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 |
|---|---|---|---|---|---|---|---|---|---|
154,209 | 21.02.2019 13:08:07 | -32,400 | fd35381191171bcfbf111a06c00006137f3ea66d | add ModelNetLSTM | [
{
"change_type": "MODIFY",
"old_path": "example/simple_net.py",
"new_path": "example/simple_net.py",
"diff": "@@ -221,6 +221,44 @@ class VNetLSTM(nn.Module):\nreturn outs, hs\n+class ModelNetLSTM(nn.Module):\n+ def __init__(self, ob_space, ac_space, h_size=1024, cell_size=512):\n+ super(ModelNetLSTM... | Python | MIT License | deepx-inc/machina | add ModelNetLSTM |
154,209 | 21.02.2019 13:44:32 | -32,400 | bc52b1b5b91792ca545ad16ea5ff0d605f5299b9 | apply rnn mode to mpc_pol.py | [
{
"change_type": "MODIFY",
"old_path": "machina/pols/mpc_pol.py",
"new_path": "machina/pols/mpc_pol.py",
"diff": "@@ -60,7 +60,7 @@ class MPCPol(BasePol):\ndef reset(self):\nsuper(MPCPol, self).reset()\n- def forward(self, ob):\n+ def forward(self, ob, hs=None, h_masks=None):\n# randomly sample N ca... | Python | MIT License | deepx-inc/machina | apply rnn mode to mpc_pol.py |
154,209 | 21.02.2019 14:28:12 | -32,400 | e742a63a8c55813e8ca48612a7d4b178b1977918 | apply rnn mode to mpc.py | [
{
"change_type": "MODIFY",
"old_path": "machina/algos/mpc.py",
"new_path": "machina/algos/mpc.py",
"diff": "@@ -21,7 +21,7 @@ def update_dm(dm, optim_dm, batch, target='next_obs', td=True):\nreturn dm_loss.detach().cpu().numpy()\n-def train_dm(rl_traj, rand_traj, dyn_model, optim_dm, epoch=60, batch... | Python | MIT License | deepx-inc/machina | apply rnn mode to mpc.py |
154,209 | 21.02.2019 14:35:44 | -32,400 | d7a44e20be0f712f62ebbe2d323ac80061f574b2 | delete raising error by selecting rnn mode | [
{
"change_type": "MODIFY",
"old_path": "machina/pols/mpc_pol.py",
"new_path": "machina/pols/mpc_pol.py",
"diff": "@@ -42,9 +42,6 @@ class MPCPol(BasePol):\ndef __init__(self, ob_space, ac_space, net, rew_func, n_samples=1000, horizon=20,\nmean_obs=0., std_obs=1., mean_acs=0., std_acs=1., rnn=False,\... | Python | MIT License | deepx-inc/machina | delete raising error by selecting rnn mode |
154,209 | 21.02.2019 15:46:24 | -32,400 | 9140963042a187f381c50a68d4aac7c6c53010ec | change shape of obs and acs as inputs | [
{
"change_type": "MODIFY",
"old_path": "machina/pols/mpc_pol.py",
"new_path": "machina/pols/mpc_pol.py",
"diff": "@@ -70,7 +70,6 @@ class MPCPol(BasePol):\nrews_sum = torch.zeros(\n(self.n_samples), dtype=torch.float)\nobs[0] = ob.repeat(self.n_samples, 1)\n- obs[0] = self._check_obs_shape(obs[0])\n... | Python | MIT License | deepx-inc/machina | change shape of obs and acs as inputs |
154,209 | 21.02.2019 16:40:05 | -32,400 | 0fce93f9a08e746b1839abc497cc9ab8f921901c | add BaseModel | [
{
"change_type": "ADD",
"old_path": null,
"new_path": "machina/models/__init__.py",
"diff": "+from machina.models.base import BaseModel\n"
},
{
"change_type": "ADD",
"old_path": null,
"new_path": "machina/models/base.py",
"diff": "+import torch.nn as nn\n+\n+\n+class BaseModel(nn... | Python | MIT License | deepx-inc/machina | add BaseModel |
154,209 | 21.02.2019 17:07:23 | -32,400 | 574d143d9b77226f25521091b6ee810833625ae6 | add DeterministicSModel | [
{
"change_type": "ADD",
"old_path": null,
"new_path": "machina/models/deterministic_state_model.py",
"diff": "+\"\"\"\n+Deterministic State Dynamics Model\n+\"\"\"\n+\n+from machina.models.base import BaseModel\n+from machina.utils import get_device\n+\n+\n+class DeterministicSModel(BaseModel):\n+ \... | Python | MIT License | deepx-inc/machina | add DeterministicSModel |
154,209 | 21.02.2019 17:15:47 | -32,400 | c3ada2798fa20d290ff3a1ee2b6de0ad68b37a7d | apply rnn mode to dynamics loss function | [
{
"change_type": "MODIFY",
"old_path": "machina/loss_functional.py",
"new_path": "machina/loss_functional.py",
"diff": "@@ -347,7 +347,7 @@ def dynamics(dm, batch, target='next_obs', td=True):\nMSE loss for Dynamics models.\nParameters\n----------\n- model : Model\n+ dm : Dynamics Model\nbatch : dic... | Python | MIT License | deepx-inc/machina | apply rnn mode to dynamics loss function |
154,209 | 21.02.2019 20:45:51 | -32,400 | 47b607d88313650e9450955630e338a826c38f92 | delete hs in dict of return | [
{
"change_type": "MODIFY",
"old_path": "machina/pols/mpc_pol.py",
"new_path": "machina/pols/mpc_pol.py",
"diff": "@@ -102,7 +102,7 @@ class MPCPol(BasePol):\nac = sample_acs[0][best_sample_index]\nac_real = ac.cpu().numpy()\n- return ac_real, ac, dict(mean=ac, hs=hs)\n+ return ac_real, ac, dict(mean... | Python | MIT License | deepx-inc/machina | delete hs in dict of return |
154,209 | 21.02.2019 21:10:40 | -32,400 | 0eab2ed2df9d6e2f9d585a6866bbb1ca5f8901fd | fix cat dimention | [
{
"change_type": "MODIFY",
"old_path": "machina/algos/mpc.py",
"new_path": "machina/algos/mpc.py",
"diff": "@@ -84,11 +84,11 @@ def train_dm(rl_traj, rand_traj, dyn_model, optim_dm, epoch=60, batch_size=512,\nbatch['next_obs'] = rl_batch['next_obs']\nelse:\nbatch['obs'] = torch.cat(\n- [rand_batch['... | Python | MIT License | deepx-inc/machina | fix cat dimention |
154,209 | 21.02.2019 22:09:09 | -32,400 | 8bb7320ad2f855911d44a13c7b6042a29ad68455 | add h_masks to connected batch | [
{
"change_type": "MODIFY",
"old_path": "machina/algos/mpc.py",
"new_path": "machina/algos/mpc.py",
"diff": "@@ -78,10 +78,14 @@ def train_dm(rl_traj, rand_traj, dyn_model, optim_dm, epoch=60, batch_size=512,\nbatch['obs'] = rand_batch['obs']\nbatch['acs'] = rand_batch['acs']\nbatch['next_obs'] = ran... | Python | MIT License | deepx-inc/machina | add h_masks to connected batch |
154,209 | 21.02.2019 22:11:01 | -32,400 | 072e18350bcd5fe47d31c89052e8aef36da95086 | fix bug of selecting the type of iterator | [
{
"change_type": "MODIFY",
"old_path": "machina/algos/mpc.py",
"new_path": "machina/algos/mpc.py",
"diff": "@@ -60,13 +60,13 @@ def train_dm(rl_traj, rand_traj, dyn_model, optim_dm, epoch=60, batch_size=512,\nlogger.log(\"Optimizing...\")\nif dyn_model.rnn:\n- rl_iterator = rl_traj.iterate(batch_siz... | Python | MIT License | deepx-inc/machina | fix bug of selecting the type of iterator |
154,209 | 21.02.2019 22:20:48 | -32,400 | 7488873d73279c9ba904054e4e988ae03df84585 | fix connect dimention of h_masks | [
{
"change_type": "MODIFY",
