diff --git a/README.md b/README.md index 1d22ea8..1f554c2 100644 --- a/README.md +++ b/README.md @@ -34,7 +34,7 @@ Note that [neptune](https://neptune.ml/ 'machine learning lab') (experiment moni ### Run experiment To run an experiment in the Neptune cloud use this command ```bash -$ neptune send experiment_manager.py --environment keras-2.0-gpu-py3 --worker gcp-gpu-medium --config neptune_config.yaml -- train-pipeline --pipeline_name ensemble_train_pipeline +$ neptune send experiment_manager.py --environment keras-2.0-gpu-py3 --worker gcp-gpu-medium --config neptune_config.yaml -- train_evaluate_predict_pipeline --pipeline_name glove_lstm ``` Check [Neptune documentation](https://docs.neptune.ml/cli/neptune_send/) for more options. @@ -42,7 +42,7 @@ Check [Neptune documentation](https://docs.neptune.ml/cli/neptune_send/) for mor # Solution visualization Below end-to-end pipeline is visualized. You can run exactly this one! -![pipeline_001](https://github.com/neptune-ml/kaggle-toxic-starter/blob/master/pipelines_visualizations/pipeline_001.png 'our initial pipeline') +![pipeline_001](https://github.com/neptune-ml/kaggle-toxic-starter/blob/master/imgs/log_reg_ensemble.png 'our initial pipeline') # Contributing diff --git a/pipelines_visualizations/char_vdcnn.png b/imgs/char_vdcnn.png similarity index 100% rename from pipelines_visualizations/char_vdcnn.png rename to imgs/char_vdcnn.png diff --git a/pipelines_visualizations/glove_dpcnn.png b/imgs/glove_dpcnn.png similarity index 100% rename from pipelines_visualizations/glove_dpcnn.png rename to imgs/glove_dpcnn.png diff --git a/pipelines_visualizations/glove_lstm.png b/imgs/glove_lstm.png similarity index 100% rename from pipelines_visualizations/glove_lstm.png rename to imgs/glove_lstm.png diff --git a/pipelines_visualizations/glove_scnn.png b/imgs/glove_scnn.png similarity index 100% rename from pipelines_visualizations/glove_scnn.png rename to imgs/glove_scnn.png diff --git a/pipelines_visualizations/log_reg_pipeline.png b/imgs/log_reg_ensemble.png similarity index 100% rename from pipelines_visualizations/log_reg_pipeline.png rename to imgs/log_reg_ensemble.png diff --git a/pipelines_visualizations/word_lstm.png b/imgs/word_lstm.png similarity index 100% rename from pipelines_visualizations/word_lstm.png rename to imgs/word_lstm.png diff --git a/neptune_config.yaml b/neptune_config.yaml index 772b23f..f6976c7 100644 --- a/neptune_config.yaml +++ b/neptune_config.yaml @@ -1,7 +1,7 @@ project-key: TOX name: toxic -tags: [ensembling, log_reg] +tags: [glove, lstm, adam] metric: @@ -13,7 +13,7 @@ metric: exclude: - output - - pipelines_visualizations + - imgs - neptune.log - offline_job.log - .git @@ -22,19 +22,18 @@ exclude: parameters: # Cloud Environment -# data_dir: /public/toxic_comments -# experiment_dir: /output/experiments/word_lstm -# embedding_filepath: /public/models/glove/glove.840B.300d.txt - + data_dir: /public/toxic_comments + experiment_dir: /output/experiments/glove_lstm + embedding_filepath: /public/models/glove/glove.840B.300d.txt # Local Environment - data_dir: /mnt/ml-team/minerva/toxic/data - experiment_dir: /mnt/ml-team/minerva/toxic/starter_experiments/log_reg_ensemble - embedding_filepath: /mnt/ml-team/minerva/pretrained/glove.840B.300d.txt +# data_dir: /mnt/ml-team/minerva/toxic/data +# experiment_dir: /mnt/ml-team/minerva/toxic/starter_experiments/log_reg_ensemble +# embedding_filepath: /mnt/ml-team/minerva/pretrained/glove.840B.300d.txt overwrite: 0 num_workers: -2 # Preprocessing - max_features_char: 100000 #100 for vdcnnneural networks, for tfidf something like 20000 should be used + max_features_char: 100 #100 for vdcnn, for tfidf something like 20000 should be used max_features_word: 100000 maxlen_char: 512 maxlen_words: 100 @@ -63,27 +62,27 @@ parameters: # global_pooling: 1 # Char VDCNN Architecture - filter_nr: 32 - kernel_size: 3 - repeat_block: 2 - global_pooling: 1 - trainable_embedding: 1 - dense_size: 256 - repeat_dense: 2 - char_embedding_size: 32 - word_embedding_size: None - - - # Glove LSTM Architecture -# filter_nr: 64 -# kernel_size: None +# filter_nr: 32 +# kernel_size: 3 # repeat_block: 2 +# global_pooling: 1 +# trainable_embedding: 1 # dense_size: 256 # repeat_dense: 2 -# global_pooling: 1 -# trainable_embedding: 0 -# word_embedding_size: 300 -# char_embedding_size: None +# char_embedding_size: 32 +# word_embedding_size: None + + + # Glove LSTM Architecture + filter_nr: 64 + kernel_size: None + repeat_block: 2 + dense_size: 256 + repeat_dense: 2 + global_pooling: 1 + trainable_embedding: 0 + word_embedding_size: 300 + char_embedding_size: None # Word LSTM Architecture # filter_nr: 64