Upload folder using huggingface_hub
Browse files- .gitattributes +4 -0
- .summary/0/events.out.tfevents.1714361447.41662289211e +3 -0
- README.md +42 -0
- a100_100x100.mp4 +3 -0
- a20_40x40.mp4 +3 -0
- a5_25x25.mp4 +3 -0
- checkpoint_p0/best_000402516_6594822144_reward_0.764.pth +3 -0
- checkpoint_p0/checkpoint_000458168_7506624512.pth +3 -0
- checkpoint_p0/checkpoint_000458577_7513325568.pth +3 -0
- config.json +159 -0
- git.diff +0 -0
- sf_log.txt +3 -0
.gitattributes
CHANGED
@@ -33,3 +33,7 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
|
|
|
|
33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
36 |
+
a100_100x100.mp4 filter=lfs diff=lfs merge=lfs -text
|
37 |
+
a20_40x40.mp4 filter=lfs diff=lfs merge=lfs -text
|
38 |
+
a5_25x25.mp4 filter=lfs diff=lfs merge=lfs -text
|
39 |
+
sf_log.txt filter=lfs diff=lfs merge=lfs -text
|
.summary/0/events.out.tfevents.1714361447.41662289211e
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:2704d990433614c7663c1721d2aea91cd7322014b4cc93dcbea99f7d84614f4d
|
3 |
+
size 106763012
|
README.md
ADDED
@@ -0,0 +1,42 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
library_name: sample-factory
|
3 |
+
tags:
|
4 |
+
- deep-reinforcement-learning
|
5 |
+
- reinforcement-learning
|
6 |
+
- sample-factory
|
7 |
+
---
|
8 |
+
|
9 |
+
A(n) **APPO** model trained on the **GDY-PowerGrid** environment.
|
10 |
+
|
11 |
+
This model was trained using Sample-Factory 2.0: https://github.com/alex-petrenko/sample-factory.
|
12 |
+
Documentation for how to use Sample-Factory can be found at https://www.samplefactory.dev/
|
13 |
+
|
14 |
+
|
15 |
+
## Downloading the model
|
16 |
+
|
17 |
+
After installing Sample-Factory, download the model with:
|
18 |
+
```
|
19 |
+
python -m sample_factory.huggingface.load_from_hub -r metta-ai/baseline.v0.1.1
|
20 |
+
```
|
21 |
+
|
22 |
+
|
23 |
+
## Using the model
|
24 |
+
|
25 |
+
To run the model after download, use the `enjoy` script corresponding to this environment:
|
26 |
+
```
|
27 |
+
python -m <path.to.enjoy.module> --algo=APPO --env=GDY-PowerGrid --train_dir=./train_dir --experiment=baseline.v0.1.1
|
28 |
+
```
|
29 |
+
|
30 |
+
|
31 |
+
You can also upload models to the Hugging Face Hub using the same script with the `--push_to_hub` flag.
|
32 |
+
See https://www.samplefactory.dev/10-huggingface/huggingface/ for more details
|
33 |
+
|
34 |
+
## Training with this model
|
35 |
+
|
36 |
+
To continue training with this model, use the `train` script corresponding to this environment:
|
37 |
+
```
|
38 |
+
python -m <path.to.train.module> --algo=APPO --env=GDY-PowerGrid --train_dir=./train_dir --experiment=baseline.v0.1.1 --restart_behavior=resume --train_for_env_steps=10000000000
|
39 |
+
```
|
40 |
+
|
41 |
+
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.
|
42 |
+
|
a100_100x100.mp4
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f70ac6b92420b99fc470b6ae3f6f269a0355f4f0fb4a217a3fb81383423eb193
|
3 |
+
size 11696364
|
a20_40x40.mp4
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:169a49f50570f1429884251272b83eb33d3d7eb4764db1a7576858cc8c1904af
|
3 |
+
size 6315632
|
a5_25x25.mp4
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:00b3815cabba23999871e455de7c0cda46893d8628764e64a1ce98484b137263
|
3 |
+
size 2020125
|
checkpoint_p0/best_000402516_6594822144_reward_0.764.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d23f4e7b67ccddc6879aed47a4d62d8ec05a3c072bbe2e0289d7b80239c5e0fa
|
3 |
+
size 38887866
|
checkpoint_p0/checkpoint_000458168_7506624512.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:15e0c7c5feb861721f044f91e794d8bcfa99fe253b3155ab83cf8b38bdd47b7c