"old_path": "machina/algos/mpc.py",
"new_path": "machina/algos/mpc.py",
"diff": "@@ -68,10 +68,7 @@ def train_dm(rl_traj, rand_traj, dyn_model, optim_dm, epoch=60, batch_size=512,\nrl_iterator = rl_traj.iterate(batch_size, epoch)\nrand_iterator = rand_traj.iterate(batch_si... | Python | MIT License | deepx-inc/machina | fix connect dimention of h_masks |
154,209 | 21.02.2019 22:30:33 | -32,400 | dfc50f9dcc81afb3ad5ea247efcca305adcb2740 | add out_masks to connected batch | [
{
"change_type": "MODIFY",
"old_path": "machina/algos/mpc.py",
"new_path": "machina/algos/mpc.py",
"diff": "@@ -77,12 +77,14 @@ def train_dm(rl_traj, rand_traj, dyn_model, optim_dm, epoch=60, batch_size=512,\nbatch['next_obs'] = rand_batch['next_obs']\nif dyn_model.rnn:\nbatch['h_masks'] = rand_batc... | Python | MIT License | deepx-inc/machina | add out_masks to connected batch |
154,209 | 21.02.2019 23:37:03 | -32,400 | 4fadd27d326b7a7745619ceb4e4d8f97bfd60fa9 | delete hs form dict | [
{
"change_type": "MODIFY",
"old_path": "machina/models/deterministic_state_model.py",
"new_path": "machina/models/deterministic_state_model.py",
"diff": "@@ -46,4 +46,4 @@ class DeterministicSModel(BaseModel):\nself.hs = hs\nelse:\nd_ob = self.net(obs, acs)\n- return d_ob.squeeze(-1), dict(mean=d_ob... | Python | MIT License | deepx-inc/machina | delete hs form dict |
154,209 | 22.02.2019 18:28:07 | -32,400 | b14bf2d8e06024f155a3125bc5201b5cc00820c0 | applay rnn to run_mpc.py | [
{
"change_type": "MODIFY",
"old_path": "example/run_mpc.py",
"new_path": "example/run_mpc.py",
"diff": "@@ -17,6 +17,7 @@ import machina as mc\nfrom machina.pols import GaussianPol, CategoricalPol, MultiCategoricalPol, MPCPol, RandomPol\nfrom machina.algos import mpc\nfrom machina.vfuncs import Dete... | Python | MIT License | deepx-inc/machina | applay rnn to run_mpc.py |
154,209 | 22.02.2019 21:05:18 | -32,400 | 5811a23158bd74ac215f3acef408d1a3f354ead5 | combine rand_traj and rl_traj | [
{
"change_type": "MODIFY",
"old_path": "example/run_mpc.py",
"new_path": "example/run_mpc.py",
"diff": "@@ -121,19 +121,17 @@ rand_sampler = EpiSampler(\nepis = rand_sampler.sample(random_pol, max_episodes=args.num_random_rollouts)\nepis = add_noise_to_init_obs(epis, args.noise_to_init_obs)\n-rand_t... | Python | MIT License | deepx-inc/machina | combine rand_traj and rl_traj |
154,209 | 22.02.2019 21:52:07 | -32,400 | cb5ef6c0e79ca0feea4441192f162b65f0dcc56e | delete rl_batch_rate | [
{
"change_type": "MODIFY",
"old_path": "example/run_mpc.py",
"new_path": "example/run_mpc.py",
"diff": "@@ -62,14 +62,13 @@ parser.add_argument('--num_parallel', type=int, default=4)\nparser.add_argument('--cuda', type=int, default=-1)\nparser.add_argument('--data_parallel', action='store_true', def... | Python | MIT License | deepx-inc/machina | delete rl_batch_rate |
154,209 | 22.02.2019 22:07:12 | -32,400 | 2f2f91e8d4e5021719c2050029f72c375fcebc67 | delete squeeze | [
{
"change_type": "MODIFY",
"old_path": "machina/models/deterministic_state_model.py",
"new_path": "machina/models/deterministic_state_model.py",
"diff": "@@ -46,4 +46,4 @@ class DeterministicSModel(BaseModel):\nself.hs = hs\nelse:\nd_ob = self.net(obs, acs)\n- return d_ob.squeeze(-1), dict(mean=d_ob... | Python | MIT License | deepx-inc/machina | delete squeeze |
154,209 | 23.02.2019 10:56:16 | -32,400 | d68a6df82135e8a873e81345d11c2fc14d61a2a1 | apply out_masks | [
{
"change_type": "MODIFY",
"old_path": "machina/algos/mpc.py",
"new_path": "machina/algos/mpc.py",
"diff": "@@ -61,7 +61,8 @@ def train_dm(traj, dyn_model, optim_dm, epoch=60, batch_size=512, target='next_o\niterator = traj.iterate(batch_size, epoch)\nfor batch in iterator:\n- dyn_model.reset()\n+ b... | Python | MIT License | deepx-inc/machina | apply out_masks |
154,209 | 23.02.2019 12:13:00 | -32,400 | 9036a67ee5bd10f8d139a139e550000b0fb1757e | fix normalization of sample_acs | [
{
"change_type": "MODIFY",
"old_path": "machina/pols/mpc_pol.py",
"new_path": "machina/pols/mpc_pol.py",
"diff": "@@ -59,10 +59,8 @@ class MPCPol(BasePol):\ndef forward(self, ob, hs=None, h_masks=None):\n# randomly sample N candidate action sequences\n- sample_acs = np.random.uniform(\n- self.ac_spa... | Python | MIT License | deepx-inc/machina | fix normalization of sample_acs |
154,209 | 23.02.2019 13:31:44 | -32,400 | e7e4cc5d745386521bac979435715783a1f1cc62 | fix bug of obs | [
{
"change_type": "MODIFY",
"old_path": "machina/pols/mpc_pol.py",
"new_path": "machina/pols/mpc_pol.py",
"diff": "@@ -85,14 +85,13 @@ class MPCPol(BasePol):\nwith torch.no_grad():\nfor i in range(self.horizon):\n- ob = obs[i+1]\nac = normalized_acs[i]\nif self.rnn:\n- d_ob, hs = self.net(ob.unsqueez... | Python | MIT License | deepx-inc/machina | fix bug of obs |
154,212 | 25.02.2019 11:47:38 | -32,400 | 6e9a787ed4f6509a5bdec0626d672a0a23159660 | wrap map object with list | [
{
"change_type": "MODIFY",
"old_path": "example/quickstart/simple_net.py",
"new_path": "example/quickstart/simple_net.py",
"diff": "@@ -51,7 +51,7 @@ class PolNet(nn.Module):\nif self.multi:\nself.output_layers = nn.ModuleList(\n[nn.Linear(h2, vec) for vec in ac_space.nvec])\n- map(lambda x: x.apply... | Python | MIT License | deepx-inc/machina | wrap map object with list |
154,209 | 25.02.2019 14:36:30 | -32,400 | 01a75d92660431def7e77e4f0c69c6f2d76bf304 | move normalize_obs_and_acs from traj_functional to epi_functional | [
{
"change_type": "MODIFY",
"old_path": "example/run_mpc.py",
"new_path": "example/run_mpc.py",
"diff": "@@ -124,14 +124,12 @@ traj = Traj()\ntraj.add_epis(epis)\ntraj = ef.add_next_obs(traj)\ntraj = ef.compute_h_masks(traj)\n+# obs, next_obs, and acs should become mean 0, std 1\n+traj, mean_obs, std... | Python | MIT License | deepx-inc/machina | move normalize_obs_and_acs from traj_functional to epi_functional |
154,209 | 25.02.2019 16:22:34 | -32,400 | 9b10f3691e1df9d30208e352f6adbfec66695c5d | add denormalization of obs and acs in reward function | [
{
"change_type": "MODIFY",
"old_path": "example/run_mpc.py",
"new_path": "example/run_mpc.py",
"diff": "@@ -36,14 +36,11 @@ def add_noise_to_init_obs(epis, std):\nreturn epis\n-def rew_func(next_obs, acs):\n+def rew_func(next_obs, acs, mean_obs=0., std_obs=1., mean_acs=0., std_acs=1.):\n+ next_obs =... | Python | MIT License | deepx-inc/machina | add denormalization of obs and acs in reward function |
