Multiagent

.cache/v/cache/lastfailed

@@ -0,0 +1,3 @@
+{
+  "tests/test_multiagent_representations.py": true
+}

control_pcgrl/configs/config.py

@@ -32,6 +32,15 @@ class BinaryPathConfig(ProblemConfig):
     name: str = 'binary'
     # Regions weight will be 0 by default.
     weights: Dict[str, int] = field(default_factory = lambda: ({
+    #    'player': 1,
+    #    'create': 1,
+    #    'target': 1,
+    #    'regions': 1,
+    #    'ratio': 1,
+    #    'dist-win': 1,
+    #    'sol-length': 2
+
+
         'path-length': 100,
     }))
 
@@ -51,6 +60,8 @@ class BinaryControlConfig(ProblemConfig):
 @dataclass
 class MultiagentConfig:
     n_agents: int = MISSING
+    # valid values: (shared, independent, JSON string)
+    policies: str = "centralized"  # use shared weights by default
 
 
 @dataclass
@@ -107,6 +118,7 @@ class ControlPCGRLConfig:
     multiagent: MultiagentConfig = MISSING
     problem: ProblemConfig = MISSING
 
+    algorithm: str = 'PPO'
     debug: bool = False
     render: bool = False
     infer: bool = False
@@ -117,6 +129,7 @@ class ControlPCGRLConfig:
 
     exp_id: str = '0'
     representation: str = 'turtle'
+    show_agents: bool = False
     learning_rate: float = 5e-6
     gamma: float = 0.99
     map_shape: List[Any] = field(default_factory=lambda: 
@@ -158,4 +171,4 @@ class ControlPCGRLConfig:
 cs.store(name="binary_path", group="problem", node=BinaryPathConfig)
 
 cs.store(name="default_model", group="model", node=ModelConfig)
-cs.store(name="seqnca", group="model", node=SeqNCAConfig)
+cs.store(name="seqnca", group="model", node=SeqNCAConfig)

control_pcgrl/configs/config.yaml

@@ -3,22 +3,22 @@ defaults:
   - _self_
 
   # Why can't we override this on the command line?
-  # - override hydra/launcher: submitit_local
+  #- override hydra/launcher: submitit_local
   - override hydra/launcher: submitit_slurm
 
 hydra:
   sweeper:
     params:
-      exp_id: 0, 1, 2
-      learning_rate: 5e-4, 1e-4, 5e-5, 1e-5, 5e-6, 1e-6
+      #exp_id: 0, 1, 2
+      learning_rate: 5e-5
 
   launcher:
     tasks_per_node: 1
 
     #FIXME: Can't set this to 1 or 2 even when only asking from 1 ("0") worker from ray...
-    cpus_per_task: 10
+    cpus_per_task: 22
 
     gpus_per_node: 1
-    timeout_min: 1440
+    timeout_min: 1440 # 1 days of training
     mem_gb: 30
     # Emails maybe?

control_pcgrl/configs/multirun/2022-11-29/15-04-08/.submitit/27482696/27482696_submission.sh

@@ -0,0 +1,20 @@
+#!/bin/bash
+
+# Parameters
+#SBATCH --array=0-3%4
+#SBATCH --cpus-per-task=10
+#SBATCH --error=/scratch/rd2893/control-pcgrl/control_pcgrl/configs/multirun/2022-11-29/15-04-08/.submitit/%A_%a/%A_%a_0_log.err
+#SBATCH --gpus-per-node=1
+#SBATCH --job-name=train_ctrl
+#SBATCH --mem=30GB
+#SBATCH --nodes=1
+#SBATCH --ntasks-per-node=1
+#SBATCH --open-mode=append
+#SBATCH --output=/scratch/rd2893/control-pcgrl/control_pcgrl/configs/multirun/2022-11-29/15-04-08/.submitit/%A_%a/%A_%a_0_log.out
+#SBATCH --signal=USR2@120
+#SBATCH --time=1440
+#SBATCH --wckey=submitit
+
+# command
+export SUBMITIT_EXECUTOR=slurm
+srun --unbuffered --output /scratch/rd2893/control-pcgrl/control_pcgrl/configs/multirun/2022-11-29/15-04-08/.submitit/%A_%a/%A_%a_%t_log.out --error /scratch/rd2893/control-pcgrl/control_pcgrl/configs/multirun/2022-11-29/15-04-08/.submitit/%A_%a/%A_%a_%t_log.err /scratch/rd2893/miniconda3/envs/pcgrl/bin/python -u -m submitit.core._submit /scratch/rd2893/control-pcgrl/control_pcgrl/configs/multirun/2022-11-29/15-04-08/.submitit/%j

