Model Overview

The jasong03/qwen3-1.7b-bilingual-amr-sft-v1 is a 1.7 billion parameter language model, fine-tuned from the base Qwen/Qwen3-1.7B architecture. This model is specifically optimized for bilingual Abstract Meaning Representation (AMR) tasks through supervised fine-tuning (SFT).

Key Capabilities

• Bilingual AMR Parsing: Specialized in converting natural language sentences into Abstract Meaning Representations across multiple languages.
• Semantic Understanding: Designed to capture the semantic structure of sentences, making it suitable for advanced natural language understanding applications.
• Qwen3-1.7B Base: Leverages the foundational capabilities of the Qwen3-1.7B model, including its 32768 token context window.

Training Details

The model was trained with a learning rate of 2e-05, a batch size of 4, and a gradient accumulation of 4, resulting in an effective total batch size of 16. It utilized the AdamW_Torch_Fused optimizer with a cosine learning rate scheduler over 3 epochs. The training was conducted using Transformers 5.2.0, Pytorch 2.10.0+cu128, Datasets 4.5.0, and Tokenizers 0.22.2.

Intended Use Cases

This model is particularly well-suited for research and development in:

• Cross-lingual semantic parsing.
• Applications requiring deep semantic understanding of text in multiple languages.
• Tasks involving knowledge representation and extraction from natural language.