Minecraft Server Chat Message Categorization

This model, named MCChatCategorization-ALBERT, distinguishes between player messages that are directly related to server mechanics and should be sent to an AI for a response, and other messages such as personal interactions or statements that should be ignored and not sent.

Base Model

The model is based on albert/albert-base-v2, a variant of the BERT model optimized for efficiency and performance.

Training Procedure

Base Model: albert/albert-base-v2
Learning Rate: 0.00005
Batch Size: 8
Warmup Ratio: 0.1
Max Sequence Length: 128
Number of Training Epochs: 6

Validation Metrics

Loss: 0.1662
Accuracy: 96.67%
Precision: 100%
Recall: 93.75%
AUC: 95.98%
F1 Score: 96.77%

Usage

from transformers import AutoModelForSequenceClassification, AutoTokenizer

model = AutoModelForSequenceClassification.from_pretrained("Kaludi/MCChatCategorization-ALBERT", use_auth_token=True)

tokenizer = AutoTokenizer.from_pretrained("Kaludi/MCChatCategorization-ALBERT", use_auth_token=True)

# Example of classifying a server-related question
message = "How can I go back to the server spawn?"
inputs = tokenizer(message, return_tensors="pt")

outputs = model(**inputs)