control_pcgrl/configs/multirun/2022-11-29/15-04-08/.submitit/27482696_0/27482696_0_0_log.err

@@ -0,0 +1,70 @@
+/scratch/rd2893/miniconda3/envs/pcgrl/lib/python3.7/site-packages/gym/utils/passive_env_checker.py:32: UserWarning: WARN: A Box observation space has an unconventional shape (neither an image, nor a 1D vector). We recommend flattening the observation to have only a 1D vector or use a custom policy to properly process the data. Actual observation shape: (16, 16)
+  "A Box observation space has an unconventional shape (neither an image, nor a 1D vector). "
+2022-11-29 15:05:25,106	WARNING env.py:236 -- Your MultiAgentEnv <MultiAgentWrapper<ControlWrapper<CroppedImagePCGRLWrapper<ToImage<OneHotEncoding<Cropped<OrderEnforcing<PcgrlCtrlEnv<binary-turtle-v0>>>>>>>>> does not have some or all of the needed base-class attributes! Make sure you call `super().__init__` from within your MutiAgentEnv's constructor. This will raise an error in the future.
+2022-11-29 15:05:27,892	INFO worker.py:1518 -- Started a local Ray instance.
+(PPOTrainer pid=3593784) 2022-11-29 15:05:35,626	INFO ppo.py:379 -- In multi-agent mode, policies will be optimized sequentially by the multi-GPU optimizer. Consider setting simple_optimizer=True if this doesn't work for you.
+(PPOTrainer pid=3593784) 2022-11-29 15:05:35,628	INFO algorithm.py:358 -- Current log_level is WARN. For more information, set 'log_level': 'INFO' / 'DEBUG' or use the -v and -vv flags.
+(PPOTrainer pid=3593784) /scratch/rd2893/miniconda3/envs/pcgrl/lib/python3.7/site-packages/gym/utils/passive_env_checker.py:32: UserWarning: WARN: A Box observation space has an unconventional shape (neither an image, nor a 1D vector). We recommend flattening the observation to have only a 1D vector or use a custom policy to properly process the data. Actual observation shape: (16, 16)
+(PPOTrainer pid=3593784)   "A Box observation space has an unconventional shape (neither an image, nor a 1D vector). "
+(PPOTrainer pid=3593784) 2022-11-29 15:05:42,133	WARNING deprecation.py:48 -- DeprecationWarning: `simple_optimizer` has been deprecated. This will raise an error in the future!
+(PPOTrainer pid=3593784) 2022-11-29 15:05:42,337	WARNING util.py:66 -- Install gputil for GPU system monitoring.
+(PPOTrainer pid=3593784) 2022-11-29 15:05:42,421	WARNING deprecation.py:48 -- DeprecationWarning: `policy_mapping_fn(agent_id)` has been deprecated. Use `policy_mapping_fn(agent_id, episode, worker, **kwargs)` instead. This will raise an error in the future!
+(PPOTrainer pid=3593784) 2022-11-29 15:05:44,433	WARNING deprecation.py:48 -- DeprecationWarning: `concat_samples` has been deprecated. Use `concat_samples() from rllib.policy.sample_batch` instead. This will raise an error in the future!
+(PPOTrainer pid=3593784) /scratch/rd2893/miniconda3/envs/pcgrl/lib/python3.7/site-packages/ray/rllib/utils/metrics/learner_info.py:110: RuntimeWarning: Mean of empty slice
+(PPOTrainer pid=3593784)   return np.nanmean(tower_data)
+2022-11-30 01:46:43,745	WARNING util.py:244 -- The `callbacks.on_trial_result` operation took 7.095 s, which may be a performance bottleneck.
