yangfy0627 commited on
Commit
342ee4c
·
verified ·
1 Parent(s): 07cc887

Archive complete RL run 92c7dc10 with replay, checkpoints, diagnostics and provenance

Browse files
Files changed (33) hide show
  1. RL/ToolHang100/DSRL-NA/diverse100/seed_1/best/source_metadata.json +27 -0
  2. RL/ToolHang100/DSRL-NA/diverse100/seed_1/best/source_resolved_config.json +548 -0
  3. RL/ToolHang100/DSRL-NA/diverse100/seed_1/best/state.pkl +3 -0
  4. RL/ToolHang100/DSRL-NA/diverse100/seed_1/comparison_evaluations.jsonl +81 -0
  5. RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/README.md +5 -0
  6. RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/SHA256SUMS +32 -0
  7. RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/archive.json +58 -0
  8. RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/full_run.tar.zst.part000 +3 -0
  9. RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/full_run.tar.zst.part001 +3 -0
  10. RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/full_run.tar.zst.part002 +3 -0
  11. RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/full_run.tar.zst.part003 +3 -0
  12. RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/full_run.tar.zst.part004 +3 -0
  13. RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/full_run.tar.zst.part005 +3 -0
  14. RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/full_run.tar.zst.part006 +3 -0
  15. RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/full_run.tar.zst.part007 +3 -0
  16. RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/full_run.tar.zst.part008 +3 -0
  17. RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/full_run.tar.zst.part009 +3 -0
  18. RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/local_archive_verification.json +6 -0
  19. RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/manifest.json +0 -0
  20. RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/publication.json +62 -0
  21. RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/restore.py +38 -0
  22. RL/ToolHang100/DSRL-NA/diverse100/seed_1/convergence.json +367 -0
  23. RL/ToolHang100/DSRL-NA/diverse100/seed_1/convergence.jsonl +81 -0
  24. RL/ToolHang100/DSRL-NA/diverse100/seed_1/eval/final.parquet +3 -0
  25. RL/ToolHang100/DSRL-NA/diverse100/seed_1/final/source_metadata.json +314 -0
  26. RL/ToolHang100/DSRL-NA/diverse100/seed_1/final/source_resolved_config.json +548 -0
  27. RL/ToolHang100/DSRL-NA/diverse100/seed_1/final/state.pkl +3 -0
  28. RL/ToolHang100/DSRL-NA/diverse100/seed_1/frozen_base.json +6 -0
  29. RL/ToolHang100/DSRL-NA/diverse100/seed_1/offline_buffer.json +847 -0
  30. RL/ToolHang100/DSRL-NA/diverse100/seed_1/resolved_config.json +548 -0
  31. RL/ToolHang100/DSRL-NA/diverse100/seed_1/start_protocol.json +153 -0
  32. RL/ToolHang100/DSRL-NA/diverse100/seed_1/status.json +170 -0
  33. RL/ToolHang100/DSRL-NA/diverse100/seed_1/training_accounting.json +51 -0
RL/ToolHang100/DSRL-NA/diverse100/seed_1/best/source_metadata.json ADDED
@@ -0,0 +1,27 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "checkpoint_sha256": "29913d8cf3eac8fcde1aa0445b84cf990630e4553b8d366339fe93fc1433d152",
3
+ "evaluation": {
4
+ "chunk_transitions": 131000,
5
+ "episodes": 200,
6
+ "eval_index": 52,
7
+ "gain": 0.038500000000000034,
8
+ "gradient_updates": 393000,
9
+ "ma": 0.3825,
10
+ "n_success": 90,
11
+ "policy_version": 131,
12
+ "rl_env_steps": 1046596,
13
+ "round": 131,
14
+ "seed_start": 960000,
15
+ "sr": 0.45,
16
+ "success_requested": 0,
17
+ "success_requested_imitation": 0,
18
+ "success_rounds_with_skips": 0,
19
+ "success_skipped": 0,
20
+ "success_skipped_imitation": 0,
21
+ "success_skipped_td": 0,
22
+ "target_step": 1040000,
23
+ "update_index": 393000
24
+ },
25
+ "includes": "complete learner state, optimizer, targets, RNG and replay",
26
+ "selection": "observed_periodic_sr"
27
+ }
RL/ToolHang100/DSRL-NA/diverse100/seed_1/best/source_resolved_config.json ADDED
@@ -0,0 +1,548 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "decision_ref": "docs/scope.md [0928-I], [0930-S2DEC]",
3
+ "env": {
4
+ "chunk_length": 8,
5
+ "n_envs": 4
6
+ },
7
+ "learner": {
8
+ "convention": "unified_env_step",
9
+ "gamma_env": 0.999
10
+ },
11
+ "overrides": [
12
+ {
13
+ "decision_ref": "[1020-S2TOOLHANG5] configs/study2_protocol.yaml",
14
+ "key": "n_step_chunks",
15
+ "reason": "task-specific unified Study2 setting",
16
+ "upstream": 3,
17
+ "value": 5
18
+ },
19
+ {
20
+ "decision_ref": "[1023-OFFLINE100] user 2026-10-23: 100-demo BC -> all 100 demos as offline D_demo",
21
+ "key": "offline.cohort_size",
22
+ "reason": "tool_hang regime diverse100 BC cohort has 100 demos; the offline demo buffer uses the checkpoint's entire BC cohort",
23
+ "upstream": 50,
24
+ "value": 100
25
+ },
26
+ {
27
+ "decision_ref": "[1023-OFFLINE100] user 2026-10-23: 100-demo BC -> all 100 demos as offline D_demo",
28
+ "key": "offline.episodes",
29
+ "reason": "tool_hang regime diverse100 BC cohort has 100 demos; the offline demo buffer uses the checkpoint's entire BC cohort",
30
+ "upstream": 50,
31
+ "value": 100
32
+ }
33
+ ],
34
+ "protocol": {
35
+ "checkpoint_every_rounds": 5,
36
+ "checkpoint_retention": {
37
+ "best": true,
38
+ "collapse_drop": {
39
+ "min_abs": 0.2,
40
+ "ratio": 0.5
41
+ },
42
+ "diagnostics": true,
43
+ "every_evals": 3
44
+ },
45
+ "collection": {
46
+ "n_envs": 4,
47
+ "train_seed_base": 100000000,
48
+ "train_seed_stride": 1000000,
49
+ "training_freq_chunks": 1000
50
+ },
51
+ "convergence": {
52
+ "early_stop": false,
53
+ "gain_lag": 10,
54
+ "high_sr": 0.95,
55
+ "high_sr_evals": 5,
56
+ "ma_window": 10,
57
+ "max_env_steps": 1606196,
58
+ "min_env_steps": 200000,
59
+ "min_gain": 0.02,
60
+ "zero_success_evals": 8
61
+ },
62
+ "critic_ensemble": {
63
+ "num": 10,
64
+ "policy": "mean",
65
+ "target": {
66
+ "kind": "subset_min",
67
+ "replace": false,
68
+ "size": 2
69
+ }
70
+ },
71
+ "eval": {
72
+ "best_of_n": 16,
73
+ "episodes": 200,
74
+ "every_env_steps": 20000,
75
+ "include_step0": true,
76
+ "seed_start": 960000,
77
+ "workers": 4
78
+ },
79
+ "final_eval": {
80
+ "episodes": 300,
81
+ "on_zero_stop": true,
82
+ "seed_start": 800000
83
+ },
84
+ "n_step_chunks": 5,
85
+ "offline": {
86
+ "adaptive_ratio": false,
87
+ "cohort_size": 100,
88
+ "episodes": 100,
89
+ "learners_without_offline": [
90
+ "dsrl_sac"
91
+ ],
92
+ "ratio": {
93
+ "decay_env_steps": {
94
+ "can": 160000,
95
+ "square": 320000,
96
+ "tool_hang": 640000,
97
+ "transport": 320000
98
+ },
99
+ "end": 0.1,
100
+ "start": {
101
+ "can": 0.7,
102
+ "square": 0.5,
103
+ "tool_hang": 0.5,
104
+ "transport": 0.9
105
+ }
106
+ },
107
+ "selection": "ascending_demo_index",
108
+ "source": "bc_regime",
109
+ "unsupported_learners": [
110
+ "zprl"
111
+ ]
112
+ },
113
+ "replay": {
114
+ "capacity_chunks": 1000000,
115
+ "pin_start_successes": false,
116
+ "success_buffer": "success_mask_on_shared_fifo",
117
+ "success_skip_rule": "skip_below_success_batch_size",
118
+ "unit": "executed_chunk"
119
+ },
120
+ "start": {
121
+ "max_episodes": 200,
122
+ "mode": "offline_buffer",
123
+ "policy": {
124
+ "default": "base_cfm_ode",
125
+ "zprl": "vib_base"
126
+ },
127
+ "target_successes": 20
128
+ },
129
+ "unified": {
130
+ "actor_target_tau": 0.005,
131
+ "critic_hidden": [
132
+ 512,
133
+ 512,
134
+ 512
135
+ ],
136
+ "critic_target_every_critic_step": true,
137
+ "critic_target_tau": 0.005,
138
+ "dsrl_noise_imitation_reference_dims": 28,
139
+ "gamma_env": 0.999,
140
+ "lr": {
141
+ "actor": 0.0001,
142
+ "critic": 0.0003,
143
+ "decay_end_env_steps": 1606196,
144
+ "min_ratio": 0.1,
145
+ "schedule": "warmup_cosine",
146
+ "temperature": 0.0003,
147
+ "warmup_env_steps": 20000
148
+ },
149
+ "reward": "binary_first_success",
150
+ "success_td_polyak": true,
151
+ "timeout_bootstrap": false,
152
+ "zprl_imitation": "mse_pretanh_mean"
153
+ },
154
+ "updates": {
155
+ "batch_size": 256,
156
+ "critic_per_actor": 2,
157
+ "success_batch_size": 256,
158
+ "utd_per_chunk": 3
159
+ },
160
+ "version": "study2_protocol_v10"
161
+ },
162
+ "protocol_hash": "903b284201350b7c879df61bb6860ec2d7fc9e4537fb5abd2aba8bcf7a60ee0f",
163
+ "provenance_extra": {
164
+ "agent_library": {
165
+ "framework": "jax",
166
+ "sources": {
167
+ "__init__.py": "0356b394bbae9d0738649e4864a034c4ee5d6f329c8a7f862717b3428e52313c",
168
+ "_vendor/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
169
+ "_vendor/ogpo/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
170
+ "_vendor/ogpo/agents/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
171
+ "_vendor/ogpo/agents/modules/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
172
+ "_vendor/ogpo/agents/modules/bc_helper.py": "8913392cc0434057323b2f5b28aaeb36d2c31e6b129ef09d21352353fe276c14",
173
+ "_vendor/ogpo/agents/modules/dist_helper.py": "c0f429c6c70082f5f5258ffffdf4bf0e6b2e1587c26528d3ee937a4e302d9951",
174
+ "_vendor/ogpo/agents/modules/flax_utils.py": "36f27885a086a4ae561171de4650da86330dcd68f17694e3f71278677c614037",
175
+ "_vendor/ogpo/agents/modules/igp_targets_helper.py": "b0343a236a856ec734cc73a1e73186d7d6520dd9efc34b3597327d21b55c9522",
176
+ "_vendor/ogpo/agents/modules/maintenance.py": "7b40d670c277783e118a0669f57aa10370c76b7b8fe3c1d27d16af1a3e9f0865",
177
+ "_vendor/ogpo/agents/modules/pg_helper.py": "cb2e0e83e0c95af408c9bac86fc9ccf1d7e427b3edb315c5d042de5d0a9328d4",
178
+ "_vendor/ogpo/agents/modules/q_helper.py": "71d874b74af5d724d82bd9c4a0a6027d8795e5dc26755ae7838abc3cda54718e",
179
+ "_vendor/ogpo/agents/ogpo.py": "3393282fd4a972c97951715a5776a506e37102cbe5e9c21e5e1ed1e6288696b2",
180
+ "_vendor/ogpo/agents/qc.py": "55906d8b598ce0a2607782e886697c3766dd281a0664ec05bf96173fc8c22c56",
181
+ "_vendor/ogpo/networks/__init__.py": "089130459e91d2b43f5bca45e2659507379fe62351ce6d2d91a49a0387ed2e90",
182
+ "_vendor/ogpo/networks/actors.py": "c0ad2b12439f5da82128e35907314741c22ccc70692de33fed90e67a32f8d86a",
183
+ "_vendor/ogpo/networks/actors_fql.py": "0d9f8646f1230b442bc80457252656c007431ca62458afa5fdecd15150ed9c4d",
184
+ "_vendor/ogpo/networks/critics.py": "77e40f7ea9d6837f8f89d889bfcd7a70e0ef20076f65720d79a9bef3db5130b4",
185
+ "_vendor/ogpo/networks/encoders.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
186
+ "_vendor/ogpo/networks/encoders/__init__.py": "41ae7973f560af5539cff23643843a0c56c253722bb17d7372a9ade029abb670",
187
+ "_vendor/ogpo/networks/encoders/impala.py": "62b7f1dab0aefe6662d3fa0fc59417d125cb9687af71d7875d533e24ac3e89d2",
188
+ "_vendor/ogpo/networks/encoders/minvit.py": "93f0af908450588cdbc9fc18929247b55447bb7972833388244707628328ed33",
189
+ "_vendor/ogpo/networks/encoders/paligemma.py": "0f2a448fd52a2c161f921e7fd7d79be462f98db34c09bf68877f8ed67de9e82c",
190
+ "_vendor/ogpo/networks/encoders/vision_proprio.py": "01ade56e774aeb29c3e7cd1d59eff5e43b7827d102769a450f8d7cc5a17666c8",
191
+ "_vendor/ogpo/networks/modules/__init__.py": "39a4fe9455c1b58c505fa1d9c0a90179dd64cdd459e8341d9ef7fcf6b7af9aae",
192
+ "_vendor/ogpo/networks/modules/attention.py": "a835876c6269ba0c5c63f49005027e5524a750f0c6e7af8697093b9bb8ff4c55",
193
+ "_vendor/ogpo/networks/modules/distributions.py": "5430f1bc6ee7fa71aac65664c9c3f13500d1fec3c5b2b1d80fc6590cca8ac71d",
194
+ "_vendor/ogpo/networks/modules/misc.py": "527793d4e64be2bbc8c22e73a574d5cb604e4672eabd0db9843853107502f620",
195
+ "_vendor/ogpo/networks/modules/mlp.py": "e2cf4d9923d95df8747c1554cf94f20000e108b1a9b0d9b0729a759b597a1316",
196
+ "_vendor/ogpo/networks/modules/time_embedding.py": "54d1a400b1b179a75c5aa55501031aa5d6513da12715ab75b89eaacdbe9f76e1",
197
+ "_vendor/ogpo/networks/modules/two_tier.py": "aa5182a5932cd500f95f4fddcd16429d66c968988639ddf3d6b33a2297953166",
198
+ "dice.py": "6989a71a178af904061e418a94c06615acb7576d9d32548566cc0032e273f22b",
199
+ "dsrl_na.py": "973970bfb02b87ad995f04572bc4d8443cc1e92d6447fbcbcfb80d8fa97cd9ce",
200
+ "dsrl_sac.py": "292d55b2d6f314ec12d17c77c2da2badef6a3fbdde89ba953044e074da6fbc6c",
201
+ "factory.py": "0117708b7cdab69cfaa83bba70f1101aff7b33866117946c377a5cb2fcc6ea49",
202
+ "flow.py": "76dce89f145b4194885524731523eada50e86f8d06b86f7cfaffd0fb3e3b779a",
203
+ "native_config.py": "e9dbf33c0ec15a654c2dde7f504d4e2492c29cbdf84c4eedf42cda94c158e4de",
204
+ "native_flow.py": "e9a9dc59e4e2ccaa7355bebb3c98cc9c1e0080c493889d58bfefdc7bc341c763",
205
+ "native_replay.py": "4f5922a41c5f4e93c599794143c0a26d8b22cb6c38d675ede9538a7bd2b30402",
206
+ "native_residual.py": "00936a650eed103e380f201c9917a9bc414d7b48ba244d0018ebd7b032b3af67",
207
+ "native_train.py": "83bd25879b1096afcb8c1e7a4c34ab77c652dafde10d86f7fdea221c98f6d4a4",
208
+ "networks.py": "81460ac49a7b67ef66b0efd04e26e86c320456c190493143ea77b57c8fbc195d",
209
+ "ogpo.py": "3501a14eb1e1dd0e7194de541dc1af2dd0b4584f102b7e23b0b625d490c79093",
210
+ "optim.py": "41f5c0e4907e822449154959067a84f09820010bc02c6a01a74b5d5478a76192",
211
+ "qc.py": "38ef1b0ed7a4d0833ee50a182262a3bd668931d9499483fa58f7cad43b3f9274",
212
+ "registry.py": "0e7928dac526397186290fe1a914403daa69c39c2c9da9aa395bacfbca9dc617",
213
+ "residual.py": "2dfda94aee07537924912d14693b7d0a41a76c9aea02852a006e5f4877c082ee",
214
+ "telemetry.py": "e5716864ea9b2126b50b86fec78d56ecfd96f17832d5006f83cb2763e6092feb",
215
+ "train.py": "af62bcfe092ab39e79eae9bc1f297430865cd31f61212faf17863e2b6de64ad8",
216
+ "vib.py": "ab3a6721c7c051d21e7955940942bc0f6d064635dd9c73b7275589c150f19f5e",
217
+ "vib_workflow.py": "35b264975fbbcb7f1142af4284f245c9de043e3dab6689ca40657e7be47090f2",
218
+ "zprl.py": "fefe27f991d1a482a57b04738d5b22b9879f0ea405e1935e94dd58ba72c30609"
219
+ },
220
+ "version": "study2_jax_agents_v2"
221
+ },
222
+ "base_framework": "jax",
223
+ "bc_identity": {
224
+ "architecture": "dice_vision_resnet_flow_unet_v1",
225
+ "family": "StandardBaseBC",
226
+ "framework": "jax"
227
+ },
228
+ "checkpoint": "/ssd/WorkSpace/RL-Postrain/runs/operations/20261024_toolhang_diverse100_v10_three/base/diverse100/epoch_300.pkl",
229
+ "checkpoint_sha256": "c5381a3e2ff4f4eeeb899a9d14286d9d42e0816c6431f1208fa39c60309c48f8",
230
+ "experiment": "tool_hang",
231
+ "frozen_hash": "719a7185a0d6a4877c76a5e1d9b1adb163a3d863d28bced54e79d534e8db6873",
232
+ "learner": {
233
+ "action_dim": 7,
234
+ "action_magnitude": null,
235
+ "actor_hidden": [
236
+ 2048,
237
+ 2048,
238
+ 2048
239
+ ],
240
+ "base_hash": "719a7185a0d6a4877c76a5e1d9b1adb163a3d863d28bced54e79d534e8db6873",
241
+ "critic_hidden": [
242
+ 512,
243
+ 512,
244
+ 512
245
+ ],
246
+ "ensemble": 10,
247
+ "feature_dim": 137,
248
+ "horizon": 8,
249
+ "normalizer_hash": "89c504c576b47f752f04402b67f3618b7b4087dd5073f15c5c79cc5a710b7f30",
250
+ "protocol": {
251
+ "checkpoint_every_rounds": 5,
252
+ "checkpoint_retention": {
253
+ "best": true,
254
+ "collapse_drop": {
255
+ "min_abs": 0.2,
256
+ "ratio": 0.5
257
+ },
258
+ "diagnostics": true,
259
+ "every_evals": 3
260
+ },
261
+ "collection": {
262
+ "n_envs": 4,
263
+ "train_seed_base": 100000000,
264
+ "train_seed_stride": 1000000,
265
+ "training_freq_chunks": 1000
266
+ },
267
+ "convergence": {
268
+ "early_stop": false,
269
+ "gain_lag": 10,
270
+ "high_sr": 0.95,
271
+ "high_sr_evals": 5,
272
+ "ma_window": 10,
273
+ "max_env_steps": 1606196,
274
+ "min_env_steps": 200000,
275
+ "min_gain": 0.02,
276
+ "zero_success_evals": 8
277
+ },
278
+ "critic_ensemble": {
279
+ "num": 10,
280
+ "policy": "mean",
281
+ "target": {
282
+ "kind": "subset_min",
283
+ "replace": false,
284
+ "size": 2
285
+ }
286
+ },
287
+ "eval": {
288
+ "best_of_n": 16,
289
+ "episodes": 200,
290
+ "every_env_steps": 20000,
291
+ "include_step0": true,
292
+ "seed_start": 960000,
293
+ "workers": 4
294
+ },
295
+ "final_eval": {
296
+ "episodes": 300,
297
+ "on_zero_stop": true,
298
+ "seed_start": 800000
299
+ },
300
+ "n_step_chunks": 5,
301
+ "offline": {
302
+ "adaptive_ratio": false,
303
+ "cohort_size": 100,
304
+ "episodes": 100,
305
+ "learners_without_offline": [
306
+ "dsrl_sac"
307
+ ],
308
+ "ratio": {
309
+ "decay_env_steps": {
310
+ "can": 160000,
311
+ "square": 320000,
312
+ "tool_hang": 640000,
313
+ "transport": 320000
314
+ },
315
+ "end": 0.1,
316
+ "start": {
317
+ "can": 0.7,
318
+ "square": 0.5,
319
+ "tool_hang": 0.5,
320
+ "transport": 0.9
321
+ }
322
+ },
323
+ "selection": "ascending_demo_index",
324
+ "source": "bc_regime",
325
+ "unsupported_learners": [
326
+ "zprl"
327
+ ]
328
+ },
329
+ "replay": {
330
+ "capacity_chunks": 1000000,
331
+ "pin_start_successes": false,
332
+ "success_buffer": "success_mask_on_shared_fifo",
333
+ "success_skip_rule": "skip_below_success_batch_size",
334
+ "unit": "executed_chunk"
335
+ },
336
+ "start": {
337
+ "max_episodes": 200,
338
+ "mode": "offline_buffer",
339
+ "policy": {
340
+ "default": "base_cfm_ode",
341
+ "zprl": "vib_base"
342
+ },
343
+ "target_successes": 20
344
+ },
345
+ "unified": {
346
+ "actor_target_tau": 0.005,
347
+ "critic_hidden": [
348
+ 512,
349
+ 512,
350
+ 512
351
+ ],
352
+ "critic_target_every_critic_step": true,
353
+ "critic_target_tau": 0.005,
354
+ "dsrl_noise_imitation_reference_dims": 28,
355
+ "gamma_env": 0.999,
356
+ "lr": {
357
+ "actor": 0.0001,
358
+ "critic": 0.0003,
359
+ "decay_end_env_steps": 1606196,
360
+ "min_ratio": 0.1,
361
+ "schedule": "warmup_cosine",
362
+ "temperature": 0.0003,
363
+ "warmup_env_steps": 20000
364
+ },
365
+ "reward": "binary_first_success",
366
+ "success_td_polyak": true,
367
+ "timeout_bootstrap": false,
368
+ "zprl_imitation": "mse_pretanh_mean"
369
+ },
370
+ "updates": {
371
+ "batch_size": 256,
372
+ "critic_per_actor": 2,
373
+ "success_batch_size": 256,
374
+ "utd_per_chunk": 3
375
+ },
376
+ "version": "study2_protocol_v10"
377
+ },
378
+ "subset": 2,
379
+ "variant": {
380
+ "flags": {
381
+ "noise_loglik_weight": 0.0,
382
+ "success_td": false
383
+ },
384
+ "learner": "dsrl_na",
385
+ "name": "dsrl_na",
386
+ "phase": 1,
387
+ "spec_hash": "288e84fd5eb8196404b74bd024f80ec14a2de9bac56dae6239fa3d474968c979",
388
+ "version": "study2_variants_v4"
389
+ },
390
+ "version": "study2_jax_agents_v2",
391
+ "vib_hash": null,
392
+ "weight_decay": null
393
+ },
394
+ "observation_mode": "pixel_state",
395
+ "offline_buffer": {
396
+ "actions_sha256": "7d65f5b7df78085d4780cf6338ac032e63f7f3d6ff9b1923bcc0992c840ee2d9",
397
+ "aux_sha256": "10f0ea12b0d1e89a65e72988f979e40fe4ffc48cec8c05e3651506b6a1d34088",
398
+ "checkpoint_sha256": "c5381a3e2ff4f4eeeb899a9d14286d9d42e0816c6431f1208fa39c60309c48f8",
399
+ "chunks": 6171,
400
+ "dataset_sha256": "c5cb01b119a109a3d4842ca515906e86f1f35a8ee1a333d22b3503c1a513a8f4",
401
+ "decision_ref": "[1017-S2OFFBC]",
402
+ "env_steps": 48984,
403
+ "features_sha256": "ed0a570dcabbbba769a4ac377436caea901a9b0238e5c9890a6db1e473a723c2",
404
+ "horizon": 8,
405
+ "identity_hash": "07e7e6bd263c84a9b8ceb56b1e6e92eea9ced27f6754960f1a8b5007f4f18779",
406
+ "learner_coordinates": "ResidualLearner:dsrl_na",
407
+ "pixel_spec_hash": "7aba3118e0304ea7265915550734918674f2b9a1e6c9bc2eb1cd99b568045a11",
408
+ "ratio": {
409
+ "adaptive_ratio": false,
410
+ "rule": "D_demo kept for the whole run (never evicted), online RL data FIFO; each ordinary batch = batch_size uniform draws over the eligible D_demo + online windows, offline rows first"
411
+ },
412
+ "selection": {
413
+ "cohort_size": 100,
414
+ "demos": [
415
+ "demo_0",
416
+ "demo_2",
417
+ "demo_5",
418
+ "demo_7",
419
+ "demo_9",
420
+ "demo_10",
421
+ "demo_13",
422
+ "demo_16",
423
+ "demo_17",
424
+ "demo_18",
425
+ "demo_20",
426
+ "demo_23",
427
+ "demo_27",
428
+ "demo_28",
429
+ "demo_29",
430
+ "demo_30",
431
+ "demo_31",
432
+ "demo_32",
433
+ "demo_34",
434
+ "demo_35",
435
+ "demo_37",
436
+ "demo_40",
437
+ "demo_41",
438
+ "demo_43",
439
+ "demo_46",
440
+ "demo_47",
441
+ "demo_50",
442
+ "demo_51",
443
+ "demo_56",
444
+ "demo_57",
445
+ "demo_58",
446
+ "demo_67",
447
+ "demo_71",
448
+ "demo_74",
449
+ "demo_76",
450
+ "demo_78",
451
+ "demo_80",
452
+ "demo_82",
453
+ "demo_83",
454
+ "demo_85",
455
+ "demo_86",
456
+ "demo_87",
457
+ "demo_93",
458
+ "demo_97",
459
+ "demo_98",
460
+ "demo_101",
461
+ "demo_102",
462
+ "demo_107",
463
+ "demo_110",
464
+ "demo_111",
465
+ "demo_112",
466
+ "demo_113",
467
+ "demo_115",
468
+ "demo_116",
469
+ "demo_119",
470
+ "demo_122",
471
+ "demo_123",
472
+ "demo_126",
473
+ "demo_127",
474
+ "demo_128",
475
+ "demo_130",
476
+ "demo_131",
477
+ "demo_132",
478
+ "demo_133",
479
+ "demo_134",
480
+ "demo_135",
481
+ "demo_138",
482
+ "demo_140",
483
+ "demo_141",
484
+ "demo_145",
485
+ "demo_147",
486
+ "demo_149",
487
+ "demo_151",
488
+ "demo_152",
489
+ "demo_154",
490
+ "demo_156",
491
+ "demo_157",
492
+ "demo_159",
493
+ "demo_161",
494
+ "demo_163",
495
+ "demo_164",
496
+ "demo_171",
497
+ "demo_172",
498
+ "demo_173",
499
+ "demo_174",
500
+ "demo_175",
501
+ "demo_176",
502
+ "demo_177",
503
+ "demo_179",
504
+ "demo_181",
505
+ "demo_182",
506
+ "demo_183",
507
+ "demo_186",
508
+ "demo_188",
509
+ "demo_190",
510
+ "demo_191",
511
+ "demo_192",
512
+ "demo_193",
513
+ "demo_196",
514
+ "demo_199"
515
+ ],
516
+ "manifest": "/ssd/WorkSpace/RL-Postrain/checkpoints/huggingface/yangfy0627/RL-postTrain/e62671ed1cf646b453660b698c22ca4d273b5ccc/Robomimic/study1/ToolHang/diverse100/seed_1/data_manifest.json",
517
+ "manifest_sha256": "9233d5824205e29f1afa721d8573fe19805a1a8e7829c296af1dd0b74a1d29b4",
518
+ "regime": "diverse100",
519
+ "source": "bc_regime",
520
+ "source_ph_sha256": "c5cb01b119a109a3d4842ca515906e86f1f35a8ee1a333d22b3503c1a513a8f4",
521
+ "split": {
522
+ "training": 90,
523
+ "validation": 10
524
+ },
525
+ "task": "tool_hang"
526
+ }
527
+ },
528
+ "offline_role": "mixed",
529
+ "pixel_spec_hash": "7aba3118e0304ea7265915550734918674f2b9a1e6c9bc2eb1cd99b568045a11",
530
+ "smoke": false,
531
+ "task": "tool_hang",
532
+ "task_spec_hash": "ac2909345e9536690a9da63f06aad8dfe135e757b89e28706b2c4af641f195c2"
533
+ },
534
+ "seed": 1,
535
+ "start_policy": "base_cfm_ode",
536
+ "variant": {
537
+ "flags": {
538
+ "noise_loglik_weight": 0.0,
539
+ "success_td": false
540
+ },
541
+ "learner": "dsrl_na",
542
+ "name": "dsrl_na",
543
+ "phase": 1,
544
+ "spec_hash": "288e84fd5eb8196404b74bd024f80ec14a2de9bac56dae6239fa3d474968c979",
545
+ "version": "study2_variants_v4"
546
+ },
547
+ "variant_hash": "288e84fd5eb8196404b74bd024f80ec14a2de9bac56dae6239fa3d474968c979"
548
+ }
RL/ToolHang100/DSRL-NA/diverse100/seed_1/best/state.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:29913d8cf3eac8fcde1aa0445b84cf990630e4553b8d366339fe93fc1433d152
3
+ size 282975465
RL/ToolHang100/DSRL-NA/diverse100/seed_1/comparison_evaluations.jsonl ADDED
@@ -0,0 +1,81 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"eval_index": 0, "rl_env_steps": 0, "target_step": 0, "sr": 0.28, "round": 0, "update_index": 0, "policy_version": 0, "chunk_transitions": 0, "gradient_updates": 0, "n_success": 56, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": null, "gain": null, "observed": true, "imputed": false}
2
+ {"eval_index": 1, "rl_env_steps": 23980, "target_step": 20000, "sr": 0.14, "round": 3, "update_index": 9000, "policy_version": 3, "chunk_transitions": 3000, "gradient_updates": 9000, "n_success": 28, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": null, "gain": null, "observed": true, "imputed": false}
3
+ {"eval_index": 2, "rl_env_steps": 47942, "target_step": 40000, "sr": 0.305, "round": 6, "update_index": 18000, "policy_version": 6, "chunk_transitions": 6000, "gradient_updates": 18000, "n_success": 61, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": null, "gain": null, "observed": true, "imputed": false}
4
+ {"eval_index": 3, "rl_env_steps": 63926, "target_step": 60000, "sr": 0.125, "round": 8, "update_index": 24000, "policy_version": 8, "chunk_transitions": 8000, "gradient_updates": 24000, "n_success": 25, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": null, "gain": null, "observed": true, "imputed": false}
5
+ {"eval_index": 4, "rl_env_steps": 87915, "target_step": 80000, "sr": 0.2, "round": 11, "update_index": 33000, "policy_version": 11, "chunk_transitions": 11000, "gradient_updates": 33000, "n_success": 40, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": null, "gain": null, "observed": true, "imputed": false}
6
+ {"eval_index": 5, "rl_env_steps": 103888, "target_step": 100000, "sr": 0.235, "round": 13, "update_index": 39000, "policy_version": 13, "chunk_transitions": 13000, "gradient_updates": 39000, "n_success": 47, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": null, "gain": null, "observed": true, "imputed": false}
7
+ {"eval_index": 6, "rl_env_steps": 127883, "target_step": 120000, "sr": 0.305, "round": 16, "update_index": 48000, "policy_version": 16, "chunk_transitions": 16000, "gradient_updates": 48000, "n_success": 61, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": null, "gain": null, "observed": true, "imputed": false}
8
+ {"eval_index": 7, "rl_env_steps": 143862, "target_step": 140000, "sr": 0.335, "round": 18, "update_index": 54000, "policy_version": 18, "chunk_transitions": 18000, "gradient_updates": 54000, "n_success": 67, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": null, "gain": null, "observed": true, "imputed": false}
9
+ {"eval_index": 8, "rl_env_steps": 167829, "target_step": 160000, "sr": 0.27, "round": 21, "update_index": 63000, "policy_version": 21, "chunk_transitions": 21000, "gradient_updates": 63000, "n_success": 54, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": null, "gain": null, "observed": true, "imputed": false}
10
