Upload . with huggingface_hub

.summary/0/events.out.tfevents.1677846920.MacBook-Pro-de-Quentin.local

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19fd4e4e1c82de8301ce4c9c412f0b6dad426311d352ef041c523247b688caac
+size 2132614

README.md

@@ -5,7 +5,7 @@ tags:
 - reinforcement-learning
 - sample-factory
 model-index:
-- name: PPO
+- name: APPO
   results:
   - task:
       type: reinforcement-learning
@@ -15,12 +15,12 @@ model-index:
       type: pick-place-v2
     metrics:
     - type: mean_reward
-      value: 24.42 +/- 7.77
+      value: 33.42 +/- 41.15
       name: mean_reward
       verified: false
 ---
 
-A(n) **PPO** model trained on the **pick-place-v2** environment.
+A(n) **APPO** model trained on the **pick-place-v2** environment.
 
 This model was trained using Sample-Factory 2.0: https://github.com/alex-petrenko/sample-factory.
 Documentation for how to use Sample-Factory can be found at https://www.samplefactory.dev/
@@ -38,7 +38,7 @@ python -m sample_factory.huggingface.load_from_hub -r qgallouedec/pick-place-v2-
 
 To run the model after download, use the `enjoy` script corresponding to this environment:
 ```
-python -m enjoy --algo=PPO --env=pick-place-v2 --train_dir=./train_dir --experiment=pick-place-v2-sf
+python -m enjoy --algo=APPO --env=pick-place-v2 --train_dir=./train_dir --experiment=pick-place-v2-sf
 ```
 
 
@@ -49,7 +49,7 @@ See https://www.samplefactory.dev/10-huggingface/huggingface/ for more details
 
 To continue training with this model, use the `train` script corresponding to this environment:
 ```
-python -m train --algo=PPO --env=pick-place-v2 --train_dir=./train_dir --experiment=pick-place-v2-sf --restart_behavior=resume --train_for_env_steps=10000000000
+python -m train --algo=APPO --env=pick-place-v2 --train_dir=./train_dir --experiment=pick-place-v2-sf --restart_behavior=resume --train_for_env_steps=10000000000
 ```
 
 Note, you may have to adjust `--train_for_env_steps` to a suitably high number as the experiment will resume at the number of steps it concluded at.

checkpoint_p0/best_000003537_3621888_reward_121.338.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b5b2ed7bf97959a5fdfc962efaac7fe2b852d149752912a06bf5f92d9f45c12c
+size 22381299

checkpoint_p0/checkpoint_000008899_9112576.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ddaf2819e3a93915f2fd4f5be171075022bece916e7a6a791fee7ccba4ed9b50
+size 22381699

checkpoint_p0/checkpoint_000008996_9211904.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7626cb0fa9fccdef19cbb55a3ccfd5023aafea6402b1edf3937497fa8f403dd
+size 22381699

config.json

@@ -4,7 +4,7 @@
   "env": "pick-place-v2",
   "experiment": "pick-place-v2",
   "train_dir": "/Users/quentingallouedec/gia/data/envs/metaworld/train_dir",
-  "restart_behavior": "resume",
+  "restart_behavior": "overwrite",
   "device": "cpu",
   "seed": null,
   "num_policies": 1,
@@ -65,7 +65,7 @@
   "summaries_use_frameskip": true,
   "heartbeat_interval": 20,
   "heartbeat_reporting_interval": 180,
-  "train_for_env_steps": 10000000000,
+  "train_for_env_steps": 10000000,
   "train_for_seconds": 10000000000,
   "save_every_sec": 120,
   "keep_checkpoints": 2,
@@ -120,15 +120,18 @@
   "pbt_target_objective": "true_objective",
   "pbt_perturb_min": 1.1,
   "pbt_perturb_max": 1.5,
-  "command_line": "--env pick-place-v2 --experiment=pick-place-v2 --with_wandb=True --wandb_user=qgallouedec --wandb_project sample_facotry_metaworld",
+  "command_line": "--env pick-place-v2 --experiment=pick-place-v2 --with_wandb=True --wandb_user=qgallouedec --wandb_project sample_facotry_metaworld --train_for_env_steps=10000000 --restart_behavior overwrite --device cpu",
   "cli_args": {
     "env": "pick-place-v2",
     "experiment": "pick-place-v2",
+    "restart_behavior": "overwrite",
+    "device": "cpu",
+    "train_for_env_steps": 10000000,
     "with_wandb": true,
     "wandb_user": "qgallouedec",
     "wandb_project": "sample_facotry_metaworld"
   },
   "git_hash": "2fd3320ea3064145b39fc5f05f1959e8efbd2051",
   "git_repo_name": "https://github.com/huggingface/gia.git",
-  "wandb_unique_id": "pick-place-v2_20230303_111825_996649"
+  "wandb_unique_id": "pick-place-v2_20230303_133519_050568"
 }

