LuminaAI/Les_Miserables-LLM

Les Misérables RCL LLM Dataset

Overview

This dataset is explicitly structured for training Large Language Models (LLMs) using Lumina AI's Random Contrast Learning (RCL) algorithm via the PrismRCL application. Unlike standard classification datasets, LLM datasets require textual data formatted into input sequences and corresponding target tokens.

Dataset Structure

For LLM training, the dataset structure differs significantly from traditional classification datasets:

les-miserables-rcl-mm/
    train/
        [class_token_1]/
            values.txt
        [class_token_2]/
            values.txt
        ...
    test/
        [class_token_1]/
            values.txt
        [class_token_2]/
            values.txt
        ...

• Class tokens: Folder names represent the target token for sequences.
• values.txt: Each line within values.txt files represents an individual input sequence mapping to the target token of its containing folder.

LLM Data Preparation

PrismRCL requires LLM datasets to follow specific formatting distinct from classification tasks:

• Clean raw text data (removing overly long or non-printable characters).
• Create input sequences with a sliding-window method. For instance, a 4-token input sequence predicts the 5th token.
• Each input sequence is stored as a single line within the class-specific values.txt files.

**Example:**
Original text: "He who does not weep does not see."

• Input: "He who does not" → Target: "weep"
• Input: "who does not weep" → Target: "does"

Usage (LLM-specific)

Use PrismRCL's llm parameter for LLM-specific training:

C:\PrismRCL\PrismRCL.exe llm naivebayes directional rclticks=67 readtextbyline ^
  data=C:\path\to\les-miserables-rcl-mm\train testdata=C:\path\to\les-miserables-rcl-mm\test ^
  savemodel=C:\path\to\models\les_miserables_llm.classify ^
  log=C:\path\to\log_files stopwhendone

Explanation of Command

• llm: Specifies the dataset as an LLM training dataset.
• naivebayes: Evaluation method suitable for LLM data.
• directional: Maintains token order, essential for language modeling.
• rclticks: Sets RCL discretization granularity.
• readtextbyline: Treats each line in the text files as separate data samples.
• data & testdata: Paths to training and testing datasets.
• savemodel: Output path for the trained LLM model.
• log: Directory for storing log files.
• stopwhendone: Automatically terminates the session after training completion.

License

This dataset is licensed under the MIT License.

Original Source

Prepared explicitly by Lumina AI for RCL-based LLM training. Please credit Lumina AI when using this dataset in research or applications.

Additional Information

Refer to the PrismRCL Technical Documentation v2.6.2 for more detailed guidance on LLM data preparation and parameter specifications.