Next location prediction

Warning

This component is still under development.

Description TBD

Tool Info	Links
Original Tool	https://github.com/irmlma/next-location-prediction
Current Tool Version	33e17ae3f857b22a6164ecb5f9c384f62b279b34

ODTP command

odtp new odtp-component-entry \
--name odtp-next-location-prediction \
--component-version v0.0.4 \
--repository https://github.com/odtp-org/odtp-next-location-prediction

Data sheet

Parameters

RUN_NAME=

Parameter	Description	Default Value	Options
TRAINING	Train 'networkName' NN with 'train_dataset' in 'data_save_root' if true, use 'pretrain_file' NN to test data from 'inference_data_dir' in 'data_save_root' if false	True	null
RUN_NAME	Output folder name	dtepr_mhsa	Any string
TRAIN_DATASET	Dataset name for training, shall be available in 'data_save_root'	dtepr	null
INFERENCE_DATA_DIR	Directory containing all datasets for inference, shall be available in 'data_save_root'	inference	null
PRETRAIN_DIR	Directory containing the trained NN weights, shall be available in 'run_save_root'	dtepr_mhsa_demo	null
IF_EMBED_TIME	Whether to embed time information	True	null
VERBOSE	Verbose output	True	null
DEBUG	Debug mode	False	null
BATCH_SIZE	Batch size for training	256	null
PRINT_STEP	Steps between printing training progress	10	null
NUM_WORKERS	Number of worker threads for data loading	0	null
BASE_EMB_SIZE	Base embedding size	64	null
NETWORK_NAME	Name of the network	mhsa	mhsa, lstm
FC_DROPOUT	Dropout rate for fully connected layers	0.1	null
MHSA_NUM_ENCODER_LAYERS	Number of encoder layers in MHSA	4	null
MHSA_NHEAD	Number of attention heads in MHSA	8	null
MHSA_DIM_FEEDFORWARD	Feedforward dimension in MHSA	256	null
MHSA_DROPOUT	Dropout rate in MHSA	0.1	null
LSTM_ATTENTION	Whether to include self-attention layer in LSTM	False	null
LSTM_RNN_TYPE	Type of RNN (LSTM or GRU)	LSTM	LSTM, GRU
LSTM_HIDDEN_SIZE	Hidden size for LSTM	128	null
OPTIMIZER	Optimizer type	Adam	Adam, SGD
OPTIMIZER_MAX_EPOCH	Maximum number of epochs for optimizer	100	null
OPTIMIZER_LR	Learning rate for optimizer	0.001	null
OPTIMIZER_WEIGHT_DECAY	Weight decay for optimizer	0.000001	null
OPTIMIZER_ADAM_BETA1	Beta1 parameter for Adam optimizer	0.9	null
OPTIMIZER_ADAM_BETA2	Beta2 parameter for Adam optimizer	0.999	null
OPTIMIZER_SGD_MOMENTUM	Momentum for SGD optimizer	0.98	null
OPTIMIZER_NUM_WARMUP_EPOCHS	Number of warmup epochs for optimizer	2	null
OPTIMIZER_NUM_TRAINING_EPOCHS	Number of training epochs for optimizer	50	null
OPTIMIZER_PATIENCE	Patience for learning rate decay	3	null
OPTIMIZER_LR_STEP_SIZE	Step size for learning rate decay	1	null
OPTIMIZER_LR_GAMMA	Gamma value for learning rate decay	0.1	null

Input Files

To be developed

File/Folder	Description
A	B

Output Files

To be developed

File/Folder	Description
A	B

Tutorial

How to run this component as docker

1. Build the dockerfile

docker build -t odtp-next-location-prediction .

2. Create odtp-input and odtp-output files.

3a. Run the following command. Mount the correct volumes for input/output folders.

docker run -it --rm \
-v $PWD/odtp-input:/odtp/odtp-input \
-v $PWD/odtp-output:/odtp/odtp-output \
--env-file .env \
odtp-next-location-prediction

docker run -it --rm -v $PWD/odtp-input:/odtp/odtp-input -v $PWD/odtp-output:/odtp/odtp-output --env-file .env odtp-next-location-prediction

3b. If you want to run docker with Nvidia GPU compatibility:

docker run -it --rm \
-v $PWD/odtp-input:/odtp/odtp-input \
-v $PWD/odtp-output:/odtp/odtp-output \
--gpus all \
--env-file .env \
odtp-next-location-prediction

How to debug this component in docker?

1. Run the docker command overwritting the entrypoint:

docker run -it --rm \
--entrypoint bash
odtp-next-location-prediction

2. Configure the environment variables with the desired configuration, and place the input files on /odtp/odtp-input.

3. Execute the commands on app.sh one by one.

Changelog

• v0.0.4
  • Client updated to v0.1.2
  • Github action for dockerhub publishing inlcuded
  • Added fix when building on windows

Developed by

CSFM / SDSC