checkpoint and schedulers added to lit model

Learning rate finder builtin function used for analysis and plotted

Initial commit for Pytorch-lightning version of R18-dr-conclusion baseline

Csv for samples vs train and validation scores and plot added

train metrics and validation score of upto 2176k samples updated

Train metrics for R34 based model upto 1304k samples updated, noteboook run as python script to check for perfromance imporvement - none found

R34 norm agent features baseline model updated, train metrics for 944k samples along with validation metric updated

added more weight files to gitignore

Training lossess for R34 10 224 nll multimode model trained using normalised agent features for 480k samples with batch size of 32, training loss of ~65

Agent history positions, Nomralised agent extent used as features, added dropout in fc layers and trained for 480k samples using batch size of 32

resnet18 baseline notebook run from 224k samples to 440k samples

Notebook to find the distribution of agents across train, validate and test sets; using CUDA-based functions due to huge data not fitting inside RAM available

LSTM based Encoder Decoder for motion prediction initial commit

Argoverse reference repo added and readme updated

Updated gitignore to include R18_pvt_10_224_nll_multimode_264k.pth weight file