search

ESkripichnikov / market-making

Reinforcement Learning in Market Making is a project that explores the application of RL techniques to develop market-making strategies, comparing them with baseline approaches and conducting experiments on real-world data.
17 stars 4 forks source link

Line 274: "while len(self.trajectory['rewards']) < count:" won't trigger in evaluate/test mode. #5

Open qiangxu opened 3 weeks ago

qiangxu commented 3 weeks ago

As self.trajectory['rewards'] is not updated if the mode is "train", the loop ends only if sim.tick() is out of data. Is this a bug?

ESkripichnikov commented 1 week ago

In training mode, the simulation ends once we reach 1000 ticks (this is the default value, but there is room for experimentation to determine the optimal simulation length for training). In evaluation mode, we only stop once we have processed all the available data.

Let me know if this doesn't answer your question.