Hebrew-Mistral-7B
| Property | Value |
|---|---|
| Parameter Count | 7.5B |
| License | Apache 2.0 |
| Tensor Type | BF16 |
| Languages | Hebrew, English |
| Author | Yam Peleg |
What is Hebrew-Mistral-7B?
Hebrew-Mistral-7B is an innovative bilingual Large Language Model based on Mistral-7B-v0.1, specifically optimized for Hebrew and English language processing. This model represents a significant advancement in multilingual AI, featuring an extended Hebrew tokenizer with 64,000 tokens and continuous pretraining from the original Mistral-7B architecture.
Implementation Details
The model utilizes the Transformers architecture and can be deployed in various configurations, including CPU, GPU, and 4-bit quantization for optimized performance. It's implemented using the Hugging Face Transformers library and supports both inference endpoints and text generation tasks.
- Extended vocabulary with 64,000 tokens specifically optimized for Hebrew
- Continuous pretraining methodology from Mistral-7B
- Multiple deployment options with varying precision levels
- Compatible with text-generation-inference systems
Core Capabilities
- Bilingual text generation in Hebrew and English
- General-purpose language understanding and processing
- Flexible deployment options for different computational resources
- Support for both CPU and GPU implementations
- 4-bit quantization support for efficient deployment
Frequently Asked Questions
Q: What makes this model unique?
The model's distinctive feature is its specialized optimization for Hebrew language processing while maintaining English language capabilities, making it one of the few high-performance bilingual models with extensive Hebrew language support.
Q: What are the recommended use cases?
The model is suitable for a wide range of natural language processing tasks, particularly those requiring Hebrew language understanding and generation, including text generation, content creation, and language processing applications in both Hebrew and English contexts.
