|
Download README.md from iwillsolvehardestproblem/dream-drift-eval-frames: direct link, hf CLI and curl.
- Browser
- Download file 1.22 kB
-
https://huggingface.co/datasets/iwillsolvehardestproblem/dream-drift-eval-frames/resolve/main/README.md
- Command line
-
hf download hf://datasets/iwillsolvehardestproblem/dream-drift-eval-frames/README.md
-
curl -L -o README.md https://huggingface.co/datasets/iwillsolvehardestproblem/dream-drift-eval-frames/resolve/main/README.md
1.22 kB
metadata
license: mit
tags:
- world-model
- dream-drift
- atari
- evaluation
- dreamerv3
pretty_name: dream-drift DreamerV3 eval frames
dream-drift · DreamerV3 evaluation frames
Raw dream-vs-reality frame dumps behind the DreamerV3 rows of the dream-drift benchmark. Re-score every published number without a GPU.
One file per game and protocol. <game>id.npz = in-distribution (the model's own replay trajectories); <game>.npz = cross-policy (expert DIAMOND-policy trajectories, the "confident healing" regime).
Arrays per file:
realuint8 (32, 50, 64, 64, 3): ground-truth frames (HWC), 32 start states x 50 stepsdreamuint8 (32, 4, 50, 64, 64, 3): 4 dream samples per start, same recorded actions, open loopstarts(32,): start indices into the source trajectory;game,burnin(=4),seed(=0) scalarsrrew,drew(32, 50) / (32, 4, 50): real and predicted rewards (predicted are raw; threshold |r|<0.5 to 0, then sign)rend,dend: real and predicted termination flags
Scoring: python harness/dreamer_score.py <dir-with-raw_frames.npz> in the dream-drift repo reproduces summary.json, drift.png, and the videos.