调参

2026-05-04 07:01:33 +00:00
parent 75a5d86966
commit 4c3fdbcce6
3 changed files with 733 additions and 3372 deletions
--- a/pusht_results.txt
+++ b/pusht_results.txt
@@ -499,3 +499,732 @@ evaluation_time: 34.548478841781616 seconds
 inference_precision: fp16
 inference_compile_target: predictor
 inference_compile_mode: reduce-overhead
+
+==== CONFIG ====
+cache_dir: /mnt/ASC1637/stablewm
+solver:
+  _target_: stable_worldmodel.solver.CEMSolver
+  model: ???
+  batch_size: 8
+  num_samples: 300
+  var_scale: 1.0
+  n_steps: 30
+  topk: 30
+  device: cuda
+  seed: ${seed}
+world:
+  env_name: swm/PushT-v1
+  num_envs: ${eval.num_eval}
+  max_episode_steps: ???
+  history_size: 1
+  frame_skip: 1
+dataset:
+  stats: ${eval.dataset_name}
+  keys_to_cache:
+  - action
+  - proprio
+  - state
+seed: 42
+policy: pusht/lewm
+inference_precision: fp16
+plan_config:
+  horizon: 5
+  receding_horizon: 5
+  action_block: 5
+eval:
+  num_eval: 50
+  goal_offset_steps: 25
+  eval_budget: 50
+  img_size: 224
+  save_video: false
+  dataset_name: pusht_expert_train
+  callables:
+  - method: _set_state
+    args:
+      state:
+        value: state
+  - method: _set_goal_state
+    args:
+      goal_state:
+        value: goal_state
+output:
+  filename: pusht_results.txt
+
+==== RESULTS ====
+metrics: {'success_rate': 94.0, 'episode_successes': array([ True,  True,  True,  True,  True, False,  True,  True,  True,
+        True,  True,  True,  True,  True,  True,  True,  True,  True,
+        True, False,  True,  True,  True,  True,  True,  True,  True,
+        True,  True,  True,  True,  True,  True,  True,  True,  True,
+        True,  True,  True,  True,  True,  True, False,  True,  True,
+        True,  True,  True,  True,  True]), 'seeds': None}
+evaluation_time: 60.4364275932312 seconds
+inference_precision: fp16
+inference_compile_target: predictor
+inference_compile_mode: reduce-overhead
+
+==== CONFIG ====
+cache_dir: null
+solver:
+  _target_: stable_worldmodel.solver.CEMSolver
+  model: ???
+  batch_size: 8
+  num_samples: 300
+  var_scale: 1.0
+  n_steps: 30
+  topk: 30
+  device: cuda
+  seed: ${seed}
+world:
+  env_name: swm/PushT-v1
+  num_envs: ${eval.num_eval}
+  max_episode_steps: ???
+  history_size: 1
+  frame_skip: 1
+dataset:
+  stats: ${eval.dataset_name}
+  keys_to_cache:
+  - action
+  - proprio
+  - state
+seed: 42
+policy: pusht/lewm
+inference_precision: fp16
+plan_config:
+  horizon: 5
+  receding_horizon: 5
+  action_block: 5
+eval:
+  num_eval: 50
+  goal_offset_steps: 25
+  eval_budget: 50
+  img_size: 224
+  save_video: false
+  dataset_name: pusht_expert_train
+  callables:
+  - method: _set_state
+    args:
+      state:
+        value: state
+  - method: _set_goal_state
+    args:
+      goal_state:
+        value: goal_state
+output:
+  filename: pusht_results.txt
+
+==== RESULTS ====
+metrics: {'success_rate': 94.0, 'episode_successes': array([ True,  True,  True,  True,  True, False,  True,  True,  True,
+        True,  True,  True,  True,  True,  True,  True,  True,  True,
+        True, False,  True,  True,  True,  True,  True,  True,  True,
+        True,  True,  True,  True,  True,  True,  True,  True,  True,
+        True,  True,  True,  True,  True,  True, False,  True,  True,
+        True,  True,  True,  True,  True]), 'seeds': None}
+evaluation_time: 53.217538356781006 seconds
+inference_precision: fp16
+inference_compile_target: predictor
+inference_compile_mode: reduce-overhead
+
+==== CONFIG ====
+cache_dir: null
+solver:
+  _target_: stable_worldmodel.solver.CEMSolver
+  model: ???
