author int64 658 755k | date stringlengths 19 19 | timezone int64 -46,800 43.2k | hash stringlengths 40 40 | message stringlengths 5 490 | mods list | language stringclasses 20 values | license stringclasses 3 values | repo stringlengths 5 68 | original_message stringlengths 12 491 |
|---|---|---|---|---|---|---|---|---|---|
154,212 | 30.08.2018 18:08:58 | -32,400 | 6a2afd389e4d32bc00f4eee5f9df508fed0490cb | add argument for trpo vector | [
{
"change_type": "MODIFY",
"old_path": "example/run_trpo_vector.py",
"new_path": "example/run_trpo_vector.py",
"diff": "@@ -45,6 +45,7 @@ parser.add_argument('--episode', type=int, default=1000000)\nparser.add_argument('--seed', type=int, default=256)\nparser.add_argument('--max_episodes', type=int,... | Python | MIT License | deepx-inc/machina | add argument for trpo vector |
154,212 | 01.09.2018 16:50:15 | -32,400 | 8bda79f5be9150450dd60dded7277d1d93c14ada | fix gaussian pol | [
{
"change_type": "MODIFY",
"old_path": "machina/pols/gaussian_pol.py",
"new_path": "machina/pols/gaussian_pol.py",
"diff": "@@ -65,8 +65,8 @@ class GaussianPol(BasePol):\nmean, _, hs = self.net(obs, hs, mask)\nself.hs = hs\nelse:\n- mean, _ = self.net(obs)\n+ mean, log_std = self.net(obs)\nmean_real... | Python | MIT License | deepx-inc/machina | fix gaussian pol |
154,212 | 10.09.2018 15:02:36 | -32,400 | 26ce4a6379bbaa858579d2c3afcc90fc01bc4d93 | fix sample process | [
{
"change_type": "MODIFY",
"old_path": "machina/samplers/parallel.py",
"new_path": "machina/samplers/parallel.py",
"diff": "@@ -18,14 +18,33 @@ def one_path(env, pol, prepro=None):\no = env.reset()\nd = False\npath_length = 0\n+ if pol.rnn:\n+ hs = pol.init_hs(batch_size=1)\n+ else:\n+ hs = None\nwh... | Python | MIT License | deepx-inc/machina | fix sample process |
154,212 | 10.09.2018 20:22:38 | -32,400 | 241b35a3f62dfd914983bc6f75479e9f2869d4e1 | data parallel in pol | [
{
"change_type": "MODIFY",
"old_path": "machina/algos/ppo_clip.py",
"new_path": "machina/algos/ppo_clip.py",
"diff": "@@ -27,32 +27,21 @@ def make_pol_loss(pol, batch, clip_param, ent_beta):\nacs = batch['acs']\nadvs = batch['advs']\n- if hasattr(pol, 'rnn'):\nif pol.rnn:\ninit_hs = batch['init_hs']... | Python | MIT License | deepx-inc/machina | data parallel in pol |
154,212 | 12.09.2018 16:24:20 | -32,400 | 6ea2c6377c18b2ce9ee45167324c6fad7518a93f | add discrete | [
{
"change_type": "MODIFY",
"old_path": "example/run_ppo_vector.py",
"new_path": "example/run_ppo_vector.py",
"diff": "@@ -26,7 +26,7 @@ import gym\nimport pybullet_envs\nimport machina as mc\n-from machina.pols import GaussianPol\n+from machina.pols import GaussianPol, CategoricalPol\nfrom machina.a... | Python | MIT License | deepx-inc/machina | add discrete |
154,212 | 20.09.2018 14:58:21 | -32,400 | 806891bab5014d6595d0f725216c56d4d6368294 | downgrade pybullet for build | [
{
"change_type": "MODIFY",
"old_path": "setup.py",
"new_path": "setup.py",
"diff": "@@ -17,7 +17,7 @@ install_requires = [\n'numpy>=1.13.3',\n'terminaltables',\n'pandas',\n- 'pybullet',\n+ 'pybullet == 1.8.8',\n]\nsetup(\n@@ -28,7 +28,7 @@ setup(\nauthor='Reiji Hatsugai',\nauthor_email='reiji.hatsug... | Python | MIT License | deepx-inc/machina | downgrade pybullet for build |
154,212 | 01.10.2018 18:36:50 | -32,400 | 730a43701560eb0a47e0d149451393494f33eb91 | add deterministic sampler | [
{
"change_type": "MODIFY",
"old_path": "machina/samplers/batch.py",
"new_path": "machina/samplers/batch.py",
"diff": "@@ -9,7 +9,7 @@ class BatchSampler(BaseSampler):\ndef __init__(self, env):\nBaseSampler.__init__(self, env)\n- def one_path(self, pol, prepro=None):\n+ def one_path(self, pol, determ... | Python | MIT License | deepx-inc/machina | add deterministic sampler |
154,206 | 16.10.2018 16:21:46 | -32,400 | 7cde6b87dfd14166bc6453aea067f33391330d44 | remove DeterministicPolNet | [
{
"change_type": "MODIFY",
"old_path": "example/run_ddpg.py",
"new_path": "example/run_ddpg.py",
"diff": "@@ -32,11 +32,11 @@ from machina.algos import ddpg\nfrom machina.prepro import BasePrePro\nfrom machina.qfuncs import DeterministicQfunc\nfrom machina.envs import GymEnv\n-from machina.data impo... | Python | MIT License | deepx-inc/machina | remove DeterministicPolNet |
154,206 | 16.10.2018 16:52:33 | -32,400 | a1a6278c0cd38e12933f8d5d625b8db4dba92c6d | fix set_device | [
{
"change_type": "MODIFY",
"old_path": "example/run_ddpg.py",
"new_path": "example/run_ddpg.py",
"diff": "@@ -67,8 +67,6 @@ parser.add_argument('--gamma', type=float, default=0.99)\nparser.add_argument('--lam', type=float, default=1)\nargs = parser.parse_args()\n-args.cuda = args.cuda if torch.cuda.... | Python | MIT License | deepx-inc/machina | fix set_device |
154,212 | 05.12.2018 18:30:52 | -32,400 | 9b57b44bf7b6e46b2fb779696960c2572dab7c2a | add data class | [
{
"change_type": "MODIFY",
"old_path": "example/run_trpo.py",
"new_path": "example/run_trpo.py",
"diff": "@@ -31,7 +31,7 @@ from machina.algos import trpo\nfrom machina.prepro import BasePrePro\nfrom machina.vfuncs import NormalizedDeterministicVfunc, DeterministicVfunc\nfrom machina.envs import Gym... | Python | MIT License | deepx-inc/machina | add data class |
154,212 | 05.12.2018 19:22:15 | -32,400 | f4e1c99b44f05a133361c1f2f910f2b3fb3d3853 | add centerize and apply ppo | [
{
"change_type": "MODIFY",
"old_path": "example/run_ppo.py",
"new_path": "example/run_ppo.py",
"diff": "@@ -31,7 +31,7 @@ from machina.algos import ppo_clip, ppo_kl\nfrom machina.prepro import BasePrePro\nfrom machina.vfuncs import DeterministicVfunc\nfrom machina.envs import GymEnv\n-from machina.d... | Python | MIT License | deepx-inc/machina | add centerize and apply ppo |
154,212 | 05.12.2018 20:20:56 | -32,400 | b70edbeb05d259f59ed6a336f89857dd5d15a9a1 | add random batch sampling function in Data and apply it to ddpg | [
{
"change_type": "MODIFY",
"old_path": "example/run_ddpg.py",
"new_path": "example/run_ddpg.py",
"diff": "@@ -33,7 +33,7 @@ from machina.algos import ddpg\nfrom machina.prepro import BasePrePro\nfrom machina.qfuncs import DeterministicQfunc\nfrom machina.envs import GymEnv\n-from machina.data import... | Python | MIT License | deepx-inc/machina | add random batch sampling function in Data and apply it to ddpg |
