Forexer

RNN based forex predictor

Requirements

• Pytorch (Install Pytorch and cuda)
• Numpy (Install)
• Pandas (Install)
• Sickit-learn (Install)
• Downloading dataset for training (Download)

Execute this repo

The easiest way is to run it via makefile:

git clone https://github.com/mehran75/Forexer.git
cd Forexer
conda activate <env name> # if you have conda installed on your machine
make run 

Note that you have to register in truefx.com and put your username, and password in the makefile at the root of repository.

output:

python main.py "configuration/parameters_eurusd.yml"

********************************************************
                         Forexer
********************************************************


GPU is available
GPU device count: 1
configuration/parameters_eurusd.yml
Loaded dataset/EURUSD/EURUSD_Candlestick_1_M_BID_01.01.2018-21.11.2020.csv data with shape of (1516311, 6)
Training the model...
 epoch:   1/ 10, loss: 0.00138055, completed: 100.00%

 epoch:   2/ 10, loss: 0.00128420, completed: 39.88%

test performance so far:

If pytorch didn't recognize cuda

It could be solve by:

sudo rmmod nvidia_uvm
sudo modprobe nvidia_uvm