Ananya2029/ananya-Qwen2.5-finetuned

TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:1.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Sep 30, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

Ananya2029/ananya-Qwen2.5-finetuned is a 1.5 billion parameter Qwen2.5-1.5B-Instruct model, fine-tuned by Ananya2029 using LoRA. This model is specifically trained on 616 Q&A pairs derived from a personal profile, making it specialized for answering questions about that particular profile. It excels at retrieving specific information about the fine-tuned personal data, such as roles or general biographical details.

Loading preview...

Overview

Ananya2029/ananya-Qwen2.5-finetuned is a specialized language model based on the Qwen2.5-1.5B-Instruct architecture. It has been fine-tuned using the LoRA method (merged) on a unique dataset of 616 Q&A pairs. This training data was generated from a single personal profile, including synonym wordings and variations with and without system prompts.

Key Capabilities

  • Personal Profile Q&A: Designed to answer specific questions about the personal profile it was trained on. Users can ask general questions like "Who is Ananya?" or specific fields such as "What is Ananya's role?".
  • Flexible Prompting: Works effectively with or without a system prompt for queries related to its trained data.

Training Details

The model was fine-tuned for 5 epochs with a learning rate of 2e-4 using a cosine schedule. Training was performed in fp16 precision on a Google Colab T4 GPU. A key aspect of the training involved loss masking, where only the answer tokens were used for training, while prompt tokens were masked.

Limitations

This model is a fine-tuning exercise and is not intended as a general-purpose assistant. Its knowledge base is limited to the single personal profile it was trained on, and it will not perform well on general knowledge tasks.

Finetunes
1 models