154,209 | 25.02.2019 18:52:29 | -32,400 | 43bc08ef4c15918b6402b12bf20a4af53f38d8fc | apply outmasks | [
{
"change_type": "MODIFY",
"old_path": "machina/loss_functional.py",
"new_path": "machina/loss_functional.py",
"diff": "@@ -375,7 +375,7 @@ def dynamics(dm, batch, target='next_obs', td=True):\ndm_loss = (pred - batch[target])**2\nelse:\ndm_loss = (pred - (batch['next_obs'] - batch['obs']))**2\n- dm... | Python | MIT License | deepx-inc/machina | apply outmasks |
154,209 | 26.02.2019 11:38:38 | -32,400 | 234952ac2810f3dcbe49fce5274475d3ed32f7e1 | fix shape of out_masks | [
{
"change_type": "MODIFY",
"old_path": "machina/algos/mpc.py",
"new_path": "machina/algos/mpc.py",
"diff": "@@ -61,8 +61,6 @@ def train_dm(traj, dyn_model, optim_dm, epoch=60, batch_size=512, target='next_o\niterator = traj.iterate(batch_size, epoch)\nfor batch in iterator:\n- batch['out_masks'] = b... | Python | MIT License | deepx-inc/machina | fix shape of out_masks |
154,206 | 26.02.2019 16:28:01 | -32,400 | e5209880da5f608c5b5517af9ffe151f163f55b7 | Fix run_ddpg.py | [
{
"change_type": "MODIFY",
"old_path": "example/run_ddpg.py",
"new_path": "example/run_ddpg.py",
"diff": "@@ -91,7 +91,7 @@ qf_net = QNet(ob_space, ac_space, args.h1, args.h2)\nqf = DeterministicSAVfunc(ob_space, ac_space, qf_net)\ntarg_qf_net = QNet(ob_space, ac_space, args.h1, args.h2)\n-targ_qf_n... | Python | MIT License | deepx-inc/machina | Fix run_ddpg.py |
154,212 | 27.02.2019 11:46:31 | -32,400 | fce79d5af4251d6e4584c06640e505c2e849a54a | add rule for variable | [
{
"change_type": "MODIFY",
"old_path": "CONTRIBUTING.md",
"new_path": "CONTRIBUTING.md",
"diff": "@@ -11,3 +11,5 @@ We are welcome for any contibutions.\n# Coding style\nWe use PEP8. Test script checks whether all scripts are passing PEP8 or not. You should apply `autopep8 -i edited_script` before c... | Python | MIT License | deepx-inc/machina | add rule for variable |
154,210 | 27.02.2019 17:35:29 | -32,400 | 91a517c5d3f9db4817764841be697462604af782 | First commit with distillation backbone | [
{
"change_type": "ADD",
"old_path": null,
"new_path": "example/run_teacher_distill.py",
"diff": "+\"\"\"\n+An example of Teacher and On-policy Distillation\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 ... | Python | MIT License | deepx-inc/machina | First commit with distillation backbone |
154,210 | 28.02.2019 12:32:11 | -32,400 | 70ae77381d5e035b0e7780d0613312cf0425351a | Finish up run_teacher_distillation example and add on_policy_algo script | [
{
"change_type": "MODIFY",
"old_path": "example/run_teacher_distill.py",
"new_path": "example/run_teacher_distill.py",
"diff": "@@ -15,7 +15,7 @@ import gym\nimport pybullet as p\nimport machina as mc\nfrom machina.pols import GaussianPol, CategoricalPol, MultiCategoricalPol\n-#Import the Distillati... | Python | MIT License | deepx-inc/machina | Finish up run_teacher_distillation example and add on_policy_algo script |
154,210 | 28.02.2019 15:19:59 | -32,400 | 5253631102876401f9e3f9ef1ebfeeb8daee9425 | Finish up draft for teacher distillation | [
{
"change_type": "MODIFY",
"old_path": "example/run_teacher_distill.py",
"new_path": "example/run_teacher_distill.py",
"diff": "@@ -103,7 +103,6 @@ else:\nsampler = EpiSampler(env, t_pol, num_parallel = args.num_parallel, seed = args.seed)\noptim_pol = torch.optim.Adam(s_pol_net.parameters(), args.p... | Python | MIT License | deepx-inc/machina | Finish up draft for teacher distillation |
154,210 | 28.02.2019 23:10:21 | -32,400 | e6a29a1db009737c7a2e61ad1d2184c1cc541e6e | Apply autopep8 to scripts and move backbone of shannon-ce to loss_functional.py | [
{
"change_type": "MODIFY",
"old_path": "example/run_teacher_distill.py",
"new_path": "example/run_teacher_distill.py",
"diff": "@@ -40,7 +40,13 @@ parser.add_argument('--epoch_per_iter', type = int, default = 10)\nparser.add_argument('--batch_size', default=int, default=256)\nparser.add_argument('--... | Python | MIT License | deepx-inc/machina | Apply autopep8 to scripts and move backbone of shannon-ce to loss_functional.py |
154,210 | 01.03.2019 16:02:39 | -32,400 | 0184fc189a8e341b00164b99b92a1dee115981c5 | update shannon-ce | [
{
"change_type": "MODIFY",
"old_path": "machina/loss_functional.py",
"new_path": "machina/loss_functional.py",
"diff": "@@ -435,13 +435,12 @@ def density_ratio_adv_cross_ent(advf, pol, batch, expert_or_agent):\nreturn discrim_loss\n-def shannon_cross_entropy(student_pol, teacher_pol, batch):\n+def s... | Python | MIT License | deepx-inc/machina | update shannon-ce |
154,206 | 01.03.2019 18:57:11 | -32,400 | 17c3ed94fcd26feb831b05f92a46d141ed38a6b4 | Fix bug of check acs obs | [
{
"change_type": "MODIFY",
"old_path": "machina/models/base.py",
"new_path": "machina/models/base.py",
"diff": "@@ -48,7 +48,7 @@ class BaseModel(nn.Module):\nadditional_shape = 1\nif len(obs.shape) < additional_shape + len(self.ob_space.shape):\nfor _ in range(additional_shape + len(self.ob_space.s... | Python | MIT License | deepx-inc/machina | Fix bug of check acs obs |
154,212 | 01.03.2019 20:46:17 | -32,400 | e60dc0e2913fcf167c536e31ea4412fb7b5a31da | add cpu_mode | [
{
"change_type": "MODIFY",
"old_path": "machina/samplers/epi_sampler.py",
"new_path": "machina/samplers/epi_sampler.py",
"diff": "@@ -10,6 +10,8 @@ import numpy as np\nimport torch\nimport torch.multiprocessing as mp\n+from machina.utils import cpu_mode\n+\nLARGE_NUMBER = 100000000\n@@ -30,6 +32,7 @... | Python | MIT License | deepx-inc/machina | add cpu_mode |
154,210 | 03.03.2019 16:46:45 | -32,400 | 4e60c0c7166ee46c678b2c438e7249e1bdbad91f | Fix policy distillation - running script | [
{
"change_type": "MODIFY",
"old_path": "example/run_teacher_distill.py",
"new_path": "example/run_teacher_distill.py",
"diff": "@@ -27,7 +27,7 @@ from simple_net import PolNet, VNet, PolNetLSTM, VNetLSTM\nparser = argparse.ArgumentParser()\nparser.add_argument('--log', type=str, default='garbage')\n... | Python | MIT License | deepx-inc/machina | Fix policy distillation - running script |
154,210 | 04.03.2019 09:09:40 | -32,400 | e8ea93837464567b3353eafeb545966198c908d2 | Clean-up loss-functional.py | [
