Together Research has developed a new model architecture named StripedHyena, which aims to improve upon the Transformer architecture by offering better performance in handling long contexts and enhancing training and inference efficiency.
The StripedHyena model, particularly the StripedHyena-Nous-7B version, is competitive with leading open-source Transformer models and is designed with a hybrid architecture that includes multi-head, grouped-query attention, and gated convolutions. This model promises lower latency, faster decoding, and the ability to process longer sequences up to 32k, making it a significant advancement over traditional Transformer models.
1. Text Generation: StripedHyena can be used to generate human-like text based on the provided prompts. This can be useful for content creation in various fields such as marketing, blogging, and scriptwriting.
2. Machine Translation: Given its ability to process long prompts, StripedHyena can be used for machine translation tasks, providing more accurate and contextually correct translations.
3. Sentiment Analysis: The model can be used to analyze sentiments from user reviews, comments, or social media posts, helping businesses understand customer sentiment towards their products or services.
4. Chatbots: StripedHyena can be used to power intelligent chatbots that can understand and respond to user queries in a human-like manner.
5. Text Summarization: The model can be used to create summaries of long documents or articles, saving users the time it would take to read the entire document.6. Question Answering: StripedHyena can be used in question answering systems to provide accurate and detailed
The StripedHyena model has shown to be competitive with the best open-source Transformers of similar sizes in short and long-context evaluations.
It offers low latency, faster decoding, and higher throughput than Transformers. It also shows improvement to training and inference-optimal scaling laws, as compared to optimized Transformer architectures such as Llama-2. Furthermore, it has been trained on sequences of up to 32k, allowing it to process longer prompts.
1. Understanding the Model: Before using the StripedHyena Nous 7B model, ensure you understand what the model can do. It's essential to know its strengths and limitations to effectively utilize it.
2. Model Configuration: Try to patiently configure and calibrate the model parameters. The parameters can have a huge impact on the performance of the model.
3. Evaluation: Always evaluate your model performance. Use appropriate metrics to assess the accuracy of your model's predictions.
4. Experiment: Feel free to experiment with different settings and approaches.
OpenAI-compatible — swap the base URL and it works with your existing SDK.
| Type | Price |
|---|---|
| Input | |
| Output | |
| Model | Input | Output | Context | Best for |
|---|---|---|---|---|
StripedHyena Nous (7B) This page |
StripedHyena Nous (7B) is priced at input $0.21 / 1M, output $0.21 / 1M.
Yes, StripedHyena Nous (7B) can stream responses as they are generated.
Yes, StripedHyena Nous (7B) supports both function calling and structured outputs.
StripedHyena Nous (7B) was built by Together.
It is served via the v1 endpoint through Together.
