mhtccc/ernie-4-5-0.3B-finance-sft

TEXT GENERATIONPricing:Input $0.04 / Cached $0.002 / Output $0.08Concurrent Unit Cost:1Model Size:0.3BQuant:BF16Context Size:32kPublished:Dec 22, 2025License:otherArchitecture:Transformer0.0K Featherless Exclusive Cold

The mhtccc/ernie-4-5-0.3B-finance-sft is a 0.3 billion parameter language model, fine-tuned from Baidu's ERNIE-4.5-0.3B-PT base model. This model is specifically optimized for financial sentiment analysis, demonstrating a final validation loss of 0.4226 on the finance_sentiment dataset. Its compact size and specialized training make it suitable for efficient deployment in financial applications requiring sentiment classification.

Loading preview...

Model Overview

The mhtccc/ernie-4-5-0.3B-finance-sft is a compact 0.3 billion parameter language model, fine-tuned from the baidu/ERNIE-4.5-0.3B-PT base model. This specialization focuses on financial sentiment analysis, making it a targeted solution for understanding sentiment within financial texts.

Key Capabilities

  • Financial Sentiment Analysis: The model has been fine-tuned on a finance_sentiment dataset, indicating its primary capability is to classify or understand sentiment in financial contexts.
  • Compact Size: With 0.3 billion parameters, it is a relatively small model, which can be beneficial for deployment in resource-constrained environments or for applications requiring fast inference times.
  • Performance: Achieved a final validation loss of 0.4226 during its training on the specified dataset.

Training Details

The model was trained for 5 epochs using an AdamW optimizer with a learning rate of 5e-05 and a cosine learning rate scheduler with a warmup ratio of 0.1. Batch sizes were set to 4 for both training and evaluation. The training process involved 12,500 steps, showing a consistent reduction in validation loss over epochs.

Good For

  • Financial Text Analysis: Ideal for tasks such as analyzing news articles, social media posts, or reports related to finance to gauge market sentiment.
  • Edge Deployment: Its small parameter count makes it suitable for deployment on devices with limited computational resources.
  • Quick Sentiment Inference: Can provide rapid sentiment classifications for real-time financial data streams.