git.diff

@@ -1,3 +1,176 @@
+diff --git a/data/envs/metaworld/README.md b/data/envs/metaworld/README.md
+index c8f64f6..2b2d72e 100644
+--- a/data/envs/metaworld/README.md
++++ b/data/envs/metaworld/README.md
+@@ -10,13 +10,59 @@ Command lines:
+ Train:
+ 
+ ```sh
+-python train.py --env pick-place-v2 --with_wandb=True --wandb_user=qgallouedec --wandb_project sample_facotry_metaworld
++python train.py --env pick-place-v2 --experiment=pick-place-v2 --with_wandb=True --wandb_user=qgallouedec --wandb_project sample_facotry_metaworld --train_for_env_steps=10000000 --learning_rate 0.00005 --restart_behavior overwrite
+ ```
+ 
+ Push to hub:
+ 
+ ```sh
+-python enjoy.py --algo=APPO --env=pick-place-v2 --experiment=default_experiment --train_dir=./train_dir --max_num_episodes=10 --push_to_hub --hf_repository=qgallouedec/pick-place-v2-sf --save_video --no_render --enjoy_script=enjoy --train_script=train --load_checkpoint_kind best
++python enjoy.py --algo=PPO --env=pick-place-v2 --experiment=pick-place-v2 --train_dir=./train_dir --max_num_episodes=10 --push_to_hub --hf_repository=qgallouedec/pick-place-v2-sf --save_video --no_render --enjoy_script=enjoy --train_script=train --load_checkpoint_kind best
+ ```
+ 
+-Generate dataset:
+
++Generate dataset:
++
++
++
++usage: train.py [-h] [--algo ALGO] --env ENV [--experiment EXPERIMENT] [--train_dir TRAIN_DIR] [--restart_behavior {resume,restart,overwrite}]
++                [--device {gpu,cpu}] [--seed SEED] [--num_policies NUM_POLICIES] [--async_rl ASYNC_RL] [--serial_mode SERIAL_MODE]
++                [--batched_sampling BATCHED_SAMPLING] [--num_batches_to_accumulate NUM_BATCHES_TO_ACCUMULATE]
++                [--worker_num_splits WORKER_NUM_SPLITS] [--policy_workers_per_policy POLICY_WORKERS_PER_POLICY]
++                [--max_policy_lag MAX_POLICY_LAG] [--num_workers NUM_WORKERS] [--num_envs_per_worker NUM_ENVS_PER_WORKER]
++                [--batch_size BATCH_SIZE] [--num_batches_per_epoch NUM_BATCHES_PER_EPOCH] [--num_epochs NUM_EPOCHS] [--rollout ROLLOUT]
++                [--recurrence RECURRENCE] [--shuffle_minibatches SHUFFLE_MINIBATCHES] [--gamma GAMMA] [--reward_scale REWARD_SCALE]
++                [--reward_clip REWARD_CLIP] [--value_bootstrap VALUE_BOOTSTRAP] [--normalize_returns NORMALIZE_RETURNS]
++                [--exploration_loss_coeff EXPLORATION_LOSS_COEFF] [--value_loss_coeff VALUE_LOSS_COEFF] [--kl_loss_coeff KL_LOSS_COEFF]
++                [--exploration_loss {entropy,symmetric_kl}] [--gae_lambda GAE_LAMBDA] [--ppo_clip_ratio PPO_CLIP_RATIO]
++                [--ppo_clip_value PPO_CLIP_VALUE] [--with_vtrace WITH_VTRACE] [--vtrace_rho VTRACE_RHO] [--vtrace_c VTRACE_C]
++                [--optimizer {adam,lamb}] [--adam_eps ADAM_EPS] [--adam_beta1 ADAM_BETA1] [--adam_beta2 ADAM_BETA2]
++                [--max_grad_norm MAX_GRAD_NORM] [--learning_rate LEARNING_RATE]
++                [--lr_schedule {constant,kl_adaptive_minibatch,kl_adaptive_epoch}] [--lr_schedule_kl_threshold LR_SCHEDULE_KL_THRESHOLD]
++                [--lr_adaptive_min LR_ADAPTIVE_MIN] [--lr_adaptive_max LR_ADAPTIVE_MAX] [--obs_subtract_mean OBS_SUBTRACT_MEAN]
++                [--obs_scale OBS_SCALE] [--normalize_input NORMALIZE_INPUT] [--normalize_input_keys [NORMALIZE_INPUT_KEYS ...]]