+  batch_size: 8
+  num_samples: 300
+  var_scale: 1.0
+  n_steps: 30
+  topk: 30
+  device: cuda
+  seed: ${seed}
+world:
+  env_name: swm/PushT-v1
+  num_envs: ${eval.num_eval}
+  max_episode_steps: ???
+  history_size: 1
+  frame_skip: 1
+dataset:
+  stats: ${eval.dataset_name}
+  keys_to_cache:
+  - action
+  - proprio
+  - state
+seed: 42
+policy: pusht/lewm
+inference_precision: fp16
+plan_config:
+  horizon: 5
+  receding_horizon: 5
+  action_block: 5
+eval:
+  num_eval: 50
+  goal_offset_steps: 25
+  eval_budget: 50
+  img_size: 224
+  save_video: false
+  dataset_name: pusht_expert_train
+  callables:
+  - method: _set_state
+    args:
+      state:
+        value: state
+  - method: _set_goal_state
+    args:
+      goal_state:
+        value: goal_state
+output:
+  filename: pusht_results.txt
+
+==== RESULTS ====
+metrics: {'success_rate': 92.0, 'episode_successes': array([ True,  True,  True,  True,  True, False,  True,  True,  True,
+        True,  True,  True,  True,  True,  True,  True, False,  True,
+        True, False,  True,  True,  True,  True,  True,  True,  True,
+        True,  True,  True,  True,  True,  True,  True,  True,  True,
+        True,  True,  True,  True,  True,  True, False,  True,  True,
+        True,  True,  True,  True,  True]), 'seeds': None}
+evaluation_time: 35.28816795349121 seconds
+inference_precision: fp16
+inference_compile_target: predictor
+inference_compile_mode: reduce-overhead
+
+==== CONFIG ====
+cache_dir: null
+solver:
+  _target_: stable_worldmodel.solver.CEMSolver
+  model: ???
+  batch_size: 8
+  num_samples: 128
+  var_scale: 1.0
+  n_steps: 15
+  topk: 16
+  device: cuda
+  seed: ${seed}
+world:
+  env_name: swm/PushT-v1
+  num_envs: ${eval.num_eval}
+  max_episode_steps: ???
+  history_size: 1
+  frame_skip: 1
+dataset:
+  stats: ${eval.dataset_name}
+  keys_to_cache:
+  - action
+  - proprio
+  - state
+seed: 42
+policy: pusht/lewm
+inference_precision: fp16
+plan_config:
+  horizon: 5
+  receding_horizon: 5
+  action_block: 5
+eval:
+  num_eval: 5
+  goal_offset_steps: 25
+  eval_budget: 50
+  img_size: 224
+  save_video: false
+  dataset_name: pusht_expert_train
+  callables:
+  - method: _set_state
+    args:
+      state:
+        value: state
+  - method: _set_goal_state
+    args:
+      goal_state:
+        value: goal_state
+output:
+  filename: pusht_results.txt
+
+==== RESULTS ====
+metrics: {'success_rate': 80.0, 'episode_successes': array([ True, False,  True,  True,  True]), 'seeds': None}
+evaluation_time: 14.388008117675781 seconds
+inference_precision: fp16
+inference_compile_target: predictor
+inference_compile_mode: reduce-overhead
+
+==== CONFIG ====
+cache_dir: null
+solver:
+  _target_: stable_worldmodel.solver.CEMSolver
+  model: ???
+  batch_size: 8
+  num_samples: 128
+  var_scale: 1.0
+  n_steps: 15
+  topk: 16
+  device: cuda
+  seed: ${seed}
+world:
+  env_name: swm/PushT-v1
+  num_envs: ${eval.num_eval}
+  max_episode_steps: ???
+  history_size: 1
+  frame_skip: 1
+dataset:
+  stats: ${eval.dataset_name}
+  keys_to_cache:
+  - action
+  - proprio
+  - state
+seed: 42
+policy: pusht/lewm
+inference_precision: fp16
+plan_config:
+  horizon: 5
+  receding_horizon: 5
+  action_block: 5
+eval:
+  num_eval: 5
+  goal_offset_steps: 25
+  eval_budget: 50
+  img_size: 224
+  save_video: false
+  dataset_name: pusht_expert_train
+  callables:
+  - method: _set_state
+    args:
+      state:
+        value: state
+  - method: _set_goal_state
+    args:
+      goal_state:
+        value: goal_state
+output:
+  filename: pusht_results.txt
+
+==== RESULTS ====
+metrics: {'success_rate': 80.0, 'episode_successes': array([ True, False,  True,  True,  True]), 'seeds': None}
+evaluation_time: 5.2156901359558105 seconds
+inference_precision: fp16
+inference_compile_target: predictor
+inference_compile_mode: reduce-overhead
+
+==== CONFIG ====
+cache_dir: null
+solver:
+  _target_: stable_worldmodel.solver.CEMSolver
+  model: ???
