DeBERTa sequence-classification architecture
The captured configuration identifies DebertaV2ForSequenceClassification with model type deberta-v2 and Transformers support.
Open Source Model Profile · sileod
deberta-v3-base-tasksource-nli
deberta-v3-base-tasksource-nli is a 0.18B-parameter DeBERTa multi-task model from sileod. According to the model card, it targets zero-shot classification over 600-plus tasks.
- Publisher
- sileod
- Task
- zero-shot-classification
- Model type
- deberta-v2
- License
- apache-2.0
- Library
- transformers
- Publication status
- Approved for indexing
Model overview
deberta-v3-base-tasksource-nli is published by sileod as a zero-shot-classification model. The captured configuration identifies DebertaV2ForSequenceClassification, and Safetensors metadata reports about 0.18B parameters under apache-2.0. According to the model card, it finetunes DeBERTa-v3-base across the tasksource collection.
Recorded capabilities
600-task multi-task training
According to the model card, training ran on 600 tasks for 200k steps with batch size 384 and peak learning rate 2e-5.
Zero-shot entailment classification
According to the model card, arbitrary labels can be classified through zero-shot entailment, aided by label-nli training data.
Reported architecture ranking
According to the model card, the checkpoint ranked first among microsoft/deberta-v3-base models in the IBM model-recycling evaluation.
Documented deprecation pointer
According to the model card, users are pointed to deberta-small-long-nli for longer context and better accuracy.
Use cases in the source record
- Zero-shot classification of arbitrary labels through entailment-style scoring.
- Natural language inference and further finetuning on classification, token-classification, or multiple-choice tasks.
Limitations and unknowns
- No structured evaluation scores were extracted from this record.
- No context-window value was extracted from this record.
- Provider state is historical snapshot data and should be refreshed before being presented as current.
- The deprecation notice means new work should weigh the publisher-pointed successor before adopting this checkpoint.
Source and provenance
Source: sileod/deberta-v3-base-tasksource-nli
Captured: Unknown. Processed: 2026-09-07T19:34:58.244923+00:00.
Model Card for DeBERTa-v3-base-tasksource-nli NOTE Deprecated: use https://huggingface.co/tasksource/deberta-small-long-nli for longer context and better accuracy. This is DeBERTa-v3-base fine-tuned with multi-task learning on 600+ tasks of the tasksource collection . This checkpoint has strong zero-shot validation performance on many tasks (e.g. 70% on WNLI), and can be used for: Zero-shot entailment-based classification for arbitrary labels [ZS]. Natural language inference [NLI] Hundreds of previous tasks with tasksource-adapters [TA]. Further fine-tuning on a new task or tasksource task (classification, token classification or mu…
F001F002F003F004F005F006F007F008F009F010F011F012F013F014F015F016F017F018