Datasets:

villekuosmanen
/

pack_toothbrush_Nov19-advantages

frame_index int64 0 44.8k	advantage float64 -262.71 601	required_intervention bool 1 class
0	-7.260891	false
1	-9.208713	false
2	-5.491373	false
3	-9.694883	false
4	7.694041	false
5	2.750132	false
6	-4.560467	false
7	-2.033669	false
8	6.048166	false
9	15.086996	false
10	-17.117506	false
11	-1.878676	false
12	1.215259	false
13	6.067531	false
14	-14.492641	false
15	-26.699285	false
16	-17.96496	false
17	-7.781845	false
18	-10.379891	false
19	-11.631451	false
20	-28.109346	false
21	-29.318888	false
22	-16.623882	false
23	-15.702841	false
24	-14.214809	false
25	-17.353101	false
26	-17.441814	false
27	-20.888191	false
28	1.65656	false
29	-4.592693	false
30	-11.182842	false
31	-22.168689	false
32	-11.76496	false
33	-13.631449	false
34	10.060073	false
35	15.217727	false
36	-0.426524	false
37	-4.707932	false
38	-7.066072	false
39	-2.12114	false
40	-14.579236	false
41	-13.681871	false
42	5.355339	false
43	-5.22209	false
44	-2.068745	false
45	10.418653	false
46	6.080612	false
47	3.062529	false
48	1.31486	false
49	27.417867	false
50	50.647772	false
51	43.256133	false
52	43.376269	false
53	40.402835	false
54	39.438095	false
55	56.807873	false
56	54.473483	false
57	26.148403	false
58	30.600296	false
59	31.816269	false
60	27.358091	false
61	23.803198	false
62	28.372961	false
63	34.223734	false
64	43.050719	false
65	45.150464	false
66	42.345469	false
67	31.498391	false
68	33.384611	false
69	25.471068	false
70	23.872984	false
71	37.760213	false
72	34.20247	false
73	39.13834	false
74	39.15069	false
75	23.375561	false
76	24.366462	false
77	27.95022	false
78	25.890886	false
79	23.145593	false
80	24.795487	false
81	17.418244	false
82	1.050328	false
83	-4.765077	false
84	-38.14178	false
85	-41.51457	false
86	-31.949233	false
87	-28.383451	false
88	-28.268869	false
89	-35.899537	false
90	-22.942411	false
91	-26.831113	false
92	-44.31708	false
93	-30.449728	false
94	-28.90382	false
95	-39.524364	false
96	-16.394645	false
97	-12.343205	false
98	-27.275922	false
99	-48.867114	false

End of preview. Expand in Data Studio

Advantage Values for villekuosmanen/pack_toothbrush_Nov19

Pre-computed advantage values for offline RL training.

Source

Dataset: villekuosmanen/pack_toothbrush_Nov19
Value Model: villekuosmanen/rewact_toothbrush_pistar_1.5.0
N-step lookahead: 50

Files

This dataset contains per-episode parquet files with advantage values for each frame.

Usage

from pathlib import Path
import pandas as pd

# Load advantages for a specific episode
advantage_df = pd.read_parquet("episode_00000.parquet")

Downloads last month: 143