chore: update readme
Browse files
README.md
CHANGED
@@ -14,7 +14,9 @@ language:
|
|
14 |
|
15 |
One of the focus areas at Together Research is new architectures for long context, improved training, and inference performance over the Transformer architecture. Spinning out of a research program from our team and academic collaborators, with roots in **signal processing-inspired sequence models**, we are excited to introduce the **StripedHyena** models. StripedHyena is the **first alternative model competitive with the best open-source Transformers** of similar sizes in short and long-context evaluations.
|
16 |
|
17 |
-
-
|
|
|
|
|
18 |
- Play with the model on our playground!
|
19 |
- Dive into the details of our [standalone implementation](https://github.com/togethercomputer/stripedhyena), and our related research: [1](https://arxiv.org/abs/2302.10866), [2](https://arxiv.org/abs/2310.18780), [3](https://arxiv.org/abs/2311.05908).
|
20 |
|
@@ -23,5 +25,6 @@ One of the focus areas at Together Research is new architectures for long contex
|
|
23 |
StripedHyena is a hybrid architecture composed of multi-head, grouped-query attention and gated convolutions arranged in [Hyena](https://arxiv.org/abs/2302.10866) blocks, different from traditional decoder-only Transformers.
|
24 |
- Costant memory decoding in Hyena blocks via representation of convolutions as state-space models (modal or canonical form), or as truncated filters.
|
25 |
- Low latency, faster decoding and higher throughput than Transformers.
|
26 |
-
- Improvement to training and inference-optimal scaling laws, compared to optimized Transformer architectures such as Llama.
|
27 |
- Trained on sequences of up to 32k, allowing it to process longer prompts.
|
|
|
|
14 |
|
15 |
One of the focus areas at Together Research is new architectures for long context, improved training, and inference performance over the Transformer architecture. Spinning out of a research program from our team and academic collaborators, with roots in **signal processing-inspired sequence models**, we are excited to introduce the **StripedHyena** models. StripedHyena is the **first alternative model competitive with the best open-source Transformers** of similar sizes in short and long-context evaluations.
|
16 |
|
17 |
+
**StripedHyena-Hessian-7B (SH 7B)** is our **base model** for this release.
|
18 |
+
|
19 |
+
- Read more here in [our blog](https://www.together.ai/blog/stripedhyena-7b).
|
20 |
- Play with the model on our playground!
|
21 |
- Dive into the details of our [standalone implementation](https://github.com/togethercomputer/stripedhyena), and our related research: [1](https://arxiv.org/abs/2302.10866), [2](https://arxiv.org/abs/2310.18780), [3](https://arxiv.org/abs/2311.05908).
|
22 |
|
|
|
25 |
StripedHyena is a hybrid architecture composed of multi-head, grouped-query attention and gated convolutions arranged in [Hyena](https://arxiv.org/abs/2302.10866) blocks, different from traditional decoder-only Transformers.
|
26 |
- Costant memory decoding in Hyena blocks via representation of convolutions as state-space models (modal or canonical form), or as truncated filters.
|
27 |
- Low latency, faster decoding and higher throughput than Transformers.
|
28 |
+
- Improvement to training and inference-optimal scaling laws, compared to optimized Transformer architectures such as Llama-2.
|
29 |
- Trained on sequences of up to 32k, allowing it to process longer prompts.
|
30 |
+
|