Support new throughputs file format

stanford-futuredata / gavel

Code for "Heterogenity-Aware Cluster Scheduling Policies for Deep Learning Workloads", which appeared at OSDI 2020

MIT License

125 stars 31 forks source link

Support new throughputs file format #152

Closed santhnm2 closed 4 years ago

santhnm2 commented 4 years ago

Still need to support getting throughputs for scale factors > 1 in scheduler.py. However, this should be pretty straightforward - just have to divide the throughput for each job by its scale factor so that the multiplication by the scale factor in the policies will recover the true throughput.

deepakn94 commented 4 years ago

Can you rebase on master, and remove the first couple of commits? Will help make the diff smaller