Hi all, here's a genuine newbie question, when using tpu and tensorflow, how do I plot the confusion matrix of the validation set after training? Currently my function looks like this:
\n\n`def train(fold_number):
\n\ntraining_data_path = '../input/siim-isic-melanoma-classification/train/'\ndf = pd.read_csv(\"/kaggle/working/sgkfold.csv\")\ndf_train = df[df.fold != fold_number].reset_index(drop=True)\ndf_valid = df[df.fold == fold_number].reset_index(drop=True)\ndf_train_path = df_train.image_name.apply(format_path_train).values\ndf_val_path = df_valid.image_name.apply(format_path_train).values\ndf_train_labels = df_train.target.values\ndf_val_labels = df_valid.target.values`\n\n`AUTO = tf.data.experimental.AUTOTUNE\n# For tf.dataset\nBATCH_SIZE = 8 * strategy.num_replicas_in_sync\nEPOCHS = 3 \n\ntrain_dataset = (tf.data.Dataset\n.from_tensor_slices((df_train_path, df_train_labels))\n.map(decode_image, num_parallel_calls=AUTO)\n.map(data_augment, num_parallel_calls=AUTO)\n.repeat()\n.shuffle(512)\n.batch(BATCH_SIZE)\n.prefetch(AUTO))\n\nvalid_dataset = (\ntf.data.Dataset\n.from_tensor_slices((df_val_path, df_val_labels))\n.map(decode_image, num_parallel_calls=AUTO)\n.batch(BATCH_SIZE)\n.cache()\n.prefetch(AUTO))\n\ndef build_lrfn(lr_start=0.00001, lr_max=0.00005, \n lr_min=0.00001, lr_rampup_epochs=5, \n lr_sustain_epochs=0, lr_exp_decay=.8):\n lr_max = lr_max * strategy.num_replicas_in_sync\n\n def lrfn(epoch):\n if epoch < lr_rampup_epochs:\n lr = (lr_max - lr_start) / lr_rampup_epochs * epoch + lr_start\n elif epoch < lr_rampup_epochs + lr_sustain_epochs:\n lr = lr_max\n else:\n lr = (lr_max - lr_min) *\\\n lr_exp_decay**(epoch - lr_rampup_epochs\\\n - lr_sustain_epochs) + lr_min\n return lr\n return lrfn \n\nlrfn = build_lrfn()\nSTEPS_PER_EPOCH = df_train_labels.shape[0] // BATCH_SIZE\nlr_schedule = tf.keras.callbacks.LearningRateScheduler(lrfn, verbose=1)\nmodel_checkpoint = tf.keras.callbacks.ModelCheckpoint('GroupKFold.h5', monitor='val_loss', verbose=2, save_best_only=True)\n\n\nwith strategy.scope():\n model = tf.keras.Sequential([\n efn.EfficientNetB3(\n input_shape=(256,256, 3),\n weights=\"imagenet\",\n include_top=False\n ),\n L.GlobalAveragePooling2D(),\n L.Dense(1, activation='sigmoid')\n ])\n model.compile(\n optimizer='adam',\n loss = 'binary_crossentropy',\n metrics=[tf.keras.metrics.AUC()],\n )\n\nhistory = model.fit(train_dataset,\n epochs=EPOCHS,\n callbacks=[model_checkpoint,lr_schedule],\n steps_per_epoch=STEPS_PER_EPOCH,\n validation_data=valid_dataset)\nreturn model`\nWould appreciate some guidance on how I should go about plotting the confusion matrix on the validation set, I want to check the false positives and negatives to make sure my model isn't just blindly predicting benign.