|
3 |
+
size 38889406
|
checkpoint_p0/checkpoint_000458577_7513325568.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:8c8f02dc495b07baed962f73df68d060c28d7cf68324b76d1da6edd066bcabdf
|
3 |
+
size 38889406
|
config.json
ADDED
@@ -0,0 +1,159 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"help": false,
|
3 |
+
"algo": "APPO",
|
4 |
+
"env": "GDY-PowerGrid",
|
5 |
+
"experiment": "p2.fa.s1",
|
6 |
+
"train_dir": "/workspace/metta/train_dir",
|
7 |
+
"restart_behavior": "resume",
|
8 |
+
"device": "gpu",
|
9 |
+
"seed": 1,
|
10 |
+
"num_policies": 1,
|
11 |
+
"async_rl": true,
|
12 |
+
"serial_mode": false,
|
13 |
+
"batched_sampling": false,
|
14 |
+
"num_batches_to_accumulate": 2,
|
15 |
+
"worker_num_splits": 2,
|
16 |
+
"policy_workers_per_policy": 1,
|
17 |
+
"max_policy_lag": 50,
|
18 |
+
"num_workers": 32,
|
19 |
+
"num_envs_per_worker": 2,
|
20 |
+
"batch_size": 16384,
|
21 |
+
"num_batches_per_epoch": 1,
|
22 |
+
"num_epochs": 1,
|
23 |
+
"rollout": 256,
|
24 |
+
"recurrence": 256,
|
25 |
+
"shuffle_minibatches": false,
|
26 |
+
"gamma": 0.99,
|
27 |
+
"reward_scale": 1.0,
|
28 |
+
"reward_clip": 1000.0,
|
29 |
+
"value_bootstrap": false,
|
30 |
+
"normalize_returns": true,
|
31 |
+
"exploration_loss_coeff": 0.002,
|
32 |
+
"value_loss_coeff": 0.976,
|
33 |
+
"kl_loss_coeff": 0.0,
|
34 |
+
"aux_loss_coeff": 0.0,
|
35 |
+
"exploration_loss": "symmetric_kl",
|
36 |
+
"gae_lambda": 0.95,
|
37 |
+
"ppo_clip_ratio": 0.1,
|
38 |
+
"ppo_clip_value": 1.0,
|
39 |
+
"with_vtrace": false,
|
40 |
+
"vtrace_rho": 1.0,
|
41 |
+
"vtrace_c": 1.0,
|
42 |
+
"optimizer": "adam",
|
43 |
+
"adam_eps": 1e-06,
|
44 |
+
"adam_beta1": 0.9,
|
45 |
+
"adam_beta2": 0.999,
|
46 |
+
"max_grad_norm": 4.0,
|
47 |
+
"learning_rate": 0.0001,
|
48 |
+
"lr_schedule": "constant",
|
49 |
+
"lr_schedule_kl_threshold": 0.008,
|
50 |
+
"lr_adaptive_min": 1e-06,
|
51 |
+
"lr_adaptive_max": 0.01,
|
52 |
+
"obs_subtract_mean": 0.0,
|
53 |
+
"obs_scale": 1.0,
|
54 |
+
"normalize_input": false,
|
55 |
+
"normalize_input_keys": null,
|
56 |
+
"decorrelate_experience_max_seconds": 150,
|
57 |
+
"decorrelate_envs_on_one_worker": true,
|
58 |
+
"actor_worker_gpus": [],
|
59 |
+
"set_workers_cpu_affinity": true,
|
60 |
+
"force_envs_single_thread": false,
|
61 |
+
"default_niceness": 0,
|
62 |
+
"log_to_file": true,
|
63 |
+
"experiment_summaries_interval": 10,
|
64 |
+
"flush_summaries_interval": 30,
|
65 |
+
"stats_avg": 100,
|
66 |
+
"summaries_use_frameskip": true,
|
67 |
+
"heartbeat_interval": 20,
|
68 |
+
"heartbeat_reporting_interval": 180,
|
69 |
+
"train_for_env_steps": 9999999999999,
|
70 |
+
"train_for_seconds": 10000000000,
|
71 |
+
"save_every_sec": 120,
|
72 |
+
"keep_checkpoints": 2,
|
73 |
+
"load_checkpoint_path": null,
|
74 |
+
"init_checkpoint_path": null,
|
75 |
+
"load_checkpoint_kind": "latest",
|
76 |
+
"save_milestones_sec": -1,
|
77 |
+
"save_best_every_sec": 5,
|
78 |
+
"save_best_metric": "reward",
|
79 |
+
"save_best_after": 100000,
|
80 |
+
"benchmark": false,
|
81 |
+
"encoder_mlp_layers": [
|
82 |
+
512,
|
83 |
+
512
|
84 |
+
],
|
85 |
+
"encoder_conv_architecture": "convnet_simple",
|
86 |
+
"encoder_conv_mlp_layers": [
|
87 |
+
512
|
88 |
+
],
|
89 |
+
"use_rnn": true,
|
90 |
+
"rnn_size": 512,
|
91 |
+
"rnn_type": "gru",
|
92 |
+
"rnn_num_layers": 1,
|
93 |
+
"decoder_mlp_layers": [],
|
94 |
+
"nonlinearity": "elu",
|
95 |
+
"policy_initialization": "orthogonal",
|
96 |
+
"policy_init_gain": 1.0,
|
97 |
+
"actor_critic_share_weights": true,
