Sprint #11 (July 13)

[pagand]

Feedback: Modelling:

• Do not need too much epochs (Use Wandb to simultaneously visit your performance. Rule of thumb, around 50 is enough)
• Validate after each epoch
• Consider using t11-t16 samples for decoder input
• Double check the running time and the space of your model. Report your model parameters number
• Try implementing more efficient transformers, such as huggingface Paper:
• Try to finish and get ready the modelling part of the paper Goal:
• Start the vessel optimization: For RL approach, take a look at offline RL in the spinningup website.

[yim-fan]

Start the vessel optimization:

• [x] Read some offline RL papers
• [x] Checked spinning up website