154,212 | 05.12.2018 20:44:21 | -32,400 | d1d91e10bf086a220afb094aadc7446e5c095391 | change ppo to be able to use cuda | [
{
"change_type": "MODIFY",
"old_path": "example/run_ppo.py",
"new_path": "example/run_ppo.py",
"diff": "@@ -34,7 +34,7 @@ from machina.envs import GymEnv\nfrom machina.data import Data, compute_vs, compute_rets, compute_advs, centerize_advs\nfrom machina.samplers import BatchSampler, ParallelSampler... | Python | MIT License | deepx-inc/machina | change ppo to be able to use cuda |
154,212 | 06.12.2018 12:16:28 | -32,400 | e44806d9b098cba8a6e8a0b6da510682f2a4fef3 | fix target and speed up random batch | [
{
"change_type": "MODIFY",
"old_path": "example/run_ddpg.py",
"new_path": "example/run_ddpg.py",
"diff": "@@ -100,10 +100,13 @@ ac_space = env.action_space\npol_net = PolNet(ob_space, ac_space, args.h1, args.h2)\nnoise = OUActionNoise(ac_space.shape)\npol = DeterministicActionNoisePol(ob_space, ac_s... | Python | MIT License | deepx-inc/machina | fix target and speed up random batch |
154,212 | 06.12.2018 21:06:48 | -32,400 | f3393b9d732da874f7848bc39e5317d36b68d6d3 | make rnn's hidden state hidden | [
{
"change_type": "MODIFY",
"old_path": "example/run_ppo.py",
"new_path": "example/run_ppo.py",
"diff": "@@ -26,7 +26,7 @@ import gym\nimport pybullet_envs\nimport machina as mc\n-from machina.pols import GaussianPol, MixtureGaussianPol\n+from machina.pols import GaussianPol, CategoricalPol\nfrom mac... | Python | MIT License | deepx-inc/machina | make rnn's hidden state hidden |
154,212 | 06.12.2018 21:54:43 | -32,400 | 4d6c23249b1108ec97e1208f058f52cc9f252eb6 | add episode iterator | [
{
"change_type": "MODIFY",
"old_path": "machina/data/data.py",
"new_path": "machina/data/data.py",
"diff": "@@ -21,10 +21,10 @@ from machina.utils import get_device\nclass Data(object):\ndef __init__(self):\nself.data_map = dict()\n- self._num_epi = 0\nself._next_id = 0\nself.current_epis = None\n+ ... | Python | MIT License | deepx-inc/machina | add episode iterator |
154,212 | 07.12.2018 20:07:47 | -32,400 | f3243067fe35c17a351cc48f5770e84858ebd3b9 | add max_grad_norm argument | [
{
"change_type": "MODIFY",
"old_path": "example/run_ppo.py",
"new_path": "example/run_ppo.py",
"diff": "@@ -62,6 +62,7 @@ parser.add_argument('--ppo_type', type=str,\nchoices=['clip', 'kl'], default='clip')\nparser.add_argument('--clip_param', type=float, default=0.2)\n+parser.add_argument('--max_gr... | Python | MIT License | deepx-inc/machina | add max_grad_norm argument |
154,212 | 08.12.2018 18:11:56 | -32,400 | fffc0055135eb92e3bdcf88d24dd133546c1b194 | delete sampler and terminate mp | [
{
"change_type": "MODIFY",
"old_path": "example/run_ppo.py",
"new_path": "example/run_ppo.py",
"diff": "@@ -181,3 +181,4 @@ while args.max_episodes > total_epi:\ntorch.save(optim_vf.state_dict(), os.path.join(\nargs.log, 'models', 'optim_vf_last.pkl'))\ndel data\n+del sampler\n"
},
{
"change... | Python | MIT License | deepx-inc/machina | delete sampler and terminate mp |
154,212 | 14.12.2018 18:03:51 | -32,400 | e1be348b4965290e4fc3bff80fd723ef816b2cbd | remove samplers, add epi_sampler | [
{
"change_type": "MODIFY",
"old_path": "example/run_ddpg.py",
"new_path": "example/run_ddpg.py",
"diff": "@@ -34,7 +34,7 @@ from machina.prepro import BasePrePro\nfrom machina.qfuncs import DeterministicQfunc\nfrom machina.envs import GymEnv\nfrom machina.data import Data, add_next_obs\n-from machin... | Python | MIT License | deepx-inc/machina | remove samplers, add epi_sampler |
154,212 | 14.12.2018 18:19:22 | -32,400 | e157020ddd79e6f23f3aa958be652b0bd66cc4f4 | max_episodes or max_samples | [
{
"change_type": "MODIFY",
"old_path": "machina/samplers/epi_sampler.py",
"new_path": "machina/samplers/epi_sampler.py",
"diff": "@@ -21,6 +21,8 @@ import torch\nimport torch.multiprocessing as mp\n+LARGE_NUMBER = 100000000\n+\ndef one_epi(env, pol, deterministic=False, prepro=None):\nif prepro is N... | Python | MIT License | deepx-inc/machina | max_episodes or max_samples |
154,212 | 14.12.2018 18:21:46 | -32,400 | 42c48156b1aa3c480705f755ad4fad15fe911416 | remove max_samples | [
{
"change_type": "MODIFY",
"old_path": "example/run_ddpg.py",
"new_path": "example/run_ddpg.py",
"diff": "@@ -48,8 +48,7 @@ parser.add_argument('--seed', type=int, default=256)\nparser.add_argument('--max_episodes', type=int, default=1000000)\nparser.add_argument('--num_parallel', type=int, default=... | Python | MIT License | deepx-inc/machina | remove max_samples |
154,212 | 17.12.2018 10:26:33 | -32,400 | ffde9e1a9522f31c4b7afda519f95a0f597e41f2 | fix prepro and move simplenet | [
{
"change_type": "MODIFY",
"old_path": "example/run_ddpg.py",
"new_path": "example/run_ddpg.py",
"diff": "@@ -37,7 +37,7 @@ from machina.data import Data, add_next_obs\nfrom machina.samplers import EpiSampler\nfrom machina.misc import logger\nfrom machina.utils import set_device, measure\n-from mach... | Python | MIT License | deepx-inc/machina | fix prepro and move simplenet |
154,212 | 17.12.2018 10:33:08 | -32,400 | 24f35d9eacf95a335828095d75a159fb7dd308f8 | fix for deterministic policy | [
{
"change_type": "MODIFY",
"old_path": "machina/pols/deterministic_action_noise_pol.py",
"new_path": "machina/pols/deterministic_action_noise_pol.py",
"diff": "@@ -34,19 +34,19 @@ class DeterministicActionNoisePol(BasePol):\ndef forward(self, obs):\nmean = self.net(obs)\n- ac = mean\n+ ac = mean[0]\... | Python | MIT License | deepx-inc/machina | fix for deterministic policy |
154,212 | 17.12.2018 10:48:02 | -32,400 | 25226d3cf00febfc35cf7c01055b929fdd180806 | deterministic policy | [
{
"change_type": "MODIFY",
"old_path": "example/run_ddpg.py",
"new_path": "example/run_ddpg.py",
"diff": "@@ -91,10 +91,11 @@ env.env.seed(args.seed)\nob_space = env.observation_space\nac_space = env.action_space\n-pol_net = PolNet(ob_space, ac_space, args.h1, args.h2)\n+pol_net = PolNet(ob_space, a... | Python | MIT License | deepx-inc/machina | deterministic policy |
154,212 | 17.12.2018 12:01:16 | -32,400 | e9c8e9fd23a78cf4d8025ab2d9d0a64fea2d2fff | add multi categorical | [
{
"change_type": "MODIFY",
"old_path": "machina/pds/categorical_pd.py",
"new_path": "machina/pds/categorical_pd.py",