+2022-11-30 01:46:43,747	WARNING util.py:244 -- The `process_trial_result` operation took 7.099 s, which may be a performance bottleneck.
+2022-11-30 01:46:43,747	WARNING util.py:244 -- Processing trial results took 7.099 s, which may be a performance bottleneck. Please consider reporting results less frequently to Ray Tune.
+2022-11-30 01:46:43,747	WARNING util.py:244 -- The `process_trial_result` operation took 7.100 s, which may be a performance bottleneck.
+2022-11-30 01:48:00,324	WARNING util.py:244 -- The `callbacks.on_trial_result` operation took 9.173 s, which may be a performance bottleneck.
+2022-11-30 01:48:00,327	WARNING util.py:244 -- The `process_trial_result` operation took 9.178 s, which may be a performance bottleneck.
+2022-11-30 01:48:00,331	WARNING util.py:244 -- Processing trial results took 9.182 s, which may be a performance bottleneck. Please consider reporting results less frequently to Ray Tune.
+2022-11-30 01:48:00,332	WARNING util.py:244 -- The `process_trial_result` operation took 9.184 s, which may be a performance bottleneck.
+2022-11-30 01:50:26,182	WARNING util.py:244 -- The `callbacks.on_trial_result` operation took 6.208 s, which may be a performance bottleneck.
+2022-11-30 01:50:26,184	WARNING util.py:244 -- The `process_trial_result` operation took 6.212 s, which may be a performance bottleneck.
+2022-11-30 01:50:26,184	WARNING util.py:244 -- Processing trial results took 6.212 s, which may be a performance bottleneck. Please consider reporting results less frequently to Ray Tune.
+2022-11-30 01:50:26,184	WARNING util.py:244 -- The `process_trial_result` operation took 6.214 s, which may be a performance bottleneck.
+submitit WARNING (2022-11-30 15:02:50,873) - Caught signal SIGUSR2 on gv002.hpc.nyu.edu: this job is timed-out.
+submitit WARNING (2022-11-30 15:02:51,153) - Bypassing signal SIGCONT
+Traceback (most recent call last):
+  File "/scratch/rd2893/miniconda3/envs/pcgrl/lib/python3.7/site-packages/submitit/core/job_environment.py", line 226, in checkpoint_and_try_requeue
+    raise utils.UncompletedJobError(message)
+submitit.core.utils.UncompletedJobError: Job not requeued because: timed-out too many times.
+Exception ignored in: 'ray._raylet.check_signals'
+Traceback (most recent call last):
+  File "/scratch/rd2893/miniconda3/envs/pcgrl/lib/python3.7/site-packages/submitit/core/job_environment.py", line 226, in checkpoint_and_try_requeue
+    raise utils.UncompletedJobError(message)
+submitit.core.utils.UncompletedJobError: Job not requeued because: timed-out too many times.
+slurmstepd: error: *** JOB 27482702 ON gv002 CANCELLED AT 2022-11-30T15:05:22 DUE TO TIME LIMIT ***
+slurmstepd: error: *** STEP 27482702.0 ON gv002 CANCELLED AT 2022-11-30T15:05:22 DUE TO TIME LIMIT ***
+*** SIGTERM received at time=1669838722 on cpu 6 ***
+[failure_signal_handler.cc : 331] RAW: Signal 15 raised at PC=0x14df1d5017b0 while already in AbslFailureSignalHandler()
+*** SIGTERM received at time=1669838722 on cpu 6 ***