+  batch_size: 8
+  num_samples: 128
+  var_scale: 1.0
+  n_steps: 15
+  topk: 16
+  device: cuda
+  seed: ${seed}
+world:
+  env_name: swm/PushT-v1
+  num_envs: ${eval.num_eval}
+  max_episode_steps: ???
+  history_size: 1
+  frame_skip: 1
+dataset:
+  stats: ${eval.dataset_name}
+  keys_to_cache:
+  - action
+  - proprio
+  - state
+seed: 42
+policy: pusht/lewm
+inference_precision: fp16
+plan_config:
+  horizon: 5
+  receding_horizon: 5
+  action_block: 5
+eval:
+  num_eval: 5
+  goal_offset_steps: 25
+  eval_budget: 50
+  img_size: 224
+  save_video: false
+  dataset_name: pusht_expert_train
+  callables:
+  - method: _set_state
+    args:
+      state:
+        value: state
+  - method: _set_goal_state
+    args:
+      goal_state:
+        value: goal_state
+output:
+  filename: pusht_results.txt
+
+==== RESULTS ====
+metrics: {'success_rate': 80.0, 'episode_successes': array([ True, False,  True,  True,  True]), 'seeds': None}
+evaluation_time: 5.612526178359985 seconds
+inference_precision: fp16
+inference_compile_target: predictor
+inference_compile_mode: reduce-overhead
+
+==== CONFIG ====
+cache_dir: null
+solver:
+  _target_: stable_worldmodel.solver.CEMSolver
+  model: ???
+  batch_size: 8
+  num_samples: 160
+  var_scale: 1.0
+  n_steps: 18
+  topk: 20
+  device: cuda
+  seed: ${seed}
+world:
+  env_name: swm/PushT-v1
+  num_envs: ${eval.num_eval}
+  max_episode_steps: ???
+  history_size: 1
+  frame_skip: 1
+dataset:
+  stats: ${eval.dataset_name}
+  keys_to_cache:
+  - action
+  - proprio
+  - state
+seed: 42
+policy: pusht/lewm
+inference_precision: fp16
+plan_config:
+  horizon: 5
+  receding_horizon: 5
+  action_block: 5
+eval:
+  num_eval: 50
+  goal_offset_steps: 25
+  eval_budget: 50
+  img_size: 224
+  save_video: false
+  dataset_name: pusht_expert_train
+  callables:
+  - method: _set_state
+    args:
+      state:
+        value: state
+  - method: _set_goal_state
+    args:
+      goal_state:
+        value: goal_state
+output:
+  filename: pusht_results.txt
+
+==== RESULTS ====
+metrics: {'success_rate': 88.0, 'episode_successes': array([ True,  True,  True,  True,  True,  True,  True,  True,  True,
+        True,  True,  True,  True,  True, False, False,  True, False,
+        True, False, False,  True,  True,  True,  True,  True,  True,
+        True,  True,  True,  True,  True,  True,  True,  True,  True,
+        True,  True, False,  True,  True,  True,  True,  True,  True,
+        True,  True,  True,  True,  True]), 'seeds': None}
+evaluation_time: 37.33809852600098 seconds
+inference_precision: fp16
+inference_compile_target: predictor
+inference_compile_mode: reduce-overhead
+
+==== CONFIG ====
+cache_dir: null
+solver:
+  _target_: stable_worldmodel.solver.CEMSolver
+  model: ???
+  batch_size: 8
+  num_samples: 128
+  var_scale: 1.0
+  n_steps: 20
+  topk: 16
+  device: cuda
+  seed: ${seed}
+world:
+  env_name: swm/PushT-v1
+  num_envs: ${eval.num_eval}
+  max_episode_steps: ???