++                [--decorrelate_experience_max_seconds DECORRELATE_EXPERIENCE_MAX_SECONDS]
++                [--decorrelate_envs_on_one_worker DECORRELATE_ENVS_ON_ONE_WORKER] [--actor_worker_gpus [ACTOR_WORKER_GPUS ...]]
++                [--set_workers_cpu_affinity SET_WORKERS_CPU_AFFINITY] [--force_envs_single_thread FORCE_ENVS_SINGLE_THREAD]
++                [--default_niceness DEFAULT_NICENESS] [--log_to_file LOG_TO_FILE]
++                [--experiment_summaries_interval EXPERIMENT_SUMMARIES_INTERVAL] [--flush_summaries_interval FLUSH_SUMMARIES_INTERVAL]
++                [--stats_avg STATS_AVG] [--summaries_use_frameskip SUMMARIES_USE_FRAMESKIP] [--heartbeat_interval HEARTBEAT_INTERVAL]
++                [--heartbeat_reporting_interval HEARTBEAT_REPORTING_INTERVAL] [--train_for_env_steps TRAIN_FOR_ENV_STEPS]
++                [--train_for_seconds TRAIN_FOR_SECONDS] [--save_every_sec SAVE_EVERY_SEC] [--keep_checkpoints KEEP_CHECKPOINTS]
++                [--load_checkpoint_kind {latest,best}] [--save_milestones_sec SAVE_MILESTONES_SEC] [--save_best_every_sec SAVE_BEST_EVERY_SEC]
++                [--save_best_metric SAVE_BEST_METRIC] [--save_best_after SAVE_BEST_AFTER] [--benchmark BENCHMARK]
++                [--encoder_mlp_layers [ENCODER_MLP_LAYERS ...]]
++                [--encoder_conv_architecture {convnet_simple,convnet_impala,convnet_atari,resnet_impala}]
++                [--encoder_conv_mlp_layers [ENCODER_CONV_MLP_LAYERS ...]] [--use_rnn USE_RNN] [--rnn_size RNN_SIZE] [--rnn_type {gru,lstm}]
++                [--rnn_num_layers RNN_NUM_LAYERS] [--decoder_mlp_layers [DECODER_MLP_LAYERS ...]] [--nonlinearity {elu,relu,tanh}]
++                [--policy_initialization {orthogonal,xavier_uniform,torch_default}] [--policy_init_gain POLICY_INIT_GAIN]
++                [--actor_critic_share_weights ACTOR_CRITIC_SHARE_WEIGHTS] [--adaptive_stddev ADAPTIVE_STDDEV]
++                [--continuous_tanh_scale CONTINUOUS_TANH_SCALE] [--initial_stddev INITIAL_STDDEV] [--use_env_info_cache USE_ENV_INFO_CACHE]
++                [--env_gpu_actions ENV_GPU_ACTIONS] [--env_gpu_observations ENV_GPU_OBSERVATIONS] [--env_frameskip ENV_FRAMESKIP]
++                [--env_framestack ENV_FRAMESTACK] [--pixel_format PIXEL_FORMAT]
++                [--use_record_episode_statistics USE_RECORD_EPISODE_STATISTICS] [--with_wandb WITH_WANDB] [--wandb_user WANDB_USER]
++                [--wandb_project WANDB_PROJECT] [--wandb_group WANDB_GROUP] [--wandb_job_type WANDB_JOB_TYPE] [--wandb_tags [WANDB_TAGS ...]]
++                [--with_pbt WITH_PBT] [--pbt_mix_policies_in_one_env PBT_MIX_POLICIES_IN_ONE_ENV]
++                [--pbt_period_env_steps PBT_PERIOD_ENV_STEPS] [--pbt_start_mutation PBT_START_MUTATION]
++                [--pbt_replace_fraction PBT_REPLACE_FRACTION] [--pbt_mutation_rate PBT_MUTATION_RATE]
++                [--pbt_replace_reward_gap PBT_REPLACE_REWARD_GAP] [--pbt_replace_reward_gap_absolute PBT_REPLACE_REWARD_GAP_ABSOLUTE]
++                [--pbt_optimize_gamma PBT_OPTIMIZE_GAMMA] [--pbt_target_objective PBT_TARGET_OBJECTIVE] [--pbt_perturb_min PBT_PERTURB_MIN]
++                [--pbt_perturb_max PBT_PERTURB_MAX]
+
+diff --git a/data/envs/metaworld/enjoy.py b/data/envs/metaworld/enjoy.py