{
"change_type": "MODIFY",
"old_path": "machina/loss_functional.py",
"new_path": "machina/loss_functional.py",
"diff": "@@ -435,11 +435,22 @@ def density_ratio_adv_cross_ent(advf, pol, batch, expert_or_agent):\nreturn discrim_loss\n-def shannon_cross_entropy(student_pol, teacher_pol, batch, num_samp... | Python | MIT License | deepx-inc/machina | Clean-up loss-functional.py |
154,206 | 04.03.2019 11:05:45 | -32,400 | db0ffb07abc698a3a1c3ef4bc34c17082ad2c0cd | Fix bug of airl | [
{
"change_type": "MODIFY",
"old_path": "example/run_airl.py",
"new_path": "example/run_airl.py",
"diff": "@@ -42,7 +42,7 @@ parser.add_argument('--num_parallel', type=int, default=4)\nparser.add_argument('--expert_dir', type=str, default='../data/expert_epis')\nparser.add_argument('--expert_fname', ... | Python | MIT License | deepx-inc/machina | Fix bug of airl |
154,210 | 04.03.2019 18:03:06 | -32,400 | 27c52e7abb4c67e538f208ac99e1990fee56c40d | Autopep modified files | [
{
"change_type": "MODIFY",
"old_path": "example/run_teacher_distill.py",
"new_path": "example/run_teacher_distill.py",
"diff": "@@ -41,8 +41,10 @@ parser.add_argument('--epoch_per_iter', type=int, default=10)\nparser.add_argument('--batch_size', type=int, default=256)\nparser.add_argument('--gamma',... | Python | MIT License | deepx-inc/machina | Autopep modified files |
154,206 | 04.03.2019 18:03:56 | -32,400 | 4c0a48f701c35e713f4d3c0b3fba99823254b760 | Update expert_epis_make.py | [
{
"change_type": "MODIFY",
"old_path": "example/expert_epis_make.py",
"new_path": "example/expert_epis_make.py",
"diff": "@@ -31,7 +31,7 @@ parser.add_argument('--num_parallel', type=int, default=8)\nparser.add_argument('--h1', type=int, default=32)\nparser.add_argument('--h2', type=int, default=32)... | Python | MIT License | deepx-inc/machina | Update expert_epis_make.py |
154,206 | 05.03.2019 01:06:56 | -32,400 | 49337a68271ff4f5adc789707c1a398a0530b3c8 | Update run_ddpg.py | [
{
"change_type": "MODIFY",
"old_path": "example/run_ddpg.py",
"new_path": "example/run_ddpg.py",
"diff": "@@ -47,7 +47,6 @@ parser.add_argument('--h2', type=int, default=32)\nparser.add_argument('--tau', type=float, default=0.001)\nparser.add_argument('--gamma', type=float, default=0.99)\n-parser.ad... | Python | MIT License | deepx-inc/machina | Update run_ddpg.py |
154,206 | 05.03.2019 15:55:14 | -32,400 | a08ab3bb8fac86d2fcfe114a3bebc4f8d29ce1d5 | Add option of off traj size | [
{
"change_type": "MODIFY",
"old_path": "example/run_ddpg.py",
"new_path": "example/run_ddpg.py",
"diff": "@@ -35,6 +35,8 @@ parser.add_argument('--env_name', type=str, default='Pendulum-v0')\nparser.add_argument('--record', action='store_true', default=False)\nparser.add_argument('--seed', type=int,... | Python | MIT License | deepx-inc/machina | Add option of off traj size |
154,206 | 05.03.2019 16:32:30 | -32,400 | 57531a1d441ed6811f950414471e6ab873bad4d9 | Add script for taking movie | [
{
"change_type": "ADD",
"old_path": null,
"new_path": "example/take_movie.py",
"diff": "+import gym\n+import torch.nn.functional as F\n+import torch.nn as nn\n+import torch\n+import numpy as np\n+import pickle\n+from pprint import pprint\n+import os\n+import json\n+import argparse\n+\"\"\"\n+Script ... | Python | MIT License | deepx-inc/machina | Add script for taking movie |
154,206 | 06.03.2019 01:34:14 | -32,400 | 633b27ebdb9abae1e6affa5f7817ccea1f800ba5 | Create IMITATION.md | [
{
"change_type": "ADD",
"old_path": null,
"new_path": "example/IMITATION.md",
"diff": "+# Imitation Learning\n+* Behavioral Cloning\n+* Generative Adversarial Imitation Learning\n+\n+ Original paper: https://arxiv.org/abs/1606.03476\n+* Adversarial Inverse Reinforcement Learning\n+\n+ Original paper... | Python | MIT License | deepx-inc/machina | Create IMITATION.md |
154,210 | 06.03.2019 10:33:26 | -32,400 | e62d91280bcab8112e5fb5224e1e3e64dfb532d7 | add on-policy distillation to test-script | [
{
"change_type": "MODIFY",
"old_path": "tests/test_algos.py",
"new_path": "tests/test_algos.py",
"diff": "@@ -3,6 +3,7 @@ Test script for algorithms.\n\"\"\"\nimport unittest\n+import os\nimport numpy as np\nimport torch\n@@ -14,7 +15,7 @@ import machina as mc\nfrom machina.pols import GaussianPol, ... | Python | MIT License | deepx-inc/machina | add on-policy distillation to test-script |
154,206 | 06.03.2019 12:27:57 | -32,400 | 202ef6775eb1440149d04837b6f8e54ec3ca7009 | Add args of env_name | [
{
"change_type": "MODIFY",
"old_path": "example/make_expert_epis.py",
"new_path": "example/make_expert_epis.py",
"diff": "@@ -33,6 +33,8 @@ parser.add_argument('--epis_dir', type=str, default='../data/expert_epis',\nhelp='Directory path to store file of expert trajectory.')\nparser.add_argument('--e... | Python | MIT License | deepx-inc/machina | Add args of env_name |
154,210 | 06.03.2019 15:23:41 | -32,400 | 366a4c8ae9ad9e37b00c7f35990350c3e653ab80 | Fix test for policy distillation | [
{
"change_type": "MODIFY",
"old_path": ".travis.yml",
"new_path": ".travis.yml",
"diff": "@@ -8,7 +8,9 @@ install:\n- pip install --upgrade pip setuptools wheel\n- pip install --default-timeout=2000 -q -e .\n- pip install autopep8\n-\n+ - curl -sc /tmp/cookie \"https://drive.google.com/uc?export=dow... | Python | MIT License | deepx-inc/machina | Fix test for policy distillation |
154,210 | 06.03.2019 15:58:09 | -32,400 | a9d478cc06ce45d11188bce98f95c6962d1ad5c8 | Fix run_teacher_distillation in example folder | [
{
"change_type": "MODIFY",
"old_path": "example/run_teacher_distill.py",
"new_path": "example/run_teacher_distill.py",
"diff": "@@ -26,27 +26,37 @@ from machina.utils import measure, set_device\nfrom simple_net import PolNet, VNet, PolNetLSTM, VNetLSTM\nparser = argparse.ArgumentParser()\n-parser.ad... | Python | MIT License | deepx-inc/machina | Fix run_teacher_distillation in example folder |
154,210 | 06.03.2019 16:25:32 | -32,400 | d45f97de210243e0ce2909e3077744a0440a9149 | Fix travis.yaml file | [
{
"change_type": "MODIFY",
"old_path": ".travis.yml",
"new_path": ".travis.yml",
"diff": "@@ -12,8 +12,8 @@ install:\n- CODE=\"$(awk '/_warning_/ {print $NF}' /tmp/cookie)\"\n- export=download&confirm=${CODE}&id=1zODylHjXq2d3R8jY20yP3aOOZFqHgVRn\" -o data/expert_epis/Pendulum-v0_100epis.pkl\nscript:... | Python | MIT License | deepx-inc/machina | Fix travis.yaml file |
