Stuck on an issue?
Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.
And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.
PPO model training with habitat 2020 challenge config
See original GitHub issue@mathfac @dhruvbatra Hi! Another issue on my side I am struggling with is the right training of habitat baseline ppo model with 2020 challenge configuration for pointnav task using habitat-api.
As PPO agent configuration I use the following file ppo_pointnav.yaml
BASE_TASK_CONFIG_PATH: "configs/tasks/pointnav_gib_rgbd_2020.yaml"
TRAINER_NAME: "ppo"
ENV_NAME: "NavRLEnv"
SIMULATOR_GPU_ID: 1
TORCH_GPU_ID: 1
VIDEO_OPTION: ["disk", "tensorboard"]
TENSORBOARD_DIR: "tb"
VIDEO_DIR: "video_dir"
TEST_EPISODE_COUNT: 994
EVAL_CKPT_PATH_DIR: "data/ppo_2020_checkpoints"
NUM_PROCESSES: 4
SENSORS: ["DEPTH_SENSOR"]
CHECKPOINT_FOLDER: "data/ppo_2020_checkpoints"
NUM_UPDATES: 270000
LOG_INTERVAL: 25
CHECKPOINT_INTERVAL: 2000
RL:
PPO:
clip_param: 0.1
ppo_epoch: 4
num_mini_batch: 2
value_loss_coef: 0.5
entropy_coef: 0.01
lr: 2.5e-4
eps: 1e-5
max_grad_norm: 0.5
num_steps: 128
hidden_size: 512
use_gae: True
gamma: 0.99
tau: 0.95
use_linear_clip_decay: True
use_linear_lr_decay: True
reward_window_size: 50
And for task configuration I used the same parameters as in challenge_pointnav2020.local.rgbd.yaml file:
ENVIRONMENT:
MAX_EPISODE_STEPS: 500
SIMULATOR:
AGENT_0:
SENSORS: ['RGB_SENSOR', 'DEPTH_SENSOR']
HEIGHT: 0.88
RADIUS: 0.18
HABITAT_SIM_V0:
GPU_DEVICE_ID: 0
ALLOW_SLIDING: False
RGB_SENSOR:
WIDTH: 640
HEIGHT: 360
HFOV: 70
POSITION: [0, 0.88, 0]
NOISE_MODEL: "GaussianNoiseModel"
NOISE_MODEL_KWARGS:
intensity_constant: 0.1
DEPTH_SENSOR:
WIDTH: 640
HEIGHT: 360
HFOV: 70
MIN_DEPTH: 0.1
MAX_DEPTH: 10.0
POSITION: [0, 0.88, 0]
NOISE_MODEL: "RedwoodDepthNoiseModel"
ACTION_SPACE_CONFIG: 'pyrobotnoisy'
NOISE_MODEL:
ROBOT: "LoCoBot"
CONTROLLER: 'Proportional'
NOISE_MULTIPLIER: 0.5
TASK:
TYPE: Nav-v0
SUCCESS_DISTANCE: 0.36
SENSORS: ['POINTGOAL_SENSOR']
POINTGOAL_SENSOR:
GOAL_FORMAT: POLAR
DIMENSIONALITY: 2
GOAL_SENSOR_UUID: pointgoal
MEASUREMENTS: ['DISTANCE_TO_GOAL', "SUCCESS", 'SPL']
SUCCESS:
SUCCESS_DISTANCE: 0.36
Just changed path to the train dataset (Habitat Challenge Data for Gibson (1.5 GB)):
DATASET:
TYPE: PointNav-v1
SPLIT: train
DATA_PATH: data/datasets/pointnav/gibson/v1/{split}/{split}.json.gz
After runnig command python -u habitat_baselines/run.py --exp-config habitat_baselines/config/pointnav/ppo_pointnav.yaml --run-type I got the following error:
---
The active scene does not contain semantic annotations.
---
I0325 20:17:08.559100 8915 simulator.py:143] Loaded navmesh data/scene_datasets/gibson/Monson.navmesh
I0325 20:17:08.559392 8915 simulator.py:155] Recomputing navmesh for agent's height 0.88 and radius 0.18.
I0325 20:17:08.567361 8915 PathFinder.cpp:338] Building navmesh with 275x112 cells
I0325 20:17:08.655342 8915 PathFinder.cpp:606] Created navmesh with 137 vertices 61 polygons
I0325 20:17:08.655371 8915 Simulator.cpp:403] reconstruct navmesh successful
2020-03-25 20:17:08,720 Initializing task Nav-v0
2020-03-25 20:17:11,725 agent number of parameters: 52694149
/home/pryhoda/anaconda3/envs/habitat/lib/python3.6/site-packages/torch-1.4.0-py3.6-linux-x86_64.egg/torch/optim/lr_scheduler.py:122: UserWarning: Detected call of `lr_scheduler.step()` before `optimizer.step()`. In PyTorch 1.1.0 and later, you should call them in the opposite order: `optimizer.step()` before `lr_scheduler.step()`. Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
"https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate", UserWarning)
/pytorch/aten/src/ATen/native/cuda/MultinomialKernel.cu:243: void at::native::<unnamed>::sampleMultinomialOnce(long *, long, int, scalar_t *, scalar_t *, int, int) [with scalar_t = float, accscalar_t = float]: block: [3,0,0], thread: [0,0,0] Assertion `val >= zero` failed.
