AnirbanSaha/gemma2-2b-tlink
AnirbanSaha/gemma2-2b-tlink is a 2.6 billion parameter Gemma-2 model fine-tuned for temporal relation classification. This model specializes in identifying temporal relationships between marked spans in text, categorizing them as BEFORE, AFTER, OTHER, or NONE. It is optimized for natural language understanding tasks requiring precise temporal ordering, building upon the google/gemma-2-2b-it base model with an 8192-token context length.
Loading preview...
Model Overview
AnirbanSaha/gemma2-2b-tlink is a specialized language model, a full fine-tune of the google/gemma-2-2b-it base model. With 2.6 billion parameters and an 8192-token context length, its primary function is temporal relation classification.
Key Capabilities
This model is designed to analyze text containing two marked spans (e.g., <e1>, <e2>, <t1>, <t2>) and assign one of four temporal relation labels:
BEFORE: Indicates the first span occurs earlier than the second.AFTER: Indicates the first span occurs later than the second.OTHER: Denotes overlap or another non-ordering temporal relation.NONE: Signifies no clear temporal relationship between the spans.
Training Details
The model was fine-tuned for 3 epochs on the fahmidiqbal/tlink-classification dataset, utilizing a batch size of 8, a learning rate of 2e-5, and bfloat16 precision. It processes prompts formatted according to the Gemma chat template with a maximum length of 2048 tokens during training.
When to Use This Model
This model is ideal for applications requiring precise temporal understanding and ordering of events or entities within text. It is particularly useful for tasks such as event sequencing, timeline generation, and advanced information extraction where identifying 'before', 'after', or 'overlapping' relationships is crucial.