+index 56f08f7..c33c4db 100644
+--- a/data/envs/metaworld/enjoy.py
++++ b/data/envs/metaworld/enjoy.py
+@@ -1,34 +1,81 @@
+ import sys
+ from typing import Optional
+-from sample_factory.enjoy import enjoy
+ 
+-import metaworld
+ import gym
+-
++import metaworld
+ from sample_factory.cfg.arguments import parse_full_cfg, parse_sf_args
++from sample_factory.enjoy import enjoy
+ from sample_factory.envs.env_utils import register_env
+-from sample_factory.train import run_rl
+ 
+-def parse_args(argv=None, evaluation=False):
+-    # parse the command line arguments to build
+-    parser, partial_cfg = parse_sf_args(argv=argv, evaluation=evaluation)
+-#    add_custom_env_args(partial_cfg.env, parser, evaluation=evaluation)
+-#    custom_env_override_defaults(partial_cfg.env, parser)
+-    final_cfg = parse_full_cfg(parser, argv)
+-    return final_cfg
++ENV_NAMES = [
++    "assembly-v2",
++    "basketball-v2",
++    "bin-picking-v2",
++    "box-close-v2",
++    "button-press-topdown-v2",
++    "button-press-topdown-wall-v2",
++    "button-press-v2",
++    "button-press-wall-v2",
++    "coffee-button-v2",
++    "coffee-pull-v2",
++    "coffee-push-v2",
++    "dial-turn-v2",
++    "disassemble-v2",
++    "door-close-v2",
++    "door-lock-v2",
++    "door-open-v2",
++    "door-unlock-v2",
++    "hand-insert-v2",
++    "drawer-close-v2",
++    "drawer-open-v2",
++    "faucet-open-v2",
++    "faucet-close-v2",
++    "hammer-v2",
++    "handle-press-side-v2",
++    "handle-press-v2",
++    "handle-pull-side-v2",
++    "handle-pull-v2",
++    "lever-pull-v2",
++    "peg-insert-side-v2",
++    "pick-place-wall-v2",
++    "pick-out-of-hole-v2",
++    "reach-v2",
++    "push-back-v2",
++    "push-v2",
++    "pick-place-v2",
++    "plate-slide-v2",
++    "plate-slide-side-v2",
++    "plate-slide-back-v2",
++    "plate-slide-back-side-v2",
++    "peg-insert-side-v2",
++    "peg-unplug-side-v2",
++    "soccer-v2",
++    "stick-push-v2",
++    "stick-pull-v2",
++    "push-wall-v2",
++    "push-v2",
++    "reach-wall-v2",
++    "reach-v2",
++    "shelf-place-v2",
++    "sweep-into-v2",
++    "sweep-v2",
++    "window-open-v2",
++    "window-close-v2",
++]
+ 
+ 
+ def make_custom_env(full_env_name: str, cfg=None, env_config=None, render_mode: Optional[str] = None):
+-    # see the section below explaining arguments
+-    return gym.make("pick-place-v2")
++    return gym.make(full_env_name, render_mode=render_mode)
++
+ 
+ def main():
+-    """Script entry point."""
+-    register_env("pick-place-v2", make_custom_env)
+-    cfg = parse_args(evaluation=True)
++    for env_name in ENV_NAMES:
++        register_env(env_name, make_custom_env)
++    parser, _ = parse_sf_args(argv=None, evaluation=True)
++    cfg = parse_full_cfg(parser)
+     status = enjoy(cfg)
+     return status
+ 
+ 
+ if __name__ == "__main__":
+-    sys.exit(main())
+
++    sys.exit(main())
 diff --git a/data/envs/metaworld/train.py b/data/envs/metaworld/train.py
 index e01df51..f0d2bb9 100644
 --- a/data/envs/metaworld/train.py

replay.mp4

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8f9f8146633325df4e663a71d66827f81598ffa38b7bee2b3b8305e1f5810e00
-size 3419609
+oid sha256:0a5406d414d56ac16bcfc09b9daa07718d5e10d6c34e17831a93dd18d4236dc3
+size 3386466