+ {"eval_index": 9, "rl_env_steps": 183814, "target_step": 180000, "sr": 0.335, "round": 23, "update_index": 69000, "policy_version": 23, "chunk_transitions": 23000, "gradient_updates": 69000, "n_success": 67, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.253, "gain": null, "observed": true, "imputed": false}
11
+ {"eval_index": 10, "rl_env_steps": 207800, "target_step": 200000, "sr": 0.285, "round": 26, "update_index": 78000, "policy_version": 26, "chunk_transitions": 26000, "gradient_updates": 78000, "n_success": 57, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.2535, "gain": null, "observed": true, "imputed": false}
12
+ {"eval_index": 11, "rl_env_steps": 223782, "target_step": 220000, "sr": 0.205, "round": 28, "update_index": 84000, "policy_version": 28, "chunk_transitions": 28000, "gradient_updates": 84000, "n_success": 41, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.26000000000000006, "gain": null, "observed": true, "imputed": false}
13
+ {"eval_index": 12, "rl_env_steps": 247757, "target_step": 240000, "sr": 0.305, "round": 31, "update_index": 93000, "policy_version": 31, "chunk_transitions": 31000, "gradient_updates": 93000, "n_success": 61, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.26, "gain": null, "observed": true, "imputed": false}
14
+ {"eval_index": 13, "rl_env_steps": 263737, "target_step": 260000, "sr": 0.27, "round": 33, "update_index": 99000, "policy_version": 33, "chunk_transitions": 33000, "gradient_updates": 99000, "n_success": 54, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.2745, "gain": null, "observed": true, "imputed": false}
15
+ {"eval_index": 14, "rl_env_steps": 287727, "target_step": 280000, "sr": 0.285, "round": 36, "update_index": 108000, "policy_version": 36, "chunk_transitions": 36000, "gradient_updates": 108000, "n_success": 57, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.28300000000000003, "gain": null, "observed": true, "imputed": false}
16
+ {"eval_index": 15, "rl_env_steps": 303723, "target_step": 300000, "sr": 0.27, "round": 38, "update_index": 114000, "policy_version": 38, "chunk_transitions": 38000, "gradient_updates": 114000, "n_success": 54, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.28650000000000003, "gain": null, "observed": true, "imputed": false}
17
+ {"eval_index": 16, "rl_env_steps": 327678, "target_step": 320000, "sr": 0.305, "round": 41, "update_index": 123000, "policy_version": 41, "chunk_transitions": 41000, "gradient_updates": 123000, "n_success": 61, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.28650000000000003, "gain": null, "observed": true, "imputed": false}
18
+ {"eval_index": 17, "rl_env_steps": 343668, "target_step": 340000, "sr": 0.265, "round": 43, "update_index": 129000, "policy_version": 43, "chunk_transitions": 43000, "gradient_updates": 129000, "n_success": 53, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.27949999999999997, "gain": null, "observed": true, "imputed": false}
19
+ {"eval_index": 18, "rl_env_steps": 367651, "target_step": 360000, "sr": 0.325, "round": 46, "update_index": 138000, "policy_version": 46, "chunk_transitions": 46000, "gradient_updates": 138000, "n_success": 65, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.28500000000000003, "gain": null, "observed": true, "imputed": false}
20
+ {"eval_index": 19, "rl_env_steps": 383633, "target_step": 380000, "sr": 0.265, "round": 48, "update_index": 144000, "policy_version": 48, "chunk_transitions": 48000, "gradient_updates": 144000, "n_success": 53, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.278, "gain": 0.025000000000000022, "observed": true, "imputed": false}
21
+ {"eval_index": 20, "rl_env_steps": 407615, "target_step": 400000, "sr": 0.325, "round": 51, "update_index": 153000, "policy_version": 51, "chunk_transitions": 51000, "gradient_updates": 153000, "n_success": 65, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.28200000000000003, "gain": 0.028500000000000025, "observed": true, "imputed": false}
22
+ {"eval_index": 21, "rl_env_steps": 423596, "target_step": 420000, "sr": 0.295, "round": 53, "update_index": 159000, "policy_version": 53, "chunk_transitions": 53000, "gradient_updates": 159000, "n_success": 59, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.29100000000000004, "gain": 0.030999999999999972, "observed": true, "imputed": false}
23
+ {"eval_index": 22, "rl_env_steps": 447575, "target_step": 440000, "sr": 0.305, "round": 56, "update_index": 168000, "policy_version": 56, "chunk_transitions": 56000, "gradient_updates": 168000, "n_success": 61, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.29100000000000004, "gain": 0.031000000000000028, "observed": true, "imputed": false}
24
+ {"eval_index": 23, "rl_env_steps": 463551, "target_step": 460000, "sr": 0.27, "round": 58, "update_index": 174000, "policy_version": 58, "chunk_transitions": 58000, "gradient_updates": 174000, "n_success": 54, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.29100000000000004, "gain": 0.016500000000000015, "observed": true, "imputed": false}
25
+ {"eval_index": 24, "rl_env_steps": 487519, "target_step": 480000, "sr": 0.295, "round": 61, "update_index": 183000, "policy_version": 61, "chunk_transitions": 61000, "gradient_updates": 183000, "n_success": 59, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.292, "gain": 0.008999999999999952, "observed": true, "imputed": false}
26
+ {"eval_index": 25, "rl_env_steps": 503500, "target_step": 500000, "sr": 0.3, "round": 63, "update_index": 189000, "policy_version": 63, "chunk_transitions": 63000, "gradient_updates": 189000, "n_success": 60, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.295, "gain": 0.008499999999999952, "observed": true, "imputed": false}
27
+ {"eval_index": 26, "rl_env_steps": 527451, "target_step": 520000, "sr": 0.295, "round": 66, "update_index": 198000, "policy_version": 66, "chunk_transitions": 66000, "gradient_updates": 198000, "n_success": 59, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.294, "gain": 0.007499999999999951, "observed": true, "imputed": false}
28
+ {"eval_index": 27, "rl_env_steps": 543434, "target_step": 540000, "sr": 0.315, "round": 68, "update_index": 204000, "policy_version": 68, "chunk_transitions": 68000, "gradient_updates": 204000, "n_success": 63, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.299, "gain": 0.019500000000000017, "observed": true, "imputed": false}
29
+ {"eval_index": 28, "rl_env_steps": 567377, "target_step": 560000, "sr": 0.345, "round": 71, "update_index": 213000, "policy_version": 71, "chunk_transitions": 71000, "gradient_updates": 213000, "n_success": 69, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.301, "gain": 0.01599999999999996, "observed": true, "imputed": false}
30
+ {"eval_index": 29, "rl_env_steps": 583356, "target_step": 580000, "sr": 0.255, "round": 73, "update_index": 219000, "policy_version": 73, "chunk_transitions": 73000, "gradient_updates": 219000, "n_success": 51, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.3, "gain": 0.021999999999999964, "observed": true, "imputed": false}
31
+ {"eval_index": 30, "rl_env_steps": 607327, "target_step": 600000, "sr": 0.375, "round": 76, "update_index": 228000, "policy_version": 76, "chunk_transitions": 76000, "gradient_updates": 228000, "n_success": 75, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.305, "gain": 0.022999999999999965, "observed": true, "imputed": false}
32
+ {"eval_index": 31, "rl_env_steps": 623310, "target_step": 620000, "sr": 0.39, "round": 78, "update_index": 234000, "policy_version": 78, "chunk_transitions": 78000, "gradient_updates": 234000, "n_success": 78, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.3145, "gain": 0.023499999999999965, "observed": true, "imputed": false}
33
+ {"eval_index": 32, "rl_env_steps": 647289, "target_step": 640000, "sr": 0.31, "round": 81, "update_index": 243000, "policy_version": 81, "chunk_transitions": 81000, "gradient_updates": 243000, "n_success": 62, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.31500000000000006, "gain": 0.02400000000000002, "observed": true, "imputed": false}
34
+ {"eval_index": 33, "rl_env_steps": 663271, "target_step": 660000, "sr": 0.345, "round": 83, "update_index": 249000, "policy_version": 83, "chunk_transitions": 83000, "gradient_updates": 249000, "n_success": 69, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.32250000000000006, "gain": 0.03150000000000003, "observed": true, "imputed": false}
35
+ {"eval_index": 34, "rl_env_steps": 687236, "target_step": 680000, "sr": 0.41, "round": 86, "update_index": 258000, "policy_version": 86, "chunk_transitions": 86000, "gradient_updates": 258000, "n_success": 82, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.33399999999999996, "gain": 0.04199999999999998, "observed": true, "imputed": false}
36
+ {"eval_index": 35, "rl_env_steps": 703211, "target_step": 700000, "sr": 0.295, "round": 88, "update_index": 264000, "policy_version": 88, "chunk_transitions": 88000, "gradient_updates": 264000, "n_success": 59, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.3335, "gain": 0.038500000000000034, "observed": true, "imputed": false}
37
+ {"eval_index": 36, "rl_env_steps": 727182, "target_step": 720000, "sr": 0.34, "round": 91, "update_index": 273000, "policy_version": 91, "chunk_transitions": 91000, "gradient_updates": 273000, "n_success": 68, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.33799999999999997, "gain": 0.043999999999999984, "observed": true, "imputed": false}
38
+ {"eval_index": 37, "rl_env_steps": 743173, "target_step": 740000, "sr": 0.42, "round": 93, "update_index": 279000, "policy_version": 93, "chunk_transitions": 93000, "gradient_updates": 279000, "n_success": 84, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.3484999999999999, "gain": 0.04949999999999993, "observed": true, "imputed": false}
39
+ {"eval_index": 38, "rl_env_steps": 767118, "target_step": 760000, "sr": 0.27, "round": 96, "update_index": 288000, "policy_version": 96, "chunk_transitions": 96000, "gradient_updates": 288000, "n_success": 54, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.34099999999999997, "gain": 0.03999999999999998, "observed": true, "imputed": false}
40
+ {"eval_index": 39, "rl_env_steps": 783089, "target_step": 780000, "sr": 0.3, "round": 98, "update_index": 294000, "policy_version": 98, "chunk_transitions": 98000, "gradient_updates": 294000, "n_success": 60, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.3455, "gain": 0.045499999999999985, "observed": true, "imputed": false}
41
+ {"eval_index": 40, "rl_env_steps": 807064, "target_step": 800000, "sr": 0.41, "round": 101, "update_index": 303000, "policy_version": 101, "chunk_transitions": 101000, "gradient_updates": 303000, "n_success": 82, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.349, "gain": 0.043999999999999984, "observed": true, "imputed": false}
42
+ {"eval_index": 41, "rl_env_steps": 823026, "target_step": 820000, "sr": 0.275, "round": 103, "update_index": 309000, "policy_version": 103, "chunk_transitions": 103000, "gradient_updates": 309000, "n_success": 55, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.3375, "gain": 0.02300000000000002, "observed": true, "imputed": false}
43
+ {"eval_index": 42, "rl_env_steps": 846982, "target_step": 840000, "sr": 0.375, "round": 106, "update_index": 318000, "policy_version": 106, "chunk_transitions": 106000, "gradient_updates": 318000, "n_success": 75, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.344, "gain": 0.028999999999999915, "observed": true, "imputed": false}
44
+ {"eval_index": 43, "rl_env_steps": 862955, "target_step": 860000, "sr": 0.335, "round": 108, "update_index": 324000, "policy_version": 108, "chunk_transitions": 108000, "gradient_updates": 324000, "n_success": 67, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.34299999999999997, "gain": 0.020499999999999907, "observed": true, "imputed": false}
45
+ {"eval_index": 44, "rl_env_steps": 886908, "target_step": 880000, "sr": 0.275, "round": 111, "update_index": 333000, "policy_version": 111, "chunk_transitions": 111000, "gradient_updates": 333000, "n_success": 55, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.32949999999999996, "gain": -0.004500000000000004, "observed": true, "imputed": false}
46
+ {"eval_index": 45, "rl_env_steps": 902888, "target_step": 900000, "sr": 0.39, "round": 113, "update_index": 339000, "policy_version": 113, "chunk_transitions": 113000, "gradient_updates": 339000, "n_success": 78, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.339, "gain": 0.005500000000000005, "observed": true, "imputed": false}
47
+ {"eval_index": 46, "rl_env_steps": 926832, "target_step": 920000, "sr": 0.38, "round": 116, "update_index": 348000, "policy_version": 116, "chunk_transitions": 116000, "gradient_updates": 348000, "n_success": 76, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.343, "gain": 0.00500000000000006, "observed": true, "imputed": false}
48
+ {"eval_index": 47, "rl_env_steps": 942799, "target_step": 940000, "sr": 0.375, "round": 118, "update_index": 354000, "policy_version": 118, "chunk_transitions": 118000, "gradient_updates": 354000, "n_success": 75, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.33849999999999997, "gain": -0.009999999999999953, "observed": true, "imputed": false}
49
+ {"eval_index": 48, "rl_env_steps": 966741, "target_step": 960000, "sr": 0.385, "round": 121, "update_index": 363000, "policy_version": 121, "chunk_transitions": 121000, "gradient_updates": 363000, "n_success": 77, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.35, "gain": 0.009000000000000008, "observed": true, "imputed": false}
50
+ {"eval_index": 49, "rl_env_steps": 982720, "target_step": 980000, "sr": 0.43, "round": 123, "update_index": 369000, "policy_version": 123, "chunk_transitions": 123000, "gradient_updates": 369000, "n_success": 86, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.36300000000000004, "gain": 0.01750000000000007, "observed": true, "imputed": false}
51
+ {"eval_index": 50, "rl_env_steps": 1006671, "target_step": 1000000, "sr": 0.405, "round": 126, "update_index": 378000, "policy_version": 126, "chunk_transitions": 126000, "gradient_updates": 378000, "n_success": 81, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.3625, "gain": 0.013500000000000012, "observed": true, "imputed": false}
52
+ {"eval_index": 51, "rl_env_steps": 1022649, "target_step": 1020000, "sr": 0.4, "round": 128, "update_index": 384000, "policy_version": 128, "chunk_transitions": 128000, "gradient_updates": 384000, "n_success": 80, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.37499999999999994, "gain": 0.03749999999999992, "observed": true, "imputed": false}
53
+ {"eval_index": 52, "rl_env_steps": 1046596, "target_step": 1040000, "sr": 0.45, "round": 131, "update_index": 393000, "policy_version": 131, "chunk_transitions": 131000, "gradient_updates": 393000, "n_success": 90, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.3825, "gain": 0.038500000000000034, "observed": true, "imputed": false}
54
+ {"eval_index": 53, "rl_env_steps": 1062573, "target_step": 1060000, "sr": 0.42, "round": 133, "update_index": 399000, "policy_version": 133, "chunk_transitions": 133000, "gradient_updates": 399000, "n_success": 84, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.391, "gain": 0.04800000000000004, "observed": true, "imputed": false}
55
+ {"eval_index": 54, "rl_env_steps": 1086521, "target_step": 1080000, "sr": 0.41, "round": 136, "update_index": 408000, "policy_version": 136, "chunk_transitions": 136000, "gradient_updates": 408000, "n_success": 82, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.40449999999999997, "gain": 0.07500000000000001, "observed": true, "imputed": false}
56
+ {"eval_index": 55, "rl_env_steps": 1102504, "target_step": 1100000, "sr": 0.375, "round": 138, "update_index": 414000, "policy_version": 138, "chunk_transitions": 138000, "gradient_updates": 414000, "n_success": 75, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.403, "gain": 0.064, "observed": true, "imputed": false}
57
+ {"eval_index": 56, "rl_env_steps": 1126426, "target_step": 1120000, "sr": 0.405, "round": 141, "update_index": 423000, "policy_version": 141, "chunk_transitions": 141000, "gradient_updates": 423000, "n_success": 81, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.4055000000000001, "gain": 0.06250000000000006, "observed": true, "imputed": false}
58
+ {"eval_index": 57, "rl_env_steps": 1142380, "target_step": 1140000, "sr": 0.38, "round": 143, "update_index": 429000, "policy_version": 143, "chunk_transitions": 143000, "gradient_updates": 429000, "n_success": 76, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.40599999999999997, "gain": 0.0675, "observed": true, "imputed": false}
59
+ {"eval_index": 58, "rl_env_steps": 1166316, "target_step": 1160000, "sr": 0.405, "round": 146, "update_index": 438000, "policy_version": 146, "chunk_transitions": 146000, "gradient_updates": 438000, "n_success": 81, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.40800000000000003, "gain": 0.05800000000000005, "observed": true, "imputed": false}
60
+ {"eval_index": 59, "rl_env_steps": 1182282, "target_step": 1180000, "sr": 0.37, "round": 148, "update_index": 444000, "policy_version": 148, "chunk_transitions": 148000, "gradient_updates": 444000, "n_success": 74, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.402, "gain": 0.03899999999999998, "observed": true, "imputed": false}
61
+ {"eval_index": 60, "rl_env_steps": 1206253, "target_step": 1200000, "sr": 0.41, "round": 151, "update_index": 453000, "policy_version": 151, "chunk_transitions": 151000, "gradient_updates": 453000, "n_success": 82, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.4025, "gain": 0.040000000000000036, "observed": true, "imputed": false}
62
+ {"eval_index": 61, "rl_env_steps": 1222217, "target_step": 1220000, "sr": 0.4, "round": 153, "update_index": 459000, "policy_version": 153, "chunk_transitions": 153000, "gradient_updates": 459000, "n_success": 80, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.4025, "gain": 0.02750000000000008, "observed": true, "imputed": false}
63
+ {"eval_index": 62, "rl_env_steps": 1246178, "target_step": 1240000, "sr": 0.425, "round": 156, "update_index": 468000, "policy_version": 156, "chunk_transitions": 156000, "gradient_updates": 468000, "n_success": 85, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.39999999999999997, "gain": 0.01749999999999996, "observed": true, "imputed": false}
64
+ {"eval_index": 63, "rl_env_steps": 1262135, "target_step": 1260000, "sr": 0.39, "round": 158, "update_index": 474000, "policy_version": 158, "chunk_transitions": 158000, "gradient_updates": 474000, "n_success": 78, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.39699999999999996, "gain": 0.00599999999999995, "observed": true, "imputed": false}
65
+ {"eval_index": 64, "rl_env_steps": 1286102, "target_step": 1280000, "sr": 0.39, "round": 161, "update_index": 483000, "policy_version": 161, "chunk_transitions": 161000, "gradient_updates": 483000, "n_success": 78, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.395, "gain": -0.009499999999999953, "observed": true, "imputed": false}
66
+ {"eval_index": 65, "rl_env_steps": 1302051, "target_step": 1300000, "sr": 0.43, "round": 163, "update_index": 489000, "policy_version": 163, "chunk_transitions": 163000, "gradient_updates": 489000, "n_success": 86, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.40049999999999997, "gain": -0.0025000000000000577, "observed": true, "imputed": false}
67
+ {"eval_index": 66, "rl_env_steps": 1326008, "target_step": 1320000, "sr": 0.36, "round": 166, "update_index": 498000, "policy_version": 166, "chunk_transitions": 166000, "gradient_updates": 498000, "n_success": 72, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.396, "gain": -0.009500000000000064, "observed": true, "imputed": false}
68
+ {"eval_index": 67, "rl_env_steps": 1341972, "target_step": 1340000, "sr": 0.4, "round": 168, "update_index": 504000, "policy_version": 168, "chunk_transitions": 168000, "gradient_updates": 504000, "n_success": 80, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.39799999999999996, "gain": -0.008000000000000007, "observed": true, "imputed": false}
69
+ {"eval_index": 68, "rl_env_steps": 1365913, "target_step": 1360000, "sr": 0.36, "round": 171, "update_index": 513000, "policy_version": 171, "chunk_transitions": 171000, "gradient_updates": 513000, "n_success": 72, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.39349999999999996, "gain": -0.014500000000000068, "observed": true, "imputed": false}
70
+ {"eval_index": 69, "rl_env_steps": 1381861, "target_step": 1380000, "sr": 0.39, "round": 173, "update_index": 519000, "policy_version": 173, "chunk_transitions": 173000, "gradient_updates": 519000, "n_success": 78, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.3955, "gain": -0.006500000000000006, "observed": true, "imputed": false}
71
+ {"eval_index": 70, "rl_env_steps": 1405828, "target_step": 1400000, "sr": 0.4, "round": 176, "update_index": 528000, "policy_version": 176, "chunk_transitions": 176000, "gradient_updates": 528000, "n_success": 80, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.3945, "gain": -0.008000000000000007, "observed": true, "imputed": false}
72
+ {"eval_index": 71, "rl_env_steps": 1421810, "target_step": 1420000, "sr": 0.445, "round": 178, "update_index": 534000, "policy_version": 178, "chunk_transitions": 178000, "gradient_updates": 534000, "n_success": 89, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.39899999999999997, "gain": -0.0035000000000000586, "observed": true, "imputed": false}
73
+ {"eval_index": 72, "rl_env_steps": 1445776, "target_step": 1440000, "sr": 0.375, "round": 181, "update_index": 543000, "policy_version": 181, "chunk_transitions": 181000, "gradient_updates": 543000, "n_success": 75, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.394, "gain": -0.00599999999999995, "observed": true, "imputed": false}
74
+ {"eval_index": 73, "rl_env_steps": 1461754, "target_step": 1460000, "sr": 0.395, "round": 183, "update_index": 549000, "policy_version": 183, "chunk_transitions": 183000, "gradient_updates": 549000, "n_success": 79, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.39449999999999996, "gain": -0.0025000000000000022, "observed": true, "imputed": false}
75
+ {"eval_index": 74, "rl_env_steps": 1485706, "target_step": 1480000, "sr": 0.385, "round": 186, "update_index": 558000, "policy_version": 186, "chunk_transitions": 186000, "gradient_updates": 558000, "n_success": 77, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.394, "gain": -0.0010000000000000009, "observed": true, "imputed": false}
76
+ {"eval_index": 75, "rl_env_steps": 1501671, "target_step": 1500000, "sr": 0.44, "round": 188, "update_index": 564000, "policy_version": 188, "chunk_transitions": 188000, "gradient_updates": 564000, "n_success": 88, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.39499999999999996, "gain": -0.005500000000000005, "observed": true, "imputed": false}
77
+ {"eval_index": 76, "rl_env_steps": 1525608, "target_step": 1520000, "sr": 0.44, "round": 191, "update_index": 573000, "policy_version": 191, "chunk_transitions": 191000, "gradient_updates": 573000, "n_success": 88, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.403, "gain": 0.007000000000000006, "observed": true, "imputed": false}
78
+ {"eval_index": 77, "rl_env_steps": 1541586, "target_step": 1540000, "sr": 0.355, "round": 193, "update_index": 579000, "policy_version": 193, "chunk_transitions": 193000, "gradient_updates": 579000, "n_success": 71, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.39849999999999997, "gain": 0.0005000000000000004, "observed": true, "imputed": false}
79
+ {"eval_index": 78, "rl_env_steps": 1565558, "target_step": 1560000, "sr": 0.405, "round": 196, "update_index": 588000, "policy_version": 196, "chunk_transitions": 196000, "gradient_updates": 588000, "n_success": 81, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.403, "gain": 0.009500000000000064, "observed": true, "imputed": false}
80
+ {"eval_index": 79, "rl_env_steps": 1581544, "target_step": 1580000, "sr": 0.39, "round": 198, "update_index": 594000, "policy_version": 198, "chunk_transitions": 198000, "gradient_updates": 594000, "n_success": 78, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.4029999999999999, "gain": 0.007499999999999896, "observed": true, "imputed": false}
81
+ {"eval_index": 80, "rl_env_steps": 1605499, "target_step": 1600000, "sr": 0.38, "round": 201, "update_index": 603000, "policy_version": 201, "chunk_transitions": 201000, "gradient_updates": 603000, "n_success": 76, "episodes": 200, "seed_start": 960000, "success_requested": 0, "success_skipped": 0, "success_rounds_with_skips": 0, "success_skipped_td": 0, "success_skipped_imitation": 0, "success_requested_imitation": 0, "ma": 0.4010000000000001, "gain": 0.006500000000000061, "observed": true, "imputed": false}
RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/README.md ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ Complete run archive
2
+
3
+ See publication.json for identity and terminal state. Download all files in this complete/ directory, then run `python restore.py --archive-dir . --output /path/to/restored`. It checks part hashes and every restored file against manifest.json. zstd command is required for .zst archives.
