search

banksemi / mpquic-rl

Reinforcement Learning-based Multipath QUIC Scheduler for Multimedia Streaming
11 stars 3 forks source link

If I changed the shcduler.go code to fit my algorithm, how to deploy the scheduler to use? #1

Open QzP-QD opened 2 years ago

QzP-QD commented 2 years ago

Hello, appreciate your repository. Still I got some questions. If I changed the shcduler.go code to fit my algorithm, how to deploy the scheduler to use? I'm a new one to this field. I ran "build.sh" to rebuild caddy server. then I run the Peekaboo.ipynb to check my changes. it turns out that its not working. So how can I turn my shceduler into implement ?

banksemi commented 2 years ago

Hello, if you are implementing a new idea scheduler, I recommend MAppLE (https://github.com/vuva/MAppLE) It provides instructions to run the caddy server and dash client.

If you need to experiment with our repository (mptcp-rl), please let me know in detail where the problem is. Thank you.

QzP-QD commented 2 years ago

Thanks for the reply! I want to implement the "Peekaboo" scheculer. It seems like your repository has finished some, but not the Reward part, Relearning part and Alpha Learning part. I wanna know whether you are going to finish it or not? And your thoughts about how to implement the following parts ?

QzP-QD commented 1 year ago

Thanks for your reply! Actually, I want to run DQN scheduler on my VM.But I met question as follows. image I traced it and found out that it may because the specFile is "",so the method "json.Unmarshal"cant deal with it. image I'll appreciate it if you can tell me how to fix this.