+PC: @     0x14df1d5017b0  (unknown)  absl::lts_20211102::debugging_internal::ParseMangledName()
+    @     0x14df25374b20       3424  (unknown)
+    @     0x14df1d4fbf78         32  absl::lts_20211102::debugging_internal::DemangleInplace()
+    @     0x14df1d4fd4b2       1296  absl::lts_20211102::debugging_internal::(anonymous namespace)::Symbolizer::GetSymbol()
+    @     0x14df1d4fd7fe         80  absl::lts_20211102::Symbolize()
+    @     0x14df1d4d162d       2144  absl::lts_20211102::debugging_internal::DumpPCAndFrameSizeAndSymbol()
+    @     0x14df1d4d1751        192  absl::lts_20211102::debugging_internal::DumpPCAndFrameSizesAndStackTrace()
+    @     0x14df1d4d10c6        496  absl::lts_20211102::WriteStackTrace()
+    @     0x14df1d4d12d1         80  absl::lts_20211102::AbslFailureSignalHandler()
+    @     0x14df25374b20  (unknown)  (unknown)
+[2022-11-30 15:05:22,384 E 3592250 3592250] logging.cc:361: *** SIGTERM received at time=1669838722 on cpu 6 ***
+[2022-11-30 15:05:22,384 E 3592250 3592250] logging.cc:361: PC: @     0x14df1d5017b0  (unknown)  absl::lts_20211102::debugging_internal::ParseMangledName()
+[2022-11-30 15:05:22,384 E 3592250 3592250] logging.cc:361:     @     0x14df25374b20       3424  (unknown)
+[2022-11-30 15:05:22,384 E 3592250 3592250] logging.cc:361:     @     0x14df1d4fbf78         32  absl::lts_20211102::debugging_internal::DemangleInplace()
+[2022-11-30 15:05:22,384 E 3592250 3592250] logging.cc:361:     @     0x14df1d4fd4b2       1296  absl::lts_20211102::debugging_internal::(anonymous namespace)::Symbolizer::GetSymbol()
+[2022-11-30 15:05:22,384 E 3592250 3592250] logging.cc:361:     @     0x14df1d4fd7fe         80  absl::lts_20211102::Symbolize()
+[2022-11-30 15:05:22,384 E 3592250 3592250] logging.cc:361:     @     0x14df1d4d162d       2144  absl::lts_20211102::debugging_internal::DumpPCAndFrameSizeAndSymbol()
+[2022-11-30 15:05:22,384 E 3592250 3592250] logging.cc:361:     @     0x14df1d4d1751        192  absl::lts_20211102::debugging_internal::DumpPCAndFrameSizesAndStackTrace()
+[2022-11-30 15:05:22,384 E 3592250 3592250] logging.cc:361:     @     0x14df1d4d10c6        496  absl::lts_20211102::WriteStackTrace()
+[2022-11-30 15:05:22,384 E 3592250 3592250] logging.cc:361:     @     0x14df1d4d12d1         80  absl::lts_20211102::AbslFailureSignalHandler()
+[2022-11-30 15:05:22,385 E 3592250 3592250] logging.cc:361:     @     0x14df25374b20  (unknown)  (unknown)
+PC: @     0x14df2537064a  (unknown)  pthread_cond_timedwait@@GLIBC_2.3.2
+    @     0x14df25374b20  (unknown)  (unknown)
+[2022-11-30 15:05:22,385 E 3592250 3592250] logging.cc:361: *** SIGTERM received at time=1669838722 on cpu 6 ***
+[2022-11-30 15:05:22,385 E 3592250 3592250] logging.cc:361: PC: @     0x14df2537064a  (unknown)  pthread_cond_timedwait@@GLIBC_2.3.2
+[2022-11-30 15:05:22,385 E 3592250 3592250] logging.cc:361:     @     0x14df25374b20  (unknown)  (unknown)
+submitit WARNING (2022-11-30 15:05:22,799) - Bypassing signal SIGTERM
+submitit WARNING (2022-11-30 15:05:22,819) - Bypassing signal SIGCONT

