{
  "id": 173427,
  "title": "is freezing here a good idea for this problem ???",
  "url": "/competitions/siim-isic-melanoma-classification/discussion/173427",
  "author_name": "RoRonoA-TKO",
  "post_date": "2020-08-09T07:28:32.437000",
  "votes": 2,
  "comment_count": 0,
  "views": 0,
  "content": "<p>```</p>\n\n<h1>Efficeint Net B0</h1>\n\n<p>with strategy.scope():\n    model0 = tf.keras.Sequential([\n        efn.EfficientNetB0(\n            input_shape=(*IMAGE_NET_SIZE, 3),\n            weights='imagenet',\n            include_top=False\n        ),\n        GlobalAveragePooling2D(),\n        Dense(1, activation='sigmoid')\n    ])</p>\n\n<h1>Efficeint Net B3</h1>\n\n<p>with strategy.scope():\n    model3 = tf.keras.Sequential([\n        efn.EfficientNetB3(\n            input_shape=(*IMAGE_NET_SIZE, 3),\n            weights='imagenet',\n            include_top=False\n        ),\n        GlobalAveragePooling2D(),\n        Dense(1, activation='sigmoid')\n    ])</p>\n\n<h1>Efficeint Net B6</h1>\n\n<p>with strategy.scope():\n    model6 = tf.keras.Sequential([\n        efn.EfficientNetB6(\n            input_shape=(*IMAGE_NET_SIZE, 3),\n            weights='imagenet',\n            include_top=False\n        ),\n        GlobalAveragePooling2D(),\n        Dense(1, activation='sigmoid')\n    ])</p>\n\n<h1>Efficeint Net B7</h1>\n\n<p>with strategy.scope():\n    model7 = tf.keras.Sequential([\n        efn.EfficientNetB7(\n            input_shape=(*IMAGE_NET_SIZE, 3),\n            weights='imagenet',\n            include_top=False\n        ),\n        GlobalAveragePooling2D(),\n        Dense(1, activation='sigmoid')\n    ])</p>\n\n<p>```</p>\n\n<h3>freeze</h3>\n\n<p><code>\nmodel11 = tf.keras.Sequential()\nfor layer in model0.layers[:-2]:\n    model11.add(layer)\nfor layer in model11.layers:\n    layer.trainable = False\nmodel22 = tf.keras.Sequential()\nfor layer in model3.layers[:-2]:\n    model22.add(layer)\nfor layer in model22.layers:\n    layer.trainable = False\nmodel33 = tf.keras.Sequential()\nfor layer in model6.layers[:-2]:\n    model33.add(layer)\nfor layer in model33.layers:\n    layer.trainable = False\nmodel44 = tf.keras.Sequential()\nfor layer in model7.layers[:-2]:\n    model44.add(layer)\nfor layer in model44.layers:\n    layer.trainable = False\n</code></p>",
  "messages": [
    {
      "id": 963640,
      "postDate": "2020-08-09T07:28:32.437Z",
      "content": "<p>```</p>\n\n<h1>Efficeint Net B0</h1>\n\n<p>with strategy.scope():\n    model0 = tf.keras.Sequential([\n        efn.EfficientNetB0(\n            input_shape=(*IMAGE_NET_SIZE, 3),\n            weights='imagenet',\n            include_top=False\n        ),\n        GlobalAveragePooling2D(),\n        Dense(1, activation='sigmoid')\n    ])</p>\n\n<h1>Efficeint Net B3</h1>\n\n<p>with strategy.scope():\n    model3 = tf.keras.Sequential([\n        efn.EfficientNetB3(\n            input_shape=(*IMAGE_NET_SIZE, 3),\n            weights='imagenet',\n            include_top=False\n        ),\n        GlobalAveragePooling2D(),\n        Dense(1, activation='sigmoid')\n    ])</p>\n\n<h1>Efficeint Net B6</h1>\n\n<p>with strategy.scope():\n    model6 = tf.keras.Sequential([\n        efn.EfficientNetB6(\n            input_shape=(*IMAGE_NET_SIZE, 3),\n            weights='imagenet',\n            include_top=False\n        ),\n        GlobalAveragePooling2D(),\n        Dense(1, activation='sigmoid')\n    ])</p>\n\n<h1>Efficeint Net B7</h1>\n\n<p>with strategy.scope():\n    model7 = tf.keras.Sequential([\n        efn.EfficientNetB7(\n            input_shape=(*IMAGE_NET_SIZE, 3),\n            weights='imagenet',\n            include_top=False\n        ),\n        GlobalAveragePooling2D(),\n        Dense(1, activation='sigmoid')\n    ])</p>\n\n<p>```</p>\n\n<h3>freeze</h3>\n\n<p><code>\nmodel11 = tf.keras.Sequential()\nfor layer in model0.layers[:-2]:\n    model11.add(layer)\nfor layer in model11.layers:\n    layer.trainable = False\nmodel22 = tf.keras.Sequential()\nfor layer in model3.layers[:-2]:\n    model22.add(layer)\nfor layer in model22.layers:\n    layer.trainable = False\nmodel33 = tf.keras.Sequential()\nfor layer in model6.layers[:-2]:\n    model33.add(layer)\nfor layer in model33.layers:\n    layer.trainable = False\nmodel44 = tf.keras.Sequential()\nfor layer in model7.layers[:-2]:\n    model44.add(layer)\nfor layer in model44.layers:\n    layer.trainable = False\n</code></p>",