"diff": "# limitations under the License.\n# ==============================================================================\n+\nimport numpy as np\nimport torch\nfrom torch.distributi... | Python | MIT License | deepx-inc/machina | add multi categorical |
154,212 | 17.12.2018 13:01:13 | -32,400 | c30c14dd21f509572e60414ba7df2fd3236dedb5 | change test env | [
{
"change_type": "MODIFY",
"old_path": "tests/test_env.py",
"new_path": "tests/test_env.py",
"diff": "@@ -23,8 +23,13 @@ from machina.envs import GymEnv, C2DEnv\ndef test_continuous2discrete():\n- continuous_env = GymEnv('LunarLanderContinuous-v2', record_video=False)\n- discrete_env = C2DEnv(contin... | Python | MIT License | deepx-inc/machina | change test env |
154,212 | 17.12.2018 14:03:12 | -32,400 | ae3caa393548fe377565be97ba5419575bf202a1 | apply to ppo, trpo | [
{
"change_type": "MODIFY",
"old_path": "example/run_ppo.py",
"new_path": "example/run_ppo.py",
"diff": "@@ -29,7 +29,7 @@ from machina.pols import GaussianPol, CategoricalPol\nfrom machina.algos import ppo_clip, ppo_kl\nfrom machina.prepro import BasePrePro\nfrom machina.vfuncs import DeterministicV... | Python | MIT License | deepx-inc/machina | apply to ppo, trpo |
154,212 | 17.12.2018 16:48:25 | -32,400 | 2b49ec97f42be33b32d038c7f63221bd63796c5b | rename value function | [
{
"change_type": "MODIFY",
"old_path": "example/run_ddpg.py",
"new_path": "example/run_ddpg.py",
"diff": "@@ -30,7 +30,7 @@ from machina.pols import DeterministicActionNoisePol\nfrom machina.noise import OUActionNoise\nfrom machina.algos import ddpg\nfrom machina.prepro import BasePrePro\n-from mach... | Python | MIT License | deepx-inc/machina | rename value function |
154,212 | 17.12.2018 17:16:21 | -32,400 | d15a0e7a645d0b866de39e96d9015393f85bd85a | fix for ddpg | [
{
"change_type": "MODIFY",
"old_path": "machina/noise/ounoise.py",
"new_path": "machina/noise/ounoise.py",
"diff": "# limitations under the License.\n# ==============================================================================\n-from machina.noise.base import BaseActionNoise\n+\nimport numpy as ... | Python | MIT License | deepx-inc/machina | fix for ddpg |
154,212 | 18.12.2018 15:04:09 | -32,400 | 3d1f22070c341c6f2c83f66fe4f6c909eedd56aa | decompose loss of algorithms and apply sac | [
{
"change_type": "MODIFY",
"old_path": "example/run_ddpg.py",
"new_path": "example/run_ddpg.py",
"diff": "# limitations under the License.\n# ==============================================================================\n+\nimport argparse\n+import copy\nimport json\nimport os\n-import copy\nfrom p... | Python | MIT License | deepx-inc/machina | decompose loss of algorithms and apply sac |
154,206 | 20.12.2018 18:51:23 | -32,400 | 9f00e6958cd544e94fe7b8d74e8679dffbe2f5b5 | rename samples and remove prepro | [
{
"change_type": "MODIFY",
"old_path": "example/run_ddpg.py",
"new_path": "example/run_ddpg.py",
"diff": "@@ -30,7 +30,6 @@ import machina as mc\nfrom machina.pols import DeterministicActionNoisePol\nfrom machina.noise import OUActionNoise\nfrom machina.algos import ddpg\n-from machina.prepro import... | Python | MIT License | deepx-inc/machina | rename samples and remove prepro |
154,212 | 21.12.2018 17:45:09 | -32,400 | 3943b5ef2da982ec23195447edfea45dbe0c697f | fix updating target network | [
{
"change_type": "MODIFY",
"old_path": "machina/algos/ddpg.py",
"new_path": "machina/algos/ddpg.py",
"diff": "@@ -28,9 +28,11 @@ def train(traj,\npol_loss.backward()\noptim_pol.step()\n- for q, targ_q, p, targ_p in zip(qf.parameters(), targ_qf.parameters(), pol.parameters(), targ_pol.parameters()):\... | Python | MIT License | deepx-inc/machina | fix updating target network |
154,212 | 21.12.2018 18:47:15 | -32,400 | 11434c6cb0ee6d1fa846c4faad017a8c7ba99eb5 | remove test of discrete rnn policy with cpu test. Error is discribed at | [
{
"change_type": "MODIFY",
"old_path": "tests/test_algos.py",
"new_path": "tests/test_algos.py",
"diff": "@@ -167,6 +167,139 @@ class TestPPODiscrete(unittest.TestCase):\ndel sampler\n+class TestTRPOContinuous(unittest.TestCase):\n+ def setUp(self):\n+ self.env = GymEnv('Pendulum-v0')\n+\n+ def test... | Python | MIT License | deepx-inc/machina | remove test of discrete rnn policy with cpu test. Error is discribed at #74 |
154,209 | 05.01.2019 16:21:55 | -32,400 | 0beb48e16ddca5300b97e691428b795f967f3d74 | add prioritized experience replay | [
{
"change_type": "ADD",
"old_path": null,
"new_path": "example/run_prioritized_ddpg.py",
"diff": "+# Copyright 2018 DeepX Inc. All Rights Reserved.\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You m... | Python | MIT License | deepx-inc/machina | add prioritized experience replay |
154,209 | 06.01.2019 13:52:27 | -32,400 | 500f0bade07d1a32923600625065ccbef05b5cf1 | fix bellman loss | [
{
"change_type": "MODIFY",
"old_path": "machina/algos/prioritized_ddpg.py",
"new_path": "machina/algos/prioritized_ddpg.py",
"diff": "@@ -35,8 +35,9 @@ def train(traj,\nqf_losses = []\nlogger.log(\"Optimizing...\")\nfor batch, indices in traj.prioritized_random_batch(batch_size, epoch, return_indice... | Python | MIT License | deepx-inc/machina | fix bellman loss |
154,209 | 06.01.2019 14:47:47 | -32,400 | 15cf9d8bc9e9381e592274012d9c56e40b3925bb | fix initial priorities | [
{
"change_type": "MODIFY",
"old_path": "example/run_prioritized_ddpg.py",
"new_path": "example/run_prioritized_ddpg.py",
"diff": "@@ -122,7 +122,8 @@ while args.max_episodes > total_epi:\non_traj.add_epis(epis)\non_traj = ef.add_next_obs(on_traj)\n- on_traj = ef.compute_pris(on_traj, qf, targ_qf, ta... | Python | MIT License | deepx-inc/machina | fix initial priorities |
154,209 | 06.01.2019 17:07:44 | -32,400 | e740b3e93f2fb6c9bcb1d80e9435a20e6428fe09 | add importance-sampling weights | [
{
"change_type": "MODIFY",
"old_path": "machina/traj/traj.py",
"new_path": "machina/traj/traj.py",
"diff": "@@ -139,8 +139,20 @@ class Traj(object):\nelse:\nreturn data_map\n- def prioritized_random_batch_once(self, batch_size, return_indices=False):\n- indices = torch.utils.data.sampler.WeightedRan... | Python | MIT License | deepx-inc/machina | add importance-sampling weights |
154,209 | 06.01.2019 19:00:39 | -32,400 | b83231778187910be17e279fc849040abee2e3c3 | add rank-based priority | [