4
+
5
+ The archive contains run/ (all retained checkpoints, replay, offline buffer, metrics, evaluations, diagnostics and W&B records) and dependencies/ (base BC, manifests, exact observed source and runtime inventory). Original absolute paths remain in historical metadata. Use original directory locations or an explicit recorded path migration before training. Raw demo/image datasets are inputs referenced by their recorded hashes; they are not duplicated in this run-artifact archive. Broken external W&B service-log links are enumerated, never represented as archived bytes.
RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/SHA256SUMS ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ 3368f0482f8c6a487d37d23a927dbeedc222de6dfe1420a9cc0639c7eb63a65d best/source_metadata.json
2
+ 7459ae36c024cf53a2a799d129a91fce3a67ddc0a78386f74f30b42c876ef736 best/source_resolved_config.json
3
+ 29913d8cf3eac8fcde1aa0445b84cf990630e4553b8d366339fe93fc1433d152 best/state.pkl
4
+ 8801c851ce127366254dee275c07351254aa613b6f9a1e7a619ba50d66c6c0b2 comparison_evaluations.jsonl
5
+ 39c57050e748e4d3f5fd0314a47c05e329d07d4468132cce324735f7fef59b56 complete/README.md
6
+ 43ac2d4be1cb514af3586da730f7b20df7cb010ab76840d0829efb2ba870b46b complete/archive.json
7
+ 67856cd29fb301582addc0e2cd4968e76c983beadb20ad70f3780676fd8c472f complete/full_run.tar.zst.part000
8
+ efc28d6f1ad7b3dc966da4391f511a5841ce62089cd7b9062524650a52efa4e7 complete/full_run.tar.zst.part001
9
+ df405f84e55f2673e2f780196999a01a9bf4ff0a6bf1fdbefe0a4bbdac2c55b8 complete/full_run.tar.zst.part002
10
+ b8cdfda859dbcd29be6745a6e6d7d263e5c54ae54e5bad4b456fc41c653582c6 complete/full_run.tar.zst.part003
11
+ 43a9fa2d7ed4aba4597411980ba4efab1f39e2d1e6911cc68208587ecfafda48 complete/full_run.tar.zst.part004
12
+ c98ea371d73f82e4a03deaa02042ef301b533eb3da4d0051231d2b8c8e877f06 complete/full_run.tar.zst.part005
13
+ 9f77a2e9cbb49c1bb2b1ef925f3ba672c7d87ea35982c098386d3d3aad479836 complete/full_run.tar.zst.part006
14
+ e929ea22b68b7138beeb1157a44a4ad97b66a84f728fc22f54b2058a3f3d24eb complete/full_run.tar.zst.part007
15
+ 349ab1bb3b9a36a7516bc0dbb9e8cc144fdd2451068f20001fd43331fb8abf97 complete/full_run.tar.zst.part008
16
+ fb8ec0773fc5a17c4bf1f83433cd828ae39f2ca7828d63ddf4f31c206c602f2d complete/full_run.tar.zst.part009
17
+ f9fcaa9b3cb82eb23e635eaed77575fcc84749aa022919bfa5c9d3819043ce90 complete/local_archive_verification.json
18
+ cf22f621a43ab01f2801415369a739685df0c7c4c4a287a0e3694f34172e19d2 complete/manifest.json
19
+ ff4d0482fb8c78212d32321ae8c15031b46c5e99bea337f6fe5ac5819386740f complete/publication.json
20
+ 149a44cd0b002e83f507d1e46d813d5e2b8cf6f9e1bb51f14708343f48788ac2 complete/restore.py
21
+ a8f0072c44e03e755681953314446aad85990faa42e1c9fa69ed741d36762e14 convergence.json
22
+ 85085b9f391d8714edb0d2cafeae74722cd1ff4c99583266edaa576ae1ce80ba convergence.jsonl
23
+ 7a7afc8dbf75ac8d5eb1f5e8407e4bb769666f98a7b70f9ef1d701eb13b1058e eval/final.parquet
24
+ 5fc534e453688835f5b0b906188a49536b44e540efe4ddc9a647e3dcc2d3d3e2 final/source_metadata.json
25
+ 7459ae36c024cf53a2a799d129a91fce3a67ddc0a78386f74f30b42c876ef736 final/source_resolved_config.json
26
+ 88e0762bf71060889e6ad409345b3357e471f684857c3b54114d0fe61c982582 final/state.pkl
27
+ b49df1f510fc5f3a256ed92e202b4b4586349d2888bcfd870c32b573eed8ebb3 frozen_base.json
28
+ 1584571e842fe4ace8897303fdd885e9a6bfdd22f895ed2bac5f2ae290f2bfa5 offline_buffer.json
29
+ 7459ae36c024cf53a2a799d129a91fce3a67ddc0a78386f74f30b42c876ef736 resolved_config.json
30
+ a6075aaefc19d21a635766f8d1c4deb0ff1abd2ad05399bf9cc4613fac6d8009 start_protocol.json
31
+ 8bdc6ad619d2a22a7b42a3253af8cfcef1e9cae42329bafc770a805adfa351c7 status.json
32
+ 8f67a72fd458b492db7767ed2ddaf8f5d993719606fa84354e969be32777e6ed training_accounting.json
RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/archive.json ADDED
@@ -0,0 +1,58 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "format": "tar.zst",
3
+ "parts": [
4
+ {
5
+ "path": "full_run.tar.zst.part000",
6
+ "size": 1073741824,
7
+ "sha256": "67856cd29fb301582addc0e2cd4968e76c983beadb20ad70f3780676fd8c472f"
8
+ },
9
+ {
10
+ "path": "full_run.tar.zst.part001",
11
+ "size": 1073741824,
12
+ "sha256": "efc28d6f1ad7b3dc966da4391f511a5841ce62089cd7b9062524650a52efa4e7"
13
+ },
14
+ {
15
+ "path": "full_run.tar.zst.part002",
16
+ "size": 1073741824,
17
+ "sha256": "df405f84e55f2673e2f780196999a01a9bf4ff0a6bf1fdbefe0a4bbdac2c55b8"
18
+ },
19
+ {
20
+ "path": "full_run.tar.zst.part003",
21
+ "size": 1073741824,
22
+ "sha256": "b8cdfda859dbcd29be6745a6e6d7d263e5c54ae54e5bad4b456fc41c653582c6"
23
+ },
24
+ {
25
+ "path": "full_run.tar.zst.part004",
26
+ "size": 1073741824,
27
+ "sha256": "43a9fa2d7ed4aba4597411980ba4efab1f39e2d1e6911cc68208587ecfafda48"
28
+ },
29
+ {
30
+ "path": "full_run.tar.zst.part005",
31
+ "size": 1073741824,
32
+ "sha256": "c98ea371d73f82e4a03deaa02042ef301b533eb3da4d0051231d2b8c8e877f06"
33
+ },
34
+ {
35
+ "path": "full_run.tar.zst.part006",
36
+ "size": 1073741824,
37
+ "sha256": "9f77a2e9cbb49c1bb2b1ef925f3ba672c7d87ea35982c098386d3d3aad479836"
38
+ },
39
+ {
40
+ "path": "full_run.tar.zst.part007",
41
+ "size": 1073741824,
42
+ "sha256": "e929ea22b68b7138beeb1157a44a4ad97b66a84f728fc22f54b2058a3f3d24eb"
43
+ },
44
+ {
45
+ "path": "full_run.tar.zst.part008",
46
+ "size": 1073741824,
47
+ "sha256": "349ab1bb3b9a36a7516bc0dbb9e8cc144fdd2451068f20001fd43331fb8abf97"
48
+ },
49
+ {
50
+ "path": "full_run.tar.zst.part009",
51
+ "size": 607438663,
52
+ "sha256": "fb8ec0773fc5a17c4bf1f83433cd828ae39f2ca7828d63ddf4f31c206c602f2d"
53
+ }
54
+ ],
55
+ "manifest_sha256": "cf22f621a43ab01f2801415369a739685df0c7c4c4a287a0e3694f34172e19d2",
56
+ "hardlink_content_dedup": true,
57
+ "member_count": 6975
58
+ }
RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/full_run.tar.zst.part000 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:67856cd29fb301582addc0e2cd4968e76c983beadb20ad70f3780676fd8c472f
3
+ size 1073741824
RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/full_run.tar.zst.part001 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:efc28d6f1ad7b3dc966da4391f511a5841ce62089cd7b9062524650a52efa4e7
3
+ size 1073741824
RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/full_run.tar.zst.part002 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:df405f84e55f2673e2f780196999a01a9bf4ff0a6bf1fdbefe0a4bbdac2c55b8
3
+ size 1073741824
RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/full_run.tar.zst.part003 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b8cdfda859dbcd29be6745a6e6d7d263e5c54ae54e5bad4b456fc41c653582c6
3
+ size 1073741824
RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/full_run.tar.zst.part004 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:43a9fa2d7ed4aba4597411980ba4efab1f39e2d1e6911cc68208587ecfafda48
3
+ size 1073741824
RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/full_run.tar.zst.part005 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c98ea371d73f82e4a03deaa02042ef301b533eb3da4d0051231d2b8c8e877f06
3
+ size 1073741824
RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/full_run.tar.zst.part006 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f77a2e9cbb49c1bb2b1ef925f3ba672c7d87ea35982c098386d3d3aad479836
3
+ size 1073741824
RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/full_run.tar.zst.part007 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e929ea22b68b7138beeb1157a44a4ad97b66a84f728fc22f54b2058a3f3d24eb
3
+ size 1073741824
RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/full_run.tar.zst.part008 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:349ab1bb3b9a36a7516bc0dbb9e8cc144fdd2451068f20001fd43331fb8abf97
3
+ size 1073741824
RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/full_run.tar.zst.part009 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fb8ec0773fc5a17c4bf1f83433cd828ae39f2ca7828d63ddf4f31c206c602f2d
3
+ size 607438663
RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/local_archive_verification.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "passed": true,
3
+ "verified_members": 6974,
4
+ "verified_archive_parts": 10,
5
+ "time": 1793313396.4566998
6
+ }
RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/manifest.json ADDED
The diff for this file is too large to render. See raw diff
 
RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/publication.json ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": 2,
3
+ "run_id": "92c7dc10",
4
+ "task": "ToolHang100",
5
+ "algorithm": "DSRL-NA",
6
+ "variant": "dsrl_na",
7
+ "cohort": "diverse100",
8
+ "seed": 1,
9
+ "protocol": "study2_protocol_v10",
10
+ "status": "budget_complete",
11
+ "path": "RL/ToolHang100/DSRL-NA/diverse100/seed_1",
12
+ "checkpoints": {
13
+ "best": {
14
+ "path": "RL/ToolHang100/DSRL-NA/diverse100/seed_1/best/state.pkl",
15
+ "sha256": "29913d8cf3eac8fcde1aa0445b84cf990630e4553b8d366339fe93fc1433d152",
16
+ "bytes": 282975465,
17
+ "source_path": "/ssd/WorkSpace/RL-Postrain/runs/experiments/study2/pixel_state/20261024_toolhang_diverse100_dsrl_na_v10_seed1/checkpoints/best/eval_00052_step_1046596_session_1792820703091133051/checkpoint/state.pkl",
18
+ "evaluation": {
19
+ "chunk_transitions": 131000,
20
+ "episodes": 200,
21
+ "eval_index": 52,
22
+ "gain": 0.038500000000000034,
23
+ "gradient_updates": 393000,
24
+ "ma": 0.3825,
25
+ "n_success": 90,
26
+ "policy_version": 131,
27
+ "rl_env_steps": 1046596,
28
+ "round": 131,
29
+ "seed_start": 960000,
30
+ "sr": 0.45,
31
+ "success_requested": 0,
32
+ "success_requested_imitation": 0,
33
+ "success_rounds_with_skips": 0,
34
+ "success_skipped": 0,
35
+ "success_skipped_imitation": 0,
36
+ "success_skipped_td": 0,
37
+ "target_step": 1040000,
38
+ "update_index": 393000
39
+ }
40
+ },
41
+ "final": {
42
+ "path": "RL/ToolHang100/DSRL-NA/diverse100/seed_1/final/state.pkl",
43
+ "sha256": "88e0762bf71060889e6ad409345b3357e471f684857c3b54114d0fe61c982582",
44
+ "bytes": 282982148,
45
+ "source_path": "/ssd/WorkSpace/RL-Postrain/runs/experiments/study2/pixel_state/20261024_toolhang_diverse100_dsrl_na_v10_seed1/checkpoints/final/session_1792820703091133051/checkpoint/state.pkl",
46
+ "evaluation": {
47
+ "episodes": 300,
48
+ "imputed": false,
49
+ "n_success": 107,
50
+ "observed": true,
51
+ "seed_start": 800000,
52
+ "sr": 0.3566666666666667
53
+ }
54
+ }
55
+ },
56
+ "complete_artifacts": "RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete",
57
+ "source_host": "Paratera",
58
+ "source_run": "/ssd/WorkSpace/RL-Postrain/runs/experiments/study2/pixel_state/20261024_toolhang_diverse100_dsrl_na_v10_seed1",
59
+ "artifact_files": 6974,
60
+ "archive_bytes": 10271115079,
61
+ "omitted_broken_log_links": 0
62
+ }
RL/ToolHang100/DSRL-NA/diverse100/seed_1/complete/restore.py ADDED
@@ -0,0 +1,38 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Restore a full RL run and verify every archived member without executing its contents."""
2
+ import argparse,hashlib,json,os,subprocess,tarfile,threading,shutil
3
+ from pathlib import Path
4
+
5
+ def digest(p):
6
+ h=hashlib.sha256()
7
+ with p.open('rb') as f:
8
+ for b in iter(lambda:f.read(8*1024**2),b''):h.update(b)
9
+ return h.hexdigest()
10
+ def main():
11
+ ap=argparse.ArgumentParser();ap.add_argument('--archive-dir',type=Path,default=Path('.'));ap.add_argument('--output',type=Path,required=True);args=ap.parse_args();base=args.archive_dir;out=args.output.resolve();out.mkdir(parents=True,exist_ok=True);assert not any(out.iterdir()),'Use an empty output directory'
12
+ arc=json.loads((base/'archive.json').read_text());assert digest(base/'manifest.json')==arc['manifest_sha256'];manifest=json.loads((base/'manifest.json').read_text());expected={e['path']:e for e in manifest['entries']}
13
+ for part in arc['parts']:
14
+ p=base/part['path'];assert p.stat().st_size==part['size'] and digest(p)==part['sha256'],p
15
+ cmd=['zstd','-q','-d','-c'] if arc['format'].endswith('zst') else ['gzip','-d','-c'];proc=subprocess.Popen(cmd,stdin=subprocess.PIPE,stdout=subprocess.PIPE);errors=[]
16
+ def feed():
17
+ try:
18
+ for part in arc['parts']:
19
+ with (base/part['path']).open('rb') as f:shutil.copyfileobj(f,proc.stdin,1024**2)
20
+ except BaseException as e:errors.append(e)
21
+ finally:proc.stdin.close()
22
+ t=threading.Thread(target=feed);t.start();seen={}
23
+ with tarfile.open(fileobj=proc.stdout,mode='r|') as tar:
24
+ for m in tar:
25
+ if m.name=='MANIFEST.json':assert hashlib.sha256(tar.extractfile(m).read()).hexdigest()==arc['manifest_sha256'];continue
26
+ e=expected[m.name];dest=out/m.name;assert dest.resolve().is_relative_to(out);dest.parent.mkdir(parents=True,exist_ok=True)
27
+ if m.issym():
28
+ assert e['type']=='symlink' and m.linkname==e['target'] and (dest.parent/m.linkname).resolve().is_relative_to(out);dest.symlink_to(m.linkname);seen[m.name]=None
29
+ elif m.islnk():
30
+ assert seen[m.linkname]==e['sha256'];src=out/m.linkname;assert src.resolve().is_relative_to(out);os.link(src,dest);seen[m.name]=e['sha256']
31
+ else:
32
+ assert m.isfile() and m.size==e['size'];h=hashlib.sha256()
33
+ with dest.open('xb') as f:
34
+ inp=tar.extractfile(m)
35
+ for b in iter(lambda:inp.read(8*1024**2),b''):f.write(b);h.update(b)
36
+ assert h.hexdigest()==e['sha256'],m.name;dest.chmod(e.get('mode',0o644));seen[m.name]=h.hexdigest()
37
+ t.join();assert proc.wait()==0 and not errors and len(seen)==len(expected);shutil.copy2(base/'manifest.json',out/'MANIFEST.json');(out/'RESTORE_VERIFIED.json').write_text(json.dumps({'passed':True,'files':len(seen),'run_id':manifest['run']['run_id']},indent=2));print('RESTORE VERIFIED',manifest['run']['run_id'],len(seen),'members')
38
+ if __name__=='__main__':main()
RL/ToolHang100/DSRL-NA/diverse100/seed_1/convergence.json ADDED
@@ -0,0 +1,367 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "evals": 81,
3
+ "final_eval": {
4
+ "episodes": 300,
5
+ "imputed": false,
6
+ "n_success": 107,
7
+ "observed": true,
8
+ "seed_start": 800000,
9
+ "sr": 0.3566666666666667,
10
+ "successes": [
11
+ 0,
12
+ 0,
13
+ 0,
14
+ 1,
15
+ 1,
16
+ 1,
17
+ 0,
18
+ 0,
19
+ 1,
20
+ 1,
21
+ 1,
22
+ 0,
23
+ 1,
24
+ 0,
25
+ 1,
26
+ 0,
27
+ 0,
28
+ 1,
29
+ 0,
30
+ 0,
31
+ 0,
32
+ 0,
33
+ 1,
34
+ 0,
35
+ 0,
36
+ 0,
37
+ 1,
38
+ 0,
39
+ 0,
40
+ 1,
41
+ 0,
42
+ 0,
43
+ 0,
44
+ 1,
45
+ 0,
46
+ 0,
47
+ 1,
48
+ 0,
49
+ 1,
50
+ 0,
51
+ 1,
52
+ 0,
53
+ 1,
54
+ 1,
55
+ 0,
56
+ 1,
57
+ 0,
58
+ 0,
59
+ 0,
60
+ 1,
61
+ 0,
62
+ 0,
63
+ 0,
64
+ 0,
65
+ 0,
66
+ 1,
67
+ 0,
68
+ 0,
69
+ 0,
70
+ 1,
71
+ 1,
72
+ 0,
73
+ 1,
74
+ 0,
75
+ 1,
76
+ 0,
77
+ 0,
78
+ 1,
79
+ 1,
80
+ 0,
81
+ 1,
82
+ 1,
83
+ 1,
84
+ 1,
85
+ 1,
86
+ 0,
87
+ 0,
88
+ 0,
89
+ 0,
90
+ 1,
91
+ 0,
92
+ 0,
93
+ 0,
94
+ 0,
95
+ 0,
96
+ 0,
97
+ 0,
98
+ 0,
99
+ 0,
100
+ 0,
101
+ 0,
102
+ 0,
103
+ 0,
104
+ 1,
105
+ 0,
106
+ 0,
107
+ 0,
108
+ 0,
109
+ 0,
110
+ 0,
111
+ 1,
112
+ 1,
113
+ 0,
114
+ 0,
115
+ 0,
116
+ 0,
117
+ 0,
118
+ 0,
119
+ 1,
120
+ 0,
121
+ 0,
122
+ 0,
123
+ 0,
124
+ 0,
125
+ 0,
126
+ 1,
127
+ 1,
128
+ 0,
129
+ 0,
130
+ 0,
131
+ 0,
132
+ 0,
133
+ 0,
134
+ 1,
135
+ 0,
136
+ 1,
137
+ 0,
138
+ 0,
139
+ 0,
140
+ 1,
141
+ 0,
142
+ 1,
143
+ 1,
144
+ 0,
145
+ 0,
146
+ 1,
147
+ 1,
148
+ 0,
149
+ 0,
150
+ 0,
151
+ 0,
152
+ 1,
153
+ 0,
154
+ 0,
155
+ 1,
156
+ 0,
157
+ 0,
158
+ 0,
159
+ 0,
160
+ 0,
161
+ 0,
162
+ 0,
163
+ 1,
164
+ 1,
165
+ 1,
166
+ 0,
167
+ 0,
168
+ 0,
169
+ 0,
170
+ 0,
171
+ 0,
172
+ 1,
173
+ 0,
174
+ 1,
175
+ 0,
176
+ 0,
177
+ 0,
178
+ 1,
179
+ 1,
180
+ 0,
181
+ 0,
182
+ 0,
183
+ 0,
184
+ 1,
185
+ 0,
186
+ 0,
187
+ 1,
188
+ 0,
189
+ 1,
190
+ 0,
191
+ 1,
192
+ 1,
193
+ 0,
194
+ 0,
195
+ 0,
196
+ 0,
197
+ 0,
198
+ 0,
199
+ 0,
200
+ 0,
201
+ 1,
202
+ 0,
203
+ 1,
204
+ 0,
205
+ 0,
206
+ 0,
207
+ 0,
208
+ 0,
209
+ 0,
210
+ 1,
211
+ 0,
212
+ 0,
213
+ 1,
214
+ 1,
215
+ 1,
216
+ 0,
217
+ 1,
218
+ 1,
219
+ 1,
220
+ 1,
221
+ 0,
222
+ 0,
223
+ 1,
224
+ 1,
225
+ 0,
226
+ 0,
227
+ 1,
228
+ 1,
229
+ 0,
230
+ 1,
231
+ 0,
232
+ 0,
233
+ 0,
234
+ 0,
235
+ 1,
236
+ 1,
237
+ 0,
238
+ 1,
239
+ 0,
240
+ 0,
241
+ 0,
242
+ 0,
243
+ 0,
244
+ 0,
245
+ 0,
246
+ 0,
247
+ 0,
248
+ 1,
249
+ 0,
250
+ 1,
251
+ 0,
252
+ 1,
253
+ 1,
254
+ 1,
255
+ 0,
256
+ 0,
257
+ 0,
258
+ 0,
259
+ 1,
260
+ 1,
261
+ 0,
262
+ 1,
263
+ 1,
264
+ 1,
265
+ 1,
266
+ 1,
267
+ 1,
268
+ 0,
269
+ 0,
270
+ 1,
271
+ 0,
272
+ 0,
273
+ 1,
274
+ 0,
275
+ 0,
276
+ 1,
277
+ 0,
278
+ 0,
279
+ 0,
280
+ 0,
281
+ 1,
282
+ 1,
283
+ 1,
284
+ 1,
285
+ 0,
286
+ 1,
287
+ 0,
288
+ 1,
289
+ 0,
290
+ 0,
291
+ 0,
292
+ 1,
293
+ 0,
294
+ 1,
295
+ 0,
296
+ 0,
297
+ 0,
298
+ 0,
299
+ 1,
300
+ 0,
301
+ 0,
302
+ 0,
303
+ 0,
304
+ 0,
305
+ 1,
306
+ 0,
307
+ 0,
308
+ 1,
309
+ 1,
310
+ 1
311
+ ]
312
+ },
313
+ "last": {
314
+ "chunk_transitions": 201000,
315
+ "episodes": 200,
316
+ "eval_index": 80,
317
+ "gain": 0.006500000000000061,
318
+ "gradient_updates": 603000,
319
+ "ma": 0.4010000000000001,
320
+ "n_success": 76,
321
+ "policy_version": 201,
322
+ "rl_env_steps": 1605499,
323
+ "round": 201,
324
+ "seed_start": 960000,
325
+ "sr": 0.38,
326
+ "success_requested": 0,
327
+ "success_requested_imitation": 0,
328
+ "success_rounds_with_skips": 0,
329
+ "success_skipped": 0,
330
+ "success_skipped_imitation": 0,
331
+ "success_skipped_td": 0,
332
+ "target_step": 1600000,
333
+ "update_index": 603000
334
+ },
335
+ "rule": {
336
+ "early_stop": false,
337
+ "every_env_steps": 20000,
338
+ "gain_lag": 10,
339
+ "high_sr": 0.95,
340
+ "high_sr_evals": 5,
341
+ "include_step0": true,
342
+ "ma_window": 10,
343
+ "max_env_steps": 1606196,
344
+ "min_env_steps": 200000,
345
+ "min_gain": 0.02,
346
+ "zero_success_evals": 8
347
+ },
348
+ "status": "budget_complete",
349
+ "stop": {
350
+ "eval_index": 80,
351
+ "gain": 0.006500000000000061,
352
+ "ma": 0.4010000000000001,
353
+ "rl_env_steps": 1613492,
354
+ "sr": 0.38,
355
+ "status": "budget_complete",
356
+ "trend": "flat"
357
+ },
358
+ "would_stop": {
359
+ "eval_index": 23,
360
+ "gain": 0.016500000000000015,
361
+ "ma": 0.29100000000000004,
362
+ "rl_env_steps": 463551,
363
+ "sr": 0.27,
364
+ "status": "converged_plateau",
365
+ "trend": "flat"
366
+ }
367
+ }
RL/ToolHang100/DSRL-NA/diverse100/seed_1/convergence.jsonl ADDED
@@ -0,0 +1,81 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"chunk_transitions": 0, "episodes": 200, "eval_index": 0, "gain": null, "gradient_updates": 0, "ma": null, "n_success": 56, "policy_version": 0, "rl_env_steps": 0, "round": 0, "seed_start": 960000, "sr": 0.28, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 0, "update_index": 0}
2
+ {"chunk_transitions": 3000, "episodes": 200, "eval_index": 1, "gain": null, "gradient_updates": 9000, "ma": null, "n_success": 28, "policy_version": 3, "rl_env_steps": 23980, "round": 3, "seed_start": 960000, "sr": 0.14, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 20000, "update_index": 9000}
3
+ {"chunk_transitions": 6000, "episodes": 200, "eval_index": 2, "gain": null, "gradient_updates": 18000, "ma": null, "n_success": 61, "policy_version": 6, "rl_env_steps": 47942, "round": 6, "seed_start": 960000, "sr": 0.305, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 40000, "update_index": 18000}
4
