
 

src/hotaru/console/footprint.py

@@ -15,7 +15,7 @@ class FootprintCommand(Command):
 '''
 
     options = [
-        _option('job-dir'),
+        _option('job-dir', 'j', 'target directory'),
         option('force', 'f'),
     ]
 
@@ -33,10 +33,10 @@ def create(self, data, prev, curr, logs, la, bx, bt):
         spike = load_numpy(f'{prev}-spike')
         nk = spike.shape[0]
         lu = self.status.params['lu']
-        with self.application.strategy.scope():
-            elems = self.get_model(data, tau, nk)
-            model = FootprintModel(*elems)
-            model.compile()
+        #with self.application.strategy.scope():
+        elems = self.get_model(data, tau, nk)
+        model = FootprintModel(*elems)
+        model.compile()
         model.set_penalty(la, lu, bx, bt)
         model.fit(
             spike, stage=curr[-3:],

src/hotaru/console/history.py

@@ -11,7 +11,7 @@ class HistoryCommand(Command):
 '''
 
     options = [
-        _option('job-dir'),
+        _option('job-dir', 'j', 'target directory'),
     ]
 
     def handle(self):

src/hotaru/console/output.py

@@ -20,7 +20,7 @@ class OutputCommand(Command):
 '''
 
     options = [
-        _option('job-dir'),
+        _option('job-dir', 'j', 'target directory'),
     ]
 
     def handle(self):

src/hotaru/console/peak.py

@@ -15,7 +15,7 @@ class PeakCommand(Command):
 '''
 
     options = [
-        _option('job-dir'),
+        _option('job-dir', 'j', 'target directory'),
         option('force', 'f'),
     ]
 

src/hotaru/console/run.py

@@ -11,7 +11,7 @@ class RunCommand(Command):
 '''
 
     options = [
-        _option('job-dir'),
+        _option('job-dir', 'j', 'target directory'),
         _option('goal'),
     ]
 

src/hotaru/console/segment.py

@@ -18,7 +18,7 @@ class SegmentCommand(Command):
 '''
 
     options = [
-        _option('job-dir'),
+        _option('job-dir', 'j', 'target directory'),
         option('force', 'f'),
     ]
 

src/hotaru/console/spike.py

@@ -13,7 +13,7 @@ class SpikeCommand(Command):
 
     name = 'spike'
     options = [
-        _option('job-dir'),
+        _option('job-dir', 'j', 'target directory'),
         option('force', 'f'),
     ]
 
@@ -30,10 +30,10 @@ def create(self, data, prev, curr, logs, tau, lu, bx, bt):
         segment = load_numpy(f'{prev}-segment')
         nk = segment.shape[0]
         la = self.status.params['la']
-        with self.application.strategy.scope():
-            elems = self.get_model(data, tau, nk)
-            model = SpikeModel(*elems)
-            model.compile()
+        #with self.application.strategy.scope():
+        elems = self.get_model(data, tau, nk)
+        model = SpikeModel(*elems)
+        model.compile()
         model.set_penalty(la, lu, bx, bt)
         model.fit(
             segment, stage=curr[-3:],

src/hotaru/console/test.py

@@ -19,7 +19,7 @@ class TestCommand(Command):
 '''
 
     options = [
-        _option('job-dir', 'j', ''),
+        _option('job-dir', 'j', 'target directory'),
     ]
 
     def handle(self):

src/hotaru/train/footprint.py

@@ -1,3 +1,5 @@
+import os
+
 import tensorflow.keras.backend as K
 import tensorflow as tf
 import numpy as np
@@ -19,7 +21,7 @@ def __init__(self, footprint, spike, variance, **kwargs):
     def call(self, inputs):
         footprint = self.footprint(inputs)
         loss, footprint_penalty, spike_penalty = self.call_common(footprint)
-        self.add_metric(footprint_penalty, 'mean', 'penalty')
+        self.add_metric(footprint_penalty, 'penalty')
         return loss
 
     def fit(self, spike, lr, batch, verbose, **kwargs):
@@ -50,21 +52,21 @@ def __init__(self, stage, *args, **kwargs):
         super().__init__(*args, **kwargs)
         self.stage = stage
 
-    def on_epoch_end(self, epoch, logs=None):
-        super().on_epoch_end(logs)
+    def set_model(self, model):
+        super().set_model(model)
+        self._train_dir = os.path.join(self._log_write_dir, 'footprint')
 
+    def on_epoch_end(self, epoch, logs=None):
         stage = self.stage
-        writer = self._get_writer('footprint')
-        with writer.as_default():
+        with self._train_writer.as_default():
             val = self.model.footprint.val
             summary_stat(val, stage, step=epoch)
+        super().on_epoch_end(logs)
 
     def on_train_end(self, logs=None):
-        super().on_train_end(logs)
-
         stage = self.stage
-        writer = self._get_writer('footprint')
-        with writer.as_default():
+        with self._train_writer.as_default():
             val = self.model.footprint.val
             mask = self.model.footprint.mask
             summary_footprint_max(val, mask, stage, step=0)
+        super().on_train_end(logs)

src/hotaru/train/model.py

@@ -35,8 +35,8 @@ def call_common(self, val):
         footprint_penalty = self.footprint_penalty()
         spike_penalty = self.spike_penalty()
         me = loss + footprint_penalty + spike_penalty
-        self.add_metric(K.sqrt(variance), 'mean', 'sigma')
-        self.add_metric(me, 'mean', 'score')
+        self.add_metric(K.sqrt(variance), 'sigma')
+        self.add_metric(me, 'score')
         return loss, footprint_penalty, spike_penalty
 
     def fit_common(self, callback, log_dir=None, stage=None, callbacks=None,

src/hotaru/train/spike.py

@@ -1,3 +1,5 @@
+import os
+
 import tensorflow.keras.backend as K
 import tensorflow as tf
 import numpy as np
@@ -19,7 +21,7 @@ def call(self, inputs):
         spike = self.spike(inputs)
         calcium = self.variance.spike_to_calcium(spike)
         loss, footprint_penalty, spike_penalty = self.call_common(calcium)
-        self.add_metric(footprint_penalty, 'mean', 'penalty')
+        self.add_metric(footprint_penalty, 'penalty')
         return loss
 
     def fit(self, footprint, lr, batch, verbose, **kwargs):
@@ -50,18 +52,18 @@ def __init__(self, stage, *args, **kwargs):
         super().__init__(*args, **kwargs)
         self.stage = stage
 
-    def on_epoch_end(self, epoch, logs=None):
-        super().on_epoch_end(epoch, logs)
+    def set_model(self, model):
+        super().set_model(model)
+        self._train_dir = os.path.join(self._log_write_dir, 'spike')
 
+    def on_epoch_end(self, epoch, logs=None):
         stage = self.stage
-        writer = self._get_writer('spike')
-        with writer.as_default():
+        with self._train_writer.as_default():
             summary_stat(self.model.spike.val, stage, step=epoch)
+        super().on_epoch_end(epoch, logs)
 
     def on_train_end(self, logs=None):
-        super().on_train_end(logs)
-
         stage = self.stage
-        writer = self._get_writer('spike')
-        with writer.as_default():
+        with self._train_writer.as_default():
             summary_spike(self.model.spike.val, stage, step=0)
+        super().on_train_end(logs)