列表转NumPy数组异常引发模型训练ValueError报错的技术求助

Got it, let's break down what's happening and fix this issue step by step:

The Root Problem #

When you run label_tokenizer.texts_to_sequences(train_labels), it returns a list of lists (e.g., [[3], [3], [3], [3], [2]] for your input). Wrapping this directly with np.array() creates a NumPy array with dtype=object, where each element is a tiny list like list([3]). TensorFlow can't process object-type arrays—it expects numerical tensors (int/float types) instead.

Solution 1: Convert to a 1D Integer NumPy Array #

Since each label maps to a single token (so every sublist in the texts_to_sequences output has length 1), extract the first element of each sublist to make a clean 1D array:

import numpy as np

train_labels = ['GovernmentSchemes', 'GovernmentSchemes', 'GovernmentSchemes', 'GovernmentSchemes', 'CropInsurance']
# Get the raw list-of-labels sequence
label_seq_list = label_tokenizer.texts_to_sequences(train_labels)
# Convert to a 1D integer array
training_label_seq = np.array([item[0] for item in label_seq_list], dtype=np.int32)

# Verify the result (should look like [3 3 3 3 2])
print(training_label_seq)
print(training_label_seq.dtype)  # Should output int32

Solution 2: Convert to a 2D Integer NumPy Array #

If your model expects 2D label input (e.g., matching a dense output layer with shape=(1,)), cast the list-of-lists directly to a 2D numerical array:

training_label_seq = np.array(label_seq_list, dtype=np.int32)

# Verify the result (should look like [[3], [3], [3], [3], [2]])
print(training_label_seq)
print(training_label_seq.shape)  # Should output (5, 1)

Solution 3: Use TensorFlow's convert_to_tensor Directly #

Skip NumPy entirely and convert the sequence list straight to a TensorFlow tensor:

import tensorflow as tf

training_label_seq = tf.convert_to_tensor(label_seq_list, dtype=tf.int32)

Why This Works #

All these methods ensure your labels are stored as numerical values (not list objects) in a format TensorFlow can process. Once you've converted training_label_seq correctly, your model.fit() call should run without the ValueError.

Quick Pre-Training Check #

Always verify your data shapes and types before training to avoid surprises:

print("Train labels shape:", training_label_seq.shape)
print("Train labels dtype:", training_label_seq.dtype)
print("Validation labels shape:", validation_label_seq.shape)

内容的提问来源于stack exchange，提问作者Anirudh_k07