+ {"chunk_transitions": 8000, "episodes": 200, "eval_index": 3, "gain": null, "gradient_updates": 24000, "ma": null, "n_success": 25, "policy_version": 8, "rl_env_steps": 63926, "round": 8, "seed_start": 960000, "sr": 0.125, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 60000, "update_index": 24000}
5
+ {"chunk_transitions": 11000, "episodes": 200, "eval_index": 4, "gain": null, "gradient_updates": 33000, "ma": null, "n_success": 40, "policy_version": 11, "rl_env_steps": 87915, "round": 11, "seed_start": 960000, "sr": 0.2, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 80000, "update_index": 33000}
6
+ {"chunk_transitions": 13000, "episodes": 200, "eval_index": 5, "gain": null, "gradient_updates": 39000, "ma": null, "n_success": 47, "policy_version": 13, "rl_env_steps": 103888, "round": 13, "seed_start": 960000, "sr": 0.235, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 100000, "update_index": 39000}
7
+ {"chunk_transitions": 16000, "episodes": 200, "eval_index": 6, "gain": null, "gradient_updates": 48000, "ma": null, "n_success": 61, "policy_version": 16, "rl_env_steps": 127883, "round": 16, "seed_start": 960000, "sr": 0.305, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 120000, "update_index": 48000}
8
+ {"chunk_transitions": 18000, "episodes": 200, "eval_index": 7, "gain": null, "gradient_updates": 54000, "ma": null, "n_success": 67, "policy_version": 18, "rl_env_steps": 143862, "round": 18, "seed_start": 960000, "sr": 0.335, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 140000, "update_index": 54000}
9
+ {"chunk_transitions": 21000, "episodes": 200, "eval_index": 8, "gain": null, "gradient_updates": 63000, "ma": null, "n_success": 54, "policy_version": 21, "rl_env_steps": 167829, "round": 21, "seed_start": 960000, "sr": 0.27, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 160000, "update_index": 63000}
10
+ {"chunk_transitions": 23000, "episodes": 200, "eval_index": 9, "gain": null, "gradient_updates": 69000, "ma": 0.253, "n_success": 67, "policy_version": 23, "rl_env_steps": 183814, "round": 23, "seed_start": 960000, "sr": 0.335, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 180000, "update_index": 69000}
11
+ {"chunk_transitions": 26000, "episodes": 200, "eval_index": 10, "gain": null, "gradient_updates": 78000, "ma": 0.2535, "n_success": 57, "policy_version": 26, "rl_env_steps": 207800, "round": 26, "seed_start": 960000, "sr": 0.285, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 200000, "update_index": 78000}
12
+ {"chunk_transitions": 28000, "episodes": 200, "eval_index": 11, "gain": null, "gradient_updates": 84000, "ma": 0.26000000000000006, "n_success": 41, "policy_version": 28, "rl_env_steps": 223782, "round": 28, "seed_start": 960000, "sr": 0.205, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 220000, "update_index": 84000}
13
+ {"chunk_transitions": 31000, "episodes": 200, "eval_index": 12, "gain": null, "gradient_updates": 93000, "ma": 0.26, "n_success": 61, "policy_version": 31, "rl_env_steps": 247757, "round": 31, "seed_start": 960000, "sr": 0.305, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 240000, "update_index": 93000}
14
+ {"chunk_transitions": 33000, "episodes": 200, "eval_index": 13, "gain": null, "gradient_updates": 99000, "ma": 0.2745, "n_success": 54, "policy_version": 33, "rl_env_steps": 263737, "round": 33, "seed_start": 960000, "sr": 0.27, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 260000, "update_index": 99000}
15
+ {"chunk_transitions": 36000, "episodes": 200, "eval_index": 14, "gain": null, "gradient_updates": 108000, "ma": 0.28300000000000003, "n_success": 57, "policy_version": 36, "rl_env_steps": 287727, "round": 36, "seed_start": 960000, "sr": 0.285, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 280000, "update_index": 108000}
16
+ {"chunk_transitions": 38000, "episodes": 200, "eval_index": 15, "gain": null, "gradient_updates": 114000, "ma": 0.28650000000000003, "n_success": 54, "policy_version": 38, "rl_env_steps": 303723, "round": 38, "seed_start": 960000, "sr": 0.27, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 300000, "update_index": 114000}
17
+ {"chunk_transitions": 41000, "episodes": 200, "eval_index": 16, "gain": null, "gradient_updates": 123000, "ma": 0.28650000000000003, "n_success": 61, "policy_version": 41, "rl_env_steps": 327678, "round": 41, "seed_start": 960000, "sr": 0.305, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 320000, "update_index": 123000}
18
+ {"chunk_transitions": 43000, "episodes": 200, "eval_index": 17, "gain": null, "gradient_updates": 129000, "ma": 0.27949999999999997, "n_success": 53, "policy_version": 43, "rl_env_steps": 343668, "round": 43, "seed_start": 960000, "sr": 0.265, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 340000, "update_index": 129000}
19
+ {"chunk_transitions": 46000, "episodes": 200, "eval_index": 18, "gain": null, "gradient_updates": 138000, "ma": 0.28500000000000003, "n_success": 65, "policy_version": 46, "rl_env_steps": 367651, "round": 46, "seed_start": 960000, "sr": 0.325, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 360000, "update_index": 138000}
20
+ {"chunk_transitions": 48000, "episodes": 200, "eval_index": 19, "gain": 0.025000000000000022, "gradient_updates": 144000, "ma": 0.278, "n_success": 53, "policy_version": 48, "rl_env_steps": 383633, "round": 48, "seed_start": 960000, "sr": 0.265, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 380000, "update_index": 144000}
21
+ {"chunk_transitions": 51000, "episodes": 200, "eval_index": 20, "gain": 0.028500000000000025, "gradient_updates": 153000, "ma": 0.28200000000000003, "n_success": 65, "policy_version": 51, "rl_env_steps": 407615, "round": 51, "seed_start": 960000, "sr": 0.325, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 400000, "update_index": 153000}
22
+ {"chunk_transitions": 53000, "episodes": 200, "eval_index": 21, "gain": 0.030999999999999972, "gradient_updates": 159000, "ma": 0.29100000000000004, "n_success": 59, "policy_version": 53, "rl_env_steps": 423596, "round": 53, "seed_start": 960000, "sr": 0.295, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 420000, "update_index": 159000}
23
+ {"chunk_transitions": 56000, "episodes": 200, "eval_index": 22, "gain": 0.031000000000000028, "gradient_updates": 168000, "ma": 0.29100000000000004, "n_success": 61, "policy_version": 56, "rl_env_steps": 447575, "round": 56, "seed_start": 960000, "sr": 0.305, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 440000, "update_index": 168000}
24
+ {"chunk_transitions": 58000, "episodes": 200, "eval_index": 23, "gain": 0.016500000000000015, "gradient_updates": 174000, "ma": 0.29100000000000004, "n_success": 54, "policy_version": 58, "rl_env_steps": 463551, "round": 58, "seed_start": 960000, "sr": 0.27, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 460000, "update_index": 174000}
25
+ {"chunk_transitions": 61000, "episodes": 200, "eval_index": 24, "gain": 0.008999999999999952, "gradient_updates": 183000, "ma": 0.292, "n_success": 59, "policy_version": 61, "rl_env_steps": 487519, "round": 61, "seed_start": 960000, "sr": 0.295, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 480000, "update_index": 183000}
26
+ {"chunk_transitions": 63000, "episodes": 200, "eval_index": 25, "gain": 0.008499999999999952, "gradient_updates": 189000, "ma": 0.295, "n_success": 60, "policy_version": 63, "rl_env_steps": 503500, "round": 63, "seed_start": 960000, "sr": 0.3, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 500000, "update_index": 189000}
27
+ {"chunk_transitions": 66000, "episodes": 200, "eval_index": 26, "gain": 0.007499999999999951, "gradient_updates": 198000, "ma": 0.294, "n_success": 59, "policy_version": 66, "rl_env_steps": 527451, "round": 66, "seed_start": 960000, "sr": 0.295, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 520000, "update_index": 198000}
28
+ {"chunk_transitions": 68000, "episodes": 200, "eval_index": 27, "gain": 0.019500000000000017, "gradient_updates": 204000, "ma": 0.299, "n_success": 63, "policy_version": 68, "rl_env_steps": 543434, "round": 68, "seed_start": 960000, "sr": 0.315, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 540000, "update_index": 204000}
29
+ {"chunk_transitions": 71000, "episodes": 200, "eval_index": 28, "gain": 0.01599999999999996, "gradient_updates": 213000, "ma": 0.301, "n_success": 69, "policy_version": 71, "rl_env_steps": 567377, "round": 71, "seed_start": 960000, "sr": 0.345, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 560000, "update_index": 213000}
30
+ {"chunk_transitions": 73000, "episodes": 200, "eval_index": 29, "gain": 0.021999999999999964, "gradient_updates": 219000, "ma": 0.3, "n_success": 51, "policy_version": 73, "rl_env_steps": 583356, "round": 73, "seed_start": 960000, "sr": 0.255, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 580000, "update_index": 219000}
31
+ {"chunk_transitions": 76000, "episodes": 200, "eval_index": 30, "gain": 0.022999999999999965, "gradient_updates": 228000, "ma": 0.305, "n_success": 75, "policy_version": 76, "rl_env_steps": 607327, "round": 76, "seed_start": 960000, "sr": 0.375, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 600000, "update_index": 228000}
32
+ {"chunk_transitions": 78000, "episodes": 200, "eval_index": 31, "gain": 0.023499999999999965, "gradient_updates": 234000, "ma": 0.3145, "n_success": 78, "policy_version": 78, "rl_env_steps": 623310, "round": 78, "seed_start": 960000, "sr": 0.39, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 620000, "update_index": 234000}
33
+ {"chunk_transitions": 81000, "episodes": 200, "eval_index": 32, "gain": 0.02400000000000002, "gradient_updates": 243000, "ma": 0.31500000000000006, "n_success": 62, "policy_version": 81, "rl_env_steps": 647289, "round": 81, "seed_start": 960000, "sr": 0.31, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 640000, "update_index": 243000}
34
+ {"chunk_transitions": 83000, "episodes": 200, "eval_index": 33, "gain": 0.03150000000000003, "gradient_updates": 249000, "ma": 0.32250000000000006, "n_success": 69, "policy_version": 83, "rl_env_steps": 663271, "round": 83, "seed_start": 960000, "sr": 0.345, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 660000, "update_index": 249000}
35
+ {"chunk_transitions": 86000, "episodes": 200, "eval_index": 34, "gain": 0.04199999999999998, "gradient_updates": 258000, "ma": 0.33399999999999996, "n_success": 82, "policy_version": 86, "rl_env_steps": 687236, "round": 86, "seed_start": 960000, "sr": 0.41, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 680000, "update_index": 258000}
36
+ {"chunk_transitions": 88000, "episodes": 200, "eval_index": 35, "gain": 0.038500000000000034, "gradient_updates": 264000, "ma": 0.3335, "n_success": 59, "policy_version": 88, "rl_env_steps": 703211, "round": 88, "seed_start": 960000, "sr": 0.295, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 700000, "update_index": 264000}
37
+ {"chunk_transitions": 91000, "episodes": 200, "eval_index": 36, "gain": 0.043999999999999984, "gradient_updates": 273000, "ma": 0.33799999999999997, "n_success": 68, "policy_version": 91, "rl_env_steps": 727182, "round": 91, "seed_start": 960000, "sr": 0.34, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 720000, "update_index": 273000}
38
+ {"chunk_transitions": 93000, "episodes": 200, "eval_index": 37, "gain": 0.04949999999999993, "gradient_updates": 279000, "ma": 0.3484999999999999, "n_success": 84, "policy_version": 93, "rl_env_steps": 743173, "round": 93, "seed_start": 960000, "sr": 0.42, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 740000, "update_index": 279000}
39
+ {"chunk_transitions": 96000, "episodes": 200, "eval_index": 38, "gain": 0.03999999999999998, "gradient_updates": 288000, "ma": 0.34099999999999997, "n_success": 54, "policy_version": 96, "rl_env_steps": 767118, "round": 96, "seed_start": 960000, "sr": 0.27, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 760000, "update_index": 288000}
40
+ {"chunk_transitions": 98000, "episodes": 200, "eval_index": 39, "gain": 0.045499999999999985, "gradient_updates": 294000, "ma": 0.3455, "n_success": 60, "policy_version": 98, "rl_env_steps": 783089, "round": 98, "seed_start": 960000, "sr": 0.3, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 780000, "update_index": 294000}
41
+ {"chunk_transitions": 101000, "episodes": 200, "eval_index": 40, "gain": 0.043999999999999984, "gradient_updates": 303000, "ma": 0.349, "n_success": 82, "policy_version": 101, "rl_env_steps": 807064, "round": 101, "seed_start": 960000, "sr": 0.41, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 800000, "update_index": 303000}
42
+ {"chunk_transitions": 103000, "episodes": 200, "eval_index": 41, "gain": 0.02300000000000002, "gradient_updates": 309000, "ma": 0.3375, "n_success": 55, "policy_version": 103, "rl_env_steps": 823026, "round": 103, "seed_start": 960000, "sr": 0.275, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 820000, "update_index": 309000}
43
+ {"chunk_transitions": 106000, "episodes": 200, "eval_index": 42, "gain": 0.028999999999999915, "gradient_updates": 318000, "ma": 0.344, "n_success": 75, "policy_version": 106, "rl_env_steps": 846982, "round": 106, "seed_start": 960000, "sr": 0.375, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 840000, "update_index": 318000}
44
+ {"chunk_transitions": 108000, "episodes": 200, "eval_index": 43, "gain": 0.020499999999999907, "gradient_updates": 324000, "ma": 0.34299999999999997, "n_success": 67, "policy_version": 108, "rl_env_steps": 862955, "round": 108, "seed_start": 960000, "sr": 0.335, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 860000, "update_index": 324000}
45
+ {"chunk_transitions": 111000, "episodes": 200, "eval_index": 44, "gain": -0.004500000000000004, "gradient_updates": 333000, "ma": 0.32949999999999996, "n_success": 55, "policy_version": 111, "rl_env_steps": 886908, "round": 111, "seed_start": 960000, "sr": 0.275, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 880000, "update_index": 333000}
46
+ {"chunk_transitions": 113000, "episodes": 200, "eval_index": 45, "gain": 0.005500000000000005, "gradient_updates": 339000, "ma": 0.339, "n_success": 78, "policy_version": 113, "rl_env_steps": 902888, "round": 113, "seed_start": 960000, "sr": 0.39, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 900000, "update_index": 339000}
47
+ {"chunk_transitions": 116000, "episodes": 200, "eval_index": 46, "gain": 0.00500000000000006, "gradient_updates": 348000, "ma": 0.343, "n_success": 76, "policy_version": 116, "rl_env_steps": 926832, "round": 116, "seed_start": 960000, "sr": 0.38, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 920000, "update_index": 348000}
48
+ {"chunk_transitions": 118000, "episodes": 200, "eval_index": 47, "gain": -0.009999999999999953, "gradient_updates": 354000, "ma": 0.33849999999999997, "n_success": 75, "policy_version": 118, "rl_env_steps": 942799, "round": 118, "seed_start": 960000, "sr": 0.375, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 940000, "update_index": 354000}
49
+ {"chunk_transitions": 121000, "episodes": 200, "eval_index": 48, "gain": 0.009000000000000008, "gradient_updates": 363000, "ma": 0.35, "n_success": 77, "policy_version": 121, "rl_env_steps": 966741, "round": 121, "seed_start": 960000, "sr": 0.385, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 960000, "update_index": 363000}
50
+ {"chunk_transitions": 123000, "episodes": 200, "eval_index": 49, "gain": 0.01750000000000007, "gradient_updates": 369000, "ma": 0.36300000000000004, "n_success": 86, "policy_version": 123, "rl_env_steps": 982720, "round": 123, "seed_start": 960000, "sr": 0.43, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 980000, "update_index": 369000}
51
+ {"chunk_transitions": 126000, "episodes": 200, "eval_index": 50, "gain": 0.013500000000000012, "gradient_updates": 378000, "ma": 0.3625, "n_success": 81, "policy_version": 126, "rl_env_steps": 1006671, "round": 126, "seed_start": 960000, "sr": 0.405, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1000000, "update_index": 378000}
52
+ {"chunk_transitions": 128000, "episodes": 200, "eval_index": 51, "gain": 0.03749999999999992, "gradient_updates": 384000, "ma": 0.37499999999999994, "n_success": 80, "policy_version": 128, "rl_env_steps": 1022649, "round": 128, "seed_start": 960000, "sr": 0.4, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1020000, "update_index": 384000}
53
+ {"chunk_transitions": 131000, "episodes": 200, "eval_index": 52, "gain": 0.038500000000000034, "gradient_updates": 393000, "ma": 0.3825, "n_success": 90, "policy_version": 131, "rl_env_steps": 1046596, "round": 131, "seed_start": 960000, "sr": 0.45, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1040000, "update_index": 393000}
54
+ {"chunk_transitions": 133000, "episodes": 200, "eval_index": 53, "gain": 0.04800000000000004, "gradient_updates": 399000, "ma": 0.391, "n_success": 84, "policy_version": 133, "rl_env_steps": 1062573, "round": 133, "seed_start": 960000, "sr": 0.42, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1060000, "update_index": 399000}
55
+ {"chunk_transitions": 136000, "episodes": 200, "eval_index": 54, "gain": 0.07500000000000001, "gradient_updates": 408000, "ma": 0.40449999999999997, "n_success": 82, "policy_version": 136, "rl_env_steps": 1086521, "round": 136, "seed_start": 960000, "sr": 0.41, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1080000, "update_index": 408000}
56
+ {"chunk_transitions": 138000, "episodes": 200, "eval_index": 55, "gain": 0.064, "gradient_updates": 414000, "ma": 0.403, "n_success": 75, "policy_version": 138, "rl_env_steps": 1102504, "round": 138, "seed_start": 960000, "sr": 0.375, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1100000, "update_index": 414000}
57
+ {"chunk_transitions": 141000, "episodes": 200, "eval_index": 56, "gain": 0.06250000000000006, "gradient_updates": 423000, "ma": 0.4055000000000001, "n_success": 81, "policy_version": 141, "rl_env_steps": 1126426, "round": 141, "seed_start": 960000, "sr": 0.405, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1120000, "update_index": 423000}
58
+ {"chunk_transitions": 143000, "episodes": 200, "eval_index": 57, "gain": 0.0675, "gradient_updates": 429000, "ma": 0.40599999999999997, "n_success": 76, "policy_version": 143, "rl_env_steps": 1142380, "round": 143, "seed_start": 960000, "sr": 0.38, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1140000, "update_index": 429000}
59
+ {"chunk_transitions": 146000, "episodes": 200, "eval_index": 58, "gain": 0.05800000000000005, "gradient_updates": 438000, "ma": 0.40800000000000003, "n_success": 81, "policy_version": 146, "rl_env_steps": 1166316, "round": 146, "seed_start": 960000, "sr": 0.405, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1160000, "update_index": 438000}
60
+ {"chunk_transitions": 148000, "episodes": 200, "eval_index": 59, "gain": 0.03899999999999998, "gradient_updates": 444000, "ma": 0.402, "n_success": 74, "policy_version": 148, "rl_env_steps": 1182282, "round": 148, "seed_start": 960000, "sr": 0.37, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1180000, "update_index": 444000}
61