{
"change_type": "MODIFY",
"old_path": "machina/traj/traj.py",
"new_path": "machina/traj/traj.py",
"diff": "@@ -139,7 +139,7 @@ class Traj(object):\nelse:\nreturn data_map\n- def prioritized_random_batch_once(self, batch_size, return_indices=False, init_beta=0.4, beta_step=0.00025/4):\n+ def priorit... | Python | MIT License | deepx-inc/machina | add rank-based priority |
154,209 | 06.01.2019 19:15:49 | -32,400 | 2ffbec4a320d15a8b74a251f2c204da489b1ed9a | fix detach timing of priority | [
{
"change_type": "MODIFY",
"old_path": "machina/traj/traj.py",
"new_path": "machina/traj/traj.py",
"diff": "@@ -147,7 +147,7 @@ class Traj(object):\nelse:\nself.pri_beta += beta_step\n- pris = self.data_map['pris'].detach().cpu().numpy()\n+ pris = self.data_map['pris'].cpu().numpy()\nif mode == 'ran... | Python | MIT License | deepx-inc/machina | fix detach timing of priority |
154,212 | 08.01.2019 17:05:45 | -32,400 | bf272cd9be82056869b7c71a90a64ef0dcddd6ea | apply autopep8 to examples | [
{
"change_type": "MODIFY",
"old_path": "example/run_prioritized_ddpg.py",
"new_path": "example/run_prioritized_ddpg.py",
"diff": "@@ -81,7 +81,8 @@ set_device(device)\nscore_file = os.path.join(args.log, 'progress.csv')\nlogger.add_tabular_output(score_file)\n-env = GymEnv(args.env_name, log_dir=os.... | Python | MIT License | deepx-inc/machina | apply autopep8 to examples |
154,212 | 08.01.2019 21:31:58 | -32,400 | d8904a145b70cab219fe05e91e18d0fadda84373 | [WIP]add algo, noise, pol | [
{
"change_type": "MODIFY",
"old_path": "machina/algos/ddpg.py",
"new_path": "machina/algos/ddpg.py",
"diff": "@@ -11,8 +11,41 @@ def train(traj,\npol, targ_pol, qf, targ_qf,\noptim_pol, optim_qf,\nepoch, batch_size, # optimization hypers\n- tau, gamma, lam # advantage estimation\n+ tau, gamma # adva... | Python | MIT License | deepx-inc/machina | [WIP]add algo, noise, pol |
154,212 | 25.01.2019 13:29:38 | -32,400 | 09f058d3f2139fe8cbf750f11521a59d1603fa88 | fix for latest | [
{
"change_type": "MODIFY",
"old_path": "example/machina_quick_start.ipynb",
"new_path": "example/machina_quick_start.ipynb",
"diff": "\"source\": [\n\"# train your policy\\n\",\n\"from machina.traj import epi_functional as ef\\n\",\n- \"from machina.misc import logger\\n\",\n+ \"from machina import ... | Python | MIT License | deepx-inc/machina | fix for latest |
154,212 | 25.01.2019 13:52:18 | -32,400 | d46bc10f9b89d03a6500f4b4aa7a71624a77e598 | add simplenet | [
{
"change_type": "ADD",
"old_path": null,
"new_path": "example/quickstart/simple_net.py",
"diff": "+\"\"\"\n+Examples of network architecture.\n+\"\"\"\n+\n+import torch\n+import torch.nn as nn\n+from torch.nn.init import kaiming_uniform_, uniform_\n+import torch.nn.functional as F\n+import gym\n+\n... | Python | MIT License | deepx-inc/machina | add simplenet |
154,212 | 04.02.2019 17:53:12 | -32,400 | 91cbf55826d0121c0261e77f6cb6effeeec7dee6 | add original env | [
{
"change_type": "MODIFY",
"old_path": "machina/envs/continuous2discrete_env.py",
"new_path": "machina/envs/continuous2discrete_env.py",
"diff": "@@ -26,6 +26,10 @@ class C2DEnv(object):\nself.n_bins = n_bins\nself.ac_space = gym.spaces.MultiDiscrete(\nenv.ac_space.shape[0] * [n_bins])\n+ if hasattr... | Python | MIT License | deepx-inc/machina | add original env |
154,212 | 12.02.2019 12:16:21 | -32,400 | cb97cd057956739da5387f31dfd9dfac554d5b85 | add new axis to hs | [
{
"change_type": "MODIFY",
"old_path": "machina/pols/categorical_pol.py",
"new_path": "machina/pols/categorical_pol.py",
"diff": "@@ -43,6 +43,8 @@ class CategoricalPol(BasePol):\nif hs is None:\nif self.hs is None:\nself.hs = self.net.init_hs(batch_size)\n+ if self.dp_run:\n+ self.hs = (self.hs[0].... | Python | MIT License | deepx-inc/machina | add new axis to hs |
154,209 | 12.02.2019 16:47:50 | -32,400 | 8692d481ec09b5518cc28d5e8b47d87ded0907d9 | add ModelNet class | [
{
"change_type": "MODIFY",
"old_path": "example/simple_net.py",
"new_path": "example/simple_net.py",
"diff": "@@ -103,6 +103,23 @@ class QNet(nn.Module):\nreturn self.output_layer(h)\n+class ModelNet(nn.Module):\n+ def __init__(self, ob_space, ac_space, h1=500, h2=500):\n+ super(ModelNet, self).__in... | Python | MIT License | deepx-inc/machina | add ModelNet class |
154,209 | 13.02.2019 20:18:08 | -32,400 | 60f570345d41e527f5a8d90023b81095b5285a45 | add run_mpc.py | [
{
"change_type": "ADD",
"old_path": null,
"new_path": "example/run_mpc.py",
"diff": "+\"\"\"\n+An example of Model Predictive Control.\n+\"\"\"\n+\n+import argparse\n+import json\n+import os\n+from pprint import pprint\n+\n+import numpy as np\n+import torch\n+import torch.nn as nn\n+import torch.nn.... | Python | MIT License | deepx-inc/machina | add run_mpc.py |
154,209 | 13.02.2019 20:40:37 | -32,400 | df2a3bd1ea59b0a32946d8daa6a597a6101b1adb | add normalize_obs_and_acs() | [
{
"change_type": "MODIFY",
"old_path": "machina/traj/traj_functional.py",
"new_path": "machina/traj/traj_functional.py",
"diff": "@@ -12,7 +12,18 @@ def update_pris(traj, td_loss, indices, alpha=0.6, epsilon=1e-6):\ntraj.data_map['pris'][indices] = pris.detach()\nreturn traj\n-def add_noise_to_obs(t... | Python | MIT License | deepx-inc/machina | add normalize_obs_and_acs() |
154,209 | 13.02.2019 20:46:45 | -32,400 | c4bcaa4aac064a00733352e121a0c47f083fbc8d | add normalization of obs, next_obs and acs | [
{
"change_type": "MODIFY",
"old_path": "example/run_mpc.py",
"new_path": "example/run_mpc.py",
"diff": "@@ -27,7 +27,6 @@ from machina.utils import measure\nfrom simple_net import PolNet, VNet, PolNetLSTM, VNetLSTM\n-\nclass RandomPolicy(nn.Module):\ndef __init__(self, action_space):\nsuper(RandomPo... | Python | MIT License | deepx-inc/machina | add normalization of obs, next_obs and acs |
154,209 | 14.02.2019 14:09:51 | -32,400 | c56ecf55bf4203563cadcccd001bea70fa71728e | add mpc_pol.py | [
{
"change_type": "ADD",
"old_path": null,
"new_path": "machina/pols/mpc_pol.py",
"diff": "+import numpy as np\n+import torch\n+\n+from machina.pds import DeterministicPd\n+from machina.pols import BasePol\n+from machina.utils import get_device\n+\n+\n+class MPCPol(BasePol):\n+ \"\"\"\n+ Policy with ... | Python | MIT License | deepx-inc/machina | add mpc_pol.py |