|
98 |
+
"adaptive_stddev": true,
|
99 |
+
"continuous_tanh_scale": 0.0,
|
100 |
+
"initial_stddev": 1.0,
|
101 |
+
"use_env_info_cache": false,
|
102 |
+
"env_gpu_actions": false,
|
103 |
+
"env_gpu_observations": true,
|
104 |
+
"env_frameskip": 1,
|
105 |
+
"env_framestack": 1,
|
106 |
+
"pixel_format": "CHW",
|
107 |
+
"use_record_episode_statistics": false,
|
108 |
+
"episode_counter": false,
|
109 |
+
"with_wandb": true,
|
110 |
+
"wandb_user": "platypus",
|
111 |
+
"wandb_project": "metta",
|
112 |
+
"wandb_group": null,
|
113 |
+
"wandb_job_type": "SF",
|
114 |
+
"wandb_tags": [],
|
115 |
+
"with_pbt": false,
|
116 |
+
"pbt_mix_policies_in_one_env": true,
|
117 |
+
"pbt_period_env_steps": 5000000,
|
118 |
+
"pbt_start_mutation": 20000000,
|
119 |
+
"pbt_replace_fraction": 0.3,
|
120 |
+
"pbt_mutation_rate": 0.15,
|
121 |
+
"pbt_replace_reward_gap": 0.1,
|
122 |
+
"pbt_replace_reward_gap_absolute": 1e-06,
|
123 |
+
"pbt_optimize_gamma": false,
|
124 |
+
"pbt_target_objective": "true_objective",
|
125 |
+
"pbt_perturb_min": 1.1,
|
126 |
+
"pbt_perturb_max": 1.5,
|
127 |
+
"env_cfg": "{\"wall_density\": 0.01, \"altar:cost\": 100, \"cost:shield\": 0, \"num_agents\": 20, \"width\": 40, \"height\": 40, \"num_altars\": 5, \"num_chargers\": 10, \"num_generators\": 50}",
|
128 |
+
"command_line": "--aux_loss_coef=0 --recurrence=256 --rollout=256 --value_loss_coeff=0.976 --exploration_loss=symmetric_kl --exploration_loss_coeff=0.002 --policy_initialization=orthogonal --learning_rate=0.0001 --max_policy_lag=50 --nonlinearity=elu --load_checkpoint_kind=latest --normalize_input=False --batch_size=16384 --decorrelate_experience_max_seconds=150 --train_for_env_steps=9999999999999 --with_wandb=True --wandb_user=platypus --wandb_project=metta --experiment=p2.fa.s1 --seed=1 --rnn_num_layers=1 --rnn_size=512 --rnn_type=gru --env=GDY-PowerGrid --env_cfg={\"wall_density\": 0.01, \"altar:cost\": 100, \"cost:shield\": 0, \"num_agents\": 20, \"width\": 40, \"height\": 40, \"num_altars\": 5, \"num_chargers\": 10, \"num_generators\": 50}",
|
129 |
+
"cli_args": {
|
130 |
+
"env": "GDY-PowerGrid",
|
131 |
+
"experiment": "p2.fa.s1",
|
132 |
+
"seed": 1,
|
133 |
+
"max_policy_lag": 50,
|
134 |
+
"batch_size": 16384,
|
135 |
+
"rollout": 256,
|
136 |
+
"recurrence": 256,
|
137 |
+
"exploration_loss_coeff": 0.002,
|
138 |
+
"value_loss_coeff": 0.976,
|
139 |
+
"aux_loss_coeff": 0.0,
|
140 |
+
"exploration_loss": "symmetric_kl",
|
141 |
+
"learning_rate": 0.0001,
|
142 |
+
"normalize_input": false,
|
143 |
+
"decorrelate_experience_max_seconds": 150,
|
144 |
+
"train_for_env_steps": 9999999999999,
|
145 |
+
"load_checkpoint_kind": "latest",
|
146 |
+
"rnn_size": 512,
|
147 |
+
"rnn_type": "gru",
|
148 |
+
"rnn_num_layers": 1,
|
149 |
+
"nonlinearity": "elu",
|
150 |
+
"policy_initialization": "orthogonal",
|
151 |
+
"with_wandb": true,
|
152 |
+
"wandb_user": "platypus",
|
153 |
+
"wandb_project": "metta",
|
154 |
+
"env_cfg": "{\"wall_density\": 0.01, \"altar:cost\": 100, \"cost:shield\": 0, \"num_agents\": 20, \"width\": 40, \"height\": 40, \"num_altars\": 5, \"num_chargers\": 10, \"num_generators\": 50}"
|
155 |
+
},
|
156 |
+
"git_hash": "1e6f44b10f61fc1cdcf59ec38035cda6e4d29622",
|
157 |
+
"git_repo_name": "https://github.com/daveey/metta.git",
|
158 |
+
"wandb_unique_id": "p2.fa.s1_20240428_203044_632178"
|
159 |
+
}
|
git.diff
ADDED
File without changes
|
sf_log.txt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a61cf9fb8bdb53026edb560c0b6c2a49fc54fc7acc0ffd02ae0e69b989970a13
|
3 |
+
size 13197292
|