+ {"chunk_transitions": 151000, "episodes": 200, "eval_index": 60, "gain": 0.040000000000000036, "gradient_updates": 453000, "ma": 0.4025, "n_success": 82, "policy_version": 151, "rl_env_steps": 1206253, "round": 151, "seed_start": 960000, "sr": 0.41, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1200000, "update_index": 453000}
62
+ {"chunk_transitions": 153000, "episodes": 200, "eval_index": 61, "gain": 0.02750000000000008, "gradient_updates": 459000, "ma": 0.4025, "n_success": 80, "policy_version": 153, "rl_env_steps": 1222217, "round": 153, "seed_start": 960000, "sr": 0.4, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1220000, "update_index": 459000}
63
+ {"chunk_transitions": 156000, "episodes": 200, "eval_index": 62, "gain": 0.01749999999999996, "gradient_updates": 468000, "ma": 0.39999999999999997, "n_success": 85, "policy_version": 156, "rl_env_steps": 1246178, "round": 156, "seed_start": 960000, "sr": 0.425, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1240000, "update_index": 468000}
64
+ {"chunk_transitions": 158000, "episodes": 200, "eval_index": 63, "gain": 0.00599999999999995, "gradient_updates": 474000, "ma": 0.39699999999999996, "n_success": 78, "policy_version": 158, "rl_env_steps": 1262135, "round": 158, "seed_start": 960000, "sr": 0.39, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1260000, "update_index": 474000}
65
+ {"chunk_transitions": 161000, "episodes": 200, "eval_index": 64, "gain": -0.009499999999999953, "gradient_updates": 483000, "ma": 0.395, "n_success": 78, "policy_version": 161, "rl_env_steps": 1286102, "round": 161, "seed_start": 960000, "sr": 0.39, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1280000, "update_index": 483000}
66
+ {"chunk_transitions": 163000, "episodes": 200, "eval_index": 65, "gain": -0.0025000000000000577, "gradient_updates": 489000, "ma": 0.40049999999999997, "n_success": 86, "policy_version": 163, "rl_env_steps": 1302051, "round": 163, "seed_start": 960000, "sr": 0.43, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1300000, "update_index": 489000}
67
+ {"chunk_transitions": 166000, "episodes": 200, "eval_index": 66, "gain": -0.009500000000000064, "gradient_updates": 498000, "ma": 0.396, "n_success": 72, "policy_version": 166, "rl_env_steps": 1326008, "round": 166, "seed_start": 960000, "sr": 0.36, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1320000, "update_index": 498000}
68
+ {"chunk_transitions": 168000, "episodes": 200, "eval_index": 67, "gain": -0.008000000000000007, "gradient_updates": 504000, "ma": 0.39799999999999996, "n_success": 80, "policy_version": 168, "rl_env_steps": 1341972, "round": 168, "seed_start": 960000, "sr": 0.4, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1340000, "update_index": 504000}
69
+ {"chunk_transitions": 171000, "episodes": 200, "eval_index": 68, "gain": -0.014500000000000068, "gradient_updates": 513000, "ma": 0.39349999999999996, "n_success": 72, "policy_version": 171, "rl_env_steps": 1365913, "round": 171, "seed_start": 960000, "sr": 0.36, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1360000, "update_index": 513000}
70
+ {"chunk_transitions": 173000, "episodes": 200, "eval_index": 69, "gain": -0.006500000000000006, "gradient_updates": 519000, "ma": 0.3955, "n_success": 78, "policy_version": 173, "rl_env_steps": 1381861, "round": 173, "seed_start": 960000, "sr": 0.39, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1380000, "update_index": 519000}
71
+ {"chunk_transitions": 176000, "episodes": 200, "eval_index": 70, "gain": -0.008000000000000007, "gradient_updates": 528000, "ma": 0.3945, "n_success": 80, "policy_version": 176, "rl_env_steps": 1405828, "round": 176, "seed_start": 960000, "sr": 0.4, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1400000, "update_index": 528000}
72
+ {"chunk_transitions": 178000, "episodes": 200, "eval_index": 71, "gain": -0.0035000000000000586, "gradient_updates": 534000, "ma": 0.39899999999999997, "n_success": 89, "policy_version": 178, "rl_env_steps": 1421810, "round": 178, "seed_start": 960000, "sr": 0.445, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1420000, "update_index": 534000}
73
+ {"chunk_transitions": 181000, "episodes": 200, "eval_index": 72, "gain": -0.00599999999999995, "gradient_updates": 543000, "ma": 0.394, "n_success": 75, "policy_version": 181, "rl_env_steps": 1445776, "round": 181, "seed_start": 960000, "sr": 0.375, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1440000, "update_index": 543000}
74
+ {"chunk_transitions": 183000, "episodes": 200, "eval_index": 73, "gain": -0.0025000000000000022, "gradient_updates": 549000, "ma": 0.39449999999999996, "n_success": 79, "policy_version": 183, "rl_env_steps": 1461754, "round": 183, "seed_start": 960000, "sr": 0.395, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1460000, "update_index": 549000}
75
+ {"chunk_transitions": 186000, "episodes": 200, "eval_index": 74, "gain": -0.0010000000000000009, "gradient_updates": 558000, "ma": 0.394, "n_success": 77, "policy_version": 186, "rl_env_steps": 1485706, "round": 186, "seed_start": 960000, "sr": 0.385, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1480000, "update_index": 558000}
76
+ {"chunk_transitions": 188000, "episodes": 200, "eval_index": 75, "gain": -0.005500000000000005, "gradient_updates": 564000, "ma": 0.39499999999999996, "n_success": 88, "policy_version": 188, "rl_env_steps": 1501671, "round": 188, "seed_start": 960000, "sr": 0.44, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1500000, "update_index": 564000}
77
+ {"chunk_transitions": 191000, "episodes": 200, "eval_index": 76, "gain": 0.007000000000000006, "gradient_updates": 573000, "ma": 0.403, "n_success": 88, "policy_version": 191, "rl_env_steps": 1525608, "round": 191, "seed_start": 960000, "sr": 0.44, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1520000, "update_index": 573000}
78
+ {"chunk_transitions": 193000, "episodes": 200, "eval_index": 77, "gain": 0.0005000000000000004, "gradient_updates": 579000, "ma": 0.39849999999999997, "n_success": 71, "policy_version": 193, "rl_env_steps": 1541586, "round": 193, "seed_start": 960000, "sr": 0.355, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1540000, "update_index": 579000}
79
+ {"chunk_transitions": 196000, "episodes": 200, "eval_index": 78, "gain": 0.009500000000000064, "gradient_updates": 588000, "ma": 0.403, "n_success": 81, "policy_version": 196, "rl_env_steps": 1565558, "round": 196, "seed_start": 960000, "sr": 0.405, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1560000, "update_index": 588000}
80
+ {"chunk_transitions": 198000, "episodes": 200, "eval_index": 79, "gain": 0.007499999999999896, "gradient_updates": 594000, "ma": 0.4029999999999999, "n_success": 78, "policy_version": 198, "rl_env_steps": 1581544, "round": 198, "seed_start": 960000, "sr": 0.39, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1580000, "update_index": 594000}
81
+ {"chunk_transitions": 201000, "episodes": 200, "eval_index": 80, "gain": 0.006500000000000061, "gradient_updates": 603000, "ma": 0.4010000000000001, "n_success": 76, "policy_version": 201, "rl_env_steps": 1605499, "round": 201, "seed_start": 960000, "sr": 0.38, "success_requested": 0, "success_requested_imitation": 0, "success_rounds_with_skips": 0, "success_skipped": 0, "success_skipped_imitation": 0, "success_skipped_td": 0, "target_step": 1600000, "update_index": 603000}
RL/ToolHang100/DSRL-NA/diverse100/seed_1/eval/final.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a7afc8dbf75ac8d5eb1f5e8407e4bb769666f98a7b70f9ef1d701eb13b1058e
3
+ size 6886
RL/ToolHang100/DSRL-NA/diverse100/seed_1/final/source_metadata.json ADDED
@@ -0,0 +1,314 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "checkpoint_sha256": "88e0762bf71060889e6ad409345b3357e471f684857c3b54114d0fe61c982582",
3
+ "evaluation": {
4
+ "episodes": 300,
5
+ "imputed": false,
6
+ "n_success": 107,
7
+ "observed": true,
8
+ "seed_start": 800000,
9
+ "sr": 0.3566666666666667,
10
+ "successes": [
11
+ 0,
12
+ 0,
13
+ 0,
14
+ 1,
15
+ 1,
16
+ 1,
17
+ 0,
18
+ 0,
19
+ 1,
20
+ 1,
21
+ 1,
22
+ 0,
23
+ 1,
24
+ 0,
25
+ 1,
26
+ 0,
27
+ 0,
28
+ 1,
29
+ 0,
30
+ 0,
31
+ 0,
32
+ 0,
33
+ 1,
34
+ 0,
35
+ 0,
36
+ 0,
37
+ 1,
38
+ 0,
39
+ 0,
40
+ 1,
41
+ 0,
42
+ 0,
43
+ 0,
44
+ 1,
45
+ 0,
46
+ 0,
47
+ 1,
48
+ 0,
49
+ 1,
50
+ 0,
51
+ 1,
52
+ 0,
53
+ 1,
54
+ 1,
55
+ 0,
56
+ 1,
57
+ 0,
58
+ 0,
59
+ 0,
60
+ 1,
61
+ 0,
62
+ 0,
63
+ 0,
64
+ 0,
65
+ 0,
66
+ 1,
67
+ 0,
68
+ 0,
69
+ 0,
70
+ 1,
71
+ 1,
72
+ 0,
73
+ 1,
74
+ 0,
75
+ 1,
76
+ 0,
77
+ 0,
78
+ 1,
79
+ 1,
80
+ 0,
81
+ 1,
82
+ 1,
83
+ 1,
84
+ 1,
85
+ 1,
86
+ 0,
87
+ 0,
88
+ 0,
89
+ 0,
90
+ 1,
91
+ 0,
92
+ 0,
93
+ 0,
94
+ 0,
95
+ 0,
96
+ 0,
97
+ 0,
98
+ 0,
99
+ 0,
100
+ 0,
101
+ 0,
102
+ 0,
103
+ 0,
104
+ 1,
105
+ 0,
106
+ 0,
107
+ 0,
108
+ 0,
109
+ 0,
110
+ 0,
111
+ 1,
112
+ 1,
113
+ 0,
114
+ 0,
115
+ 0,
116
+ 0,
117
+ 0,
118
+ 0,
119
+ 1,
120
+ 0,
121
+ 0,
122
+ 0,
123
+ 0,
124
+ 0,
125
+ 0,
126
+ 1,
127
+ 1,
128
+ 0,
129
+ 0,
130
+ 0,
131
+ 0,
132
+ 0,
133
+ 0,
134
+ 1,
135
+ 0,
136
+ 1,
137
+ 0,
138
+ 0,
139
+ 0,
140
+ 1,
141
+ 0,
142
+ 1,
143
+ 1,
144
+ 0,
145
+ 0,
146
+ 1,
147
+ 1,
148
+ 0,
149
+ 0,
150
+ 0,
151
+ 0,
152
+ 1,
153
+ 0,
154
+ 0,
155
+ 1,
156
+ 0,
157
+ 0,
158
+ 0,
159
+ 0,
160
+ 0,
161
+ 0,
162
+ 0,
163
+ 1,
164
+ 1,
165
+ 1,
166
+ 0,
167
+ 0,
168
+ 0,
169
+ 0,
170
+ 0,
171
+ 0,
172
+ 1,
173
+ 0,
174
+ 1,
175
+ 0,
176
+ 0,
177
+ 0,
178
+ 1,
179
+ 1,
180
+ 0,
181
+ 0,
182
+ 0,
183
+ 0,
184
+ 1,
185
+ 0,
186
+ 0,
187
+ 1,
188
+ 0,
189
+ 1,
190
+ 0,
191
+ 1,
192
+ 1,
193
+ 0,
194
+ 0,
195
+ 0,
196
+ 0,
197
+ 0,
198
+ 0,
199
+ 0,
200
+ 0,
201
+ 1,
202
+ 0,
203
+ 1,
204
+ 0,
205
+ 0,
206
+ 0,
207
+ 0,
208
+ 0,
209
+ 0,
210
+ 1,
211
+ 0,
212
+ 0,
213
+ 1,
214
+ 1,
215
+ 1,
216
+ 0,
217
+ 1,
218
+ 1,
219
+ 1,
220
+ 1,
221
+ 0,
222
+ 0,
223
+ 1,
224
+ 1,
225
+ 0,
226
+ 0,
227
+ 1,
228
+ 1,
229
+ 0,
230
+ 1,
231
+ 0,
232
+ 0,
233
+ 0,
234
+ 0,
235
+ 1,
236
+ 1,
237
+ 0,
238
+ 1,
239
+ 0,
240
+ 0,
241
+ 0,
242
+ 0,
243
+ 0,
244
+ 0,
245
+ 0,
246
+ 0,
247
+ 0,
248
+ 1,
249
+ 0,
250
+ 1,
251
+ 0,
252
+ 1,
253
+ 1,
254
+ 1,
255
+ 0,
256
+ 0,
257
+ 0,
258
+ 0,
259
+ 1,
260
+ 1,
261
+ 0,
262
+ 1,
263
+ 1,
264
+ 1,
265
+ 1,
266
+ 1,
267
+ 1,
268
+ 0,
269
+ 0,
270
+ 1,
271
+ 0,
272
+ 0,
273
+ 1,
274
+ 0,
275
+ 0,
276
+ 1,
277
+ 0,
278
+ 0,
279
+ 0,
280
+ 0,
281
+ 1,
282
+ 1,
283
+ 1,
284
+ 1,
285
+ 0,
286
+ 1,
287
+ 0,
288
+ 1,
289
+ 0,
290
+ 0,
291
+ 0,
292
+ 1,
293
+ 0,
294
+ 1,
295
+ 0,
296
+ 0,
297
+ 0,
298
+ 0,
299
+ 1,
300
+ 0,
301
+ 0,
302
+ 0,
303
+ 0,
304
+ 0,
305
+ 1,
306
+ 0,
307
+ 0,
308
+ 1,
309
+ 1,
310
+ 1
311
+ ]
312
+ },
313
+ "status": "budget_complete"
314
+ }
RL/ToolHang100/DSRL-NA/diverse100/seed_1/final/source_resolved_config.json ADDED
@@ -0,0 +1,548 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "decision_ref": "docs/scope.md [0928-I], [0930-S2DEC]",
3
+ "env": {
4
+ "chunk_length": 8,
5
+ "n_envs": 4
6
+ },
7
+ "learner": {
8
+ "convention": "unified_env_step",
9
+ "gamma_env": 0.999
10
+ },
11
+ "overrides": [
12
+ {
13
+ "decision_ref": "[1020-S2TOOLHANG5] configs/study2_protocol.yaml",
14
+ "key": "n_step_chunks",
15
+ "reason": "task-specific unified Study2 setting",
16
+ "upstream": 3,
17
+ "value": 5
18
+ },
19
+ {
20
+ "decision_ref": "[1023-OFFLINE100] user 2026-10-23: 100-demo BC -> all 100 demos as offline D_demo",
21
+ "key": "offline.cohort_size",
22
+ "reason": "tool_hang regime diverse100 BC cohort has 100 demos; the offline demo buffer uses the checkpoint's entire BC cohort",
23
+ "upstream": 50,
24
+ "value": 100
25
+ },
26
+ {
27
+ "decision_ref": "[1023-OFFLINE100] user 2026-10-23: 100-demo BC -> all 100 demos as offline D_demo",
28
+ "key": "offline.episodes",
29
+ "reason": "tool_hang regime diverse100 BC cohort has 100 demos; the offline demo buffer uses the checkpoint's entire BC cohort",
30
+ "upstream": 50,
31
+ "value": 100
32
+ }
33
+ ],
34
+ "protocol": {
35
+ "checkpoint_every_rounds": 5,
36
+ "checkpoint_retention": {
37
+ "best": true,
38
+ "collapse_drop": {
39
+ "min_abs": 0.2,
40
+ "ratio": 0.5
41
+ },
42
+ "diagnostics": true,
43
+ "every_evals": 3
44
+ },
45
+ "collection": {
46
+ "n_envs": 4,
47
+ "train_seed_base": 100000000,
48
+ "train_seed_stride": 1000000,
49
+ "training_freq_chunks": 1000
50
+ },
51
+ "convergence": {
52
+ "early_stop": false,
53
+ "gain_lag": 10,
54
+ "high_sr": 0.95,
55
+ "high_sr_evals": 5,
56
+ "ma_window": 10,
57
+ "max_env_steps": 1606196,
58
+ "min_env_steps": 200000,
59
+ "min_gain": 0.02,
60
+ "zero_success_evals": 8
61
+ },
62
+ "critic_ensemble": {
63
+ "num": 10,
64
+ "policy": "mean",
65
+ "target": {
66
+ "kind": "subset_min",
67
+ "replace": false,
68
+ "size": 2
69
+ }
70
+ },
71
+ "eval": {
72
+ "best_of_n": 16,
73
+ "episodes": 200,
74
+ "every_env_steps": 20000,
75
+ "include_step0": true,
76
+ "seed_start": 960000,
77
+ "workers": 4
78
+ },
79
+ "final_eval": {
80
+ "episodes": 300,
81
+ "on_zero_stop": true,
82
+ "seed_start": 800000
83
+ },
84
+ "n_step_chunks": 5,
85
+ "offline": {
86
+ "adaptive_ratio": false,
87
+ "cohort_size": 100,
88
+ "episodes": 100,
89
+ "learners_without_offline": [
90
+ "dsrl_sac"
91
+ ],
92
+ "ratio": {
93
+ "decay_env_steps": {
94
+ "can": 160000,
95
+ "square": 320000,
96
+ "tool_hang": 640000,
97
+ "transport": 320000
98
+ },
99
+ "end": 0.1,
100
+ "start": {
101
+ "can": 0.7,
102
+ "square": 0.5,
103
+ "tool_hang": 0.5,
104
+ "transport": 0.9
105
+ }
106
+ },
107
+ "selection": "ascending_demo_index",
108
+ "source": "bc_regime",
109
+ "unsupported_learners": [
110
+ "zprl"
111
+ ]
112
+ },
113
+ "replay": {
114
+ "capacity_chunks": 1000000,
115
+ "pin_start_successes": false,
116
+ "success_buffer": "success_mask_on_shared_fifo",
117
+ "success_skip_rule": "skip_below_success_batch_size",
118
+ "unit": "executed_chunk"
119
+ },
120
+ "start": {
121
+ "max_episodes": 200,
122
+ "mode": "offline_buffer",
123
+ "policy": {
124
+ "default": "base_cfm_ode",
125
+ "zprl": "vib_base"
126
+ },
127
+ "target_successes": 20
128
+ },
129
+ "unified": {
130
+ "actor_target_tau": 0.005,
131
+ "critic_hidden": [
132
+ 512,
133
+ 512,
134
+ 512
135
+ ],
136
+ "critic_target_every_critic_step": true,
137
+ "critic_target_tau": 0.005,
138
+ "dsrl_noise_imitation_reference_dims": 28,
139
+ "gamma_env": 0.999,
140
+ "lr": {
141
+ "actor": 0.0001,
142
+ "critic": 0.0003,
143
+ "decay_end_env_steps": 1606196,
144
+ "min_ratio": 0.1,
145
+ "schedule": "warmup_cosine",
146
+ "temperature": 0.0003,
147
+ "warmup_env_steps": 20000
148
+ },
149
+ "reward": "binary_first_success",
150
+ "success_td_polyak": true,
151
+ "timeout_bootstrap": false,
152
+ "zprl_imitation": "mse_pretanh_mean"
153
+ },
154
+ "updates": {
155
+ "batch_size": 256,
156
+ "critic_per_actor": 2,
157
+ "success_batch_size": 256,
158
+ "utd_per_chunk": 3
159
+ },
160
+ "version": "study2_protocol_v10"
161
+ },
162
+ "protocol_hash": "903b284201350b7c879df61bb6860ec2d7fc9e4537fb5abd2aba8bcf7a60ee0f",
163
+ "provenance_extra": {
164
+ "agent_library": {
165
+ "framework": "jax",
166
+ "sources": {
167
+ "__init__.py": "0356b394bbae9d0738649e4864a034c4ee5d6f329c8a7f862717b3428e52313c",
168
+ "_vendor/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
169
+ "_vendor/ogpo/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
170
+ "_vendor/ogpo/agents/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
171
+ "_vendor/ogpo/agents/modules/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
172
+ "_vendor/ogpo/agents/modules/bc_helper.py": "8913392cc0434057323b2f5b28aaeb36d2c31e6b129ef09d21352353fe276c14",
173
+ "_vendor/ogpo/agents/modules/dist_helper.py": "c0f429c6c70082f5f5258ffffdf4bf0e6b2e1587c26528d3ee937a4e302d9951",
174
+ "_vendor/ogpo/agents/modules/flax_utils.py": "36f27885a086a4ae561171de4650da86330dcd68f17694e3f71278677c614037",
175
+ "_vendor/ogpo/agents/modules/igp_targets_helper.py": "b0343a236a856ec734cc73a1e73186d7d6520dd9efc34b3597327d21b55c9522",
176
+ "_vendor/ogpo/agents/modules/maintenance.py": "7b40d670c277783e118a0669f57aa10370c76b7b8fe3c1d27d16af1a3e9f0865",
177
+ "_vendor/ogpo/agents/modules/pg_helper.py": "cb2e0e83e0c95af408c9bac86fc9ccf1d7e427b3edb315c5d042de5d0a9328d4",
178
+ "_vendor/ogpo/agents/modules/q_helper.py": "71d874b74af5d724d82bd9c4a0a6027d8795e5dc26755ae7838abc3cda54718e",
179
+ "_vendor/ogpo/agents/ogpo.py": "3393282fd4a972c97951715a5776a506e37102cbe5e9c21e5e1ed1e6288696b2",
180
+ "_vendor/ogpo/agents/qc.py": "55906d8b598ce0a2607782e886697c3766dd281a0664ec05bf96173fc8c22c56",
181
+ "_vendor/ogpo/networks/__init__.py": "089130459e91d2b43f5bca45e2659507379fe62351ce6d2d91a49a0387ed2e90",
182
+ "_vendor/ogpo/networks/actors.py": "c0ad2b12439f5da82128e35907314741c22ccc70692de33fed90e67a32f8d86a",
183
+ "_vendor/ogpo/networks/actors_fql.py": "0d9f8646f1230b442bc80457252656c007431ca62458afa5fdecd15150ed9c4d",
184
+ "_vendor/ogpo/networks/critics.py": "77e40f7ea9d6837f8f89d889bfcd7a70e0ef20076f65720d79a9bef3db5130b4",
185
+ "_vendor/ogpo/networks/encoders.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
186
+ "_vendor/ogpo/networks/encoders/__init__.py": "41ae7973f560af5539cff23643843a0c56c253722bb17d7372a9ade029abb670",
187
+ "_vendor/ogpo/networks/encoders/impala.py": "62b7f1dab0aefe6662d3fa0fc59417d125cb9687af71d7875d533e24ac3e89d2",
188
+ "_vendor/ogpo/networks/encoders/minvit.py": "93f0af908450588cdbc9fc18929247b55447bb7972833388244707628328ed33",
189
+ "_vendor/ogpo/networks/encoders/paligemma.py": "0f2a448fd52a2c161f921e7fd7d79be462f98db34c09bf68877f8ed67de9e82c",
190
+ "_vendor/ogpo/networks/encoders/vision_proprio.py": "01ade56e774aeb29c3e7cd1d59eff5e43b7827d102769a450f8d7cc5a17666c8",
191
+ "_vendor/ogpo/networks/modules/__init__.py": "39a4fe9455c1b58c505fa1d9c0a90179dd64cdd459e8341d9ef7fcf6b7af9aae",
192
+ "_vendor/ogpo/networks/modules/attention.py": "a835876c6269ba0c5c63f49005027e5524a750f0c6e7af8697093b9bb8ff4c55",
193
+ "_vendor/ogpo/networks/modules/distributions.py": "5430f1bc6ee7fa71aac65664c9c3f13500d1fec3c5b2b1d80fc6590cca8ac71d",
194
+ "_vendor/ogpo/networks/modules/misc.py": "527793d4e64be2bbc8c22e73a574d5cb604e4672eabd0db9843853107502f620",
195
+ "_vendor/ogpo/networks/modules/mlp.py": "e2cf4d9923d95df8747c1554cf94f20000e108b1a9b0d9b0729a759b597a1316",
196
+ "_vendor/ogpo/networks/modules/time_embedding.py": "54d1a400b1b179a75c5aa55501031aa5d6513da12715ab75b89eaacdbe9f76e1",
197
+ "_vendor/ogpo/networks/modules/two_tier.py": "aa5182a5932cd500f95f4fddcd16429d66c968988639ddf3d6b33a2297953166",
198
+ "dice.py": "6989a71a178af904061e418a94c06615acb7576d9d32548566cc0032e273f22b",
199
+ "dsrl_na.py": "973970bfb02b87ad995f04572bc4d8443cc1e92d6447fbcbcfb80d8fa97cd9ce",
200
+ "dsrl_sac.py": "292d55b2d6f314ec12d17c77c2da2badef6a3fbdde89ba953044e074da6fbc6c",
201
+ "factory.py": "0117708b7cdab69cfaa83bba70f1101aff7b33866117946c377a5cb2fcc6ea49",
202
+ "flow.py": "76dce89f145b4194885524731523eada50e86f8d06b86f7cfaffd0fb3e3b779a",
203
+ "native_config.py": "e9dbf33c0ec15a654c2dde7f504d4e2492c29cbdf84c4eedf42cda94c158e4de",
204
+ "native_flow.py": "e9a9dc59e4e2ccaa7355bebb3c98cc9c1e0080c493889d58bfefdc7bc341c763",
205
+ "native_replay.py": "4f5922a41c5f4e93c599794143c0a26d8b22cb6c38d675ede9538a7bd2b30402",
206
+ "native_residual.py": "00936a650eed103e380f201c9917a9bc414d7b48ba244d0018ebd7b032b3af67",
207