154,209 | 14.02.2019 15:33:18 | -32,400 | 2b621ade4463e0aafdd6f16afaf7ac136dec3786 | add reward function | [
{
"change_type": "MODIFY",
"old_path": "example/run_mpc.py",
"new_path": "example/run_mpc.py",
"diff": "@@ -43,7 +43,7 @@ class RandomPolicy(nn.Module):\nreturn ac_real, ac, dict(mean=mean)\n-def add_noise_to_init_ob(data, std):\n+def add_noise_to_init_obs(data, std):\nepis = data.current_epis\nwith... | Python | MIT License | deepx-inc/machina | add reward function |
154,209 | 14.02.2019 15:48:37 | -32,400 | 00ed3d962f65a1d4abcd768a4710a8cceff32878 | make env correspond to str | [
{
"change_type": "MODIFY",
"old_path": "machina/pols/mpc_pol.py",
"new_path": "machina/pols/mpc_pol.py",
"diff": "@@ -40,6 +40,8 @@ class MPCPol(BasePol):\n'rnn with MPCPol is not supported now')\nBasePol.__init__(self, ob_space, ac_space, net, rnn=rnn, normalize_ac=normalize_ac,\ndata_parallel=data... | Python | MIT License | deepx-inc/machina | make env correspond to str |
154,209 | 14.02.2019 16:49:03 | -32,400 | e96f683474846f87ecfb2b7bc5468f2e545698bb | add loss function for dynamics model | [
{
"change_type": "MODIFY",
"old_path": "machina/loss_functional.py",
"new_path": "machina/loss_functional.py",
"diff": "@@ -327,3 +327,29 @@ def monte_carlo(vf, batch, clip_param=0.2, clip=False):\nelse:\nvf_loss = 0.5 * torch.mean(vfloss1 * out_masks)\nreturn vf_loss\n+\n+\n+def dynamics(dm, batch,... | Python | MIT License | deepx-inc/machina | add loss function for dynamics model |
154,209 | 14.02.2019 17:03:47 | -32,400 | 4b4831fdb5c27104e1d302be88bded64beb02703 | add train function for dynamics model | [
{
"change_type": "MODIFY",
"old_path": "machina/algos/mpc.py",
"new_path": "machina/algos/mpc.py",
"diff": "@@ -76,6 +76,49 @@ def make_kl(pol, batch):\n)\n+def update_dm(dm, optim_dm, batch):\n+ dm_loss = lf.dynamics(model, batch)\n+ optim_dm.zero_grad()\n+ dm_loss.backward()\n+ optim_dm.step()\n+\... | Python | MIT License | deepx-inc/machina | add train function for dynamics model |
154,212 | 14.02.2019 18:19:10 | -32,400 | 364a13f7aff363f93ea52a3deb175be0fa461f90 | action is included in observation | [
{
"change_type": "ADD",
"old_path": null,
"new_path": "machina/envs/ac_in_ob_env.py",
"diff": "+import gym\n+import numpy as np\n+\n+\n+class AcInObEnv(gym.Env):\n+ def __init__(self, env, dim=0, normalize=True, initial_value=0):\n+ self.env = env\n+ if hasattr(env, 'original_env'):\n+ self.original... | Python | MIT License | deepx-inc/machina | action is included in observation |
154,212 | 14.02.2019 18:20:06 | -32,400 | b29eb25735f965ae587c3d6b6f45c3af9977a9cf | inherit gym.Env | [
{
"change_type": "MODIFY",
"old_path": "machina/envs/gym_env.py",
"new_path": "machina/envs/gym_env.py",
"diff": "@@ -50,7 +50,7 @@ class NoVideoSchedule(object):\nreturn False\n-class GymEnv(object):\n+class GymEnv(gym.Env):\ndef __init__(self, env, record_video=False, video_schedule=None, log_dir=... | Python | MIT License | deepx-inc/machina | inherit gym.Env |
154,212 | 14.02.2019 18:33:49 | -32,400 | 1887c41d5971ef88507661a8c7e5bf785b188b04 | reward is included in observation 114 | [
{
"change_type": "ADD",
"old_path": null,
"new_path": "machina/envs/rew_in_ob_env.py",
"diff": "+import gym\n+import numpy as np\n+\n+\n+class RewInObEnv(gym.Env):\n+ def __init__(self, env, dim=0, normalize=True, initial_value=0, mean=0, std=1, low=-np.inf, high=np.inf):\n+ self.env = env\n+ if has... | Python | MIT License | deepx-inc/machina | reward is included in observation #113, 114 |
154,212 | 14.02.2019 19:10:43 | -32,400 | 8879429e05c454b55b2a7ceddceea0539304d813 | add ob_space to c2d | [
{
"change_type": "MODIFY",
"old_path": "machina/envs/continuous2discrete_env.py",
"new_path": "machina/envs/continuous2discrete_env.py",
"diff": "@@ -26,6 +26,7 @@ class C2DEnv(object):\nself.n_bins = n_bins\nself.ac_space = gym.spaces.MultiDiscrete(\nenv.ac_space.shape[0] * [n_bins])\n+ self.ob_spa... | Python | MIT License | deepx-inc/machina | add ob_space to c2d |
154,212 | 14.02.2019 19:11:28 | -32,400 | c913910ba3261d277a6bbf61dbe32d91af23664b | add mixed env | [
{
"change_type": "ADD",
"old_path": null,
"new_path": "example/run_mixed_env.py",
"diff": "+\"\"\"\n+An example of mixed environment with ppo.\n+\"\"\"\n+\n+import argparse\n+import json\n+import os\n+from pprint import pprint\n+\n+import numpy as np\n+import torch\n+import torch.nn as nn\n+import t... | Python | MIT License | deepx-inc/machina | add mixed env |
154,209 | 14.02.2019 19:38:32 | -32,400 | cc3ff15fe1c39c859e9a86d6fd708e0bfc918793 | add statistic to args of normalize function | [
{
"change_type": "MODIFY",
"old_path": "machina/traj/traj_functional.py",
"new_path": "machina/traj/traj_functional.py",
"diff": "@@ -13,21 +13,31 @@ def update_pris(traj, td_loss, indices, alpha=0.6, epsilon=1e-6):\nreturn traj\n-def normalize_obs_and_acs(traj):\n+def normalize_obs_and_acs(traj, me... | Python | MIT License | deepx-inc/machina | add statistic to args of normalize function |
154,209 | 14.02.2019 19:44:13 | -32,400 | 10dcac7a9fd904f73522e7887ba4690c8e817862 | add normalization of rl_traj | [
{
"change_type": "MODIFY",
"old_path": "example/run_mpc.py",
"new_path": "example/run_mpc.py",
"diff": "@@ -153,7 +153,7 @@ rand_traj_val = ef.add_next_obs(rand_traj_val)\nrand_traj_val.register_epis()\n# obs, next_obs, and acs should become mean 0, std 1\n-traj, mean_obs, std_obs, mean_next_obs, st... | Python | MIT License | deepx-inc/machina | add normalization of rl_traj |
154,209 | 14.02.2019 20:15:46 | -32,400 | 9c673ea34619fac0a6e71e627f9568a5a41220e0 | add normalization of inputs and outputs of dynamic model | [
{
"change_type": "MODIFY",
"old_path": "machina/pols/mpc_pol.py",
"new_path": "machina/pols/mpc_pol.py",
"diff": "@@ -34,7 +34,9 @@ class MPCPol(BasePol):\nSplitted dimension in data parallel.\n\"\"\"\n- def __init__(self, ob_space, ac_space, net, rew_func, env, n_samples=1000, horizon=20, rnn=False... | Python | MIT License | deepx-inc/machina | add normalization of inputs and outputs of dynamic model |
154,209 | 14.02.2019 20:24:48 | -32,400 | 6d71dd7f8c261a4c28cf5f841f8186a0029954f8 | inf to mean after normalization | [
{
"change_type": "MODIFY",