+  history_size: 1
+  frame_skip: 1
+dataset:
+  stats: ${eval.dataset_name}
+  keys_to_cache:
+  - action
+  - proprio
+  - state
+seed: 42
+policy: pusht/lewm
+inference_precision: fp16
+plan_config:
+  horizon: 5
+  receding_horizon: 5
+  action_block: 5
+eval:
+  num_eval: 50
+  goal_offset_steps: 25
+  eval_budget: 50
+  img_size: 224
+  save_video: false
+  dataset_name: pusht_expert_train
+  callables:
+  - method: _set_state
+    args:
+      state:
+        value: state
+  - method: _set_goal_state
+    args:
+      goal_state:
+        value: goal_state
+output:
+  filename: pusht_results.txt
+
+==== RESULTS ====
+metrics: {'success_rate': 90.0, 'episode_successes': array([ True,  True,  True,  True,  True,  True,  True,  True,  True,
+        True,  True,  True,  True,  True, False,  True,  True,  True,
+        True,  True,  True,  True,  True,  True,  True,  True,  True,
+        True,  True,  True,  True,  True,  True,  True,  True,  True,
+        True,  True,  True, False,  True,  True, False,  True,  True,
+        True,  True, False, False,  True]), 'seeds': None}
+evaluation_time: 37.4112982749939 seconds
+inference_precision: fp16
+inference_compile_target: predictor
+inference_compile_mode: reduce-overhead
+
+==== CONFIG ====
+cache_dir: null
+solver:
+  _target_: stable_worldmodel.solver.CEMSolver
+  model: ???
+  batch_size: 16
+  num_samples: 96
+  var_scale: 1.0
+  n_steps: 12
+  topk: 12
+  device: cuda
+  seed: ${seed}
+world:
+  env_name: swm/PushT-v1
+  num_envs: ${eval.num_eval}
+  max_episode_steps: ???
+  history_size: 1
+  frame_skip: 1
+dataset:
+  stats: ${eval.dataset_name}
+  keys_to_cache:
+  - action
+  - proprio
+  - state
+seed: 42
+policy: pusht/lewm
+inference_precision: fp16
+plan_config:
+  horizon: 5
+  receding_horizon: 5
+  action_block: 5
+eval:
+  num_eval: 50
+  goal_offset_steps: 25
+  eval_budget: 50
+  img_size: 224
+  save_video: false
+  dataset_name: pusht_expert_train
+  callables:
+  - method: _set_state
+    args:
+      state:
+        value: state
+  - method: _set_goal_state
+    args:
+      goal_state:
+        value: goal_state
+output:
+  filename: pusht_results.txt
+
+==== RESULTS ====
+metrics: {'success_rate': 90.0, 'episode_successes': array([ True, False,  True,  True,  True,  True,  True,  True,  True,
+        True,  True,  True,  True,  True, False, False,  True,  True,
+        True,  True,  True,  True,  True,  True,  True,  True,  True,
+        True,  True,  True,  True,  True,  True,  True,  True, False,
+        True,  True,  True,  True,  True,  True,  True,  True,  True,
+        True,  True,  True, False,  True]), 'seeds': None}
+evaluation_time: 33.814064264297485 seconds
+inference_precision: fp16
+inference_compile_target: predictor
+inference_compile_mode: reduce-overhead
+
+==== CONFIG ====
+cache_dir: null
+solver:
+  _target_: stable_worldmodel.solver.CEMSolver
+  model: ???
+  batch_size: 8
+  num_samples: 64
+  var_scale: 1.0
+  n_steps: 10
+  topk: 8
+  device: cuda
+  seed: ${seed}
+world:
+  env_name: swm/PushT-v1
+  num_envs: ${eval.num_eval}
+  max_episode_steps: ???