/pytorch/aten/src/ATen/native/cuda/MultinomialKernel.cu:243: void at::native::<unnamed>::sampleMultinomialOnce(long *, long, int, scalar_t *, scalar_t *, int, int) [with scalar_t = float, accscalar_t = float]: block: [3,0,0], thread: [1,0,0] Assertion `val >= zero` failed.
/pytorch/aten/src/ATen/native/cuda/MultinomialKernel.cu:243: void at::native::<unnamed>::sampleMultinomialOnce(long *, long, int, scalar_t *, scalar_t *, int, int) [with scalar_t = float, accscalar_t = float]: block: [3,0,0], thread: [2,0,0] Assertion `val >= zero` failed.
/pytorch/aten/src/ATen/native/cuda/MultinomialKernel.cu:243: void at::native::<unnamed>::sampleMultinomialOnce(long *, long, int, scalar_t *, scalar_t *, int, int) [with scalar_t = float, accscalar_t = float]: block: [3,0,0], thread: [3,0,0] Assertion `val >= zero` failed.
/pytorch/aten/src/ATen/native/cuda/MultinomialKernel.cu:243: void at::native::<unnamed>::sampleMultinomialOnce(long *, long, int, scalar_t *, scalar_t *, int, int) [with scalar_t = float, accscalar_t = float]: block: [0,0,0], thread: [0,0,0] Assertion `val >= zero` failed.
/pytorch/aten/src/ATen/native/cuda/MultinomialKernel.cu:243: void at::native::<unnamed>::sampleMultinomialOnce(long *, long, int, scalar_t *, scalar_t *, int, int) [with scalar_t = float, accscalar_t = float]: block: [0,0,0], thread: [1,0,0] Assertion `val >= zero` failed.
/pytorch/aten/src/ATen/native/cuda/MultinomialKernel.cu:243: void at::native::<unnamed>::sampleMultinomialOnce(long *, long, int, scalar_t *, scalar_t *, int, int) [with scalar_t = float, accscalar_t = float]: block: [0,0,0], thread: [2,0,0] Assertion `val >= zero` failed.
/pytorch/aten/src/ATen/native/cuda/MultinomialKernel.cu:243: void at::native::<unnamed>::sampleMultinomialOnce(long *, long, int, scalar_t *, scalar_t *, int, int) [with scalar_t = float, accscalar_t = float]: block: [0,0,0], thread: [3,0,0] Assertion `val >= zero` failed.
/pytorch/aten/src/ATen/native/cuda/MultinomialKernel.cu:243: void at::native::<unnamed>::sampleMultinomialOnce(long *, long, int, scalar_t *, scalar_t *, int, int) [with scalar_t = float, accscalar_t = float]: block: [2,0,0], thread: [0,0,0] Assertion `val >= zero` failed.
/pytorch/aten/src/ATen/native/cuda/MultinomialKernel.cu:243: void at::native::<unnamed>::sampleMultinomialOnce(long *, long, int, scalar_t *, scalar_t *, int, int) [with scalar_t = float, accscalar_t = float]: block: [2,0,0], thread: [1,0,0] Assertion `val >= zero` failed.
/pytorch/aten/src/ATen/native/cuda/MultinomialKernel.cu:243: void at::native::<unnamed>::sampleMultinomialOnce(long *, long, int, scalar_t *, scalar_t *, int, int) [with scalar_t = float, accscalar_t = float]: block: [2,0,0], thread: [2,0,0] Assertion `val >= zero` failed.
/pytorch/aten/src/ATen/native/cuda/MultinomialKernel.cu:243: void at::native::<unnamed>::sampleMultinomialOnce(long *, long, int, scalar_t *, scalar_t *, int, int) [with scalar_t = float, accscalar_t = float]: block: [2,0,0], thread: [3,0,0] Assertion `val >= zero` failed.
/pytorch/aten/src/ATen/native/cuda/MultinomialKernel.cu:243: void at::native::<unnamed>::sampleMultinomialOnce(long *, long, int, scalar_t *, scalar_t *, int, int) [with scalar_t = float, accscalar_t = float]: block: [1,0,0], thread: [0,0,0] Assertion `val >= zero` failed.
/pytorch/aten/src/ATen/native/cuda/MultinomialKernel.cu:243: void at::native::<unnamed>::sampleMultinomialOnce(long *, long, int, scalar_t *, scalar_t *, int, int) [with scalar_t = float, accscalar_t = float]: block: [1,0,0], thread: [1,0,0] Assertion `val >= zero` failed.
