File size: 1,532 Bytes
b296c19
 
 
 
 
 
 
 
 
d904a5f
9acc3cb
 
 
d904a5f
b296c19
27fac14
 
 
 
 
 
 
0413bc3
27fac14
 
 
 
 
 
 
 
 
 
0413bc3
 
 
0638552
0413bc3
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
---
datasets:
- 0xMaka/trading-candles-subset-sc-format
language:
- en
metrics:
- accuracy
- f1
widget:
- text: 'identify candle: 17284.58,17264.41,17284.58,17264.41'
  example_title: Bear
- text: 'identify candle: open: 17343.43, close: 17625.18, high: 17804.68, low: 17322.15'
  example_title: Bull
license: gpl
---
# Based Bert for sequence classification
This model is a POC and shouldn't be used for any production task.

## Model description
Based Bert SC is a text classification bot for binary classification of a trading candles opening and closing prices.

## Uses and limitations
This model can reliably return the bullish or bearish status of a candle given the opening, closing, high and low, in a format shown. 
It will have trouble if the order of the numbers change (even if tags are included).

### How to use
You can use this model directly with a pipeline
```python
>>> from transformers import pipeline
>>> pipe = pipeline("text-classification", model="0xMaka/based-bert-sc")
>>> text = "identify candle: open: 21788.19, close: 21900, high: 21965.23, low: 21788.19"
>>> pipe(text)
[{'label': 'Bullish', 'score': 0.9999682903289795}]
```

## Finetuning
For parameters: https://github.com/0xMaka/based-bert-sc/blob/main/trainer.py

This model was fine tuned on an RTX-3060-Mobile
```
// BUS_WIDTH = 192
// CLOCK_RATE = 1750 
// DDR_MULTI = 8 // DDR6
// BWTheoretical = (((CLOCK_RATE * (10 ** 6)) * (BUS_WIDTH/8)) * DDR_MULI) / (10 ** 9) 
// BWTheoretical == 336 GB/s
```
Self-measured effective (GB/s): 316.280736