+ "native_train.py": "83bd25879b1096afcb8c1e7a4c34ab77c652dafde10d86f7fdea221c98f6d4a4",
208
+ "networks.py": "81460ac49a7b67ef66b0efd04e26e86c320456c190493143ea77b57c8fbc195d",
209
+ "ogpo.py": "3501a14eb1e1dd0e7194de541dc1af2dd0b4584f102b7e23b0b625d490c79093",
210
+ "optim.py": "41f5c0e4907e822449154959067a84f09820010bc02c6a01a74b5d5478a76192",
211
+ "qc.py": "38ef1b0ed7a4d0833ee50a182262a3bd668931d9499483fa58f7cad43b3f9274",
212
+ "registry.py": "0e7928dac526397186290fe1a914403daa69c39c2c9da9aa395bacfbca9dc617",
213
+ "residual.py": "2dfda94aee07537924912d14693b7d0a41a76c9aea02852a006e5f4877c082ee",
214
+ "telemetry.py": "e5716864ea9b2126b50b86fec78d56ecfd96f17832d5006f83cb2763e6092feb",
215
+ "train.py": "af62bcfe092ab39e79eae9bc1f297430865cd31f61212faf17863e2b6de64ad8",
216
+ "vib.py": "ab3a6721c7c051d21e7955940942bc0f6d064635dd9c73b7275589c150f19f5e",
217
+ "vib_workflow.py": "35b264975fbbcb7f1142af4284f245c9de043e3dab6689ca40657e7be47090f2",
218
+ "zprl.py": "fefe27f991d1a482a57b04738d5b22b9879f0ea405e1935e94dd58ba72c30609"
219
+ },
220
+ "version": "study2_jax_agents_v2"
221
+ },
222
+ "base_framework": "jax",
223
+ "bc_identity": {
224
+ "architecture": "dice_vision_resnet_flow_unet_v1",
225
+ "family": "StandardBaseBC",
226
+ "framework": "jax"
227
+ },
228
+ "checkpoint": "/ssd/WorkSpace/RL-Postrain/runs/operations/20261024_toolhang_diverse100_v10_three/base/diverse100/epoch_300.pkl",
229
+ "checkpoint_sha256": "c5381a3e2ff4f4eeeb899a9d14286d9d42e0816c6431f1208fa39c60309c48f8",
230
+ "experiment": "tool_hang",
231
+ "frozen_hash": "719a7185a0d6a4877c76a5e1d9b1adb163a3d863d28bced54e79d534e8db6873",
232
+ "learner": {
233
+ "action_dim": 7,
234
+ "action_magnitude": null,
235
+ "actor_hidden": [
236
+ 2048,
237
+ 2048,
238
+ 2048
239
+ ],
240
+ "base_hash": "719a7185a0d6a4877c76a5e1d9b1adb163a3d863d28bced54e79d534e8db6873",
241
+ "critic_hidden": [
242
+ 512,
243
+ 512,
244
+ 512
245
+ ],
246
+ "ensemble": 10,
247
+ "feature_dim": 137,
248
+ "horizon": 8,
249
+ "normalizer_hash": "89c504c576b47f752f04402b67f3618b7b4087dd5073f15c5c79cc5a710b7f30",
250
+ "protocol": {
251
+ "checkpoint_every_rounds": 5,
252
+ "checkpoint_retention": {
253
+ "best": true,
254
+ "collapse_drop": {
255
+ "min_abs": 0.2,
256
+ "ratio": 0.5
257
+ },
258
+ "diagnostics": true,
259
+ "every_evals": 3
260
+ },
261
+ "collection": {
262
+ "n_envs": 4,
263
+ "train_seed_base": 100000000,
264
+ "train_seed_stride": 1000000,
265
+ "training_freq_chunks": 1000
266
+ },
267
+ "convergence": {
268
+ "early_stop": false,
269
+ "gain_lag": 10,
270
+ "high_sr": 0.95,
271
+ "high_sr_evals": 5,
272
+ "ma_window": 10,
273
+ "max_env_steps": 1606196,
274
+ "min_env_steps": 200000,
275
+ "min_gain": 0.02,
276
+ "zero_success_evals": 8
277
+ },
278
+ "critic_ensemble": {
279
+ "num": 10,
280
+ "policy": "mean",
281
+ "target": {
282
+ "kind": "subset_min",
283
+ "replace": false,
284
+ "size": 2
285
+ }
286
+ },
287
+ "eval": {
288
+ "best_of_n": 16,
289
+ "episodes": 200,
290
+ "every_env_steps": 20000,
291
+ "include_step0": true,
292
+ "seed_start": 960000,
293
+ "workers": 4
294
+ },
295
+ "final_eval": {
296
+ "episodes": 300,
297
+ "on_zero_stop": true,
298
+ "seed_start": 800000
299
+ },
300
+ "n_step_chunks": 5,
301
+ "offline": {
302
+ "adaptive_ratio": false,
303
+ "cohort_size": 100,
304
+ "episodes": 100,
305
+ "learners_without_offline": [
306
+ "dsrl_sac"
307
+ ],
308
+ "ratio": {
309
+ "decay_env_steps": {
310
+ "can": 160000,
311
+ "square": 320000,
312
+ "tool_hang": 640000,
313
+ "transport": 320000
314
+ },
315
+ "end": 0.1,
316
+ "start": {
317
+ "can": 0.7,
318
+ "square": 0.5,
319
+ "tool_hang": 0.5,
320
+ "transport": 0.9
321
+ }
322
+ },
323
+ "selection": "ascending_demo_index",
324
+ "source": "bc_regime",
325
+ "unsupported_learners": [
326
+ "zprl"
327
+ ]
328
+ },
329
+ "replay": {
330
+ "capacity_chunks": 1000000,
331
+ "pin_start_successes": false,
332
+ "success_buffer": "success_mask_on_shared_fifo",
333
+ "success_skip_rule": "skip_below_success_batch_size",
334
+ "unit": "executed_chunk"
335
+ },
336
+ "start": {
337
+ "max_episodes": 200,
338
+ "mode": "offline_buffer",
339
+ "policy": {
340
+ "default": "base_cfm_ode",
341
+ "zprl": "vib_base"
342
+ },
343
+ "target_successes": 20
344
+ },
345
+ "unified": {
346
+ "actor_target_tau": 0.005,
347
+ "critic_hidden": [
348
+ 512,
349
+ 512,
350
+ 512
351
+ ],
352
+ "critic_target_every_critic_step": true,
353
+ "critic_target_tau": 0.005,
354
+ "dsrl_noise_imitation_reference_dims": 28,
355
+ "gamma_env": 0.999,
356
+ "lr": {
357
+ "actor": 0.0001,
358
+ "critic": 0.0003,
359
+ "decay_end_env_steps": 1606196,
360
+ "min_ratio": 0.1,
361
+ "schedule": "warmup_cosine",
362
+ "temperature": 0.0003,
363
+ "warmup_env_steps": 20000
364
+ },
365
+ "reward": "binary_first_success",
366
+ "success_td_polyak": true,
367
+ "timeout_bootstrap": false,
368
+ "zprl_imitation": "mse_pretanh_mean"
369
+ },
370
+ "updates": {
371
+ "batch_size": 256,
372
+ "critic_per_actor": 2,
373
+ "success_batch_size": 256,
374
+ "utd_per_chunk": 3
375
+ },
376
+ "version": "study2_protocol_v10"
377
+ },
378
+ "subset": 2,
379
+ "variant": {
380
+ "flags": {
381
+ "noise_loglik_weight": 0.0,
382
+ "success_td": false
383
+ },
384
+ "learner": "dsrl_na",
385
+ "name": "dsrl_na",
386
+ "phase": 1,
387
+ "spec_hash": "288e84fd5eb8196404b74bd024f80ec14a2de9bac56dae6239fa3d474968c979",
388
+ "version": "study2_variants_v4"
389
+ },
390
+ "version": "study2_jax_agents_v2",
391
+ "vib_hash": null,
392
+ "weight_decay": null
393
+ },
394
+ "observation_mode": "pixel_state",
395
+ "offline_buffer": {
396
+ "actions_sha256": "7d65f5b7df78085d4780cf6338ac032e63f7f3d6ff9b1923bcc0992c840ee2d9",
397
+ "aux_sha256": "10f0ea12b0d1e89a65e72988f979e40fe4ffc48cec8c05e3651506b6a1d34088",
398
+ "checkpoint_sha256": "c5381a3e2ff4f4eeeb899a9d14286d9d42e0816c6431f1208fa39c60309c48f8",
399
+ "chunks": 6171,
400
+ "dataset_sha256": "c5cb01b119a109a3d4842ca515906e86f1f35a8ee1a333d22b3503c1a513a8f4",
401
+ "decision_ref": "[1017-S2OFFBC]",
402
+ "env_steps": 48984,
403
+ "features_sha256": "ed0a570dcabbbba769a4ac377436caea901a9b0238e5c9890a6db1e473a723c2",
404
+ "horizon": 8,
405
+ "identity_hash": "07e7e6bd263c84a9b8ceb56b1e6e92eea9ced27f6754960f1a8b5007f4f18779",
406
+ "learner_coordinates": "ResidualLearner:dsrl_na",
407
+ "pixel_spec_hash": "7aba3118e0304ea7265915550734918674f2b9a1e6c9bc2eb1cd99b568045a11",
408
+ "ratio": {
409
+ "adaptive_ratio": false,
410
+ "rule": "D_demo kept for the whole run (never evicted), online RL data FIFO; each ordinary batch = batch_size uniform draws over the eligible D_demo + online windows, offline rows first"
411
+ },
412
+ "selection": {
413
+ "cohort_size": 100,
414
+ "demos": [
415
+ "demo_0",
416
+ "demo_2",
417
+ "demo_5",
418
+ "demo_7",
419
+ "demo_9",
420
+ "demo_10",
421
+ "demo_13",
422
+ "demo_16",
423
+ "demo_17",
424
+ "demo_18",
425
+ "demo_20",
426
+ "demo_23",
427
+ "demo_27",
428
+ "demo_28",
429
+ "demo_29",
430
+ "demo_30",
431
+ "demo_31",
432
+ "demo_32",
433
+ "demo_34",
434
+ "demo_35",
435
+ "demo_37",
436
+ "demo_40",
437
+ "demo_41",
438
+ "demo_43",
439
+ "demo_46",
440
+ "demo_47",
441
+ "demo_50",
442
+ "demo_51",
443
+ "demo_56",
444
+ "demo_57",
445
+ "demo_58",
446
+ "demo_67",
447
+ "demo_71",
448
+ "demo_74",
449
+ "demo_76",
450
+ "demo_78",
451
+ "demo_80",
452
+ "demo_82",
453
+ "demo_83",
454
+ "demo_85",
455
+ "demo_86",
456
+ "demo_87",
457
+ "demo_93",
458
+ "demo_97",
459
+ "demo_98",
460
+ "demo_101",
461
+ "demo_102",
462
+ "demo_107",
463
+ "demo_110",
464
+ "demo_111",
465
+ "demo_112",
466
+ "demo_113",
467
+ "demo_115",
468
+ "demo_116",
469
+ "demo_119",
470
+ "demo_122",
471
+ "demo_123",
472
+ "demo_126",
473
+ "demo_127",
474
+ "demo_128",
475
+ "demo_130",
476
+ "demo_131",
477
+ "demo_132",
478
+ "demo_133",
479
+ "demo_134",
480
+ "demo_135",
481
+ "demo_138",
482
+ "demo_140",
483
+ "demo_141",
484
+ "demo_145",
485
+ "demo_147",
486
+ "demo_149",
487
+ "demo_151",
488
+ "demo_152",
489
+ "demo_154",
490
+ "demo_156",
491
+ "demo_157",
492
+ "demo_159",
493
+ "demo_161",
494
+ "demo_163",
495
+ "demo_164",
496
+ "demo_171",
497
+ "demo_172",
498
+ "demo_173",
499
+ "demo_174",
500
+ "demo_175",
501
+ "demo_176",
502
+ "demo_177",
503
+ "demo_179",
504
+ "demo_181",
505
+ "demo_182",
506
+ "demo_183",
507
+ "demo_186",
508
+ "demo_188",
509
+ "demo_190",
510
+ "demo_191",
511
+ "demo_192",
512
+ "demo_193",
513
+ "demo_196",
514
+ "demo_199"
515
+ ],
516
+ "manifest": "/ssd/WorkSpace/RL-Postrain/checkpoints/huggingface/yangfy0627/RL-postTrain/e62671ed1cf646b453660b698c22ca4d273b5ccc/Robomimic/study1/ToolHang/diverse100/seed_1/data_manifest.json",
517
+ "manifest_sha256": "9233d5824205e29f1afa721d8573fe19805a1a8e7829c296af1dd0b74a1d29b4",
518
+ "regime": "diverse100",
519
+ "source": "bc_regime",
520
+ "source_ph_sha256": "c5cb01b119a109a3d4842ca515906e86f1f35a8ee1a333d22b3503c1a513a8f4",
521
+ "split": {
522
+ "training": 90,
523
+ "validation": 10
524
+ },
525
+ "task": "tool_hang"
526
+ }
527
+ },
528
+ "offline_role": "mixed",
529
+ "pixel_spec_hash": "7aba3118e0304ea7265915550734918674f2b9a1e6c9bc2eb1cd99b568045a11",
530
+ "smoke": false,
531
+ "task": "tool_hang",
532
+ "task_spec_hash": "ac2909345e9536690a9da63f06aad8dfe135e757b89e28706b2c4af641f195c2"
533
+ },
534
+ "seed": 1,
535
+ "start_policy": "base_cfm_ode",
536
+ "variant": {
537
+ "flags": {
538
+ "noise_loglik_weight": 0.0,
539
+ "success_td": false
540
+ },
541
+ "learner": "dsrl_na",
542
+ "name": "dsrl_na",
543
+ "phase": 1,
544
+ "spec_hash": "288e84fd5eb8196404b74bd024f80ec14a2de9bac56dae6239fa3d474968c979",
545
+ "version": "study2_variants_v4"
546
+ },
547
+ "variant_hash": "288e84fd5eb8196404b74bd024f80ec14a2de9bac56dae6239fa3d474968c979"
548
+ }
RL/ToolHang100/DSRL-NA/diverse100/seed_1/final/state.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:88e0762bf71060889e6ad409345b3357e471f684857c3b54114d0fe61c982582
3
+ size 282982148
RL/ToolHang100/DSRL-NA/diverse100/seed_1/frozen_base.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "after": "719a7185a0d6a4877c76a5e1d9b1adb163a3d863d28bced54e79d534e8db6873",
3
+ "before": "719a7185a0d6a4877c76a5e1d9b1adb163a3d863d28bced54e79d534e8db6873",
4
+ "checkpoint_unchanged": true,
5
+ "unchanged": true
6
+ }
RL/ToolHang100/DSRL-NA/diverse100/seed_1/offline_buffer.json ADDED
@@ -0,0 +1,847 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "identity": {
3
+ "actions_sha256": "7d65f5b7df78085d4780cf6338ac032e63f7f3d6ff9b1923bcc0992c840ee2d9",
4
+ "aux_sha256": "10f0ea12b0d1e89a65e72988f979e40fe4ffc48cec8c05e3651506b6a1d34088",
5
+ "checkpoint_sha256": "c5381a3e2ff4f4eeeb899a9d14286d9d42e0816c6431f1208fa39c60309c48f8",
6
+ "chunks": 6171,
7
+ "dataset_sha256": "c5cb01b119a109a3d4842ca515906e86f1f35a8ee1a333d22b3503c1a513a8f4",
8
+ "decision_ref": "[1017-S2OFFBC]",
9
+ "env_steps": 48984,
10
+ "features_sha256": "ed0a570dcabbbba769a4ac377436caea901a9b0238e5c9890a6db1e473a723c2",
11
+ "horizon": 8,
12
+ "learner_coordinates": "ResidualLearner:dsrl_na",
13
+ "pixel_spec_hash": "7aba3118e0304ea7265915550734918674f2b9a1e6c9bc2eb1cd99b568045a11",
14
+ "ratio": {
15
+ "adaptive_ratio": false,
16
+ "rule": "D_demo kept for the whole run (never evicted), online RL data FIFO; each ordinary batch = batch_size uniform draws over the eligible D_demo + online windows, offline rows first"
17
+ },
18
+ "selection": {
19
+ "cohort_size": 100,
20
+ "demos": [
21
+ "demo_0",
22
+ "demo_2",
23
+ "demo_5",
24
+ "demo_7",
25
+ "demo_9",
26
+ "demo_10",
27
+ "demo_13",
28
+ "demo_16",
29
+ "demo_17",
30
+ "demo_18",
31
+ "demo_20",
32
+ "demo_23",
33
+ "demo_27",
34
+ "demo_28",
35
+ "demo_29",
36
+ "demo_30",
37
+ "demo_31",
38
+ "demo_32",
39
+ "demo_34",
40
+ "demo_35",
41
+ "demo_37",
42
+ "demo_40",
43
+ "demo_41",
44
+ "demo_43",
45
+ "demo_46",
46
+ "demo_47",
47
+ "demo_50",
48
+ "demo_51",
49
+ "demo_56",
50
+ "demo_57",
51
+ "demo_58",
52
+ "demo_67",
53
+ "demo_71",
54
+ "demo_74",
55
+ "demo_76",
56
+ "demo_78",
57
+ "demo_80",
58
+ "demo_82",
59
+ "demo_83",
60
+ "demo_85",
61
+ "demo_86",
62
+ "demo_87",
63
+ "demo_93",
64
+ "demo_97",
65
+ "demo_98",
66
+ "demo_101",
67
+ "demo_102",
68
+ "demo_107",
69
+ "demo_110",
70
+ "demo_111",
71
+ "demo_112",
72
+ "demo_113",
73
+ "demo_115",
74
+ "demo_116",
75
+ "demo_119",
76
+ "demo_122",
77
+ "demo_123",
78
+ "demo_126",
79
+ "demo_127",
80
+ "demo_128",
81
+ "demo_130",
82
+ "demo_131",
83
+ "demo_132",
84
+ "demo_133",
85
+ "demo_134",
86
+ "demo_135",
87
+ "demo_138",
88
+ "demo_140",
89
+ "demo_141",
90
+ "demo_145",
91
+ "demo_147",
92
+ "demo_149",
93
+ "demo_151",
94
+ "demo_152",
95
+ "demo_154",
96
+ "demo_156",
97
+ "demo_157",
98
+ "demo_159",
99
+ "demo_161",
100
+ "demo_163",
101
+ "demo_164",
102
+ "demo_171",
103
+ "demo_172",
104
+ "demo_173",
105
+ "demo_174",
106
+ "demo_175",
107
+ "demo_176",
108
+ "demo_177",
109
+ "demo_179",
110
+ "demo_181",
111
+ "demo_182",
112
+ "demo_183",
113
+ "demo_186",
114
+ "demo_188",
115
+ "demo_190",
116
+ "demo_191",
117
+ "demo_192",
118
+ "demo_193",
119
+ "demo_196",
120
+ "demo_199"
121
+ ],
122
+ "manifest": "/ssd/WorkSpace/RL-Postrain/checkpoints/huggingface/yangfy0627/RL-postTrain/e62671ed1cf646b453660b698c22ca4d273b5ccc/Robomimic/study1/ToolHang/diverse100/seed_1/data_manifest.json",
123
+ "manifest_sha256": "9233d5824205e29f1afa721d8573fe19805a1a8e7829c296af1dd0b74a1d29b4",
124
+ "regime": "diverse100",
125
+ "source": "bc_regime",
126
+ "source_ph_sha256": "c5cb01b119a109a3d4842ca515906e86f1f35a8ee1a333d22b3503c1a513a8f4",
127
+ "split": {
128
+ "training": 90,
129
+ "validation": 10
130
+ },
131
+ "task": "tool_hang"
132
+ }
133
+ },
134
+ "identity_hash": "07e7e6bd263c84a9b8ceb56b1e6e92eea9ced27f6754960f1a8b5007f4f18779",
135
+ "stats": {
136
+ "chunks": 6171,
137
+ "dataset": "/ssd/WorkSpace/RL-Postrain/data/robomimic/tool_hang/ph/low_dim_v15.hdf5",
138
+ "dataset_sha256": "c5cb01b119a109a3d4842ca515906e86f1f35a8ee1a333d22b3503c1a513a8f4",
139
+ "env_steps": 48984,
140
+ "per_demo": [
141
+ {
142
+ "chunks": 85,
143
+ "demo": "demo_0",
144
+ "env_steps": 677,
145
+ "frames": 681,
146
+ "terminal_next_frame": 677
147
+ },
148
+ {
149
+ "chunks": 92,
150
+ "demo": "demo_2",
151
+ "env_steps": 733,
152
+ "frames": 737,
153
+ "terminal_next_frame": 733
154
+ },
155
+ {
156
+ "chunks": 85,
157
+ "demo": "demo_5",
158
+ "env_steps": 677,
159
+ "frames": 681,
160
+ "terminal_next_frame": 677
161
+ },
162
+ {
163
+ "chunks": 78,
164
+ "demo": "demo_7",
165
+ "env_steps": 618,
166
+ "frames": 622,
167
+ "terminal_next_frame": 618
168
+ },
169
+ {
170
+ "chunks": 78,
171
+ "demo": "demo_9",
172
+ "env_steps": 619,
173
+ "frames": 623,
174
+ "terminal_next_frame": 619
175
+ },
176
+ {
177
+ "chunks": 62,
178
+ "demo": "demo_10",
179
+ "env_steps": 493,
180
+ "frames": 497,
181
+ "terminal_next_frame": 493
182
+ },
183
+ {
184
+ "chunks": 64,
185
+ "demo": "demo_13",
186
+ "env_steps": 510,
187
+ "frames": 514,
188
+ "terminal_next_frame": 510
189
+ },
190
+ {
191
+ "chunks": 84,
192
+ "demo": "demo_16",
193
+ "env_steps": 665,
194
+ "frames": 676,
195
+ "terminal_next_frame": 665
196
+ },
197
+ {
198
+ "chunks": 84,
199
+ "demo": "demo_17",
200
+ "env_steps": 668,
201
+ "frames": 677,
202
+ "terminal_next_frame": 668
203
+ },
204
+ {
205
+ "chunks": 76,
206
+ "demo": "demo_18",
207
+ "env_steps": 605,
208
+ "frames": 632,
209
+ "terminal_next_frame": 605
210
+ },
211
+ {
212
+ "chunks": 71,
213
+ "demo": "demo_20",
214
+ "env_steps": 561,
215
+ "frames": 565,
216
+ "terminal_next_frame": 561
217
+ },
218
+ {
219
+ "chunks": 93,
220
+ "demo": "demo_23",
221
+ "env_steps": 738,
222
+ "frames": 742,
223
+ "terminal_next_frame": 738
224
+ },
225
+ {
226
+ "chunks": 59,
227
+ "demo": "demo_27",
228
+ "env_steps": 465,
229
+ "frames": 469,
230
+ "terminal_next_frame": 465
231
+ },
232
+ {
233
+ "chunks": 64,
234
+ "demo": "demo_28",
235
+ "env_steps": 505,
236
+ "frames": 512,
237
+ "terminal_next_frame": 505
238
+ },
239
+ {
240
+ "chunks": 66,
241
+ "demo": "demo_29",
242
+ "env_steps": 526,
243
+ "frames": 532,
244
+ "terminal_next_frame": 526
245
+ },
246
+ {
247
+ "chunks": 51,
248
+ "demo": "demo_30",
249
+ "env_steps": 405,
250
+ "frames": 409,
251
+ "terminal_next_frame": 405
252
+ },
253
+ {
254
+ "chunks": 58,
255
+ "demo": "demo_31",
256
+ "env_steps": 458,
257
+ "frames": 476,
258
+ "terminal_next_frame": 458
259
+ },
260
+ {
261
+ "chunks": 59,
262
+ "demo": "demo_32",
263
+ "env_steps": 467,
264
+ "frames": 473,
265
+ "terminal_next_frame": 467
266
+ },
267
+ {
268
+ "chunks": 72,
269
+ "demo": "demo_34",
270
+ "env_steps": 570,
271
+ "frames": 574,
272
+ "terminal_next_frame": 570
273
+ },
274
+ {
275
+ "chunks": 74,
276
+ "demo": "demo_35",
277
+ "env_steps": 585,
278
+ "frames": 589,
279
+ "terminal_next_frame": 585
280
+ },
281
+ {
282
+ "chunks": 71,
283
+ "demo": "demo_37",
284
+ "env_steps": 563,
285
+ "frames": 567,
286
+ "terminal_next_frame": 563
287
+ },
288
+ {
289
+ "chunks": 81,
290
+ "demo": "demo_40",
291
+ "env_steps": 645,
292
+ "frames": 649,
293
+ "terminal_next_frame": 645
294
+ },
295
+ {
296
+ "chunks": 68,
297
+ "demo": "demo_41",
298
+ "env_steps": 538,
299
+ "frames": 538,
300
+ "terminal_next_frame": 537
301
+ },
302
+ {
303
+ "chunks": 62,
304
+ "demo": "demo_43",
305
+ "env_steps": 491,
306
+ "frames": 495,
307
+ "terminal_next_frame": 491
308
+ },
309
+ {
310
+ "chunks": 56,
311
+ "demo": "demo_46",
312
+ "env_steps": 446,
313
+ "frames": 450,
314
+ "terminal_next_frame": 446
315
+ },
316
+ {
317
+ "chunks": 67,
318
+ "demo": "demo_47",
319
+ "env_steps": 535,
320
+ "frames": 539,
321
+ "terminal_next_frame": 535
322
+ },
323
+ {
324
+ "chunks": 54,
325
+ "demo": "demo_50",
326
+ "env_steps": 432,
327
+ "frames": 509,
328
+ "terminal_next_frame": 432
329
+ },
330
+ {
331
+ "chunks": 63,
332
+ "demo": "demo_51",
333
+ "env_steps": 498,
334
+ "frames": 502,
335
+ "terminal_next_frame": 498
336
+ },
337
+ {
338
+ "chunks": 59,
339
+ "demo": "demo_56",
340
+ "env_steps": 465,
341
+ "frames": 469,
342
+ "terminal_next_frame": 465
343
+ },
344
+ {
345
+ "chunks": 59,
346
+ "demo": "demo_57",
347
+ "env_steps": 472,
348
+ "frames": 476,
349
+ "terminal_next_frame": 472
350
+ },
351
+ {
352
+ "chunks": 54,
353
+ "demo": "demo_58",
354
+ "env_steps": 429,
355
+ "frames": 433,
356
+ "terminal_next_frame": 429
357
+ },
358
+ {
359
+ "chunks": 59,
360
+ "demo": "demo_67",
361
+ "env_steps": 470,
362
+ "frames": 474,
363
+ "terminal_next_frame": 470
364
+ },
365
+ {
366
+ "chunks": 70,
367
+ "demo": "demo_71",
368
+ "env_steps": 555,
369
+ "frames": 562,
370
+ "terminal_next_frame": 555
371
+ },
372
+ {
373
+ "chunks": 59,
374
+ "demo": "demo_74",
375
+ "env_steps": 468,
376
+ "frames": 470,
377
+ "terminal_next_frame": 468
378
+ },
379
+ {
380
+ "chunks": 52,
381
+ "demo": "demo_76",
382
+ "env_steps": 412,
383
+ "frames": 439,
384
+ "terminal_next_frame": 412
385
+ },
386
+ {
387
+ "chunks": 56,
388
+ "demo": "demo_78",
389
+ "env_steps": 448,
390
+ "frames": 452,
391
+ "terminal_next_frame": 448
392
+ },
393
+ {
394
+ "chunks": 67,
395
+ "demo": "demo_80",
396
+ "env_steps": 533,
397
+ "frames": 537,
398
+ "terminal_next_frame": 533
399
+ },
400
