rollback to py3.7

utils.py

@@ -1,13 +1,8 @@
-from subprocess import run
-# For TF > 2.15, use a tfimm branch that uses tf_keras instead of tf.keras
-run('pip install git+https://github.com/pajotarthur/tensorflow-image-models.git@update_python_3.11'.split())
-
 import math
 import json
 
 import numpy as np
 import tensorflow as tf
-import tf_keras
 import tfimm
 import efficientnet.tfkeras as efnv1
 import keras_efficientnet_v2 as efnv2
@@ -131,7 +126,7 @@ def get_confidence(similarity, threshold):
     return sigmoid(abs(logit_sim - logit_threshold))
 
 
-class ArcMarginProductSubCenter(tf_keras.layers.Layer):
+class ArcMarginProductSubCenter(tf.keras.layers.Layer):
     '''
     Implements large margin arc distance.
 
@@ -256,43 +251,43 @@ def get_model(cfg):
             name=f'head/{cfg.head}',
             dtype='float32')
 
-        inp = tf_keras.layers.Input(shape=[*cfg.IMAGE_SIZE, 3], name='inp1')
-        label = tf_keras.layers.Input(shape=(), name='inp2')
+        inp = tf.keras.layers.Input(shape=[*cfg.IMAGE_SIZE, 3], name='inp1')
+        label = tf.keras.layers.Input(shape=(), name='inp2')
         if aux_arcface:
-            label2 = tf_keras.layers.Input(shape=(), name='inp3')
+            label2 = tf.keras.layers.Input(shape=(), name='inp3')
 
         if cfg.arch_name.startswith('efnv1'):
             x = EFN[cfg.arch_name](weights=cfg.pretrained, include_top=False)(inp)
             if cfg.pool == 'flatten':
-                embed = tf_keras.layers.Flatten()(x)
+                embed = tf.keras.layers.Flatten()(x)
             elif cfg.pool == 'fc':
-                embed = tf_keras.layers.Flatten()(x)
-                embed = tf_keras.layers.Dropout(0.1)(embed)
-                embed = tf_keras.layers.Dense(1024)(embed)
+                embed = tf.keras.layers.Flatten()(x)
+                embed = tf.keras.layers.Dropout(0.1)(embed)
+                embed = tf.keras.layers.Dense(1024)(embed)
             elif cfg.pool == 'concat':
-                embed = tf_keras.layers.concatenate([tf_keras.layers.GlobalAveragePooling2D()(x),
-                                                     tf_keras.layers.GlobalAveragePooling2D()(x)])
+                embed = tf.keras.layers.concatenate([tf.keras.layers.GlobalAveragePooling2D()(x),
+                                                     tf.keras.layers.GlobalAveragePooling2D()(x)])
             elif cfg.pool == 'max':
-                embed = tf_keras.layers.GlobalMaxPooling2D()(x)
+                embed = tf.keras.layers.GlobalMaxPooling2D()(x)
             else:
-                embed = tf_keras.layers.GlobalAveragePooling2D()(x)
+                embed = tf.keras.layers.GlobalAveragePooling2D()(x)
 
         elif cfg.arch_name.startswith('efnv2'):
             x = EFN[cfg.arch_name](input_shape=(None, None, 3), num_classes=0,
                                    pretrained=cfg.pretrained)(inp)
             if cfg.pool == 'flatten':
-                embed = tf_keras.layers.Flatten()(x)
+                embed = tf.keras.layers.Flatten()(x)
             elif cfg.pool == 'fc':
-                embed = tf_keras.layers.Flatten()(x)
-                embed = tf_keras.layers.Dropout(0.1)(embed)
-                embed = tf_keras.layers.Dense(1024)(embed)
+                embed = tf.keras.layers.Flatten()(x)
+                embed = tf.keras.layers.Dropout(0.1)(embed)
+                embed = tf.keras.layers.Dense(1024)(embed)
             elif cfg.pool == 'concat':
-                embed = tf_keras.layers.concatenate([tf_keras.layers.GlobalAveragePooling2D()(x),
-                                                     tf_keras.layers.GlobalAveragePooling2D()(x)])
+                embed = tf.keras.layers.concatenate([tf.keras.layers.GlobalAveragePooling2D()(x),
+                                                     tf.keras.layers.GlobalAveragePooling2D()(x)])
             elif cfg.pool == 'max':
-                embed = tf_keras.layers.GlobalMaxPooling2D()(x)
+                embed = tf.keras.layers.GlobalMaxPooling2D()(x)
             else:
-                embed = tf_keras.layers.GlobalAveragePooling2D()(x)
+                embed = tf.keras.layers.GlobalAveragePooling2D()(x)
 
         elif cfg.arch_name in TFHUB:
             # tfhub models cannot be modified => Pooling cannot be changed!
@@ -306,21 +301,21 @@ def get_model(cfg):
 
         if len(cfg.dropout_ps) > 0:
             # Chris Deotte posted model code without Dropout/FC1 after pooling
-            embed = tf_keras.layers.Dropout(cfg.dropout_ps[0])(embed)
-            embed = tf_keras.layers.Dense(1024)(embed)       # tunable embedding size
-        embed = tf_keras.layers.BatchNormalization()(embed)  # missing in public notebooks
+            embed = tf.keras.layers.Dropout(cfg.dropout_ps[0])(embed)
+            embed = tf.keras.layers.Dense(1024)(embed)       # tunable embedding size
+        embed = tf.keras.layers.BatchNormalization()(embed)  # missing in public notebooks
         x = margin([embed, label])
 
-        output = tf_keras.layers.Softmax(dtype='float32', name='arc' if cfg.aux_loss else None)(x)
+        output = tf.keras.layers.Softmax(dtype='float32', name='arc' if cfg.aux_loss else None)(x)
 
         if cfg.aux_loss:
-            aux_features = tf_keras.layers.Dense(cfg.n_species)(embed)
-            aux_output = tf_keras.layers.Softmax(dtype='float32', name='aux')(aux_features)
+            aux_features = tf.keras.layers.Dense(cfg.n_species)(embed)
+            aux_output = tf.keras.layers.Softmax(dtype='float32', name='aux')(aux_features)
         inputs = [inp, label, label2] if (cfg.aux_loss and aux_arcface) else [inp, label]
         outputs = (output, aux_output) if cfg.aux_loss else [output]
 
-        model = tf_keras.models.Model(inputs=inputs, outputs=outputs)
-        embed_model = tf_keras.models.Model(inputs=inp, outputs=embed)
+        model = tf.keras.models.Model(inputs=inputs, outputs=outputs)
+        embed_model = tf.keras.models.Model(inputs=inp, outputs=embed)
 
         if cfg.FREEZE_BATCH_NORM:
             raise NotImplementedError