"old_path": "machina/pols/mpc_pol.py",
"new_path": "machina/pols/mpc_pol.py",
"diff": "@@ -75,6 +75,9 @@ class MPCPol(BasePol):\nfor i in range(self.horizon):\nob = (obs[i] - mean_obs) / std_obs\nac = (sample_acs[i] - mean_acs) / std_acs\n+ # inf to mean\n+ ob[ob == float(... | Python | MIT License | deepx-inc/machina | inf to mean after normalization |
154,212 | 14.02.2019 20:27:51 | -32,400 | ffe7dbcdd4ab1007d0118a06ac66d0f50d4f4d98 | remove roboschool | [
{
"change_type": "MODIFY",
"old_path": "example/run_ppo.py",
"new_path": "example/run_ppo.py",
"diff": "@@ -30,18 +30,17 @@ parser = argparse.ArgumentParser()\nparser.add_argument('--log', type=str, default='garbage')\nparser.add_argument('--env_name', type=str, default='Pendulum-v0')\nparser.add_ar... | Python | MIT License | deepx-inc/machina | remove roboschool |
154,212 | 14.02.2019 20:31:37 | -32,400 | 107cbf39f28e39375868c897ded5443bdfb94746 | add data parallel to ppo | [
{
"change_type": "MODIFY",
"old_path": "example/run_ppo.py",
"new_path": "example/run_ppo.py",
"diff": "@@ -35,6 +35,7 @@ parser.add_argument('--seed', type=int, default=256)\nparser.add_argument('--max_episodes', type=int, default=1000000)\nparser.add_argument('--num_parallel', type=int, default=4)... | Python | MIT License | deepx-inc/machina | add data parallel to ppo |
154,212 | 14.02.2019 20:37:24 | -32,400 | 965c83750b7883d9d331254d721979f85d4c40cd | add dp to sac | [
{
"change_type": "MODIFY",
"old_path": "example/run_sac.py",
"new_path": "example/run_sac.py",
"diff": "@@ -36,6 +36,7 @@ parser.add_argument('--seed', type=int, default=256)\nparser.add_argument('--max_episodes', type=int, default=1000000)\nparser.add_argument('--num_parallel', type=int, default=4)... | Python | MIT License | deepx-inc/machina | add dp to sac |
154,212 | 14.02.2019 20:51:11 | -32,400 | 021bc13f03c3f3bc1a6b34b7ef2a0d358f9450a2 | add ppo and sac | [
{
"change_type": "ADD",
"old_path": null,
"new_path": "example/run_ppo_sac.py",
"diff": "+\"\"\"\n+An example of ppo and sac\n+\"\"\"\n+\n+import argparse\n+import copy\n+import json\n+import os\n+from pprint import pprint\n+\n+import numpy as np\n+import torch\n+import torch.nn as nn\n+import torch... | Python | MIT License | deepx-inc/machina | add ppo and sac |
154,209 | 14.02.2019 20:54:24 | -32,400 | 5495784df271b833c368b7a68b416737cf3314de | add MPCPol to initialization of algos | [
{
"change_type": "MODIFY",
"old_path": "machina/pols/__init__.py",
"new_path": "machina/pols/__init__.py",
"diff": "@@ -4,3 +4,4 @@ from machina.pols.mixture_gaussian_pol import MixtureGaussianPol\nfrom machina.pols.deterministic_action_noise_pol import DeterministicActionNoisePol\nfrom machina.pols... | Python | MIT License | deepx-inc/machina | add MPCPol to initialization of algos |
154,209 | 14.02.2019 20:59:39 | -32,400 | b639705177708b6e6f490d252fd58f2394cbedbe | fix syntax and spell | [
{
"change_type": "MODIFY",
"old_path": "example/run_mpc.py",
"new_path": "example/run_mpc.py",
"diff": "@@ -15,7 +15,7 @@ import gym\nimport machina as mc\nfrom machina.pols import GaussianPol, CategoricalPol, MultiCategoricalPol, MPCPol\n-from machina.algos import mpo\n+from machina.algos import mp... | Python | MIT License | deepx-inc/machina | fix syntax and spell |
154,209 | 14.02.2019 21:01:28 | -32,400 | 1a6d4854583a89337e8367f6d3fb0aec493ee414 | correspond to roboschool | [
{
"change_type": "MODIFY",
"old_path": "example/run_mpc.py",
"new_path": "example/run_mpc.py",
"diff": "@@ -105,6 +105,9 @@ if not os.path.exists(os.path.join(args.log, 'models')):\nnp.random.seed(args.seed)\ntorch.manual_seed(args.seed)\n+if args.roboschool:\n+ import roboschool\n+\nscore_file = os... | Python | MIT License | deepx-inc/machina | correspond to roboschool |
154,209 | 14.02.2019 21:20:14 | -32,400 | 4df50a85bce63ed5b01c375fac45232e980876b1 | mave RandomPol from run_mpo.py to random_pol.py | [
{
"change_type": "MODIFY",
"old_path": "example/run_mpc.py",
"new_path": "example/run_mpc.py",
"diff": "@@ -14,7 +14,7 @@ import torch.nn.functional as F\nimport gym\nimport machina as mc\n-from machina.pols import GaussianPol, CategoricalPol, MultiCategoricalPol, MPCPol\n+from machina.pols import G... | Python | MIT License | deepx-inc/machina | mave RandomPol from run_mpo.py to random_pol.py |
154,209 | 14.02.2019 21:42:27 | -32,400 | 6ff0a293dd2b3909adc658a9edd5bca812457777 | fix std of noise to init obs | [
{
"change_type": "MODIFY",
"old_path": "example/run_mpc.py",
"new_path": "example/run_mpc.py",
"diff": "@@ -28,13 +28,11 @@ from machina.utils import measure\nfrom simple_net import PolNet, VNet, ModelNet, PolNetLSTM, VNetLSTM\n-def add_noise_to_init_obs(data, std):\n- epis = data.current_epis\n+def... | Python | MIT License | deepx-inc/machina | fix std of noise to init obs |
154,209 | 14.02.2019 21:46:00 | -32,400 | bbeb8d23aaf093d6421c1b317ef1d27a6e95473d | fix name of traj | [
{
"change_type": "MODIFY",
"old_path": "example/run_mpc.py",
"new_path": "example/run_mpc.py",
"diff": "@@ -131,21 +131,21 @@ sampler = EpiSampler(\nepis = sampler.sample(random_pol, max_episodes=args.num_rollouts_train)\nepis = add_noise_to_init_obs(epis, args.noise_to_init_obs)\n-rand_traj_train =... | Python | MIT License | deepx-inc/machina | fix name of traj |
154,209 | 14.02.2019 22:02:38 | -32,400 | d522cf27c154ac0d610f0773df12219095ebcc45 | fix size of inf mask | [
{
"change_type": "MODIFY",
"old_path": "machina/traj/traj_functional.py",
"new_path": "machina/traj/traj_functional.py",
"diff": "@@ -38,14 +38,20 @@ def normalize_obs_and_acs(traj, mean_obs=None, std_obs=None, mean_acs=None, std_\ntraj.data_map['next_obs'] - mean_next_obs) / std_next_obs\n# inf to ... | Python | MIT License | deepx-inc/machina | fix size of inf mask |
154,209 | 15.02.2019 10:19:58 | -32,400 | c76528b7e08cd6fc534c639e2ae6203338fcdc00 | enable cuda mode | [
{
"change_type": "MODIFY",
"old_path": "example/run_mpc.py",
"new_path": "example/run_mpc.py",
"diff": "@@ -23,7 +23,7 @@ from machina.traj import epi_functional as ef\nfrom machina.traj import traj_functional as tf\nfrom machina.samplers import EpiSampler\nfrom machina import logger\n-from machina.... | Python | MIT License | deepx-inc/machina | enable cuda mode |
