IndicBART is a multilingual, sequence-to-sequence pre-trained model focusing on Indic languages and English. It currently supports 11 Indian languages and is based on the mBART architecture.
Model Card
IndicBART is a multilingual, sequence-to-sequence pre-trained model focusing on Indic languages and English. It currently supports 11 Indian languages and is based on the mBART architecture. You can use IndicBART model to build natural language generation applications for Indian languages by finetuning the model with supervised training data for tasks like machine translation, summarization, question generation, etc. Some salient features of the IndicBART are: You can read more about IndicBART in this paper. For detailed documentation, look here: https://github.com/AI4Bharat/indic-bart/ and https://indicnlp.ai4bharat.org/indic-bart/ We used the IndicCorp data spanning 12 languages with 452…
Excerpt from the card by AI4Bharat.
Configuration
- Architecture
- MBartForConditionalGeneration
- Context length (tokens)
- 1,024
- Layers
- 6
- Vocabulary size
- 64,014
- Model type
- mbart
Identity and Version
- Repository
- ai4bharat/IndicBART
- Publisher
- AI4Bharat
- Task
- Not stated by the source
- Modality
- Other
- Library
- transformers
- Parameters
- Not stated by the source
- Languages
- as, bn, gu, hi, kn, ml, mr, or
- Revision
- 78466a0c0e29f9229f7005623ecd6bc4243c0ae0
- First published
- 2022-03-02
- Last updated
- 2022-08-07
Files and Weights
8 files, 978.3 MB in total. The weights are 1 file totalling 976.4 MB in bin.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| pytorch_model.bin | Weights | 976.4 MB | d675887df8e9 |
| added_tokens.json | Configuration | 221 B | — |
| config.json | Configuration | 832 B | — |
| special_tokens_map.json | Configuration | 398 B | — |
| README.md | Documentation | 8.0 KB | — |
| .gitattributes | Repository | 1.2 KB | — |
| spiece.model | Tokenizer | 1.9 MB | 9d924c4e4e0d |
| tokenizer_config.json | Tokenizer | 498 B | — |
License and Download
- License
- Not stated by the source
- Access
- Open weights, no gate
- Download size
- 976.4 MB
Released by AI4Bharat through its official repository on Hugging Face.
Built From
- Described by arXiv:2109.02903
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 976.4 MB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About IndicBART
What is IndicBART's context length?
1,024 tokens, from the maximum position embeddings in its published configuration.