dolly
Free10.8k
FreeFree tier
Inputs: textOutputs: text
About dolly
Dolly is an open-source, instruction-following large language model developed by Databricks. It is based on EleutherAI's Pythia-12b architecture and fine-tuned on a dataset of approximately 15,000 instruction/response pairs generated by Databricks employees. The model is designed to perform a variety of natural language tasks such as brainstorming, classification, closed question answering, generation, information extraction, open question answering, and summarization. Dolly is available for commercial use and can be accessed via Hugging Face as databricks/dolly-v2-12b. The project is hosted on GitHub under the databrickslabs organization, providing code for training and inference, as well as the dataset used for fine-tuning.
Key Features
Open-source large language model based on Pythia-12b
Fine-tuned on ~15k instruction/response pairs (databricks-dolly-15k)
Licensed for commercial use
Supports multiple instruction-following tasks: brainstorming, classification, closed QA, generation, information extraction, open QA, summarization
Available on Hugging Face for easy integration
Trained on the Databricks Machine Learning Platform
Pros & Cons
Pros
- Open-source and free to use, with commercial licensing
- Trained on a diverse set of instruction types, enabling versatile applications
- Based on a well-known architecture (Pythia-12b) with available model weights
- Community-supported with active development on GitHub
Cons
- Not a state-of-the-art model; performance may be lower than larger or more recent models
- Requires significant computational resources for inference (12 billion parameters)
- Fine-tuning dataset is relatively small (~15k records), which may limit generalization
- Documentation and usage examples may be limited to the GitHub repository
Best For
Building custom chatbots and virtual assistantsAutomating content generation and summarizationEnhancing search and information retrieval systemsAssisting with data classification and extractionEducational tools for question answering and brainstorming
FAQ
What is Dolly?
Dolly is an open-source, instruction-following large language model created by Databricks, based on the Pythia-12b architecture and fine-tuned on a dataset of ~15k instruction/response pairs.
Is Dolly free to use?
Yes, Dolly is open-source and licensed for commercial use. The model weights and training code are freely available on Hugging Face and GitHub.
What tasks can Dolly perform?
Based on available information, Dolly can handle tasks such as brainstorming, classification, closed QA, generation, information extraction, open QA, and summarization.
How do I use Dolly?
Dolly can be accessed via Hugging Face (databricks/dolly-v2-12b) or by cloning the GitHub repository and running the provided training or inference scripts.
What are the system requirements for running Dolly?
Running Dolly requires significant computational resources due to its 12 billion parameters. A GPU with sufficient memory (e.g., 16GB or more) is recommended for inference.
Is Dolly state-of-the-art?
According to the project description, Dolly is not a state-of-the-art model but exhibits surprisingly high quality instruction-following behavior for its size.