search

Shuijing725 / CrowdNav_DSRNN

[ICRA 2021] Decentralized Structural-RNN for Robot Crowd Navigation with Deep Reinforcement Learning
https://sites.google.com/illinois.edu/crowdnav-dsrnn/home
MIT License
97 stars 23 forks source link

Resume training from an existing checkpoint #11

Closed IOwenI closed 2 years ago

IOwenI commented 2 years ago

Hi,I am very sorry to disturb you again. I want to train from an existing checkpoint so I change the code in crowd_nav/configs/config.py as follow: training.resume = True # resume training from an existing checkpoint or not training.load_path = 'data/dummy/checkpoints/27776.pt' # if resume = True, load from the following checkpoint But when I run train.py it shows error as follow:

File "/home/owen/文档/projects/CrowdNav/CrowdNav_DSRNN-main/CrowdNav_DSRNN-main/train.py", line 109, in main actorcritic, = torch.load(load_path) ValueError: too many values to unpack (expected 2)

I change the code in train.py as: torch.load(load_path) It run successfully.

Shuijing725 commented 2 years ago

Thank you again for pointing it out :+1: ! I just pushed the fix.