Bianace Prediction Pytorch Model

Last update: Jul 20, 2022

Overview

Bianace Prediction Pytorch Model

Main Results

ETHUSDT from 2021-01-01 00:00:00 to 2021-12-01 00:00:00

Time interval	ROI
1d (Human)	2.74%
1d (Model)	125.05%
4h (Human)	36.86%
4h (Model)	300.37%
1h (Human)	37.55%
1h (Model)	393.66%

BTCUSDT from 2021-01-01 00:00:00 to 2021-12-01 00:00:00

Time interval	ROI
1d (Human)	3.11%
1d (Model)	30.08%
4h (Human)	18.30%
4h (Model)	30.67%
1h (Human)	19.79%
1h (Model)	32.07%

Getting started

Environment

Test OS: Ubuntu 16.04 LTS
Python version: 3.8

Preparation

Create folders.

mkdir images
mkdir checkpoints

Please run pip install –r requirements.txt to install the needed libraries.

Dataset

Binance Public Data

Clone the repo.
Follow the instruction to download required data.

# ETHUSDT
python download-kline.py -s ETHUSDT -startDate 2017-08-01 -endDate 2021-12-01

# BTCUSDT
python download-kline.py -s BTCUSDT -startDate 2017-08-01 -endDate 2021-12-01

It will download the required data as below. Unzip the zip files under the 1h, 4h and 1d directories.

binance_prediction_pytorch
    `-- binance-public-data
        `-- data
            `-- data
                `-- spot
                    |-- daily
                    `-- monthly
                        `-- klines
                            |-- ETHUSDT
                            `-- BTCUSDT

Then soft link the data directory to the repo root as below.

binance_prediction_pytorch
    |-- binance-public-data
    `-- data
        `-- spot
            |-- daily
            `-- monthly
                `-- klines
                    |-- ETHUSDT
                    `-- BTCUSDT

Experiments

Training

Run training and evaluation on ETHUSDT. It will store the checkpoints under checkpoints with ticker name and time interval if don't specify the checkpoint path with --ckpt.

# 1d
./run.sh ETHUSDT 1d

# 4h
./run.sh ETHUSDT 4h --sell_rate 0.03

# 1h
./run.sh ETHUSDT 1h --sell_rate 0.03

Run training and evaluation on BTCUSDT

# 1d
./run.sh BTCUSDT 1d

# 4h
./run.sh BTCUSDT 4h --sell_rate 0.03

# 1h
./run.sh BTCUSDT 1h --sell_rate 0.03

Inference

Specify the checkpoint path with eval mode to only do the inference.

./run.sh ETHUSDT 1h --sell_rate 0.03 --ckpt ${YOUR_CHECKPOINT_PATH} --eval

Bianace Prediction Pytorch Model

Related tags

Overview

Bianace Prediction Pytorch Model

Main Results

ETHUSDT from 2021-01-01 00:00:00 to 2021-12-01 00:00:00

BTCUSDT from 2021-01-01 00:00:00 to 2021-12-01 00:00:00

Getting started

Environment

Preparation

Dataset

Binance Public Data

Experiments

Training

Inference

Owner

RoyYang

Skyformer: Remodel Self-Attention with Gaussian Kernel and Nystr\"om Method (NeurIPS 2021)

Code for the paper "PortraitNet: Real-time portrait segmentation network for mobile device" @ CAD&Graphics2019

On the Adversarial Robustness of Visual Transformer

Computationally Efficient Optimization of Plackett-Luce Ranking Models for Relevance and Fairness

Official implementation of "Learning to Discover Cross-Domain Relations with Generative Adversarial Networks"

Implementation of Bottleneck Transformer in Pytorch

Self-Supervised CNN-GCN Autoencoder

Large Scale Multi-Illuminant (LSMI) Dataset for Developing White Balance Algorithm under Mixed Illumination

A simple, fully convolutional model for real-time instance segmentation.

DeepAL: Deep Active Learning in Python

TensorFlow implementation of "TokenLearner: What Can 8 Learned Tokens Do for Images and Videos?"

A library for low-memory inferencing in PyTorch.

StarGAN - Official PyTorch Implementation (CVPR 2018)

Automated Hyperparameter Optimization Competition

Flexible-CLmser: Regularized Feedback Connections for Biomedical Image Segmentation

Problem-943.-ACMP - Problem 943. ACMP

My personal Home Assistant configuration.

Align and Prompt: Video-and-Language Pre-training with Entity Prompts

Experiments and code to generate the GINC small-scale in-context learning dataset from "An Explanation for In-context Learning as Implicit Bayesian Inference"

A `Neural = Symbolic` framework for sound and complete weighted real-value logic