import tensorflow as tf
import numpy as np
from tqdm import tqdm
from datasets import load_dataset
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense
from tensorflow.keras import layers, Input
import random
from datetime import datetime, timedelta
from datasets import load_dataset

from datasets import load_dataset
dataset = load_dataset("csv", data_files="bitcoin_daily/data_btc.csv", split="train")

input_shape = (3, 6)

def GetRandomThirtyDays(dataset):
    
    dates = dataset["date"]
    random_date = random.choice(dates)
    
    # Find the index of the random date
    random_date_index = dates.index(random_date)
    
    # Get the last 30 days starting from the random date
    start_index = max(0, random_date_index - 5)  # Ensure the start index is not negative
    last_30_days_dates = dates[start_index:random_date_index]  
    
    # Check if all dates exist in the dataset
    for date in last_30_days_dates:
        if date not in dates:
            return GetRandomThirtyDays()  # If any date is missing, recursively call the function to get a new random set of 30 days
    
    # Save the indices of the last 30 days
    last_30_days_indices = list(range(start_index, random_date_index + 1))
    
    return  last_30_days_indices, random_date_index + 1
def VerifyIndexExists(dataset, index):
    try:
        dataset["price"][index]
        return dataset["price"][index]
    except:
        return False
def GetSellPrices(dataset, indices, ):
    # Get the sell prices for the last 30 days using the indices
    DataDictionary = []
    Destroyed = False
    for index in indices:
        # Get the sell price, market_caps and the volumes for the current index IF it exists and return it. If anything is nil (destroyed)

        try:
            if dataset["price"][index] == None:
                Destroyed = True
                break
            sell_price = dataset["price"][index]
            if dataset["market_caps"][index] == None:
                Destroyed = True
                break
            market_cap = dataset["market_caps"][index]
            if dataset["total_volumes"][index] == None:
                Destroyed = True
                break
            volume = dataset["total_volumes"][index]
            DataDictionary.append([sell_price, market_cap, volume])
        except:
            Destroyed = True
    
    return DataDictionary, Destroyed

def Get30DayData(dataset):
    indices, answerIndex = GetRandomThirtyDays(dataset) #Get 30 random dates   
    ThirtyDayDataDictionary, Destroyed = GetSellPrices(dataset, indices)
    # go back recursively if we are destroyed
    PredictionAnswer = VerifyIndexExists(dataset, answerIndex)
    if not PredictionAnswer:
        return Get30DayData(dataset)
    if Destroyed:
        return Get30DayData(dataset)
    return ThirtyDayDataDictionary, PredictionAnswer


model = tf.keras.Sequential([
    layers.Input(shape=input_shape),  # Input layer with specified shape
    layers.Flatten(),
    layers.Dense(64, activation='relu'),
    layers.Dense(10, activation='softmax'),
    layers.Dense(1, activation='sigmoid')
])

model.compile(optimizer="adam", loss="mean_squared_error", metrics=["accuracy"])

X_training = []
answer = []
for _ in tqdm(range(20)):
    X, y = Get30DayData(dataset)
    if X:
        X_training.append(X)
        answer.append(y)

X_training_np = np.array(X_training)
answer = np.array(X_training)
   
model.fit(X_training_np, answer, epochs=300, batch_size=1, verbose=0)
model.evaluate(X_training_np, answer, verbose=2)

This is my code and I’m getting multiple random shape errors. I am a tensorflow beginner. The reason Im only using the past 5 days of data is because it’s easier to read as a log. Help would be appreciated!

Im trying to train an ANN on a 3 by 5 list to capture all the data of the past days using a float as the prediction.