+  history_size: 1
+  frame_skip: 1
+dataset:
+  stats: ${eval.dataset_name}
+  keys_to_cache:
+  - action
+  - proprio
+  - state
+seed: 42
+policy: pusht/lewm
+inference_precision: fp16
+plan_config:
+  horizon: 5
+  receding_horizon: 5
+  action_block: 5
+eval:
+  num_eval: 50
+  goal_offset_steps: 25
+  eval_budget: 50
+  img_size: 224
+  save_video: false
+  dataset_name: pusht_expert_train
+  callables:
+  - method: _set_state
+    args:
+      state:
+        value: state
+  - method: _set_goal_state
+    args:
+      goal_state:
+        value: goal_state
+output:
+  filename: pusht_results.txt
+
+==== RESULTS ====
+metrics: {'success_rate': 90.0, 'episode_successes': array([ True,  True,  True,  True,  True,  True,  True,  True,  True,
+        True,  True,  True,  True,  True, False, False,  True,  True,
+        True,  True,  True,  True,  True,  True,  True,  True,  True,
+        True,  True,  True,  True,  True,  True,  True,  True, False,
+        True,  True,  True,  True,  True,  True, False,  True,  True,
+        True,  True,  True, False,  True]), 'seeds': None}
+evaluation_time: 31.93920063972473 seconds
+inference_precision: fp16
+inference_compile_target: predictor
+inference_compile_mode: reduce-overhead
+
+==== CONFIG ====
+cache_dir: null
+solver:
+  _target_: stable_worldmodel.solver.CEMSolver
+  model: ???
+  batch_size: 8
+  num_samples: 48
+  var_scale: 1.0
+  n_steps: 8
+  topk: 6
+  device: cuda
+  seed: ${seed}
+world:
+  env_name: swm/PushT-v1
+  num_envs: ${eval.num_eval}
+  max_episode_steps: ???
+  history_size: 1
+  frame_skip: 1
+dataset:
+  stats: ${eval.dataset_name}
+  keys_to_cache:
+  - action
+  - proprio
+  - state
+seed: 42
+policy: pusht/lewm
+inference_precision: fp16
+plan_config:
+  horizon: 5
+  receding_horizon: 5
+  action_block: 5
+eval:
+  num_eval: 50
+  goal_offset_steps: 25
+  eval_budget: 50
+  img_size: 224
+  save_video: false
+  dataset_name: pusht_expert_train
+  callables:
+  - method: _set_state
+    args:
+      state:
+        value: state
+  - method: _set_goal_state
+    args:
+      goal_state:
+        value: goal_state
+output:
+  filename: pusht_results.txt
+
+==== RESULTS ====
+metrics: {'success_rate': 80.0, 'episode_successes': array([ True,  True, False,  True,  True,  True,  True,  True,  True,
+        True,  True,  True,  True,  True,  True, False, False,  True,
+        True,  True, False,  True, False,  True,  True,  True,  True,
+        True,  True,  True,  True,  True,  True,  True,  True, False,
+        True,  True,  True, False,  True, False, False,  True,  True,
+        True,  True,  True, False,  True]), 'seeds': None}
+evaluation_time: 31.703332662582397 seconds
+inference_precision: fp16
+inference_compile_target: predictor
+inference_compile_mode: reduce-overhead
+
+==== CONFIG ====
+cache_dir: null
+solver:
+  _target_: stable_worldmodel.solver.CEMSolver
+  model: ???
+  batch_size: 8
+  num_samples: 64
+  var_scale: 1.0
+  n_steps: 10
+  topk: 8
+  device: cuda
+  seed: ${seed}
+world:
+  env_name: swm/PushT-v1
+  num_envs: ${eval.num_eval}
+  max_episode_steps: ???
+  history_size: 1
+  frame_skip: 1
+dataset:
+  stats: ${eval.dataset_name}
+  keys_to_cache:
+  - action
+  - proprio
+  - state
+seed: 42
+policy: pusht/lewm
+inference_precision: fp16
+plan_config:
+  horizon: 5
+  receding_horizon: 5
+  action_block: 5
+eval:
+  num_eval: 50
+  goal_offset_steps: 25
+  eval_budget: 50
+  img_size: 224
+  save_video: false
+  dataset_name: pusht_expert_train
+  callables:
+  - method: _set_state
+    args:
+      state:
+        value: state
+  - method: _set_goal_state
+    args:
+      goal_state:
+        value: goal_state
+output:
+  filename: pusht_results.txt
+
+==== RESULTS ====
+metrics: {'success_rate': 90.0, 'episode_successes': array([ True,  True,  True,  True,  True,  True,  True,  True,  True,
+        True,  True,  True,  True,  True, False, False,  True,  True,
+        True,  True,  True,  True,  True,  True,  True,  True,  True,
+        True,  True,  True,  True,  True,  True,  True,  True, False,
+        True,  True,  True,  True,  True,  True, False,  True,  True,
+        True,  True,  True, False,  True]), 'seeds': None}
+evaluation_time: 16.38284730911255 seconds
+inference_precision: fp16
+inference_compile_target: predictor
+inference_compile_mode: reduce-overhead