154,209 | 15.02.2019 16:36:34 | -32,400 | 70bebddbb5682639cf6a88b17340fdca6e29a637 | fix inputs of reward function | [
{
"change_type": "MODIFY",
"old_path": "machina/pols/mpc_pol.py",
"new_path": "machina/pols/mpc_pol.py",
"diff": "@@ -79,8 +79,8 @@ class MPCPol(BasePol):\nob[ob == float('inf')] = mean_obs[ob == float('inf')]\nac[ac == float('inf')] = mean_acs[ac == float('inf')]\nnext_ob = ob + self.net(ob, ac)\n-... | Python | MIT License | deepx-inc/machina | fix inputs of reward function |
154,209 | 15.02.2019 17:33:46 | -32,400 | 492a89d49c18b5619de2d7d0559cd72d63860031 | fix hyperperameter | [
{
"change_type": "MODIFY",
"old_path": "example/run_mpc.py",
"new_path": "example/run_mpc.py",
"diff": "@@ -61,9 +61,10 @@ parser.add_argument('--num_rollouts_val', type=int, default=20)\nparser.add_argument('--noise_to_init_obs', type=float, default=0.001)\nparser.add_argument('--n_samples', type=i... | Python | MIT License | deepx-inc/machina | fix hyperperameter |
154,209 | 15.02.2019 18:12:50 | -32,400 | 247f179d2faf7aa468470780c35fd0162e9607fa | split traj to rand_traj and rl_traj | [
{
"change_type": "MODIFY",
"old_path": "example/run_mpc.py",
"new_path": "example/run_mpc.py",
"diff": "@@ -132,21 +132,23 @@ sampler = EpiSampler(\nepis = sampler.sample(random_pol, max_episodes=args.num_rollouts_train)\nepis = add_noise_to_init_obs(epis, args.noise_to_init_obs)\n-traj_train = Traj... | Python | MIT License | deepx-inc/machina | split traj to rand_traj and rl_traj |
154,212 | 15.02.2019 19:46:39 | -32,400 | 666cfdc370c0b796331c9df06526172f25b32d00 | normalize pg_weight | [
{
"change_type": "MODIFY",
"old_path": "machina/loss_functional.py",
"new_path": "machina/loss_functional.py",
"diff": "@@ -165,7 +165,7 @@ def bellman(qf, targ_qf, targ_pol, batch, gamma, continuous=True, deterministic=\n\"Only Q function with continuous action space is supported now.\")\n-def sac(... | Python | MIT License | deepx-inc/machina | normalize pg_weight |
154,209 | 16.02.2019 11:08:03 | -32,400 | 9c2b2900f7daf01f06a30b2bd28f31755f50d791 | delete rand_traj_val | [
{
"change_type": "MODIFY",
"old_path": "example/run_mpc.py",
"new_path": "example/run_mpc.py",
"diff": "@@ -124,13 +124,6 @@ rand_traj.add_epis(epis)\nrand_traj = ef.add_next_obs(rand_traj)\nrand_traj.register_epis()\n-epis = sampler.sample(random_pol, max_episodes=args.num_rollouts_val)\n-epis = ad... | Python | MIT License | deepx-inc/machina | delete rand_traj_val |
154,209 | 16.02.2019 11:15:43 | -32,400 | 72955e0809c2cd76880f9a1ec9467b17ff493c68 | change architecture of ModelNet | [
{
"change_type": "MODIFY",
"old_path": "example/simple_net.py",
"new_path": "example/simple_net.py",
"diff": "@@ -106,16 +106,16 @@ class QNet(nn.Module):\nclass ModelNet(nn.Module):\ndef __init__(self, ob_space, ac_space, h1=500, h2=500):\nsuper(ModelNet, self).__init__()\n- self.fc1 = nn.Linear(ob... | Python | MIT License | deepx-inc/machina | change architecture of ModelNet |
154,209 | 16.02.2019 13:45:45 | -32,400 | 93cd94b52d781a3f5b0a8bb974ef4a42d7f175ea | fix trainig epoch | [
{
"change_type": "MODIFY",
"old_path": "example/run_mpc.py",
"new_path": "example/run_mpc.py",
"diff": "@@ -38,9 +38,15 @@ def add_noise_to_init_obs(epis, std):\ndef rew_func(next_obs, acs):\n# HarfCheetah\nindex_of_velx = 3\n+ if isinstance(next_obs, torch.Tensor):\nrews = next_obs[:, index_of_velx... | Python | MIT License | deepx-inc/machina | fix trainig epoch |
154,209 | 16.02.2019 15:51:10 | -32,400 | 54be502c5b8f8e52630053c3dc23d9954dec9838 | fix bug of denormalization | [
{
"change_type": "MODIFY",
"old_path": "machina/pols/mpc_pol.py",
"new_path": "machina/pols/mpc_pol.py",
"diff": "@@ -79,7 +79,7 @@ class MPCPol(BasePol):\nob[ob == float('inf')] = mean_obs[ob == float('inf')]\nac[ac == float('inf')] = mean_acs[ac == float('inf')]\nnext_ob = ob + self.net(ob, ac)\n-... | Python | MIT License | deepx-inc/machina | fix bug of denormalization |
154,209 | 16.02.2019 20:10:06 | -32,400 | 60b5a3a7f890400ac332f859017ffde7e5a5c6e0 | split sampler to rand_rampler and rl_sampler, and fix bugs | [
{
"change_type": "MODIFY",
"old_path": "example/run_mpc.py",
"new_path": "example/run_mpc.py",
"diff": "@@ -39,12 +39,12 @@ def rew_func(next_obs, acs):\n# HarfCheetah\nindex_of_velx = 3\nif isinstance(next_obs, torch.Tensor):\n- rews = next_obs[:, index_of_velx] + 0.05 * \\\n- torch.sum(acs**2, dim... | Python | MIT License | deepx-inc/machina | split sampler to rand_rampler and rl_sampler, and fix bugs |
154,209 | 16.02.2019 21:22:49 | -32,400 | 28be487965cc5865290b463e25e372ea8ef38691 | fix init obs on the mpc | [
{
"change_type": "MODIFY",
"old_path": "example/run_mpc.py",
"new_path": "example/run_mpc.py",
"diff": "@@ -39,12 +39,12 @@ def rew_func(next_obs, acs):\n# HarfCheetah\nindex_of_velx = 3\nif isinstance(next_obs, torch.Tensor):\n- rews = next_obs[:, index_of_velx] # - 0.05 * \\\n- #torch.sum(acs**2, ... | Python | MIT License | deepx-inc/machina | fix init obs on the mpc |
154,209 | 16.02.2019 22:08:44 | -32,400 | b44709763efb1163f4cbb63ac862cacc41066377 | change nan to mean to nan to zero | [
{
"change_type": "MODIFY",
"old_path": "machina/pols/mpc_pol.py",
"new_path": "machina/pols/mpc_pol.py",
"diff": "@@ -80,8 +80,10 @@ class MPCPol(BasePol):\nob = (obs[i] - self.mean_obs) / self.std_obs\nac = (sample_acs[i] - self.mean_acs) / self.std_acs\n# inf to mean\n- ob[ob == float('inf')] = se... | Python | MIT License | deepx-inc/machina | change nan to mean to nan to zero |
154,209 | 17.02.2019 01:35:27 | -32,400 | 3500365e64ffbb64ae22dc8a15efd534f74ed5cc | apply mpc_pol to cpu | [
{
"change_type": "MODIFY",
"old_path": "machina/pols/mpc_pol.py",
"new_path": "machina/pols/mpc_pol.py",
"diff": "@@ -48,13 +48,13 @@ class MPCPol(BasePol):\nself.rew_func = rew_func\nself.n_samples = n_samples\nself.horizon = horizon\n- self.mean_obs = mean_obs.repeat(n_samples, 1)\n- self.std_obs ... | Python | MIT License | deepx-inc/machina | apply mpc_pol to cpu |
154,209 | 17.02.2019 03:49:27 | -32,400 | 692401647d90ecf2fa6a09bbecf2564f56216484 | move train after sample | [
{
"change_type": "MODIFY",
"old_path": "example/run_mpc.py",
"new_path": "example/run_mpc.py",
