| model_details: | |
| name: Kirim1/Bert | |
| description: A multilingual instruction-following language model supporting 100+ languages with primary focus on English | |
| version: 1.0.0 | |
| author: Kirim1 | |
| license: apache-2.0 | |
| model_type: bert | |
| base_model: BERT | |
| model_characteristics: | |
| architecture: BertForCausalLM | |
| parameters: ~110M | |
| hidden_size: 768 | |
| num_layers: 12 | |
| num_attention_heads: 12 | |
| vocab_size: 119547 | |
| max_position_embeddings: 512 | |
| context_window: 512 | |
| capabilities: | |
| primary_language: English | |
| supported_languages: 100+ | |
| instruction_following: true | |
| text_generation: true | |
| multilingual_understanding: true | |
| cross_lingual_tasks: true | |
| training: | |
| training_data: Multilingual corpus with English emphasis | |
| instruction_tuning: true | |
| optimization: Instruction-following and task completion | |
| use_cases: | |
| - Natural language understanding | |
| - Text generation | |
| - Question answering | |
| - Text classification | |
| - Multilingual text processing | |
| - Instruction-following tasks | |
| - Cross-lingual information retrieval | |
| limitations: | |
| - Performance varies across language families | |
| - Best results with English inputs | |
| - May require domain-specific fine-tuning | |
| - Limited performance on low-resource languages | |
| ethical_considerations: | |
| - May reflect biases in training data | |
| - Requires careful deployment in production | |
| - Not suitable for sensitive decision-making without oversight | |
| - Regular monitoring recommended | |
| technical_specifications: | |
| framework: PyTorch | |
| library: transformers | |
| minimum_transformers_version: 4.36.0 | |
| precision: float32 | |
| deployment: CPU and GPU compatible |