/pytorch/aten/src/ATen/native/cuda/MultinomialKernel.cu:243: void at::native::<unnamed>::sampleMultinomialOnce(long *, long, int, scalar_t *, scalar_t *, int, int) [with scalar_t = float, accscalar_t = float]: block: [1,0,0], thread: [2,0,0] Assertion `val >= zero` failed.
/pytorch/aten/src/ATen/native/cuda/MultinomialKernel.cu:243: void at::native::<unnamed>::sampleMultinomialOnce(long *, long, int, scalar_t *, scalar_t *, int, int) [with scalar_t = float, accscalar_t = float]: block: [1,0,0], thread: [3,0,0] Assertion `val >= zero` failed.
Traceback (most recent call last):
File "habitat_baselines/run.py", line 70, in <module>
main()
File "habitat_baselines/run.py", line 40, in main
run_exp(**vars(args))
File "habitat_baselines/run.py", line 64, in run_exp
trainer.train()
File "/home/pryhoda/HabitatProject/habitat-api/habitat_baselines/rl/ppo/ppo_trainer.py", line 346, in train
rollouts, current_episode_reward, running_episode_stats
File "/home/pryhoda/HabitatProject/habitat-api/habitat_baselines/rl/ppo/ppo_trainer.py", line 181, in _collect_rollout_step
outputs = self.envs.step([a[0].item() for a in actions])
File "/home/pryhoda/HabitatProject/habitat-api/habitat_baselines/rl/ppo/ppo_trainer.py", line 181, in <listcomp>
outputs = self.envs.step([a[0].item() for a in actions])
RuntimeError: CUDA error: device-side assert triggered
Exception ignored in: <bound method VectorEnv.__del__ of <habitat.core.vector_env.VectorEnv object at 0x7f8dfa79ea58>>
Traceback (most recent call last):
File "/home/pryhoda/HabitatProject/habitat-api/habitat/core/vector_env.py", line 468, in __del__
self.close()
File "/home/pryhoda/HabitatProject/habitat-api/habitat/core/vector_env.py", line 350, in close
write_fn((CLOSE_COMMAND, None))
File "/home/pryhoda/anaconda3/envs/habitat/lib/python3.6/multiprocessing/connection.py", line 206, in send
self._send_bytes(_ForkingPickler.dumps(obj))
File "/home/pryhoda/anaconda3/envs/habitat/lib/python3.6/multiprocessing/connection.py", line 404, in _send_bytes
self._send(header + buf)
File "/home/pryhoda/anaconda3/envs/habitat/lib/python3.6/multiprocessing/connection.py", line 368, in _send
n = write(self._handle, buf)
BrokenPipeError: [Errno 32] Broken pipe
I am wondering if PPO agent is not adapted to train with 2020 challenge config (it run ok for me with 2019 challenge config - pointnav_gibson_rgbd.yaml ) or it is some issues on my side ? Thanks in advance!
Issue Analytics
- State:
- Created 3 years ago
- Comments:20 (10 by maintainers)
Top Results From Across the Web
Habitat 2.0 Overview
Here we will detail how to run the Task-Planning with Skills trained via reinforcement learning (TP-SRL) baseline from the Habitat 2.0 paper. This...
Read more >PointGoal Navigation | Papers With Code
We accelerate deep reinforcement learning-based training in visually complex 3D environments by two orders of magnitude over prior work, realizing end-to-end ...
Read more >DD-PPO: LEARNING NEAR-PERFECT POINTGOAL
of an agent equipped with RGB-D and GPS+Compass sensors on the Habitat Challenge 2019 (Savva et al., 2019) train & val sets. Using...
Read more >Zero-Shot Object-Goal Navigation using Multimodal ... - arXiv
similar or better than the 5% improvement in success between the Habitat 2020 and 2021 ObjectNav challenge winners. In an open-world setting ......
Read more >Does Evaluation in Simulation Predict Real-World Performance
[1] and the Habitat. Challenge 2019 [13] model the agent as an idealized cylinder of radius 0.1m and height 1.5m. As shown in...
Read more >
Top Related Medium Post
No results found
Top Related StackOverflow Question
No results found
Troubleshoot Live Code
Lightrun enables developers to add logs, metrics and snapshots to live code - no restarts or redeploys required.
Start Free
Top Related Reddit Thread
No results found
Top Related Hackernoon Post
No results found
Top Related Tweet
No results found
Top Related Dev.to Post
No results found
Top Related Hashnode Post
No results found
@erikwijmans As you suggested I trained DD-PPO model with resnet18 backbone.
When I tried to evaluate it, I got the following error:
Looks like it loads config for model with resnet50 backbone, but here is my config file:
I am wondering where could be the problem.
You can change resnet50 to resnet18 in the config, that will improve training speed.