google-deepmind / google-deepmind/pysc2

Extracted features from replays are different from expected

Open
#285 1 comment 0 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
8.3k
Forks
1.2k
PR merge metrics
No merged PRs in 30d

Description

Hi,
I am working to extract features from replays. However, some observations are quite different from expected. Specifically:
1. The buildable and pathable feature layers are always empty(filled with zeros). But when I was playing via PySC2 with the latest version of StarCraftII(4.11.2), these two feature layers have valid displays on the screen. Could someone please tell me whether these two features are available during replays?

2. When reading observations from a replay, the unit_type minimap feature layer is always valid whatever the value of SpatialCameraSetup.allow_cheating_layers is;

3. obs.observation.player.larva_count is 0 for most of the time. Is it only valid when a Hatchery, Lair, or Hive is selected ?

FYI, I'm using the version 4.7.1 Linux package and replays.

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.