StarChat2 Models by Hugging Face H4
About
The StarChat family of large language models (LLMs) specializes in serving as adept coding assistants. StarChat2, the latest model in this lineup, is a refined iteration of the 15-billion parameter StarCoder2 model. It leverages supervised fine-tuning and direct preference optimization on synthetic datasets to enhance both chat and coding functions. Despite its English-centric training, StarChat2 supports over 600 programming languages and exhibits strong performance on benchmarks such as MT Bench, IFEval for chat, and HumanEval for Python code tasks. However, the model lacks reinforcement learning from human feedback, which might lead to occasional problematic outputs. To accommodate various deployment needs, it also comes in several quantized forms that balance performance and resource consumption.
Current Variants
Use-when guidance is derived from seed capabilities, context, release, and replacement fields.
Use when the workload needs 16k context and 15B parameters.
| Model | Use when | Released | Signals | Status |
|---|---|---|---|---|
| StarChat2 15B | Use when the workload needs 16k context and 15B parameters. | 2024-07 | 16k context15B parameters | Current |
Release Timeline
1 release groupSpecifications(1 models)
| Model | Released | Context | Parameters |
|---|---|---|---|
| StarChat2 15B | 2024-07 | 16k | 15B |
Frequently Asked Questions
- What is StarChat2 used for?
- StarChat2 is used for coding. The family description and listed model capabilities point to those workloads as the best fit.
- How does StarChat2 compare to Zephyr?
- StarChat2 by Hugging Face H4 is strongest where you need coding, while Zephyr by Hugging Face H4 is the closest related family to check for structured outputs. StarChat2 has 1 listed variant and reaches up to 16k context, while Zephyr reaches up to 8k context, so compare the specs and pricing tables before choosing a production model.
- Which StarChat2 model should I use?
- If price is the main constraint, use the pricing table first because StarChat2 does not have complete provider pricing in the local data. For the most capable/latest local choice, evaluate StarChat2 15B with 16k context.