154,206 | 06.03.2019 18:32:20 | -32,400 | 95edb67981131ec38d7e0166c0f24dbabd0a34c3 | Simplify num_step method | [
{
"change_type": "MODIFY",
"old_path": "machina/traj/traj.py",
"new_path": "machina/traj/traj.py",
"diff": "@@ -35,9 +35,7 @@ class Traj(object):\n@property\ndef num_step(self):\n- if self.data_map:\n- return list(self.data_map.values())[0].size(0)\n- return 0\n+ return self._epis_index[-1]\n@proper... | Python | MIT License | deepx-inc/machina | Simplify num_step method |
154,206 | 06.03.2019 21:35:54 | -32,400 | 76547fca7fc5b35d4b903c770dbb13f3f6607732 | Change max_epis to max_steps | [
{
"change_type": "MODIFY",
"old_path": "example/run_ddpg.py",
"new_path": "example/run_ddpg.py",
"diff": "@@ -35,8 +35,8 @@ parser.add_argument('--env_name', type=str, default='Pendulum-v0')\nparser.add_argument('--record', action='store_true', default=False)\nparser.add_argument('--seed', type=int,... | Python | MIT License | deepx-inc/machina | Change max_epis to max_steps |
154,206 | 06.03.2019 22:39:29 | -32,400 | c854cf367f3f39000f2ffe7c8fda7b1bdb120c44 | Change condition of error | [
{
"change_type": "MODIFY",
"old_path": "machina/traj/traj.py",
"new_path": "machina/traj/traj.py",
"diff": "@@ -95,6 +95,9 @@ class Traj(object):\nadd_num_step = traj.num_step\nif pre_num_step + add_num_step <= self.max_steps:\nself._concat_data_map(traj.data_map)\n+ epis_index = epis_index + self._... | Python | MIT License | deepx-inc/machina | Change condition of error |
154,209 | 06.03.2019 23:23:22 | -32,400 | 08bd4d938745d068ed5d4b10502bbc3949b96333 | add help to run_mpc.py | [
{
"change_type": "MODIFY",
"old_path": "example/run_mpc.py",
"new_path": "example/run_mpc.py",
"diff": "@@ -67,14 +67,18 @@ parser.add_argument('--data_parallel', action='store_true', default=False,\nhelp='If True, inference is done in parallel on gpus.')\nparser.add_argument('--pybullet_env', actio... | Python | MIT License | deepx-inc/machina | add help to run_mpc.py |
154,206 | 07.03.2019 00:33:10 | -32,400 | 0e37f209518a33d0ebbc4f2e5108e36b766d104f | Remove lambda from prioritized ddpg | [
{
"change_type": "MODIFY",
"old_path": "example/run_prioritized_ddpg.py",
"new_path": "example/run_prioritized_ddpg.py",
"diff": "@@ -63,8 +63,6 @@ parser.add_argument('--tau', type=float, default=0.001,\nhelp='Coefficient of target function.')\nparser.add_argument('--gamma', type=float, default=0.9... | Python | MIT License | deepx-inc/machina | Remove lambda from prioritized ddpg |
154,206 | 07.03.2019 10:17:25 | -32,400 | c3f371663b48580192277ade7bba0c74bb9d3402 | Update machina/traj/traj.py | [
{
"change_type": "MODIFY",
"old_path": "machina/traj/traj.py",
"new_path": "machina/traj/traj.py",
"diff": "@@ -108,17 +108,16 @@ class Traj(object):\nepis_index = epis_index + self._epis_index[-1]\nself._epis_index = np.concatenate(\n[self._epis_index, epis_index[1:]])\n- else:\n- remain_index = 1\... | Python | MIT License | deepx-inc/machina | Update machina/traj/traj.py |
154,206 | 07.03.2019 11:12:36 | -32,400 | fa60d8db947b4c2257e962253f9dcdb80760034c | Fix mistakes of removing help of argparse | [
{
"change_type": "MODIFY",
"old_path": "example/run_ddpg.py",
"new_path": "example/run_ddpg.py",
"diff": "@@ -39,13 +39,16 @@ parser.add_argument('--c2d', action='store_true',\nparser.add_argument('--record', action='store_true',\ndefault=False, help='If True, movie is saved.')\nparser.add_argument(... | Python | MIT License | deepx-inc/machina | Fix mistakes of removing help of argparse |
154,206 | 04.03.2019 11:00:29 | -32,400 | 1ddd9544da42fec1b01f4a58c004cf29b3c96091 | add test of bc airl gail | [
{
"change_type": "MODIFY",
"old_path": "tests/simple_net.py",
"new_path": "tests/simple_net.py",
"diff": "@@ -202,3 +202,17 @@ class VNetLSTM(nn.Module):\nouts = self.output_layer(hiddens)\nreturn outs, hs\n+\n+\n+class DiscrimNet(nn.Module):\n+ def __init__(self, ob_space, ac_space, h1=32, h2=32):\... | Python | MIT License | deepx-inc/machina | add test of bc airl gail |
154,209 | 05.03.2019 19:32:59 | -32,400 | 99c94b03040bc14b7403ef662369ba7b85500144 | add test code of MPC | [
{
"change_type": "MODIFY",
"old_path": "tests/simple_net.py",
"new_path": "tests/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).__init__... | Python | MIT License | deepx-inc/machina | add test code of MPC |
154,206 | 05.03.2019 21:47:08 | -32,400 | e0a0f5ed32beed331ec7bf80de3e593d3d8d20b7 | Chagne file path | [
{
"change_type": "MODIFY",
"old_path": "tests/test_algos.py",
"new_path": "tests/test_algos.py",
"diff": "@@ -527,7 +527,7 @@ class TestBehaviorClone(unittest.TestCase):\noptim_pol = torch.optim.Adam(pol_net.parameters(), 3e-4)\n- with open(os.path.join('../data/expert_epis', 'Pendulum-v0_100epis.pk... | Python | MIT License | deepx-inc/machina | Chagne file path |
154,209 | 06.03.2019 22:59:19 | -32,400 | 9fa20a1ab451eae356d4aefa6b9ef3911e1cd4b7 | fix bug of device of out_masks | [
{
"change_type": "MODIFY",
"old_path": "machina/loss_functional.py",
"new_path": "machina/loss_functional.py",
"diff": "@@ -426,7 +426,8 @@ def dynamics(dm, batch, target='next_obs', td=True):\nout_masks = batch['out_masks']\npred, _ = dm(obs, acs, h_masks=h_masks)\nelse:\n- out_masks = torch.ones(o... | Python | MIT License | deepx-inc/machina | fix bug of device of out_masks |
154,206 | 07.03.2019 00:39:43 | -32,400 | ece4e9dba75d8ab97145ed3bc374da97b0b477ed | Fixed arg for consistency with other pr | [
{
"change_type": "MODIFY",
"old_path": "tests/test_algos.py",
"new_path": "tests/test_algos.py",
"diff": "@@ -639,7 +639,7 @@ class TestAIRL(unittest.TestCase):\nagent_traj.register_epis()\nresult_dict = airl.train(agent_traj, expert_traj, pol, vf, optim_vf, optim_discrim,\n- rewf=rewf, shaping_vf=s... | Python | MIT License | deepx-inc/machina | Fixed arg for consistency with other pr |
154,209 | 07.03.2019 14:49:19 | -32,400 | b9d81e43cd6486eed5c489ca44c654101279f875 | change iterate to random_batch | [
{
"change_type": "MODIFY",
"old_path": "machina/algos/mpc.py",
"new_path": "machina/algos/mpc.py",
