flexflow / FlexFlow

FlexFlow Serve: Low-Latency, High-Performance LLM Serving
https://flexflow.readthedocs.io
Apache License 2.0
1.6k stars 219 forks source link

Add primitive for estimating cost of a fully-mapped PCG #1323

Open wmdi opened 4 months ago

wmdi commented 4 months ago

This involves 1) supporting parallelization in the cost model; 2) interacting with profiling part.

lockshaw commented 4 months ago

Should be similar to the original FlexFlow simulator (which has also been asked about a lot, so it will be really nice to have it back up and running)