", "messages": [ { "id": 888597, "postDate": "2020-06-16T13:00:30.187Z", "content": "Hi all, here's a genuine newbie question, when using tpu and tensorflow, how do I plot the confusion matrix of the validation set after training? Currently my function looks like this:
\n\n`def train(fold_number):
\n\ntraining_data_path = '../input/siim-isic-melanoma-classification/train/'\ndf = pd.read_csv(\"/kaggle/working/sgkfold.csv\")\ndf_train = df[df.fold != fold_number].reset_index(drop=True)\ndf_valid = df[df.fold == fold_number].reset_index(drop=True)\ndf_train_path = df_train.image_name.apply(format_path_train).values\ndf_val_path = df_valid.image_name.apply(format_path_train).values\ndf_train_labels = df_train.target.values\ndf_val_labels = df_valid.target.values`\n\n`AUTO = tf.data.experimental.AUTOTUNE\n# For tf.dataset\nBATCH_SIZE = 8 * strategy.num_replicas_in_sync\nEPOCHS = 3 \n\ntrain_dataset = (tf.data.Dataset\n.from_tensor_slices((df_train_path, df_train_labels))\n.map(decode_image, num_parallel_calls=AUTO)\n.map(data_augment, num_parallel_calls=AUTO)\n.repeat()\n.shuffle(512)\n.batch(BATCH_SIZE)\n.prefetch(AUTO))\n\nvalid_dataset = (\ntf.data.Dataset\n.from_tensor_slices((df_val_path, df_val_labels))\n.map(decode_image, num_parallel_calls=AUTO)\n.batch(BATCH_SIZE)\n.cache()\n.prefetch(AUTO))\n\ndef build_lrfn(lr_start=0.00001, lr_max=0.00005, \n lr_min=0.00001, lr_rampup_epochs=5, \n lr_sustain_epochs=0, lr_exp_decay=.8):\n lr_max = lr_max * strategy.num_replicas_in_sync\n\n def lrfn(epoch):\n if epoch < lr_rampup_epochs:\n lr = (lr_max - lr_start) / lr_rampup_epochs * epoch + lr_start\n elif epoch < lr_rampup_epochs + lr_sustain_epochs:\n lr = lr_max\n else:\n lr = (lr_max - lr_min) *\\\n lr_exp_decay**(epoch - lr_rampup_epochs\\\n - lr_sustain_epochs) + lr_min\n return lr\n return lrfn \n\nlrfn = build_lrfn()\nSTEPS_PER_EPOCH = df_train_labels.shape[0] // BATCH_SIZE\nlr_schedule = tf.keras.callbacks.LearningRateScheduler(lrfn, verbose=1)\nmodel_checkpoint = tf.keras.callbacks.ModelCheckpoint('GroupKFold.h5', monitor='val_loss', verbose=2, save_best_only=True)\n\n\nwith strategy.scope():\n model = tf.keras.Sequential([\n efn.EfficientNetB3(\n input_shape=(256,256, 3),\n weights=\"imagenet\",\n include_top=False\n ),\n L.GlobalAveragePooling2D(),\n L.Dense(1, activation='sigmoid')\n ])\n model.compile(\n optimizer='adam',\n loss = 'binary_crossentropy',\n metrics=[tf.keras.metrics.AUC()],\n )\n\nhistory = model.fit(train_dataset,\n epochs=EPOCHS,\n callbacks=[model_checkpoint,lr_schedule],\n steps_per_epoch=STEPS_PER_EPOCH,\n validation_data=valid_dataset)\nreturn model`\nWould appreciate some guidance on how I should go about plotting the confusion matrix on the validation set, I want to check the false positives and negatives to make sure my model isn't just blindly predicting benign.
", "rawMarkdown": "Hi all, here's a genuine newbie question, when using tpu and tensorflow, how do I plot the confusion matrix of the validation set after training? Currently my function looks like this:\n\n`def train(fold_number):\n\n training_data_path = '../input/siim-isic-melanoma-classification/train/'\n df = pd.read_csv(\"/kaggle/working/sgkfold.csv\")\n df_train = df[df.fold != fold_number].reset_index(drop=True)\n df_valid = df[df.fold == fold_number].reset_index(drop=True)\n df_train_path = df_train.image_name.apply(format_path_train).values\n df_val_path = df_valid.image_name.apply(format_path_train).values\n df_train_labels = df_train.target.values\n df_val_labels = df_valid.target.values`\n \n `AUTO = tf.data.experimental.AUTOTUNE\n # For tf.dataset\n BATCH_SIZE = 8 * strategy.num_replicas_in_sync\n EPOCHS = 3 \n \n train_dataset = (tf.data.Dataset\n .from_tensor_slices((df_train_path, df_train_labels))\n .map(decode_image, num_parallel_calls=AUTO)\n .map(data_augment, num_parallel_calls=AUTO)\n .repeat()\n .shuffle(512)\n .batch(BATCH_SIZE)\n .prefetch(AUTO))\n \n valid_dataset = (\n tf.data.Dataset\n .from_tensor_slices((df_val_path, df_val_labels))\n .map(decode_image, num_parallel_calls=AUTO)\n .batch(BATCH_SIZE)\n .cache()\n .prefetch(AUTO))\n \n def build_lrfn(lr_start=0.00001, lr_max=0.00005, \n lr_min=0.00001, lr_rampup_epochs=5, \n lr_sustain_epochs=0, lr_exp_decay=.8):\n lr_max = lr_max * strategy.num_replicas_in_sync\n\n def lrfn(epoch):\n if epoch < lr_rampup_epochs:\n lr = (lr_max - lr_start) / lr_rampup_epochs * epoch + lr_start\n elif epoch < lr_rampup_epochs + lr_sustain_epochs:\n lr = lr_max\n else:\n lr = (lr_max - lr_min) *\\\n lr_exp_decay**(epoch - lr_rampup_epochs\\\n - lr_sustain_epochs) + lr_min\n return lr\n return lrfn \n \n lrfn = build_lrfn()\n STEPS_PER_EPOCH = df_train_labels.shape[0] // BATCH_SIZE\n lr_schedule = tf.keras.callbacks.LearningRateScheduler(lrfn, verbose=1)\n model_checkpoint = tf.keras.callbacks.ModelCheckpoint('GroupKFold.h5', monitor='val_loss', verbose=2, save_best_only=True)\n \n \n with strategy.scope():\n model = tf.keras.Sequential([\n efn.EfficientNetB3(\n input_shape=(256,256, 3),\n weights=\"imagenet\",\n include_top=False\n ),\n L.GlobalAveragePooling2D(),\n L.Dense(1, activation='sigmoid')\n ])\n model.compile(\n optimizer='adam',\n loss = 'binary_crossentropy',\n metrics=[tf.keras.metrics.AUC()],\n )\n \n history = model.fit(train_dataset,\n epochs=EPOCHS,\n callbacks=[model_checkpoint,lr_schedule],\n steps_per_epoch=STEPS_PER_EPOCH,\n validation_data=valid_dataset)\n return model`\n\n\nWould appreciate some guidance on how I should go about plotting the confusion matrix on the validation set, I want to check the false positives and negatives to make sure my model isn't just blindly predicting benign.", "votes": 1 }, { "id": 888677, "postDate": "2020-06-16T13:55:03.430Z", "rawMarkdown": "", "votes": 1, "isDeleted": true } ], "comments": [ { "id": 888677, "author_name": "", "author_url": "", "post_date": "2020-06-16T13:55:03.430000", "content": "", "votes": 1, "replies": [] } ], "raw_markdown_by_id": { "888597": "Hi all, here's a genuine newbie question, when using tpu and tensorflow, how do I plot the confusion matrix of the validation set after training? Currently my function looks like this:\n\n`def train(fold_number):\n\n training_data_path = '../input/siim-isic-melanoma-classification/train/'\n df = pd.read_csv(\"/kaggle/working/sgkfold.csv\")\n df_train = df[df.fold != fold_number].reset_index(drop=True)\n df_valid = df[df.fold == fold_number].reset_index(drop=True)\n df_train_path = df_train.image_name.apply(format_path_train).values\n df_val_path = df_valid.image_name.apply(format_path_train).values\n df_train_labels = df_train.target.values\n df_val_labels = df_valid.target.values`\n \n `AUTO = tf.data.experimental.AUTOTUNE\n # For tf.dataset\n BATCH_SIZE = 8 * strategy.num_replicas_in_sync\n EPOCHS = 3 \n \n train_dataset = (tf.data.Dataset\n .from_tensor_slices((df_train_path, df_train_labels))\n .map(decode_image, num_parallel_calls=AUTO)\n .map(data_augment, num_parallel_calls=AUTO)\n .repeat()\n .shuffle(512)\n .batch(BATCH_SIZE)\n .prefetch(AUTO))\n \n valid_dataset = (\n tf.data.Dataset\n .from_tensor_slices((df_val_path, df_val_labels))\n .map(decode_image, num_parallel_calls=AUTO)\n .batch(BATCH_SIZE)\n .cache()\n .prefetch(AUTO))\n \n def build_lrfn(lr_start=0.00001, lr_max=0.00005, \n lr_min=0.00001, lr_rampup_epochs=5, \n lr_sustain_epochs=0, lr_exp_decay=.8):\n lr_max = lr_max * strategy.num_replicas_in_sync\n\n def lrfn(epoch):\n if epoch < lr_rampup_epochs:\n lr = (lr_max - lr_start) / lr_rampup_epochs * epoch + lr_start\n elif epoch < lr_rampup_epochs + lr_sustain_epochs:\n lr = lr_max\n else:\n lr = (lr_max - lr_min) *\\\n lr_exp_decay**(epoch - lr_rampup_epochs\\\n - lr_sustain_epochs) + lr_min\n return lr\n return lrfn \n \n lrfn = build_lrfn()\n STEPS_PER_EPOCH = df_train_labels.shape[0] // BATCH_SIZE\n lr_schedule = tf.keras.callbacks.LearningRateScheduler(lrfn, verbose=1)\n model_checkpoint = tf.keras.callbacks.ModelCheckpoint('GroupKFold.h5', monitor='val_loss', verbose=2, save_best_only=True)\n \n \n with strategy.scope():\n model = tf.keras.Sequential([\n efn.EfficientNetB3(\n input_shape=(256,256, 3),\n weights=\"imagenet\",\n include_top=False\n ),\n L.GlobalAveragePooling2D(),\n L.Dense(1, activation='sigmoid')\n ])\n model.compile(\n optimizer='adam',\n loss = 'binary_crossentropy',\n metrics=[tf.keras.metrics.AUC()],\n )\n \n history = model.fit(train_dataset,\n epochs=EPOCHS,\n callbacks=[model_checkpoint,lr_schedule],\n steps_per_epoch=STEPS_PER_EPOCH,\n validation_data=valid_dataset)\n return model`\n\n\nWould appreciate some guidance on how I should go about plotting the confusion matrix on the validation set, I want to check the false positives and negatives to make sure my model isn't just blindly predicting benign.", "888677": "" } }