"diff": "@@ -58,7 +58,7 @@ def train_dm(traj, dyn_model, optim_dm, epoch=60, batch_size=512, target='next_o\niterator = traj.iterate_rnn(\nbatch_size=batch_size, num_epi_per_seq=num_epi_per_seq, epoch=... | Python | MIT License | deepx-inc/machina | change iterate to random_batch |
154,209 | 07.03.2019 16:58:13 | -32,400 | d73d6dfb94559e825007e20e0116435753befbb6 | add __deepcopy__ to mpc_pol | [
{
"change_type": "MODIFY",
"old_path": "machina/pols/mpc_pol.py",
"new_path": "machina/pols/mpc_pol.py",
"diff": "@@ -58,6 +58,18 @@ class MPCPol(BasePol):\nself.std_acs = torch.tensor(\nstd_acs, dtype=torch.float).repeat(n_samples, 1)\n+ def __deepcopy__(self, memo):\n+ mpc_pol = MPCPol(self.ob_spa... | Python | MIT License | deepx-inc/machina | add __deepcopy__ to mpc_pol |
154,209 | 07.03.2019 17:06:29 | -32,400 | d1599985e4bc5f2d54cdd5815bbfa815f55408d2 | add deepcopy of self.net in __deepcopy__ | [
{
"change_type": "MODIFY",
"old_path": "machina/pols/mpc_pol.py",
"new_path": "machina/pols/mpc_pol.py",
"diff": "import numpy as np\nimport torch\n+import copy\nfrom machina.pds import DeterministicPd\nfrom machina.pols import BasePol\n@@ -59,10 +60,10 @@ class MPCPol(BasePol):\nstd_acs, dtype=torc... | Python | MIT License | deepx-inc/machina | add deepcopy of self.net in __deepcopy__ |
154,206 | 07.03.2019 18:27:43 | -32,400 | 5a7b4056c2e9d5aaa2167311a12d868b65a81e83 | Rename from pd_shape to a_i_shape | [
{
"change_type": "MODIFY",
"old_path": "machina/pols/base.py",
"new_path": "machina/pols/base.py",
"diff": "@@ -45,15 +45,14 @@ class BasePol(nn.Module):\nself.multi = isinstance(ac_space, gym.spaces.MultiDiscrete)\nif not self.discrete:\n- self.pd_shape = ac_space.shape\n+ self.a_i_shape = ac_space... | Python | MIT License | deepx-inc/machina | Rename from pd_shape to a_i_shape |
154,212 | 07.03.2019 18:58:36 | -32,400 | cfb311664ace04aa578c78658347cbc1f1abcdeb | comment out __deepcopy__ and define rew_func in learning script | [
{
"change_type": "MODIFY",
"old_path": "machina/pols/mpc_pol.py",
"new_path": "machina/pols/mpc_pol.py",
"diff": "@@ -59,17 +59,17 @@ class MPCPol(BasePol):\nself.std_acs = torch.tensor(\nstd_acs, dtype=torch.float).repeat(n_samples, 1)\n- def __deepcopy__(self, memo):\n- mpc_pol = MPCPol(self.ob_sp... | Python | MIT License | deepx-inc/machina | comment out __deepcopy__ and define rew_func in learning script |
154,209 | 07.03.2019 21:03:38 | -32,400 | 6beca494411d2c876646385a8fb6f866aac568da | delete random sample and fix hyperparams | [
{
"change_type": "MODIFY",
"old_path": "tests/test_algos.py",
"new_path": "tests/test_algos.py",
"diff": "@@ -669,98 +669,59 @@ class TestMPC(unittest.TestCase):\nreturn rews\ndef test_learning(self):\n- # sample with random policy\n- random_pol = RandomPol(self.env.ob_space, self.env.ac_space)\n- r... | Python | MIT License | deepx-inc/machina | delete random sample and fix hyperparams |
154,209 | 07.03.2019 21:24:41 | -32,400 | ccc1a7760da9d7449e7e5d37b5484ea91c0d847f | delete self of rew_func | [
{
"change_type": "MODIFY",
"old_path": "tests/test_algos.py",
"new_path": "tests/test_algos.py",
"diff": "@@ -659,7 +659,7 @@ class TestMPC(unittest.TestCase):\nreturn epis\ndef test_learning(self):\n- def rew_func(self, next_obs, acs, mean_obs=0., std_obs=1., mean_acs=0., std_acs=1.):\n+ def rew_fu... | Python | MIT License | deepx-inc/machina | delete self of rew_func |
154,212 | 08.03.2019 10:58:59 | -32,400 | a5ab3985c165dc7032af971e637108ca05ab0965 | delete deepcopy | [
{
"change_type": "MODIFY",
"old_path": "machina/pols/mpc_pol.py",
"new_path": "machina/pols/mpc_pol.py",
"diff": "@@ -59,18 +59,6 @@ class MPCPol(BasePol):\nself.std_acs = torch.tensor(\nstd_acs, dtype=torch.float).repeat(n_samples, 1)\n- # def __deepcopy__(self, memo):\n- # mpc_pol = MPCPol(self.ob... | Python | MIT License | deepx-inc/machina | delete deepcopy |
154,212 | 10.03.2019 12:39:16 | -32,400 | 2087d324676c70d21d3ab68802c1fc98bb49185d | change episode to epi | [
{
"change_type": "MODIFY",
"old_path": "README.md",
"new_path": "README.md",
"diff": "@@ -32,8 +32,8 @@ env1 = GymEnv('HumanoidBulletEnv-v0')\nenv2 = GymEnv('HumanoidFlagrunBulletEnv-v0')\n-epis1 = sampler1.sample(pol, max_episodes=args.max_episodes_per_iter)\n-epis2 = sampler2.sample(pol, max_episo... | Python | MIT License | deepx-inc/machina | change episode to epi |
154,206 | 10.03.2019 07:25:40 | -3,600 | 4ba714a1e4b14042d81e202ebac0082db3089844 | Rename lf.likelihood | [
{
"change_type": "MODIFY",
"old_path": "machina/algos/behavior_clone.py",
"new_path": "machina/algos/behavior_clone.py",
"diff": "@@ -10,7 +10,7 @@ from machina import loss_functional as lf\ndef update_pol(pol, optim_pol, batch):\n- pol_loss = lf.likelihood(pol, batch)\n+ pol_loss = lf.log_likelihoo... | Python | MIT License | deepx-inc/machina | Rename lf.likelihood |
154,213 | 10.03.2019 15:59:47 | -32,400 | 65343c04441c58cfc5adaa1056ea25749f4c3c1c | fix softmax twice in cross_ent_diayn | [
{
"change_type": "MODIFY",
"old_path": "machina/loss_functional.py",
"new_path": "machina/loss_functional.py",
"diff": "@@ -465,7 +465,7 @@ def cross_ent_diayn(discrim, batch, ob_dim):\nskill = batch['obs'][:, ob_dim:]\n_, targets = skill.max(dim=1)\nlogits = discrim.net.learn(obs)\n- discrim_loss =... | Python | MIT License | deepx-inc/machina | fix softmax twice in cross_ent_diayn |
154,213 | 10.03.2019 23:56:30 | -32,400 | c7ef75d7bfbcd71142b84c09d27fb30eec678591 | fix F to torch | [
{
"change_type": "MODIFY",
"old_path": "example/simple_net.py",
"new_path": "example/simple_net.py",
"diff": "@@ -287,13 +287,13 @@ class DiaynDiscrimNet(nn.Module):\nob = obskill[:, :-self.num_skill]\nfeat = self.discrim_f(ob)\nskill = obskill[:, -self.num_skill:]\n- logit = self.output_layer(F.rel... | Python | MIT License | deepx-inc/machina | fix F to torch |
154,213 | 11.03.2019 00:02:34 | -32,400 | 4f4cb590fcf27e99d4d2fbe37fee4f511d41e22c | fix name log* to logit | [
{
"change_type": "MODIFY",
"old_path": "example/simple_net.py",
"new_path": "example/simple_net.py",
"diff": "@@ -288,9 +288,9 @@ class DiaynDiscrimNet(nn.Module):\nfeat = self.discrim_f(ob)\nskill = obskill[:, -self.num_skill:]\nlogit = self.output_layer(torch.relu(self.fc1(feat)))\n- logqz = torch... | Python | MIT License | deepx-inc/machina | fix name log* to logit |
