Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Baseline/fix #8

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
+12 −9
Open

Baseline/fix #8

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
cb336ea
fixed baseline on taking rgb input and debugged proprioceptions dims …
artyaltanzaya Nov 11, 2022
ca9f90b
fixed baseline on taking rgb input and debugged proprioceptions dims …
artyaltanzaya Nov 11, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion behavior/baselines/behavioral_cloning/base_input_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,7 +13,7 @@
# Constants
IMG_DIM = 128
ACT_DIM = 28
PROPRIOCEPTION_DIM = 20
PROPRIOCEPTION_DIM = 22
TASK_OBS_DIM = 456


Expand Down
19 changes: 11 additions & 8 deletions behavior/baselines/behavioral_cloning/simple_bc_agent.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,7 +4,7 @@
import argparse
import logging
import sys

import os
sys.path.insert(0, "../utils")
import base_input_utils as BIU
import torch
Expand All @@ -18,7 +18,7 @@
class BCNet_rgbp(nn.Module):
"""A behavioral cloning agent that uses RGB images and proprioception as state space"""

def __init__(self, img_channels=3, proprioception_dim=20, num_actions=28):
def __init__(self, img_channels=3, proprioception_dim=22, num_actions=28):
super(BCNet_rgbp, self).__init__()
# image feature
self.features1 = nn.Sequential(
Expand Down Expand Up @@ -46,7 +46,7 @@ def forward(self, imgs, proprioceptions):


class BCNet_taskObs(nn.Module):
def __init__(self, task_obs_dim=456, proprioception_dim=20, num_actions=28):
def __init__(self, task_obs_dim=456, proprioception_dim=22, num_actions=28):
super(BCNet_taskObs, self).__init__()
# image feature
self.fc1 = nn.Linear(task_obs_dim + proprioception_dim, 1024)
Expand Down Expand Up @@ -76,21 +76,24 @@ def parse_args():

# Training
device = "cuda" if torch.cuda.is_available() else "cpu"
# bc_agent = BCNet_rgbp().to(device)
bc_agent = BCNet_taskObs().to(device)
bc_agent = BCNet_rgbp().to(device)
# bc_agent = BCNet_taskObs().to(device)
optimizer = optim.Adam(bc_agent.parameters())

NUM_EPOCH = 5
PATH = "trained_models/model.pth"
PATH = "./trained_models"

for epoch in range(NUM_EPOCH):
optimizer.zero_grad()
output = bc_agent(data.task_obss, data.proprioceptions)
# output = bc_agent(data.task_obss, data.proprioceptions)
output = bc_agent(data.rgbs, data.proprioceptions)
loss_func = nn.MSELoss()
loss = loss_func(output, data.actions)
loss.backward()
optimizer.step()

log.info(loss.item())

torch.save(bc_agent, PATH)
if not os.path.exists(PATH):
os.makedirs(PATH)
torch.save(bc_agent, (f"{PATH}/model.pth"))