I'm having a couple different issues so far with my downloaded version of The Deep Reinforcement Learning course.I got fairly far but now I'm running into issues trying to reference the path for the data_modules. First, when I try to open a file in the data_modules in the jupyter notebook it gives me an error: stating the path and then saying… is not UTF-8 encoded. So then I tried continuing on with the code by openeing the quantra_reinforcement_learning py file and just pulling the run and the trade analytics info directly from there, the only problem is that it references under the 'WEIGHTS_FILE' indicator_model_mixed_2.h5 which isn't even a model in the data_modules folder. There are other ones referenced that are also not included and even the one I try to include by referencing the direct pathway it doesn't work. Would love any help.
Thank you,
Hi Emmanuel,
I suggest you fix the path for the data_modules and then try rerunning the notebooks.
Also, some files get auto-generated when you run the code. For example, in the code snippet attached below, the WEIGHTS_FILE, TRADE_FILE and REPLAY_FILE are saved every 10th iteration. So if the model runs for at least 10 iterations, these files will get generated, and you will be able to find them in the data_modules folder.
So basically, there are two things in this, if you set the PRELOAD=False in rl_config, the model is trained from scratch, and the files get generated and saved as shown above.
For the other scenerio, let's say you want to use some pre-trained model. Then you can copy these files generated using your model to the data_modules folder and set the PRELOAD=True. By doing this, you can directly use the pre-trained model. A good use case for this is live trading, where you do not want to waste time training the model and want to take trades based on the pre-trained model.
Hope this helps! Please do let us know if you still face any issues.
Thanks,
Akshay