154,213 | 11.03.2019 00:06:17 | -32,400 | 3f6ba1f9a80d148b35a82b6a8b42244782439013 | fix logit to logits | [
{
"change_type": "MODIFY",
"old_path": "example/simple_net.py",
"new_path": "example/simple_net.py",
"diff": "@@ -288,9 +288,9 @@ class DiaynDiscrimNet(nn.Module):\nfeat = self.discrim_f(ob)\nskill = obskill[:, -self.num_skill:]\nlogit = self.output_layer(torch.relu(self.fc1(feat)))\n- discrim_logit... | Python | MIT License | deepx-inc/machina | fix logit to logits |
154,206 | 11.03.2019 09:25:13 | -3,600 | 4953eca087ab5271b0ebcb331c4e2c27791ea144 | Add device to tensors | [
{
"change_type": "MODIFY",
"old_path": "machina/vfuncs/state_action_vfuncs/cem_state_action_vfunc.py",
"new_path": "machina/vfuncs/state_action_vfuncs/cem_state_action_vfunc.py",
"diff": "@@ -62,7 +62,7 @@ class CEMDeterministicSAVfunc(DeterministicSAVfunc):\ndtype=torch.float, device=obs.device)\nl... | Python | MIT License | deepx-inc/machina | Add device to tensors |
154,206 | 11.03.2019 13:15:58 | -3,600 | fc7077fcde46f35ea61a96669a7ff10a791e67be | Change ac_space from tuple to Box | [
{
"change_type": "MODIFY",
"old_path": "example/run_ddpg.py",
"new_path": "example/run_ddpg.py",
"diff": "@@ -93,7 +93,7 @@ ob_space = env.observation_space\nac_space = env.action_space\npol_net = PolNet(ob_space, ac_space, args.h1, args.h2, deterministic=True)\n-noise = OUActionNoise(ac_space.shape... | Python | MIT License | deepx-inc/machina | Change ac_space from tuple to Box |
154,206 | 11.03.2019 13:16:43 | -3,600 | 6cf84abacf82c493c0deee0d3ac9264cd4dc34a3 | Utilize no_noise | [
{
"change_type": "MODIFY",
"old_path": "machina/algos/ddpg.py",
"new_path": "machina/algos/ddpg.py",
"diff": "@@ -59,7 +59,7 @@ def train(traj,\nqf_bellman_loss.backward()\noptim_qf.step()\n- pol_loss = lf.ag(pol, qf, batch)\n+ pol_loss = lf.ag(pol, qf, batch, no_noise=True)\noptim_pol.zero_grad()\n... | Python | MIT License | deepx-inc/machina | Utilize no_noise |
154,206 | 11.03.2019 14:03:14 | -3,600 | 12e9ea14aa02c7678553d9575874af9396a124e9 | Use get_device | [
{
"change_type": "MODIFY",
"old_path": "machina/vfuncs/state_action_vfuncs/cem_state_action_vfunc.py",
"new_path": "machina/vfuncs/state_action_vfuncs/cem_state_action_vfunc.py",
"diff": "@@ -59,10 +59,10 @@ class CEMDeterministicSAVfunc(DeterministicSAVfunc):\nself.batch_size = obs.shape[0]\nself.d... | Python | MIT License | deepx-inc/machina | Use get_device |
154,206 | 12.03.2019 19:07:23 | -3,600 | 7d04b3d55a45b2f5a21ece55591126c4efd8ae5b | Add classification of algorithms | [
{
"change_type": "MODIFY",
"old_path": "README.md",
"new_path": "README.md",
"diff": "@@ -109,17 +109,21 @@ To obtain this composability, machina's sampling method is deliberatly restricte\n## Implemented Algorithms\n-\n+### Model-Free On-Policy RL\n+ [Proximal Policy Optimization](https://arxiv.org... | Python | MIT License | deepx-inc/machina | Add classification of algorithms |
154,206 | 12.03.2019 21:35:21 | -3,600 | 26f80e033ab6d42ca37e0c68832eb1f3783b548a | Add explanation of algorithm class | [
{
"change_type": "MODIFY",
"old_path": "README.md",
"new_path": "README.md",
"diff": "@@ -109,21 +109,27 @@ To obtain this composability, machina's sampling method is deliberatly restricte\n## Implemented Algorithms\n+All classes of algorithms are important for real-world Deep Reinforcement Learning... | Python | MIT License | deepx-inc/machina | Add explanation of algorithm class |
154,213 | 13.03.2019 08:36:10 | -32,400 | 81585ebb9c4881f29ad1ded32ee42e2a6001d1f9 | fix net.learn to net.logits | [
{
"change_type": "MODIFY",
"old_path": "example/simple_net.py",
"new_path": "example/simple_net.py",
"diff": "@@ -288,12 +288,12 @@ class DiaynDiscrimNet(nn.Module):\nfeat = self.discrim_f(ob)\nskill = obskill[:, -self.num_skill:]\nlogit = self.output_layer(torch.relu(self.fc1(feat)))\n- discrim_log... | Python | MIT License | deepx-inc/machina | fix net.learn to net.logits |
154,213 | 13.03.2019 10:25:03 | -32,400 | 97311dcfba00248210e4ae8c49ce0b15faede0bd | fix discrim.net.logit to discrim.net | [
{
"change_type": "MODIFY",
"old_path": "example/run_diayn.py",
"new_path": "example/run_diayn.py",
"diff": "@@ -189,11 +189,12 @@ while args.max_episodes > total_epi:\npol, qfs, targ_qfs, log_alpha,\noptim_pol, optim_qfs, optim_alpha,\nstep, args.batch_size,\n- args.tau, args.gamma, args.sampling, d... | Python | MIT License | deepx-inc/machina | fix discrim.net.logit to discrim.net |
154,212 | 13.03.2019 11:30:04 | -32,400 | 3d895502ccc3586dbd3aed2f490ae5427cae5e63 | fix from pierre's comment | [
{
"change_type": "MODIFY",
"old_path": "README.md",
"new_path": "README.md",
"diff": "@@ -109,22 +109,22 @@ To obtain this composability, machina's sampling method is deliberatly restricte\n## Implemented Algorithms\n-All classes of algorithms are important for real-world Deep Reinforcement Learning... | Python | MIT License | deepx-inc/machina | fix from pierre's comment |
154,206 | 13.03.2019 07:15:09 | -3,600 | 62e0f99a675022c3d536f9e80ef240bd9970c7a7 | Add explanation about rnn | [
{
"change_type": "MODIFY",
"old_path": "README.md",
"new_path": "README.md",
"diff": "@@ -131,8 +131,8 @@ This class removes the need for reward designing.\n### Policy Distillation\nThis class enables fast forward computation and reduces necessary computation resources during deployment of policy.\n... | Python | MIT License | deepx-inc/machina | Add explanation about rnn |
154,212 | 13.03.2019 16:10:51 | -32,400 | 36f6e39a7ef9d6a5b6071091009a4950c28707ac | add rnn support | [
{
"change_type": "MODIFY",
"old_path": "README.md",
"new_path": "README.md",
"diff": "@@ -112,8 +112,8 @@ To obtain this composability, machina's sampling method is deliberatly restricte\nThe algorithms classes described below are useful for real-world Deep Reinforcement Learning.\n### Model-Free On... | Python | MIT License | deepx-inc/machina | add rnn support |
154,212 | 13.03.2019 18:07:15 | -32,400 | 9e51a09e29103c5b835353bd26ceedb519598714 | meaningless update for activating travis | [
{
"change_type": "MODIFY",
"old_path": "README.md",
"new_path": "README.md",
