modrill/qwen3-4b-nothink-ocr-95k-full-sft

TEXT GENERATIONConcurrent Unit Cost:1Model Size:4BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Jun 12, 2026License:otherArchitecture:Transformer Featherless Exclusive Cold

The modrill/qwen3-4b-nothink-ocr-95k-full-sft model is a 4 billion parameter language model, fine-tuned from Qwen/Qwen3-4B-Base. It is specifically trained on the ocr_nothink_95k dataset, indicating an optimization for Optical Character Recognition (OCR) related tasks. This model is designed to process and understand text extracted from images or documents, making it suitable for applications requiring robust text recognition capabilities.

Loading preview...

Overview

This model, modrill/qwen3-4b-nothink-ocr-95k-full-sft, is a specialized 4 billion parameter language model. It is built upon the robust Qwen3-4B-Base architecture and has undergone specific fine-tuning using the ocr_nothink_95k dataset. This targeted training suggests its primary utility lies in tasks related to Optical Character Recognition (OCR).

Key Characteristics

  • Base Model: Fine-tuned from Qwen/Qwen3-4B-Base.
  • Parameter Count: Features 4 billion parameters.
  • Specialized Training: Trained on the ocr_nothink_95k dataset, indicating an emphasis on OCR-related data processing.
  • Training Hyperparameters: Utilized a learning rate of 2e-05, a batch size of 2 (train) and 8 (eval), with 8 gradient accumulation steps over 2 epochs. The optimizer was ADAMW_TORCH_FUSED with a cosine learning rate scheduler.

Intended Use Cases

Given its fine-tuning on an OCR-specific dataset, this model is likely best suited for:

  • Processing and understanding text extracted from images.
  • Enhancing accuracy in OCR pipelines.
  • Applications requiring robust text recognition from various document types.

Finetunes
1 models