      "rawMarkdown": "```\n# Efficeint Net B0\n\nwith strategy.scope():\n    model0 = tf.keras.Sequential([\n        efn.EfficientNetB0(\n            input_shape=(*IMAGE_NET_SIZE, 3),\n            weights='imagenet',\n            include_top=False\n        ),\n        GlobalAveragePooling2D(),\n        Dense(1, activation='sigmoid')\n    ])\n    \n\n# Efficeint Net B3\n\nwith strategy.scope():\n    model3 = tf.keras.Sequential([\n        efn.EfficientNetB3(\n            input_shape=(*IMAGE_NET_SIZE, 3),\n            weights='imagenet',\n            include_top=False\n        ),\n        GlobalAveragePooling2D(),\n        Dense(1, activation='sigmoid')\n    ])\n    \n# Efficeint Net B6\n\nwith strategy.scope():\n    model6 = tf.keras.Sequential([\n        efn.EfficientNetB6(\n            input_shape=(*IMAGE_NET_SIZE, 3),\n            weights='imagenet',\n            include_top=False\n        ),\n        GlobalAveragePooling2D(),\n        Dense(1, activation='sigmoid')\n    ])\n    \n# Efficeint Net B7\n\n\nwith strategy.scope():\n    model7 = tf.keras.Sequential([\n        efn.EfficientNetB7(\n            input_shape=(*IMAGE_NET_SIZE, 3),\n            weights='imagenet',\n            include_top=False\n        ),\n        GlobalAveragePooling2D(),\n        Dense(1, activation='sigmoid')\n    ])\n\n```\n\n\n### freeze \n\n```\nmodel11 = tf.keras.Sequential()\nfor layer in model0.layers[:-2]:\n    model11.add(layer)\nfor layer in model11.layers:\n    layer.trainable = False\nmodel22 = tf.keras.Sequential()\nfor layer in model3.layers[:-2]:\n    model22.add(layer)\nfor layer in model22.layers:\n    layer.trainable = False\nmodel33 = tf.keras.Sequential()\nfor layer in model6.layers[:-2]:\n    model33.add(layer)\nfor layer in model33.layers:\n    layer.trainable = False\nmodel44 = tf.keras.Sequential()\nfor layer in model7.layers[:-2]:\n    model44.add(layer)\nfor layer in model44.layers:\n    layer.trainable = False\n```",
      "votes": 2
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "963640": "```\n# Efficeint Net B0\n\nwith strategy.scope():\n    model0 = tf.keras.Sequential([\n        efn.EfficientNetB0(\n            input_shape=(*IMAGE_NET_SIZE, 3),\n            weights='imagenet',\n            include_top=False\n        ),\n        GlobalAveragePooling2D(),\n        Dense(1, activation='sigmoid')\n    ])\n    \n\n# Efficeint Net B3\n\nwith strategy.scope():\n    model3 = tf.keras.Sequential([\n        efn.EfficientNetB3(\n            input_shape=(*IMAGE_NET_SIZE, 3),\n            weights='imagenet',\n            include_top=False\n        ),\n        GlobalAveragePooling2D(),\n        Dense(1, activation='sigmoid')\n    ])\n    \n# Efficeint Net B6\n\nwith strategy.scope():\n    model6 = tf.keras.Sequential([\n        efn.EfficientNetB6(\n            input_shape=(*IMAGE_NET_SIZE, 3),\n            weights='imagenet',\n            include_top=False\n        ),\n        GlobalAveragePooling2D(),\n        Dense(1, activation='sigmoid')\n    ])\n    \n# Efficeint Net B7\n\n\nwith strategy.scope():\n    model7 = tf.keras.Sequential([\n        efn.EfficientNetB7(\n            input_shape=(*IMAGE_NET_SIZE, 3),\n            weights='imagenet',\n            include_top=False\n        ),\n        GlobalAveragePooling2D(),\n        Dense(1, activation='sigmoid')\n    ])\n\n```\n\n\n### freeze \n\n```\nmodel11 = tf.keras.Sequential()\nfor layer in model0.layers[:-2]:\n    model11.add(layer)\nfor layer in model11.layers:\n    layer.trainable = False\nmodel22 = tf.keras.Sequential()\nfor layer in model3.layers[:-2]:\n    model22.add(layer)\nfor layer in model22.layers:\n    layer.trainable = False\nmodel33 = tf.keras.Sequential()\nfor layer in model6.layers[:-2]:\n    model33.add(layer)\nfor layer in model33.layers:\n    layer.trainable = False\nmodel44 = tf.keras.Sequential()\nfor layer in model7.layers[:-2]:\n    model44.add(layer)\nfor layer in model44.layers:\n    layer.trainable = False\n```"
  }
}