"diff": "+import pybullet_envs\n\"\"\"\nAn example of Model Predictive Control.\n\"\"\"\n@@ -38,28 +39,28 @@ def add_noise_to_init_obs(epis, std):\ndef rew_func(next_obs, acs):\n# HarfCheetah\nindex_of_vel... | Python | MIT License | deepx-inc/machina | move train after sample |
154,209 | 17.02.2019 20:59:01 | -32,400 | a189d55a5d0f18e2d8bc99ec6db41e044e00e47c | delete env in policy | [
{
"change_type": "MODIFY",
"old_path": "example/run_mpc.py",
"new_path": "example/run_mpc.py",
"diff": "@@ -136,7 +136,7 @@ rl_traj = Traj()\n# initialize dynamics model and mpc policy\ndyn_model = ModelNet(ob_space, ac_space)\n-mpc_pol = MPCPol(ob_space, ac_space, dyn_model, rew_func, env,\n+mpc_po... | Python | MIT License | deepx-inc/machina | delete env in policy |
154,209 | 17.02.2019 21:05:16 | -32,400 | bb795c30d3f32fc609cec38c3fcf2349863dc3dd | replace std tp mean | [
{
"change_type": "MODIFY",
"old_path": "machina/traj/traj_functional.py",
"new_path": "machina/traj/traj_functional.py",
"diff": "@@ -25,7 +25,7 @@ def normalize_obs_and_acs(traj, mean_obs=None, std_obs=None, mean_acs=None, std_\nif std_acs is None:\nstd_acs = torch.std(traj.data_map['acs'], dim=0, ... | Python | MIT License | deepx-inc/machina | replace std tp mean |
154,206 | 18.02.2019 08:52:08 | -32,400 | 6ae8bf4d17f366b90edc02f03322e5b102d22727 | Delete airl.py | [
{
"change_type": "DELETE",
"old_path": "machina/pds/airl.py",
"new_path": null,
"diff": "-# Copyright 2018 DeepX Inc. All Rights Reserved.\n-#\n-# Licensed under the Apache License, Version 2.0 (the \"License\");\n-# you may not use this file except in compliance with the License.\n-# You may obtain... | Python | MIT License | deepx-inc/machina | Delete airl.py |
154,209 | 18.02.2019 10:27:03 | -32,400 | 02023f46fb5e1d725ad549c36bdf28612952e57a | delete mean and atd of next_obs | [
{
"change_type": "MODIFY",
"old_path": "example/run_mpc.py",
"new_path": "example/run_mpc.py",
"diff": "@@ -39,12 +39,12 @@ def rew_func(next_obs, acs):\n# HarfCheetah\nindex_of_velx = 3\nif isinstance(next_obs, np.ndarray):\n- rews = next_obs[:, index_of_velx] # - 0.05 * \\\n- #np.sum(acs**2, axis=... | Python | MIT License | deepx-inc/machina | delete mean and atd of next_obs |
154,206 | 18.02.2019 12:43:21 | -32,400 | f775364b36d4db4a4c66b78def576440b37d45de | Use DeterministicPd | [
{
"change_type": "MODIFY",
"old_path": ".gitignore",
"new_path": ".gitignore",
"diff": "@@ -7,3 +7,7 @@ docs/_build/\n.ipynb_checkpoints\n.python-version\n.idea/\n+*.png\n+*.json\n+*.pkl\n+*.csv\n"
},
{
"change_type": "MODIFY",
"old_path": "machina/algos/behavior_clone.py",
"new_path... | Python | MIT License | deepx-inc/machina | Use DeterministicPd |
154,206 | 18.02.2019 12:43:45 | -32,400 | 0599fe4de368565a55c6c9fa8d769260f0ac1d9c | add record_results for bc | [
{
"change_type": "MODIFY",
"old_path": "example/run_behavior_clone.py",
"new_path": "example/run_behavior_clone.py",
"diff": "@@ -22,16 +22,10 @@ import pickle\nimport numpy as np\nimport torch\n-import torch.nn as nn\n-import torch.nn.functional as F\nimport gym\n-import pybullet_envs\n-import mach... | Python | MIT License | deepx-inc/machina | add record_results for bc |
154,206 | 18.02.2019 12:49:19 | -32,400 | 4ea69bc4e904193ea966d70966511c1df4165fb9 | Rename the method 'step' to 'iterate_step' | [
{
"change_type": "MODIFY",
"old_path": "machina/traj/traj.py",
"new_path": "machina/traj/traj.py",
"diff": "@@ -156,7 +156,7 @@ class Traj(object):\nyield self._next_batch(batch_size, indices)\nself._next_id = 0\n- def step(self, batch_size, step=1, indices=None, shuffle=True):\n+ def iterate_step(s... | Python | MIT License | deepx-inc/machina | Rename the method 'step' to 'iterate_step' |
154,206 | 18.02.2019 14:30:18 | -32,400 | c28d5c0e91f80881cae628935380825c9ce6f635 | Remove Apache Liscense and autopep | [
{
"change_type": "MODIFY",
"old_path": "example/run_behavior_clone.py",
"new_path": "example/run_behavior_clone.py",
"diff": "-# Copyright 2018 DeepX Inc. All Rights Reserved.\n-#\n-# Licensed under the Apache License, Version 2.0 (the \"License\");\n-# you may not use this file except in compliance... | Python | MIT License | deepx-inc/machina | Remove Apache Liscense and autopep |
154,209 | 18.02.2019 15:59:33 | -32,400 | 0ae6dd26988003c7a8df5f5239cabacaca0eb3e1 | add small number eps | [
{
"change_type": "MODIFY",
"old_path": "machina/pols/mpc_pol.py",
"new_path": "machina/pols/mpc_pol.py",
"diff": "@@ -80,9 +80,6 @@ class MPCPol(BasePol):\nfor i in range(self.horizon):\nob = (obs[i] - self.mean_obs) / self.std_obs\nac = (sample_acs[i] - self.mean_acs) / self.std_acs\n- # inf to zer... | Python | MIT License | deepx-inc/machina | add small number eps |
154,206 | 18.02.2019 17:46:47 | -32,400 | ab6c871929ba4c8d033071635c777d15a9fdf838 | Fix name traj.step to traj.iterate_step | [
{
"change_type": "MODIFY",
"old_path": "machina/algos/gail.py",
"new_path": "machina/algos/gail.py",
"diff": "@@ -106,9 +106,9 @@ def train(agent_traj, expert_traj, pol, vf, discrim,\nelse:\nnew_kl_beta = kl_beta\n- agent_iterator = agent_traj.step(\n+ agent_iterator = agent_traj.iterate_step(\nbatc... | Python | MIT License | deepx-inc/machina | Fix name traj.step to traj.iterate_step |
154,206 | 18.02.2019 18:04:43 | -32,400 | 947be465487d2d82f28c75feb475785efc4a1c44 | Implement AIRL | [
{
"change_type": "ADD",
"old_path": null,
"new_path": "example/run_airl.py",
"diff": "+\"\"\"\n+An example of Generative Adversarial Imitation Learning.\n+\"\"\"\n+\n+import argparse\n+import copy\n+import json\n+import os\n+from pprint import pprint\n+import pickle\n+\n+import numpy as np\n+import ... | Python | MIT License | deepx-inc/machina | Implement AIRL |
154,212 | 18.02.2019 21:40:46 | -32,400 | c793beb78f1d04671984688d3ecc9e0a4e51395a | add reparam, add twin | [
{
"change_type": "MODIFY",
"old_path": "example/run_sac.py",
"new_path": "example/run_sac.py",
"diff": "@@ -17,7 +17,7 @@ import gym\nimport machina as mc\nfrom machina.pols import GaussianPol\nfrom machina.algos import sac\n-from machina.vfuncs import DeterministicSVfunc, DeterministicSAVfunc\n+fro... | Python | MIT License | deepx-inc/machina | add reparam, add twin |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.