Training Terminated Due to missing task_id
- Lenguaje dominante
- Python
- Estrellas
- 36
- Forks
- 7
- Métricas de merge de PR
- Sin PR fusionados en 30 d
Descripción
Hi there! thank you for opening sourcing this work.
I am following the training and evaluation instructions, running
```
python -m procthor_objectnav.main \
experiment=procthor_objectnav/experiments/rgb_clipresnet50gru_ddppo \
agent=locobot \
target_object_types=robothor_habitat2022 \
machine.num_train_processes=12 \
machine.num_val_processes=1 \
ai2thor.platform=CloudRendering \
model.add_prev_actions_embedding=true \
callbacks=wandb_logging_callback \
seed=100 \
procthor.p_randomize_materials=0.8 \
wandb.project=$WANDB_PROJECT \
wandb.name=$WANDB_RUN_NAME \
wandb.project=$WANDB_PROJECT \
wandb.entity=$WANDB_ENTITY \
```
The training went well for 10 minutes and sudden stop because of
```
[05/05 13:28:56 ERROR:] Traceback (most recent call last):
File "/home/tsaisplus/miniconda3/envs/procthor-rl/lib/python3.9/site-packages/allenact/algorithms/onpolicy_sync/vector_sampled_tasks.py", line 986, in _task_sampling_loop_generator_fn
task_callback_data = callback_sensor_suite.get_observations(
File "/home/tsaisplus/miniconda3/envs/procthor-rl/lib/python3.9/site-packages/allenact/base_abstractions/sensor.py", line 131, in get_observations
return {
File "/home/tsaisplus/miniconda3/envs/procthor-rl/lib/python3.9/site-packages/allenact/base_abstractions/sensor.py", line 132, in
uuid: sensor.get_observation(env=env, task=task, **kwargs) # type: ignore
File "/home/tsaisplus/projects/procthor-rl/procthor_objectnav/callbacks/local_logging.py", line 78, in get_observation
f"{d}/ac-data/{task.task_info['id']}.txt",
KeyError: 'id'
[vector_sampled_tasks.py: 1083]
[05/05 13:28:56 INFO:] SingleProcessVectorSampledTask 0 closing. [vector_sampled_tasks.py: 1087]
```
It seems that 'id' key is missing from the task_info dictionary. Any intuition or solution on that?
Best regards
Yuxin
Guía de contribución
No hay ninguna guía de contribución indexada para este repositorio
Evaluación
Este issue todavía no se ha evaluado.