admin/lerobot
1
0
Fork 0
mirror of https://github.com/huggingface/lerobot.git synced 2026-07-23 01:41:54 +00:00
Code Issues Packages Projects Releases Wiki Activity

debug

Browse Source
This commit is contained in:
Pepijn
2026-01-09 16:54:11 +01:00
parent 480ee3299f
commit feedababd2
1 changed files with 67 additions and 54 deletions
Download Patch File Download Diff File Expand all files Collapse all files
examples/rac/rac_data_collection_openarms_rtc.py
+67 -54
View File
@@ -356,77 +356,90 @@ def rtc_inference_thread(
get_actions_threshold = 0 get_actions_threshold = 0
inference_count = 0 inference_count = 0
wait_logged = False
while not shutdown_event.is_set(): while not shutdown_event.is_set():
if not policy_active.is_set(): if not policy_active.is_set():
if not wait_logged:
logger.info("[RTC] Waiting for policy_active...")
wait_logged = True
time.sleep(0.01) time.sleep(0.01)
continue continue
wait_logged = False
action_queue = queue_holder["queue"] action_queue = queue_holder["queue"]
if action_queue is None: if action_queue is None:
logger.warning("[RTC] queue_holder['queue'] is None!")
time.sleep(0.01) time.sleep(0.01)
continue continue
# Get observation from shared holder (set by main loop) # Get observation from shared holder (set by main loop)
obs_filtered = obs_holder.get("obs") obs_filtered = obs_holder.get("obs")
if obs_filtered is None: if obs_filtered is None:
if inference_count == 0: logger.warning("[RTC] obs_holder['obs'] is None - main loop not setting it?")
logger.warning("[RTC] Waiting for observation from main loop...")
time.sleep(0.01) time.sleep(0.01)
continue continue
if action_queue.qsize() <= get_actions_threshold: qsize = action_queue.qsize()
if inference_count == 0: if qsize <= get_actions_threshold:
logger.info(f"[RTC] Starting first inference, obs has {len(obs_filtered)} keys") try:
current_time = time.perf_counter() if inference_count == 0:
action_index_before_inference = action_queue.get_action_index() logger.info(f"[RTC] Starting first inference, obs has {len(obs_filtered)} keys, qsize={qsize}, threshold={get_actions_threshold}")
prev_actions = action_queue.get_left_over() current_time = time.perf_counter()
action_index_before_inference = action_queue.get_action_index()
inference_latency = latency_tracker.max() prev_actions = action_queue.get_left_over()
inference_delay = math.ceil(inference_latency / time_per_chunk) if inference_latency else 0
inference_latency = latency_tracker.max()
# Build observation for policy (using obs from main loop) inference_delay = math.ceil(inference_latency / time_per_chunk) if inference_latency else 0
obs_with_policy_features = build_dataset_frame(hw_features, obs_filtered, prefix="observation")
# Build observation for policy (using obs from main loop)
# Convert to tensors (like evaluate_with_rtc.py) obs_with_policy_features = build_dataset_frame(hw_features, obs_filtered, prefix="observation")
for name in obs_with_policy_features:
obs_with_policy_features[name] = torch.from_numpy(obs_with_policy_features[name]) # Convert to tensors (like evaluate_with_rtc.py)
if "image" in name: for name in obs_with_policy_features:
obs_with_policy_features[name] = ( obs_with_policy_features[name] = torch.from_numpy(obs_with_policy_features[name])
obs_with_policy_features[name].type(torch.float32) / 255 if "image" in name:
) obs_with_policy_features[name] = (
obs_with_policy_features[name] = ( obs_with_policy_features[name].type(torch.float32) / 255
obs_with_policy_features[name].permute(2, 0, 1).contiguous() )
) obs_with_policy_features[name] = (
obs_with_policy_features[name] = obs_with_policy_features[name].unsqueeze(0) obs_with_policy_features[name].permute(2, 0, 1).contiguous()
obs_with_policy_features[name] = obs_with_policy_features[name].to(policy_device) )
obs_with_policy_features[name] = obs_with_policy_features[name].unsqueeze(0)
obs_with_policy_features["task"] = [cfg.dataset.single_task] obs_with_policy_features[name] = obs_with_policy_features[name].to(policy_device)
obs_with_policy_features["robot_type"] = obs_holder.get("robot_type", "openarms_follower")
obs_with_policy_features["task"] = [cfg.dataset.single_task]
# Preprocess and run inference obs_with_policy_features["robot_type"] = obs_holder.get("robot_type", "openarms_follower")
preprocessed_obs = preprocessor(obs_with_policy_features)
# Preprocess and run inference
actions = policy.predict_action_chunk( preprocessed_obs = preprocessor(obs_with_policy_features)
preprocessed_obs,
inference_delay=inference_delay, actions = policy.predict_action_chunk(
prev_chunk_left_over=prev_actions, preprocessed_obs,
) inference_delay=inference_delay,
prev_chunk_left_over=prev_actions,
original_actions = actions.squeeze(0).clone() )
postprocessed_actions = postprocessor(actions).squeeze(0)
original_actions = actions.squeeze(0).clone()
new_latency = time.perf_counter() - current_time postprocessed_actions = postprocessor(actions).squeeze(0)
new_delay = math.ceil(new_latency / time_per_chunk)
latency_tracker.add(new_latency) new_latency = time.perf_counter() - current_time
new_delay = math.ceil(new_latency / time_per_chunk)
# Put actions in queue! latency_tracker.add(new_latency)
action_queue.merge(
original_actions, postprocessed_actions, new_delay, action_index_before_inference # Put actions in queue!
) action_queue.merge(
original_actions, postprocessed_actions, new_delay, action_index_before_inference
inference_count += 1 )
logger.info(f"[RTC] Inference #{inference_count}, latency={new_latency:.2f}s, queue={action_queue.qsize()}, shape={postprocessed_actions.shape}")
inference_count += 1
logger.info(f"[RTC] Inference #{inference_count}, latency={new_latency:.2f}s, queue={action_queue.qsize()}, shape={postprocessed_actions.shape}")
except Exception as e:
logger.error(f"[RTC] Inference failed: {e}")
import traceback
traceback.print_exc()
time.sleep(1.0) # Don't spam errors
else: else:
time.sleep(0.01) time.sleep(0.01)