"diff": "@@ -132,6 +132,7 @@ This class removes the need for reward designing.\nThis class enables fast forward computation and reduces necessary computation resources during deployment of policy.\n+ [Teacher Distillation](... | Python | MIT License | deepx-inc/machina | meaningless update for activating travis |
154,206 | 13.03.2019 11:08:17 | -3,600 | 111f7211cb6b1f9b890dcb4ea251d6f783c5e348 | Change num expert epis from 100 to 2 | [
{
"change_type": "MODIFY",
"old_path": ".travis.yml",
"new_path": ".travis.yml",
"diff": "@@ -11,9 +11,9 @@ install:\n- curl -sc /tmp/cookie \"https://drive.google.com/uc?export=download&id=1zODylHjXq2d3R8jY20yP3aOOZFqHgVRn\" > /dev/null\n- CODE=\"$(awk '/_warning_/ {print $NF}' /tmp/cookie)\"\n- cu... | Python | MIT License | deepx-inc/machina | Change num expert epis from 100 to 2 |
154,206 | 13.03.2019 11:39:47 | -3,600 | 15eb8b0b35482985bd62770e58be4960c0c22bcd | Update tests/test_algos.py | [
{
"change_type": "MODIFY",
"old_path": "tests/test_algos.py",
"new_path": "tests/test_algos.py",
"diff": "@@ -526,7 +526,7 @@ class TestBehaviorClone(unittest.TestCase):\noptim_pol = torch.optim.Adam(pol_net.parameters(), 3e-4)\n- with open(os.path.join('data/expert_epis', 'Pendulum-v0_100epis.pkl')... | Python | MIT License | deepx-inc/machina | Update tests/test_algos.py |
154,210 | 13.03.2019 22:56:20 | -32,400 | 3486072ff695f80ec2a587a25fc3daebace72730 | Update test_algos.py | [
{
"change_type": "MODIFY",
"old_path": "tests/test_algos.py",
"new_path": "tests/test_algos.py",
"diff": "@@ -488,10 +488,6 @@ class TestOnpolicyDistillation(unittest.TestCase):\ns_pol = GaussianPol(\nself.env.ob_space, self.env.ac_space, s_pol_net)\n- # Please import your own teacher-policy here\n-... | Python | MIT License | deepx-inc/machina | Update test_algos.py |
154,209 | 15.03.2019 17:22:53 | -32,400 | 644693638fe967893d5ab251113823cbcb59a751 | add random_batch_rnn | [
{
"change_type": "MODIFY",
"old_path": "machina/algos/mpc.py",
"new_path": "machina/algos/mpc.py",
"diff": "@@ -55,8 +55,8 @@ def train_dm(traj, dyn_model, optim_dm, epoch=60, batch_size=512, target='next_o\nbatch_size = min(batch_size, traj.num_epi)\nif dyn_model.rnn:\n- iterator = traj.iterate_rnn... | Python | MIT License | deepx-inc/machina | add random_batch_rnn |
154,209 | 15.03.2019 17:24:20 | -32,400 | 2d4f31ecc5de7a2acc4df8f999e2b942f4538c0a | fix hidden size of ModelNets | [
{
"change_type": "MODIFY",
"old_path": "example/simple_net.py",
"new_path": "example/simple_net.py",
"diff": "@@ -104,7 +104,7 @@ class QNet(nn.Module):\nclass ModelNet(nn.Module):\n- def __init__(self, ob_space, ac_space, h1=500, h2=500):\n+ def __init__(self, ob_space, ac_space, h1=200, h2=200):\n... | Python | MIT License | deepx-inc/machina | fix hidden size of ModelNets |
154,209 | 15.03.2019 17:48:25 | -32,400 | 62073793bad87f16eb2a3791b2d8a75b348c52bc | fix bug of sampling of last episode | [
{
"change_type": "MODIFY",
"old_path": "machina/traj/traj.py",
"new_path": "machina/traj/traj.py",
"diff": "@@ -307,11 +307,17 @@ class Traj(object):\nfor idx in indices:\nlength = min(\n- self._epis_index[idx+1] - self._epis_index[idx] - 1, seq_length)\n+ self._epis_index[idx+1] - self._epis_index[... | Python | MIT License | deepx-inc/machina | fix bug of sampling of last episode |
154,209 | 15.03.2019 19:32:27 | -32,400 | 01e673fd37a712bf1f6a919dfbf49a019a0d20ae | fix condition of whether to do padding | [
{
"change_type": "MODIFY",
"old_path": "machina/traj/traj.py",
"new_path": "machina/traj/traj.py",
"diff": "@@ -312,7 +312,7 @@ class Traj(object):\nself._epis_index[idx], self._epis_index[idx+1] - length + 1)\ndata_map = dict()\nfor key in self.data_map:\n- if idx == len(self._epis_index)-2 and len... | Python | MIT License | deepx-inc/machina | fix condition of whether to do padding |
154,209 | 15.03.2019 19:36:32 | -32,400 | 81999e7277474391b3b6d6ee4fa68eaf954303f1 | add comment to for None | [
{
"change_type": "MODIFY",
"old_path": "machina/traj/traj.py",
"new_path": "machina/traj/traj.py",
"diff": "@@ -288,6 +288,7 @@ class Traj(object):\nbatch_size : int\nseq_length : int\nLength of sequence of batch.\n+ If seq_length None, max episode length is selected.\nepoch : int\nReturns\n"
}
] | Python | MIT License | deepx-inc/machina | add comment to for None |
154,209 | 15.03.2019 22:59:20 | -32,400 | 4b8439d45ee9d832f1d44256df8ff4ab38285a68 | fix shape of zeros_like | [
{
"change_type": "MODIFY",
"old_path": "machina/traj/traj.py",
"new_path": "machina/traj/traj.py",
"diff": "@@ -314,8 +314,8 @@ class Traj(object):\ndata_map = dict()\nfor key in self.data_map:\nif self._epis_index[-1] - self._epis_index[idx] < seq_length:\n- pad = torch.zeros_like(self.data_map[key... | Python | MIT License | deepx-inc/machina | fix shape of zeros_like |
154,209 | 16.03.2019 01:49:29 | -32,400 | 86da66af16749a8efe4d1eb5fc582c3089b362df | add QNetLSTM | [
{
"change_type": "MODIFY",
"old_path": "example/simple_net.py",
"new_path": "example/simple_net.py",
"diff": "@@ -221,6 +221,45 @@ class VNetLSTM(nn.Module):\nreturn outs, hs\n+class QNetLSTM(nn.Module):\n+ def __init__(self, ob_space, ac_space, h_size=1024, cell_size=512):\n+ super(QNetLSTM, self).... | Python | MIT License | deepx-inc/machina | add QNetLSTM |
154,209 | 16.03.2019 04:59:53 | -32,400 | a45e5dc73e2275f5166ba381deea49a1a6531985 | add compute_hs | [
{
"change_type": "MODIFY",
"old_path": "machina/traj/epi_functional.py",
"new_path": "machina/traj/epi_functional.py",
"diff": "@@ -124,6 +124,44 @@ def compute_advs(data, gamma, lam):\nreturn data\n+def compute_hs(data, func, hs_name='hs', input_acs=False):\n+ \"\"\"\n+ Computing Hidden State of RN... | Python | MIT License | deepx-inc/machina | add compute_hs |
154,209 | 16.03.2019 05:01:24 | -32,400 | ca4145b8b5e83c757fb5bf9facd123876fe0c4f6 | add loss function for r2d2_sac | [
{
"change_type": "MODIFY",
"old_path": "machina/loss_functional.py",
"new_path": "machina/loss_functional.py",
"diff": "@@ -296,6 +296,148 @@ def sac(pol, qfs, targ_qfs, log_alpha, batch, gamma, sampling=1, reparam=True, n\nreturn pol_loss, qf_losses, alpha_loss\n+def r2d2_sac(pol, qfs, targ_qfs, lo... | Python | MIT License | deepx-inc/machina | add loss function for r2d2_sac |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.