+ {
401
+ "chunks": 52,
402
+ "demo": "demo_82",
403
+ "env_steps": 416,
404
+ "frames": 422,
405
+ "terminal_next_frame": 416
406
+ },
407
+ {
408
+ "chunks": 51,
409
+ "demo": "demo_83",
410
+ "env_steps": 408,
411
+ "frames": 440,
412
+ "terminal_next_frame": 408
413
+ },
414
+ {
415
+ "chunks": 61,
416
+ "demo": "demo_85",
417
+ "env_steps": 484,
418
+ "frames": 488,
419
+ "terminal_next_frame": 484
420
+ },
421
+ {
422
+ "chunks": 60,
423
+ "demo": "demo_86",
424
+ "env_steps": 475,
425
+ "frames": 479,
426
+ "terminal_next_frame": 475
427
+ },
428
+ {
429
+ "chunks": 55,
430
+ "demo": "demo_87",
431
+ "env_steps": 438,
432
+ "frames": 443,
433
+ "terminal_next_frame": 438
434
+ },
435
+ {
436
+ "chunks": 47,
437
+ "demo": "demo_93",
438
+ "env_steps": 374,
439
+ "frames": 378,
440
+ "terminal_next_frame": 374
441
+ },
442
+ {
443
+ "chunks": 76,
444
+ "demo": "demo_97",
445
+ "env_steps": 601,
446
+ "frames": 608,
447
+ "terminal_next_frame": 601
448
+ },
449
+ {
450
+ "chunks": 49,
451
+ "demo": "demo_98",
452
+ "env_steps": 390,
453
+ "frames": 394,
454
+ "terminal_next_frame": 390
455
+ },
456
+ {
457
+ "chunks": 52,
458
+ "demo": "demo_101",
459
+ "env_steps": 410,
460
+ "frames": 414,
461
+ "terminal_next_frame": 410
462
+ },
463
+ {
464
+ "chunks": 74,
465
+ "demo": "demo_102",
466
+ "env_steps": 586,
467
+ "frames": 590,
468
+ "terminal_next_frame": 586
469
+ },
470
+ {
471
+ "chunks": 47,
472
+ "demo": "demo_107",
473
+ "env_steps": 371,
474
+ "frames": 375,
475
+ "terminal_next_frame": 371
476
+ },
477
+ {
478
+ "chunks": 47,
479
+ "demo": "demo_110",
480
+ "env_steps": 373,
481
+ "frames": 377,
482
+ "terminal_next_frame": 373
483
+ },
484
+ {
485
+ "chunks": 51,
486
+ "demo": "demo_111",
487
+ "env_steps": 401,
488
+ "frames": 405,
489
+ "terminal_next_frame": 401
490
+ },
491
+ {
492
+ "chunks": 61,
493
+ "demo": "demo_112",
494
+ "env_steps": 482,
495
+ "frames": 486,
496
+ "terminal_next_frame": 482
497
+ },
498
+ {
499
+ "chunks": 55,
500
+ "demo": "demo_113",
501
+ "env_steps": 436,
502
+ "frames": 440,
503
+ "terminal_next_frame": 436
504
+ },
505
+ {
506
+ "chunks": 59,
507
+ "demo": "demo_115",
508
+ "env_steps": 466,
509
+ "frames": 470,
510
+ "terminal_next_frame": 466
511
+ },
512
+ {
513
+ "chunks": 55,
514
+ "demo": "demo_116",
515
+ "env_steps": 434,
516
+ "frames": 472,
517
+ "terminal_next_frame": 434
518
+ },
519
+ {
520
+ "chunks": 67,
521
+ "demo": "demo_119",
522
+ "env_steps": 531,
523
+ "frames": 535,
524
+ "terminal_next_frame": 531
525
+ },
526
+ {
527
+ "chunks": 45,
528
+ "demo": "demo_122",
529
+ "env_steps": 356,
530
+ "frames": 370,
531
+ "terminal_next_frame": 356
532
+ },
533
+ {
534
+ "chunks": 48,
535
+ "demo": "demo_123",
536
+ "env_steps": 383,
537
+ "frames": 415,
538
+ "terminal_next_frame": 383
539
+ },
540
+ {
541
+ "chunks": 46,
542
+ "demo": "demo_126",
543
+ "env_steps": 365,
544
+ "frames": 371,
545
+ "terminal_next_frame": 365
546
+ },
547
+ {
548
+ "chunks": 50,
549
+ "demo": "demo_127",
550
+ "env_steps": 396,
551
+ "frames": 421,
552
+ "terminal_next_frame": 396
553
+ },
554
+ {
555
+ "chunks": 71,
556
+ "demo": "demo_128",
557
+ "env_steps": 563,
558
+ "frames": 567,
559
+ "terminal_next_frame": 563
560
+ },
561
+ {
562
+ "chunks": 60,
563
+ "demo": "demo_130",
564
+ "env_steps": 474,
565
+ "frames": 478,
566
+ "terminal_next_frame": 474
567
+ },
568
+ {
569
+ "chunks": 42,
570
+ "demo": "demo_131",
571
+ "env_steps": 332,
572
+ "frames": 336,
573
+ "terminal_next_frame": 332
574
+ },
575
+ {
576
+ "chunks": 54,
577
+ "demo": "demo_132",
578
+ "env_steps": 427,
579
+ "frames": 431,
580
+ "terminal_next_frame": 427
581
+ },
582
+ {
583
+ "chunks": 62,
584
+ "demo": "demo_133",
585
+ "env_steps": 494,
586
+ "frames": 498,
587
+ "terminal_next_frame": 494
588
+ },
589
+ {
590
+ "chunks": 41,
591
+ "demo": "demo_134",
592
+ "env_steps": 328,
593
+ "frames": 332,
594
+ "terminal_next_frame": 328
595
+ },
596
+ {
597
+ "chunks": 52,
598
+ "demo": "demo_135",
599
+ "env_steps": 411,
600
+ "frames": 415,
601
+ "terminal_next_frame": 411
602
+ },
603
+ {
604
+ "chunks": 70,
605
+ "demo": "demo_138",
606
+ "env_steps": 554,
607
+ "frames": 580,
608
+ "terminal_next_frame": 554
609
+ },
610
+ {
611
+ "chunks": 71,
612
+ "demo": "demo_140",
613
+ "env_steps": 568,
614
+ "frames": 572,
615
+ "terminal_next_frame": 568
616
+ },
617
+ {
618
+ "chunks": 45,
619
+ "demo": "demo_141",
620
+ "env_steps": 357,
621
+ "frames": 361,
622
+ "terminal_next_frame": 357
623
+ },
624
+ {
625
+ "chunks": 48,
626
+ "demo": "demo_145",
627
+ "env_steps": 379,
628
+ "frames": 384,
629
+ "terminal_next_frame": 379
630
+ },
631
+ {
632
+ "chunks": 41,
633
+ "demo": "demo_147",
634
+ "env_steps": 328,
635
+ "frames": 333,
636
+ "terminal_next_frame": 328
637
+ },
638
+ {
639
+ "chunks": 59,
640
+ "demo": "demo_149",
641
+ "env_steps": 466,
642
+ "frames": 474,
643
+ "terminal_next_frame": 466
644
+ },
645
+ {
646
+ "chunks": 87,
647
+ "demo": "demo_151",
648
+ "env_steps": 694,
649
+ "frames": 698,
650
+ "terminal_next_frame": 694
651
+ },
652
+ {
653
+ "chunks": 93,
654
+ "demo": "demo_152",
655
+ "env_steps": 740,
656
+ "frames": 744,
657
+ "terminal_next_frame": 740
658
+ },
659
+ {
660
+ "chunks": 50,
661
+ "demo": "demo_154",
662
+ "env_steps": 393,
663
+ "frames": 397,
664
+ "terminal_next_frame": 393
665
+ },
666
+ {
667
+ "chunks": 85,
668
+ "demo": "demo_156",
669
+ "env_steps": 678,
670
+ "frames": 682,
671
+ "terminal_next_frame": 678
672
+ },
673
+ {
674
+ "chunks": 55,
675
+ "demo": "demo_157",
676
+ "env_steps": 433,
677
+ "frames": 437,
678
+ "terminal_next_frame": 433
679
+ },
680
+ {
681
+ "chunks": 57,
682
+ "demo": "demo_159",
683
+ "env_steps": 456,
684
+ "frames": 465,
685
+ "terminal_next_frame": 456
686
+ },
687
+ {
688
+ "chunks": 53,
689
+ "demo": "demo_161",
690
+ "env_steps": 421,
691
+ "frames": 425,
692
+ "terminal_next_frame": 421
693
+ },
694
+ {
695
+ "chunks": 59,
696
+ "demo": "demo_163",
697
+ "env_steps": 466,
698
+ "frames": 469,
699
+ "terminal_next_frame": 466
700
+ },
701
+ {
702
+ "chunks": 67,
703
+ "demo": "demo_164",
704
+ "env_steps": 529,
705
+ "frames": 533,
706
+ "terminal_next_frame": 529
707
+ },
708
+ {
709
+ "chunks": 68,
710
+ "demo": "demo_171",
711
+ "env_steps": 537,
712
+ "frames": 541,
713
+ "terminal_next_frame": 537
714
+ },
715
+ {
716
+ "chunks": 69,
717
+ "demo": "demo_172",
718
+ "env_steps": 550,
719
+ "frames": 554,
720
+ "terminal_next_frame": 550
721
+ },
722
+ {
723
+ "chunks": 54,
724
+ "demo": "demo_173",
725
+ "env_steps": 432,
726
+ "frames": 441,
727
+ "terminal_next_frame": 432
728
+ },
729
+ {
730
+ "chunks": 56,
731
+ "demo": "demo_174",
732
+ "env_steps": 448,
733
+ "frames": 452,
734
+ "terminal_next_frame": 448
735
+ },
736
+ {
737
+ "chunks": 65,
738
+ "demo": "demo_175",
739
+ "env_steps": 515,
740
+ "frames": 525,
741
+ "terminal_next_frame": 515
742
+ },
743
+ {
744
+ "chunks": 49,
745
+ "demo": "demo_176",
746
+ "env_steps": 391,
747
+ "frames": 395,
748
+ "terminal_next_frame": 391
749
+ },
750
+ {
751
+ "chunks": 54,
752
+ "demo": "demo_177",
753
+ "env_steps": 429,
754
+ "frames": 433,
755
+ "terminal_next_frame": 429
756
+ },
757
+ {
758
+ "chunks": 57,
759
+ "demo": "demo_179",
760
+ "env_steps": 449,
761
+ "frames": 454,
762
+ "terminal_next_frame": 449
763
+ },
764
+ {
765
+ "chunks": 53,
766
+ "demo": "demo_181",
767
+ "env_steps": 422,
768
+ "frames": 426,
769
+ "terminal_next_frame": 422
770
+ },
771
+ {
772
+ "chunks": 73,
773
+ "demo": "demo_182",
774
+ "env_steps": 582,
775
+ "frames": 586,
776
+ "terminal_next_frame": 582
777
+ },
778
+ {
779
+ "chunks": 56,
780
+ "demo": "demo_183",
781
+ "env_steps": 443,
782
+ "frames": 447,
783
+ "terminal_next_frame": 443
784
+ },
785
+ {
786
+ "chunks": 56,
787
+ "demo": "demo_186",
788
+ "env_steps": 447,
789
+ "frames": 451,
790
+ "terminal_next_frame": 447
791
+ },
792
+ {
793
+ "chunks": 89,
794
+ "demo": "demo_188",
795
+ "env_steps": 705,
796
+ "frames": 709,
797
+ "terminal_next_frame": 705
798
+ },
799
+ {
800
+ "chunks": 52,
801
+ "demo": "demo_190",
802
+ "env_steps": 416,
803
+ "frames": 421,
804
+ "terminal_next_frame": 416
805
+ },
806
+ {
807
+ "chunks": 52,
808
+ "demo": "demo_191",
809
+ "env_steps": 411,
810
+ "frames": 415,
811
+ "terminal_next_frame": 411
812
+ },
813
+ {
814
+ "chunks": 52,
815
+ "demo": "demo_192",
816
+ "env_steps": 413,
817
+ "frames": 416,
818
+ "terminal_next_frame": 413
819
+ },
820
+ {
821
+ "chunks": 62,
822
+ "demo": "demo_193",
823
+ "env_steps": 494,
824
+ "frames": 498,
825
+ "terminal_next_frame": 494
826
+ },
827
+ {
828
+ "chunks": 53,
829
+ "demo": "demo_196",
830
+ "env_steps": 419,
831
+ "frames": 423,
832
+ "terminal_next_frame": 419
833
+ },
834
+ {
835
+ "chunks": 59,
836
+ "demo": "demo_199",
837
+ "env_steps": 469,
838
+ "frames": 473,
839
+ "terminal_next_frame": 469
840
+ }
841
+ ],
842
+ "pixel_cache": "/ssd/WorkSpace/RL-Postrain/data/pixels/tool_hang/v1",
843
+ "raw_coords": 342888,
844
+ "raw_coords_clipped": 0
845
+ },
846
+ "stats_hash": "919708648ac0650eea14b505ff433b77c7affe19a4514c64f3f32d6458412715"
847
+ }
RL/ToolHang100/DSRL-NA/diverse100/seed_1/resolved_config.json ADDED
@@ -0,0 +1,548 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "decision_ref": "docs/scope.md [0928-I], [0930-S2DEC]",
3
+ "env": {
4
+ "chunk_length": 8,
5
+ "n_envs": 4
6
+ },
7
+ "learner": {
8
+ "convention": "unified_env_step",
9
+ "gamma_env": 0.999
10
+ },
11
+ "overrides": [
12
+ {
13
+ "decision_ref": "[1020-S2TOOLHANG5] configs/study2_protocol.yaml",
14
+ "key": "n_step_chunks",
15
+ "reason": "task-specific unified Study2 setting",
16
+ "upstream": 3,
17
+ "value": 5
18
+ },
19
+ {
20
+ "decision_ref": "[1023-OFFLINE100] user 2026-10-23: 100-demo BC -> all 100 demos as offline D_demo",
21
+ "key": "offline.cohort_size",
22
+ "reason": "tool_hang regime diverse100 BC cohort has 100 demos; the offline demo buffer uses the checkpoint's entire BC cohort",
23
+ "upstream": 50,
24
+ "value": 100
25
+ },
26
+ {
27
+ "decision_ref": "[1023-OFFLINE100] user 2026-10-23: 100-demo BC -> all 100 demos as offline D_demo",
28
+ "key": "offline.episodes",
29
+ "reason": "tool_hang regime diverse100 BC cohort has 100 demos; the offline demo buffer uses the checkpoint's entire BC cohort",
30
+ "upstream": 50,
31
+ "value": 100
32
+ }
33
+ ],
34
+ "protocol": {
35
+ "checkpoint_every_rounds": 5,
36
+ "checkpoint_retention": {
37
+ "best": true,
38
+ "collapse_drop": {
39
+ "min_abs": 0.2,
40
+ "ratio": 0.5
41
+ },
42
+ "diagnostics": true,
43
+ "every_evals": 3
44
+ },
45
+ "collection": {
46
+ "n_envs": 4,
47
+ "train_seed_base": 100000000,
48
+ "train_seed_stride": 1000000,
49
+ "training_freq_chunks": 1000
50
+ },
51
+ "convergence": {
52
+ "early_stop": false,
53
+ "gain_lag": 10,
54
+ "high_sr": 0.95,
55
+ "high_sr_evals": 5,
56
+ "ma_window": 10,
57
+ "max_env_steps": 1606196,
58
+ "min_env_steps": 200000,
59
+ "min_gain": 0.02,
60
+ "zero_success_evals": 8
61
+ },
62
+ "critic_ensemble": {
63
+ "num": 10,
64
+ "policy": "mean",
65
+ "target": {
66
+ "kind": "subset_min",
67
+ "replace": false,
68
+ "size": 2
69
+ }
70
+ },
71
+ "eval": {
72
+ "best_of_n": 16,
73
+ "episodes": 200,
74
+ "every_env_steps": 20000,
75
+ "include_step0": true,
76
+ "seed_start": 960000,
77
+ "workers": 4
78
+ },
79
+ "final_eval": {
80
+ "episodes": 300,
81
+ "on_zero_stop": true,
82
+ "seed_start": 800000
83
+ },
84
+ "n_step_chunks": 5,
85
+ "offline": {
86
+ "adaptive_ratio": false,
87
+ "cohort_size": 100,
88
+ "episodes": 100,
89
+ "learners_without_offline": [
90
+ "dsrl_sac"
91
+ ],
92
+ "ratio": {
93
+ "decay_env_steps": {
94
+ "can": 160000,
95
+ "square": 320000,
96
+ "tool_hang": 640000,
97
+ "transport": 320000
98
+ },
99
+ "end": 0.1,
100
+ "start": {
101
+ "can": 0.7,
102
+ "square": 0.5,
103
+ "tool_hang": 0.5,
104
+ "transport": 0.9
105
+ }
106
+ },
107
+ "selection": "ascending_demo_index",
108
+ "source": "bc_regime",
109
+ "unsupported_learners": [
110
+ "zprl"
111
+ ]
112
+ },
113
+ "replay": {
114
+ "capacity_chunks": 1000000,
115
+ "pin_start_successes": false,
116
+ "success_buffer": "success_mask_on_shared_fifo",
117
+ "success_skip_rule": "skip_below_success_batch_size",
118
+ "unit": "executed_chunk"
119
+ },
120
+ "start": {
121
+ "max_episodes": 200,
122
+ "mode": "offline_buffer",
123
+ "policy": {
124
+ "default": "base_cfm_ode",
125
+ "zprl": "vib_base"
126
+ },
127
+ "target_successes": 20
128
+ },
129
+ "unified": {
130
+ "actor_target_tau": 0.005,
131
+ "critic_hidden": [
132
+ 512,
133
+ 512,
134
+ 512
135
+ ],
136
+ "critic_target_every_critic_step": true,
137
+ "critic_target_tau": 0.005,
138
+ "dsrl_noise_imitation_reference_dims": 28,
139
+ "gamma_env": 0.999,
140
+ "lr": {
141
+ "actor": 0.0001,
142
+ "critic": 0.0003,
143
+ "decay_end_env_steps": 1606196,
144
+ "min_ratio": 0.1,
145
+ "schedule": "warmup_cosine",
146
+ "temperature": 0.0003,
147
+ "warmup_env_steps": 20000
148
+ },
149
+ "reward": "binary_first_success",
150
+ "success_td_polyak": true,
151
+ "timeout_bootstrap": false,
152
+ "zprl_imitation": "mse_pretanh_mean"
153
+ },
154
+ "updates": {
155
+ "batch_size": 256,
156
+ "critic_per_actor": 2,
157
+ "success_batch_size": 256,
158
+ "utd_per_chunk": 3
159
+ },
160
+ "version": "study2_protocol_v10"
161
+ },
162
+ "protocol_hash": "903b284201350b7c879df61bb6860ec2d7fc9e4537fb5abd2aba8bcf7a60ee0f",
163
+ "provenance_extra": {
164
+ "agent_library": {
165
+ "framework": "jax",
166
+ "sources": {
167
+ "__init__.py": "0356b394bbae9d0738649e4864a034c4ee5d6f329c8a7f862717b3428e52313c",
168
+ "_vendor/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
169
+ "_vendor/ogpo/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
170
+ "_vendor/ogpo/agents/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
171
+ "_vendor/ogpo/agents/modules/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
172
+ "_vendor/ogpo/agents/modules/bc_helper.py": "8913392cc0434057323b2f5b28aaeb36d2c31e6b129ef09d21352353fe276c14",
173
+ "_vendor/ogpo/agents/modules/dist_helper.py": "c0f429c6c70082f5f5258ffffdf4bf0e6b2e1587c26528d3ee937a4e302d9951",
174
+ "_vendor/ogpo/agents/modules/flax_utils.py": "36f27885a086a4ae561171de4650da86330dcd68f17694e3f71278677c614037",
175
+ "_vendor/ogpo/agents/modules/igp_targets_helper.py": "b0343a236a856ec734cc73a1e73186d7d6520dd9efc34b3597327d21b55c9522",
176
+ "_vendor/ogpo/agents/modules/maintenance.py": "7b40d670c277783e118a0669f57aa10370c76b7b8fe3c1d27d16af1a3e9f0865",
177
+ "_vendor/ogpo/agents/modules/pg_helper.py": "cb2e0e83e0c95af408c9bac86fc9ccf1d7e427b3edb315c5d042de5d0a9328d4",
178
+ "_vendor/ogpo/agents/modules/q_helper.py": "71d874b74af5d724d82bd9c4a0a6027d8795e5dc26755ae7838abc3cda54718e",
179
+ "_vendor/ogpo/agents/ogpo.py": "3393282fd4a972c97951715a5776a506e37102cbe5e9c21e5e1ed1e6288696b2",
180
+ "_vendor/ogpo/agents/qc.py": "55906d8b598ce0a2607782e886697c3766dd281a0664ec05bf96173fc8c22c56",
181
+ "_vendor/ogpo/networks/__init__.py": "089130459e91d2b43f5bca45e2659507379fe62351ce6d2d91a49a0387ed2e90",
182
+ "_vendor/ogpo/networks/actors.py": "c0ad2b12439f5da82128e35907314741c22ccc70692de33fed90e67a32f8d86a",
183
+ "_vendor/ogpo/networks/actors_fql.py": "0d9f8646f1230b442bc80457252656c007431ca62458afa5fdecd15150ed9c4d",
184
+ "_vendor/ogpo/networks/critics.py": "77e40f7ea9d6837f8f89d889bfcd7a70e0ef20076f65720d79a9bef3db5130b4",
185
+ "_vendor/ogpo/networks/encoders.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
186
+ "_vendor/ogpo/networks/encoders/__init__.py": "41ae7973f560af5539cff23643843a0c56c253722bb17d7372a9ade029abb670",
187
+ "_vendor/ogpo/networks/encoders/impala.py": "62b7f1dab0aefe6662d3fa0fc59417d125cb9687af71d7875d533e24ac3e89d2",
188
+ "_vendor/ogpo/networks/encoders/minvit.py": "93f0af908450588cdbc9fc18929247b55447bb7972833388244707628328ed33",
189
+ "_vendor/ogpo/networks/encoders/paligemma.py": "0f2a448fd52a2c161f921e7fd7d79be462f98db34c09bf68877f8ed67de9e82c",
190
+ "_vendor/ogpo/networks/encoders/vision_proprio.py": "01ade56e774aeb29c3e7cd1d59eff5e43b7827d102769a450f8d7cc5a17666c8",
191
+ "_vendor/ogpo/networks/modules/__init__.py": "39a4fe9455c1b58c505fa1d9c0a90179dd64cdd459e8341d9ef7fcf6b7af9aae",
192
+ "_vendor/ogpo/networks/modules/attention.py": "a835876c6269ba0c5c63f49005027e5524a750f0c6e7af8697093b9bb8ff4c55",
193
+ "_vendor/ogpo/networks/modules/distributions.py": "5430f1bc6ee7fa71aac65664c9c3f13500d1fec3c5b2b1d80fc6590cca8ac71d",
194
+ "_vendor/ogpo/networks/modules/misc.py": "527793d4e64be2bbc8c22e73a574d5cb604e4672eabd0db9843853107502f620",
195
+ "_vendor/ogpo/networks/modules/mlp.py": "e2cf4d9923d95df8747c1554cf94f20000e108b1a9b0d9b0729a759b597a1316",
196
+ "_vendor/ogpo/networks/modules/time_embedding.py": "54d1a400b1b179a75c5aa55501031aa5d6513da12715ab75b89eaacdbe9f76e1",
197
+ "_vendor/ogpo/networks/modules/two_tier.py": "aa5182a5932cd500f95f4fddcd16429d66c968988639ddf3d6b33a2297953166",
198
+ "dice.py": "6989a71a178af904061e418a94c06615acb7576d9d32548566cc0032e273f22b",
199
+ "dsrl_na.py": "973970bfb02b87ad995f04572bc4d8443cc1e92d6447fbcbcfb80d8fa97cd9ce",
200
+ "dsrl_sac.py": "292d55b2d6f314ec12d17c77c2da2badef6a3fbdde89ba953044e074da6fbc6c",
201
+ "factory.py": "0117708b7cdab69cfaa83bba70f1101aff7b33866117946c377a5cb2fcc6ea49",
202
+ "flow.py": "76dce89f145b4194885524731523eada50e86f8d06b86f7cfaffd0fb3e3b779a",
203
+ "native_config.py": "e9dbf33c0ec15a654c2dde7f504d4e2492c29cbdf84c4eedf42cda94c158e4de",
204
+ "native_flow.py": "e9a9dc59e4e2ccaa7355bebb3c98cc9c1e0080c493889d58bfefdc7bc341c763",
205
+ "native_replay.py": "4f5922a41c5f4e93c599794143c0a26d8b22cb6c38d675ede9538a7bd2b30402",
206
+ "native_residual.py": "00936a650eed103e380f201c9917a9bc414d7b48ba244d0018ebd7b032b3af67",
207
+ "native_train.py": "83bd25879b1096afcb8c1e7a4c34ab77c652dafde10d86f7fdea221c98f6d4a4",
208
+ "networks.py": "81460ac49a7b67ef66b0efd04e26e86c320456c190493143ea77b57c8fbc195d",
209
+ "ogpo.py": "3501a14eb1e1dd0e7194de541dc1af2dd0b4584f102b7e23b0b625d490c79093",
210
+ "optim.py": "41f5c0e4907e822449154959067a84f09820010bc02c6a01a74b5d5478a76192",
211
+ "qc.py": "38ef1b0ed7a4d0833ee50a182262a3bd668931d9499483fa58f7cad43b3f9274",
212
+ "registry.py": "0e7928dac526397186290fe1a914403daa69c39c2c9da9aa395bacfbca9dc617",
213
+ "residual.py": "2dfda94aee07537924912d14693b7d0a41a76c9aea02852a006e5f4877c082ee",
214
+ "telemetry.py": "e5716864ea9b2126b50b86fec78d56ecfd96f17832d5006f83cb2763e6092feb",
215
+ "train.py": "af62bcfe092ab39e79eae9bc1f297430865cd31f61212faf17863e2b6de64ad8",
216
+ "vib.py": "ab3a6721c7c051d21e7955940942bc0f6d064635dd9c73b7275589c150f19f5e",
217
+ "vib_workflow.py": "35b264975fbbcb7f1142af4284f245c9de043e3dab6689ca40657e7be47090f2",
218
+ "zprl.py": "fefe27f991d1a482a57b04738d5b22b9879f0ea405e1935e94dd58ba72c30609"
219
+ },
220
+ "version": "study2_jax_agents_v2"
221
+ },
222
+ "base_framework": "jax",
223
+ "bc_identity": {
224
+ "architecture": "dice_vision_resnet_flow_unet_v1",
225
+ "family": "StandardBaseBC",
226
+ "framework": "jax"
227
+ },
228
+ "checkpoint": "/ssd/WorkSpace/RL-Postrain/runs/operations/20261024_toolhang_diverse100_v10_three/base/diverse100/epoch_300.pkl",
229
+ "checkpoint_sha256": "c5381a3e2ff4f4eeeb899a9d14286d9d42e0816c6431f1208fa39c60309c48f8",