control_pcgrl/configs/multirun/2022-11-29/15-04-08/.submitit/27482696_1/27482696_1_0_log.err

@@ -0,0 +1,20 @@
+/scratch/rd2893/miniconda3/envs/pcgrl/lib/python3.7/site-packages/gym/utils/passive_env_checker.py:32: UserWarning: WARN: A Box observation space has an unconventional shape (neither an image, nor a 1D vector). We recommend flattening the observation to have only a 1D vector or use a custom policy to properly process the data. Actual observation shape: (16, 16)
+  "A Box observation space has an unconventional shape (neither an image, nor a 1D vector). "
+2022-11-29 15:05:25,086	WARNING env.py:236 -- Your MultiAgentEnv <MultiAgentWrapper<ControlWrapper<CroppedImagePCGRLWrapper<ToImage<OneHotEncoding<Cropped<OrderEnforcing<PcgrlCtrlEnv<binary-turtle-v0>>>>>>>>> does not have some or all of the needed base-class attributes! Make sure you call `super().__init__` from within your MutiAgentEnv's constructor. This will raise an error in the future.
+2022-11-29 15:05:27,858	INFO worker.py:1518 -- Started a local Ray instance.
+(PPOTrainer pid=532048) 2022-11-29 15:05:35,162	INFO ppo.py:379 -- In multi-agent mode, policies will be optimized sequentially by the multi-GPU optimizer. Consider setting simple_optimizer=True if this doesn't work for you.
+(PPOTrainer pid=532048) 2022-11-29 15:05:35,163	INFO algorithm.py:358 -- Current log_level is WARN. For more information, set 'log_level': 'INFO' / 'DEBUG' or use the -v and -vv flags.
+(PPOTrainer pid=532048) /scratch/rd2893/miniconda3/envs/pcgrl/lib/python3.7/site-packages/gym/utils/passive_env_checker.py:32: UserWarning: WARN: A Box observation space has an unconventional shape (neither an image, nor a 1D vector). We recommend flattening the observation to have only a 1D vector or use a custom policy to properly process the data. Actual observation shape: (16, 16)
+(PPOTrainer pid=532048)   "A Box observation space has an unconventional shape (neither an image, nor a 1D vector). "
+(PPOTrainer pid=532048) 2022-11-29 15:05:41,998	WARNING deprecation.py:48 -- DeprecationWarning: `simple_optimizer` has been deprecated. This will raise an error in the future!
+(PPOTrainer pid=532048) 2022-11-29 15:05:42,202	WARNING util.py:66 -- Install gputil for GPU system monitoring.
+(PPOTrainer pid=532048) 2022-11-29 15:05:42,269	WARNING deprecation.py:48 -- DeprecationWarning: `policy_mapping_fn(agent_id)` has been deprecated. Use `policy_mapping_fn(agent_id, episode, worker, **kwargs)` instead. This will raise an error in the future!
+(PPOTrainer pid=532048) 2022-11-29 15:05:44,548	WARNING deprecation.py:48 -- DeprecationWarning: `concat_samples` has been deprecated. Use `concat_samples() from rllib.policy.sample_batch` instead. This will raise an error in the future!
+(PPOTrainer pid=532048) /scratch/rd2893/miniconda3/envs/pcgrl/lib/python3.7/site-packages/ray/rllib/utils/metrics/learner_info.py:110: RuntimeWarning: Mean of empty slice
+(PPOTrainer pid=532048)   return np.nanmean(tower_data)
+2022-11-30 01:48:00,310	WARNING util.py:244 -- The `callbacks.on_trial_result` operation took 9.325 s, which may be a performance bottleneck.
+2022-11-30 01:48:00,312	WARNING util.py:244 -- The `process_trial_result` operation took 9.327 s, which may be a performance bottleneck.
+2022-11-30 01:48:00,312	WARNING util.py:244 -- Processing trial results took 9.328 s, which may be a performance bottleneck. Please consider reporting results less frequently to Ray Tune.
+2022-11-30 01:48:00,312	WARNING util.py:244 -- The `process_trial_result` operation took 9.329 s, which may be a performance bottleneck.
+submitit WARNING (2022-11-30 15:02:50,633) - Caught signal SIGUSR2 on gv005.hpc.nyu.edu: this job is timed-out.
+submitit WARNING (2022-11-30 15:02:50,633) - Bypassing signal SIGCONT