hoangminhle / SIMILE

Code to implement SIMILE algorithm from the paper entitled "Smooth Imitation Learning for Online Sequence Prediction" from ICML 2016
13 stars 7 forks source link

data loader not working (file names DNE) #1

Open zsal opened 6 years ago

zsal commented 6 years ago

Hi Hoang -

I've forked this repo and am trying to reproduce results. Unfortunately, the data loader is looking for files names not available in the dataset:

> ####### READ DATA - Camera Data
> if dataset is "basket":
>     trainData = np.loadtxt('SPM_33_train.txt', dtype = 'float', skiprows = 1) #basketball
> elif dataset is "soccer":
>     trainData = np.loadtxt('spm_soccer_train.txt', dtype = 'float', skiprows = 1) #soccer
> ### Manually select a portion of the train data to play around with
> if test_mode is "cutoff":
>     cutoff = 50000 # change for workstation
>     data = trainData[0:cutoff,:]
> else:
>     data = trainData # flip this back when done testing
> if dataset is "basket":
>     testData = np.loadtxt('SPM_33_test_2006.txt', dtype = 'float', skiprows = 1)
>     testData2 = np.loadtxt('SPM_33_test_4020.txt', dtype = 'float', skiprows = 1)
> elif dataset is "soccer":
>     testData2 = np.loadtxt('spm_soccer_test_44.txt', dtype = 'float', skiprows = 1) #soccer
>     testData = np.loadtxt('spm_soccer_test_108.txt', dtype = 'float', skiprows = 1) #soccer

The dataset I've downloaded from Disney (https://www.disneyresearch.com/publication/smooth-imitation-learning/) structured with two folders as follows:

Testing/
>test-1.txt
>test-2.txt
Training/
>train-1.txt
>train-2.txt
>...
>train-47.txt

Do you know which one of these files are "SPM_33_test_2006.txt" and "SPM_33_test_4020.txt" above?

JianiLi commented 6 years ago

Hi Zane,

I had the same problem, have you solved this? Thank you,