230
+ "experiment": "tool_hang",
231
+ "frozen_hash": "719a7185a0d6a4877c76a5e1d9b1adb163a3d863d28bced54e79d534e8db6873",
232
+ "learner": {
233
+ "action_dim": 7,
234
+ "action_magnitude": null,
235
+ "actor_hidden": [
236
+ 2048,
237
+ 2048,
238
+ 2048
239
+ ],
240
+ "base_hash": "719a7185a0d6a4877c76a5e1d9b1adb163a3d863d28bced54e79d534e8db6873",
241
+ "critic_hidden": [
242
+ 512,
243
+ 512,
244
+ 512
245
+ ],
246
+ "ensemble": 10,
247
+ "feature_dim": 137,
248
+ "horizon": 8,
249
+ "normalizer_hash": "89c504c576b47f752f04402b67f3618b7b4087dd5073f15c5c79cc5a710b7f30",
250
+ "protocol": {
251
+ "checkpoint_every_rounds": 5,
252
+ "checkpoint_retention": {
253
+ "best": true,
254
+ "collapse_drop": {
255
+ "min_abs": 0.2,
256
+ "ratio": 0.5
257
+ },
258
+ "diagnostics": true,
259
+ "every_evals": 3
260
+ },
261
+ "collection": {
262
+ "n_envs": 4,
263
+ "train_seed_base": 100000000,
264
+ "train_seed_stride": 1000000,
265
+ "training_freq_chunks": 1000
266
+ },
267
+ "convergence": {
268
+ "early_stop": false,
269
+ "gain_lag": 10,
270
+ "high_sr": 0.95,
271
+ "high_sr_evals": 5,
272
+ "ma_window": 10,
273
+ "max_env_steps": 1606196,
274
+ "min_env_steps": 200000,
275
+ "min_gain": 0.02,
276
+ "zero_success_evals": 8
277
+ },
278
+ "critic_ensemble": {
279
+ "num": 10,
280
+ "policy": "mean",
281
+ "target": {
282
+ "kind": "subset_min",
283
+ "replace": false,
284
+ "size": 2
285
+ }
286
+ },
287
+ "eval": {
288
+ "best_of_n": 16,
289
+ "episodes": 200,
290
+ "every_env_steps": 20000,
291
+ "include_step0": true,
292
+ "seed_start": 960000,
293
+ "workers": 4
294
+ },
295
+ "final_eval": {
296
+ "episodes": 300,
297
+ "on_zero_stop": true,
298
+ "seed_start": 800000
299
+ },
300
+ "n_step_chunks": 5,
301
+ "offline": {
302
+ "adaptive_ratio": false,
303
+ "cohort_size": 100,
304
+ "episodes": 100,
305
+ "learners_without_offline": [
306
+ "dsrl_sac"
307
+ ],
308
+ "ratio": {
309
+ "decay_env_steps": {
310
+ "can": 160000,
311
+ "square": 320000,
312
+ "tool_hang": 640000,
313
+ "transport": 320000
314
+ },
315
+ "end": 0.1,
316
+ "start": {
317
+ "can": 0.7,
318
+ "square": 0.5,
319
+ "tool_hang": 0.5,
320
+ "transport": 0.9
321
+ }
322
+ },
323
+ "selection": "ascending_demo_index",
324
+ "source": "bc_regime",
325
+ "unsupported_learners": [
326
+ "zprl"
327
+ ]
328
+ },
329
+ "replay": {
330
+ "capacity_chunks": 1000000,
331
+ "pin_start_successes": false,
332
+ "success_buffer": "success_mask_on_shared_fifo",
333
+ "success_skip_rule": "skip_below_success_batch_size",
334
+ "unit": "executed_chunk"
335
+ },
336
+ "start": {
337
+ "max_episodes": 200,
338
+ "mode": "offline_buffer",
339
+ "policy": {
340
+ "default": "base_cfm_ode",
341
+ "zprl": "vib_base"
342
+ },
343
+ "target_successes": 20
344
+ },
345
+ "unified": {
346
+ "actor_target_tau": 0.005,
347
+ "critic_hidden": [
348
+ 512,
349
+ 512,
350
+ 512
351
+ ],
352
+ "critic_target_every_critic_step": true,
353
+ "critic_target_tau": 0.005,
354
+ "dsrl_noise_imitation_reference_dims": 28,
355
+ "gamma_env": 0.999,
356
+ "lr": {
357
+ "actor": 0.0001,
358
+ "critic": 0.0003,
359
+ "decay_end_env_steps": 1606196,
360
+ "min_ratio": 0.1,
361
+ "schedule": "warmup_cosine",
362
+ "temperature": 0.0003,
363
+ "warmup_env_steps": 20000
364
+ },
365
+ "reward": "binary_first_success",
366
+ "success_td_polyak": true,
367
+ "timeout_bootstrap": false,
368
+ "zprl_imitation": "mse_pretanh_mean"
369
+ },
370
+ "updates": {
371
+ "batch_size": 256,
372
+ "critic_per_actor": 2,
373
+ "success_batch_size": 256,
374
+ "utd_per_chunk": 3
375
+ },
376
+ "version": "study2_protocol_v10"
377
+ },
378
+ "subset": 2,
379
+ "variant": {
380
+ "flags": {
381
+ "noise_loglik_weight": 0.0,
382
+ "success_td": false
383
+ },
384
+ "learner": "dsrl_na",
385
+ "name": "dsrl_na",
386
+ "phase": 1,
387
+ "spec_hash": "288e84fd5eb8196404b74bd024f80ec14a2de9bac56dae6239fa3d474968c979",
388
+ "version": "study2_variants_v4"
389
+ },
390
+ "version": "study2_jax_agents_v2",
391
+ "vib_hash": null,
392
+ "weight_decay": null
393
+ },
394
+ "observation_mode": "pixel_state",
395
+ "offline_buffer": {
396
+ "actions_sha256": "7d65f5b7df78085d4780cf6338ac032e63f7f3d6ff9b1923bcc0992c840ee2d9",
397
+ "aux_sha256": "10f0ea12b0d1e89a65e72988f979e40fe4ffc48cec8c05e3651506b6a1d34088",
398
+ "checkpoint_sha256": "c5381a3e2ff4f4eeeb899a9d14286d9d42e0816c6431f1208fa39c60309c48f8",
399
+ "chunks": 6171,
400
+ "dataset_sha256": "c5cb01b119a109a3d4842ca515906e86f1f35a8ee1a333d22b3503c1a513a8f4",
401
+ "decision_ref": "[1017-S2OFFBC]",
402
+ "env_steps": 48984,
403
+ "features_sha256": "ed0a570dcabbbba769a4ac377436caea901a9b0238e5c9890a6db1e473a723c2",
404
+ "horizon": 8,
405
+ "identity_hash": "07e7e6bd263c84a9b8ceb56b1e6e92eea9ced27f6754960f1a8b5007f4f18779",
406
+ "learner_coordinates": "ResidualLearner:dsrl_na",
407
+ "pixel_spec_hash": "7aba3118e0304ea7265915550734918674f2b9a1e6c9bc2eb1cd99b568045a11",
408
+ "ratio": {
409
+ "adaptive_ratio": false,
410
+ "rule": "D_demo kept for the whole run (never evicted), online RL data FIFO; each ordinary batch = batch_size uniform draws over the eligible D_demo + online windows, offline rows first"
411
+ },
412
+ "selection": {
413
+ "cohort_size": 100,
414
+ "demos": [
415
+ "demo_0",
416
+ "demo_2",
417
+ "demo_5",
418
+ "demo_7",
419
+ "demo_9",
420
+ "demo_10",
421
+ "demo_13",
422
+ "demo_16",
423
+ "demo_17",
424
+ "demo_18",
425
+ "demo_20",
426
+ "demo_23",
427
+ "demo_27",
428
+ "demo_28",
429
+ "demo_29",
430
+ "demo_30",
431
+ "demo_31",
432
+ "demo_32",
433
+ "demo_34",
434
+ "demo_35",
435
+ "demo_37",
436
+ "demo_40",
437
+ "demo_41",
438
+ "demo_43",
439
+ "demo_46",
440
+ "demo_47",
441
+ "demo_50",
442
+ "demo_51",
443
+ "demo_56",
444
+ "demo_57",
445
+ "demo_58",
446
+ "demo_67",
447
+ "demo_71",
448
+ "demo_74",
449
+ "demo_76",
450
+ "demo_78",
451
+ "demo_80",
452
+ "demo_82",
453
+ "demo_83",
454
+ "demo_85",
455
+ "demo_86",
456
+ "demo_87",
457
+ "demo_93",
458
+ "demo_97",
459
+ "demo_98",
460
+ "demo_101",
461
+ "demo_102",
462
+ "demo_107",
463
+ "demo_110",
464
+ "demo_111",
465
+ "demo_112",
466
+ "demo_113",
467
+ "demo_115",
468
+ "demo_116",
469
+ "demo_119",
470
+ "demo_122",
471
+ "demo_123",
472
+ "demo_126",
473
+ "demo_127",
474
+ "demo_128",
475
+ "demo_130",
476
+ "demo_131",
477
+ "demo_132",
478
+ "demo_133",
479
+ "demo_134",
480
+ "demo_135",
481
+ "demo_138",
482
+ "demo_140",
483
+ "demo_141",
484
+ "demo_145",
485
+ "demo_147",
486
+ "demo_149",
487
+ "demo_151",
488
+ "demo_152",
489
+ "demo_154",
490
+ "demo_156",
491
+ "demo_157",
492
+ "demo_159",
493
+ "demo_161",
494
+ "demo_163",
495
+ "demo_164",
496
+ "demo_171",
497
+ "demo_172",
498
+ "demo_173",
499
+ "demo_174",
500
+ "demo_175",
501
+ "demo_176",
502
+ "demo_177",
503
+ "demo_179",
504
+ "demo_181",
505
+ "demo_182",
506
+ "demo_183",
507
+ "demo_186",
508
+ "demo_188",
509
+ "demo_190",
510
+ "demo_191",
511
+ "demo_192",
512
+ "demo_193",
513
+ "demo_196",
514
+ "demo_199"
515
+ ],
516
+ "manifest": "/ssd/WorkSpace/RL-Postrain/checkpoints/huggingface/yangfy0627/RL-postTrain/e62671ed1cf646b453660b698c22ca4d273b5ccc/Robomimic/study1/ToolHang/diverse100/seed_1/data_manifest.json",
517
+ "manifest_sha256": "9233d5824205e29f1afa721d8573fe19805a1a8e7829c296af1dd0b74a1d29b4",
518
+ "regime": "diverse100",
519
+ "source": "bc_regime",
520
+ "source_ph_sha256": "c5cb01b119a109a3d4842ca515906e86f1f35a8ee1a333d22b3503c1a513a8f4",
521
+ "split": {
522
+ "training": 90,
523
+ "validation": 10
524
+ },
525
+ "task": "tool_hang"
526
+ }
527
+ },
528
+ "offline_role": "mixed",
529
+ "pixel_spec_hash": "7aba3118e0304ea7265915550734918674f2b9a1e6c9bc2eb1cd99b568045a11",
530
+ "smoke": false,
531
+ "task": "tool_hang",
532
+ "task_spec_hash": "ac2909345e9536690a9da63f06aad8dfe135e757b89e28706b2c4af641f195c2"
533
+ },
534
+ "seed": 1,
535
+ "start_policy": "base_cfm_ode",
536
+ "variant": {
537
+ "flags": {
538
+ "noise_loglik_weight": 0.0,
539
+ "success_td": false
540
+ },
541
+ "learner": "dsrl_na",
542
+ "name": "dsrl_na",
543
+ "phase": 1,
544
+ "spec_hash": "288e84fd5eb8196404b74bd024f80ec14a2de9bac56dae6239fa3d474968c979",
545
+ "version": "study2_variants_v4"
546
+ },
547
+ "variant_hash": "288e84fd5eb8196404b74bd024f80ec14a2de9bac56dae6239fa3d474968c979"
548
+ }
RL/ToolHang100/DSRL-NA/diverse100/seed_1/start_protocol.json ADDED
@@ -0,0 +1,153 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "committed_episodes": [],
3
+ "mode": "offline_buffer",
4
+ "note": "[1006-S2V5] no BC start rollout; RL collects from step 0",
5
+ "offline": {
6
+ "adaptive_ratio": false,
7
+ "batches": 0,
8
+ "decay_env_steps": null,
9
+ "identity_hash": "07e7e6bd263c84a9b8ceb56b1e6e92eea9ced27f6754960f1a8b5007f4f18779",
10
+ "last_ratio": null,
11
+ "offline_rows": 0,
12
+ "online_rows": 0,
13
+ "ratio_end": null,
14
+ "ratio_start": null,
15
+ "records": 6171
16
+ },
17
+ "offline_identity": {
18
+ "actions_sha256": "7d65f5b7df78085d4780cf6338ac032e63f7f3d6ff9b1923bcc0992c840ee2d9",
19
+ "aux_sha256": "10f0ea12b0d1e89a65e72988f979e40fe4ffc48cec8c05e3651506b6a1d34088",
20
+ "checkpoint_sha256": "c5381a3e2ff4f4eeeb899a9d14286d9d42e0816c6431f1208fa39c60309c48f8",
21
+ "chunks": 6171,
22
+ "dataset_sha256": "c5cb01b119a109a3d4842ca515906e86f1f35a8ee1a333d22b3503c1a513a8f4",
23
+ "decision_ref": "[1017-S2OFFBC]",
24
+ "env_steps": 48984,
25
+ "features_sha256": "ed0a570dcabbbba769a4ac377436caea901a9b0238e5c9890a6db1e473a723c2",
26
+ "horizon": 8,
27
+ "learner_coordinates": "ResidualLearner:dsrl_na",
28
+ "pixel_spec_hash": "7aba3118e0304ea7265915550734918674f2b9a1e6c9bc2eb1cd99b568045a11",
29
+ "ratio": {
30
+ "adaptive_ratio": false,
31
+ "rule": "D_demo kept for the whole run (never evicted), online RL data FIFO; each ordinary batch = batch_size uniform draws over the eligible D_demo + online windows, offline rows first"
32
+ },
33
+ "selection": {
34
+ "cohort_size": 100,
35
+ "demos": [
36
+ "demo_0",
37
+ "demo_2",
38
+ "demo_5",
39
+ "demo_7",
40
+ "demo_9",
41
+ "demo_10",
42
+ "demo_13",
43
+ "demo_16",
44
+ "demo_17",
45
+ "demo_18",
46
+ "demo_20",
47
+ "demo_23",
48
+ "demo_27",
49
+ "demo_28",
50
+ "demo_29",
51
+ "demo_30",
52
+ "demo_31",
53
+ "demo_32",
54
+ "demo_34",
55
+ "demo_35",
56
+ "demo_37",
57
+ "demo_40",
58
+ "demo_41",
59
+ "demo_43",
60
+ "demo_46",
61
+ "demo_47",
62
+ "demo_50",
63
+ "demo_51",
64
+ "demo_56",
65
+ "demo_57",
66
+ "demo_58",
67
+ "demo_67",
68
+ "demo_71",
69
+ "demo_74",
70
+ "demo_76",
71
+ "demo_78",
72
+ "demo_80",
73
+ "demo_82",
74
+ "demo_83",
75
+ "demo_85",
76
+ "demo_86",
77
+ "demo_87",
78
+ "demo_93",
79
+ "demo_97",
80
+ "demo_98",
81
+ "demo_101",
82
+ "demo_102",
83
+ "demo_107",
84
+ "demo_110",
85
+ "demo_111",
86
+ "demo_112",
87
+ "demo_113",
88
+ "demo_115",
89
+ "demo_116",
90
+ "demo_119",
91
+ "demo_122",
92
+ "demo_123",
93
+ "demo_126",
94
+ "demo_127",
95
+ "demo_128",
96
+ "demo_130",
97
+ "demo_131",
98
+ "demo_132",
99
+ "demo_133",
100
+ "demo_134",
101
+ "demo_135",
102
+ "demo_138",
103
+ "demo_140",
104
+ "demo_141",
105
+ "demo_145",
106
+ "demo_147",
107
+ "demo_149",
108
+ "demo_151",
109
+ "demo_152",
110
+ "demo_154",
111
+ "demo_156",
112
+ "demo_157",
113
+ "demo_159",
114
+ "demo_161",
115
+ "demo_163",
116
+ "demo_164",
117
+ "demo_171",
118
+ "demo_172",
119
+ "demo_173",
120
+ "demo_174",
121
+ "demo_175",
122
+ "demo_176",
123
+ "demo_177",
124
+ "demo_179",
125
+ "demo_181",
126
+ "demo_182",
127
+ "demo_183",
128
+ "demo_186",
129
+ "demo_188",
130
+ "demo_190",
131
+ "demo_191",
132
+ "demo_192",
133
+ "demo_193",
134
+ "demo_196",
135
+ "demo_199"
136
+ ],
137
+ "manifest": "/ssd/WorkSpace/RL-Postrain/checkpoints/huggingface/yangfy0627/RL-postTrain/e62671ed1cf646b453660b698c22ca4d273b5ccc/Robomimic/study1/ToolHang/diverse100/seed_1/data_manifest.json",
138
+ "manifest_sha256": "9233d5824205e29f1afa721d8573fe19805a1a8e7829c296af1dd0b74a1d29b4",
139
+ "regime": "diverse100",
140
+ "source": "bc_regime",
141
+ "source_ph_sha256": "c5cb01b119a109a3d4842ca515906e86f1f35a8ee1a333d22b3503c1a513a8f4",
142
+ "split": {
143
+ "training": 90,
144
+ "validation": 10
145
+ },
146
+ "task": "tool_hang"
147
+ }
148
+ },
149
+ "policy": "base_cfm_ode",
150
+ "start_chunks": 0,
151
+ "start_env_steps": 0,
152
+ "status": "complete"
153
+ }
RL/ToolHang100/DSRL-NA/diverse100/seed_1/status.json ADDED
@@ -0,0 +1,170 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "accounting": {
3
+ "evaluation": {
4
+ "calls": 82,
5
+ "episodes": 16500
6
+ },
7
+ "history_complete": true,
8
+ "normalized": {
9
+ "checkpoint_seconds_per_env_step": 6.398828653153809e-05,
10
+ "chunks_per_10000_env_steps": 1251.9429907306637,
11
+ "denominator_env_steps": 1613492,
12
+ "diagnostics_seconds_per_env_step": 0.00014297916185665408,
13
+ "episodes_per_10000_env_steps": 12.24053171630228,
14
+ "evaluation_seconds_per_env_step": 0.04909627107300482,
15
+ "history_complete": true,
16
+ "projected_chunks": 201086.58239396292,
17
+ "projected_episodes": 1966.069308059786,
18
+ "projected_training_seconds": 24841.36105879986,
19
+ "projected_wall_seconds": 104076.39008927111,
20
+ "projection_budget_env_steps": 1606196,
21
+ "projection_is_estimate": true,
22
+ "rollout_seconds_per_env_step": 0.005770759973359839,
23
+ "training_seconds_per_env_step": 0.01546595873654265,
24
+ "updates_seconds_per_env_step": 0.00969519876318281,
25
+ "wall_seconds_per_env_step": 0.06479681812759533
26
+ },
27
+ "rollout": {
28
+ "chunks": 202000,
29
+ "env_steps": 1613492,
30
+ "episodes": 1975,
31
+ "start_chunks": 0,
32
+ "start_env_steps": 0,
33
+ "successes": 734,
34
+ "truncated_episodes": 1241
35
+ },
36
+ "schema_version": 1,
37
+ "sequence": 1385,
38
+ "sessions": 1,
39
+ "time": {
40
+ "checkpoint_seconds": 103.24458841234446,
41
+ "diagnostics_seconds": 230.6957338224165,
42
+ "evaluation_seconds": 79216.4406061247,
43
+ "preparation_seconds": 39.746325477957726,
44
+ "restore_seconds": 0.0,
45
+ "rollout_seconds": 9311.075050936313,
46
+ "start_rollout_seconds": 0.0,
47
+ "training_seconds": 24954.200693741674,
48
+ "updates_seconds": 15643.12564280536,
49
+ "wall_hours": 29.041429909536124,
50
+ "wall_seconds": 104549.14767433004
51
+ }
52
+ },
53
+ "detail": null,
54
+ "evals": 81,
55
+ "final_eval": {
56
+ "episodes": 300,
57
+ "imputed": false,
58
+ "n_success": 107,
59
+ "observed": true,
60
+ "seed_start": 800000,
61
+ "sr": 0.3566666666666667
62
+ },
63
+ "learner": "dsrl_na",
64
+ "learner_counters": {
65
+ "actor_updates": 303000,
66
+ "alpha_updates": 606000,
67
+ "clip_stats": {
68
+ "clip_mag_sum": 0.0,
69
+ "coords_clipped_frac": 0.0,
70
+ "max_clip_mag": 0.0,
71
+ "mean_clip_mag": 0.0,
72
+ "n_clipped": 0,
73
+ "n_coords": 11312000
74
+ },
75
+ "clip_stats_start": null,
76
+ "critic_updates": 606000,
77
+ "imitation_skipped": 0,
78
+ "imitation_updates": 0,
79
+ "noise_updates": 303000,
80
+ "success_td_skipped": 0,
81
+ "success_td_updates": 0,
82
+ "target_updates": 606000
83
+ },
84
+ "phase": 1,
85
+ "protocol_hash": "903b284201350b7c879df61bb6860ec2d7fc9e4537fb5abd2aba8bcf7a60ee0f",
86
+ "resumes": [],
87
+ "schedule": {
88
+ "actor_updates": 303000,
89
+ "chunk_transitions": 202000,
90
+ "critic_updates": 606000,
91
+ "gradient_updates": 606000,
92
+ "rl_chunks": 202000,
93
+ "rl_env_steps": 1613492,
94
+ "round_index": 202,
95
+ "update_index": 606000
96
+ },
97
+ "seed": 1,
98
+ "start": {
99
+ "mode": "offline_buffer",
100
+ "note": "[1006-S2V5] no BC start rollout; RL collects from step 0",
101
+ "offline": {
102
+ "adaptive_ratio": false,
103
+ "batches": 909000,
104
+ "decay_env_steps": null,
105
+ "identity_hash": "07e7e6bd263c84a9b8ceb56b1e6e92eea9ced27f6754960f1a8b5007f4f18779",
106
+ "last_ratio": 0.029646746832828093,
107
+ "offline_rows": 24486886,
108
+ "online_rows": 208217114,
109
+ "ratio_end": null,
110
+ "ratio_start": null,
111
+ "records": 6171
112
+ },
113
+ "start_chunks": 0,
114
+ "start_env_steps": 0,
115
+ "status": "complete"
116
+ },
117
+ "status": "budget_complete",
118
+ "stop": {
119
+ "eval_index": 80,
120
+ "gain": 0.006500000000000061,
121
+ "ma": 0.4010000000000001,
122
+ "rl_env_steps": 1613492,
123
+ "sr": 0.38,
124
+ "status": "budget_complete",
125
+ "trend": "flat"
126
+ },
127
+ "success_sampling": {
128
+ "imitation": {
129
+ "requested": 0,
130
+ "rounds_with_skips": 0,
131
+ "sampled": 0,
132
+ "skipped": 0
133
+ },
134
+ "last_round": {
135
+ "imitation": {
136
+ "requested": 0,
137
+ "sampled": 0,
138
+ "skipped": 0
139
+ },
140
+ "requested": 0,
141
+ "round": 201,
142
+ "sampled": 0,
143
+ "skipped": 0
144
+ },
145
+ "requested": 0,
146
+ "rounds_with_skips": 0,
147
+ "sampled": 0,
148
+ "skipped": 0,
149
+ "td": {
150
+ "requested": 0,
151
+ "rounds_with_skips": 0,
152
+ "sampled": 0,
153
+ "skipped": 0
154
+ }
155
+ },
156
+ "variant": "dsrl_na",
157
+ "variant_hash": "288e84fd5eb8196404b74bd024f80ec14a2de9bac56dae6239fa3d474968c979",
158
+ "wandb": {
159
+ "enabled": false,
160
+ "error": null,
161
+ "finished": true,
162
+ "mode": "online",
163
+ "name": "ToolHang-diverse100-DSRL-NA",
164
+ "project": "RL-postTrain",
165
+ "reason": null,
166
+ "run_id": "92c7dc10",
167
+ "transport": "relay",
168
+ "url": "https://wandb.ai/f70331658-university-of-chinese-academy-of-sciences/RL-postTrain/runs/92c7dc10"
169
+ }
170
+ }
RL/ToolHang100/DSRL-NA/diverse100/seed_1/training_accounting.json ADDED
@@ -0,0 +1,51 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "evaluation": {
3
+ "calls": 82,
4
+ "episodes": 16500
5
+ },
6
+ "history_complete": true,
7
+ "normalized": {
8
+ "checkpoint_seconds_per_env_step": 6.398828653153809e-05,
9
+ "chunks_per_10000_env_steps": 1251.9429907306637,
10
+ "denominator_env_steps": 1613492,
11
+ "diagnostics_seconds_per_env_step": 0.00014297916185665408,
12
+ "episodes_per_10000_env_steps": 12.24053171630228,
13
+ "evaluation_seconds_per_env_step": 0.04909627107300482,
14
+ "history_complete": true,
15
+ "projected_chunks": 201086.58239396292,
16
+ "projected_episodes": 1966.069308059786,
17
+ "projected_training_seconds": 24841.36105879986,
18
+ "projected_wall_seconds": 104076.39008927111,
19
+ "projection_budget_env_steps": 1606196,
20
+ "projection_is_estimate": true,
21
+ "rollout_seconds_per_env_step": 0.005770759973359839,
22
+ "training_seconds_per_env_step": 0.01546595873654265,
23
+ "updates_seconds_per_env_step": 0.00969519876318281,
24
+ "wall_seconds_per_env_step": 0.06479681812759533
25
+ },
26
+ "rollout": {
27
+ "chunks": 202000,
28
+ "env_steps": 1613492,
29
+ "episodes": 1975,
30
+ "start_chunks": 0,
31
+ "start_env_steps": 0,
32
+ "successes": 734,
33
+ "truncated_episodes": 1241
34
+ },
35
+ "schema_version": 1,
36
+ "sequence": 1386,
37
+ "sessions": 1,
38
+ "time": {
39
+ "checkpoint_seconds": 103.24458841234446,
40
+ "diagnostics_seconds": 230.6957338224165,
41
+ "evaluation_seconds": 79216.4406061247,
42
+ "preparation_seconds": 39.746325477957726,
43
+ "restore_seconds": 0.0,
44
+ "rollout_seconds": 9311.075050936313,
45
+ "start_rollout_seconds": 0.0,
46
+ "training_seconds": 24954.200693741674,
47
+ "updates_seconds": 15643.12564280536,
48
+ "wall_hours": 29.041429909536124,
49
+ "wall_seconds": 104549.14767433004
50
+ }
51
+ }