Background: The Tutorials\word_embeddigns_tensorflow_com.py is based on the tutorial https://www.tensorflow.org/text/guide/word_embeddings. The original code downloads a dataset of IMBD reviews which takes a long time to extract. One improvement is to make downloading a flag and use the dataset on disk in all but the original runs while making improvements to the model.
Issue: Currently, the code is able to set the name of the file that is downloaded, but the extracted file ends up a directory level above in
Deep-Learning-And-Code-Samples\ rather than
Deep-Learning-And-Code-Samples\Tutorials where the .gz file is. So, the path to retrieve the data on disk is hard-coded as a temporary fix.
To Fix: make extract location specifiable in the code, or otherwise add a stable reference to the extracted dataset.
Background: The Tutorials\word_embeddigns_tensorflow_com.py is based on the tutorial https://www.tensorflow.org/text/guide/word_embeddings. The original code downloads a dataset of IMBD reviews which takes a long time to extract. One improvement is to make downloading a flag and use the dataset on disk in all but the original runs while making improvements to the model. Issue: Currently, the code is able to set the name of the file that is downloaded, but the extracted file ends up a directory level above in Deep-Learning-And-Code-Samples\ rather than Deep-Learning-And-Code-Samples\Tutorials where the .gz file is. So, the path to retrieve the data on disk is hard-coded as a temporary fix. To Fix: make extract location specifiable in the code, or otherwise add a stable reference to the extracted dataset.