Add NIAH eval results
Browse files
README.md
CHANGED
|
@@ -13,17 +13,21 @@ license: llama3
|
|
| 13 |
|
| 14 |
Join our custom agent and long context (262k-1M+) waitlist: https://forms.gle/L6TDY7dozx8TuoUv7
|
| 15 |
|
| 16 |
-
Gradient incorporates your data to deploy autonomous assistants that power critical operations across your business. If you're looking to build custom AI models or agents, email us a message [email protected].
|
| 17 |
-
|
| 18 |
-
For more info see our [End-to-end development service for custom LLMs and AI systems](https://gradient.ai/development-lab)
|
| 19 |
|
| 20 |
[Join our Discord](https://discord.com/invite/2QVy2qt2mf)
|
| 21 |
|
| 22 |
This model extends LLama-3 8B's context length from 8k to > 1040K, developed by Gradient, sponsored by compute from [Crusoe Energy](https://huggingface.co/crusoeai). It demonstrates that SOTA LLMs can learn to operate on long context with minimal training by appropriately adjusting RoPE theta. We trained on 830M tokens for this stage, and 1.4B tokens total for all stages, which is < 0.01% of Llama-3's original pre-training data.
|
| 23 |
|
| 24 |
-
**Update (5/3): We further fine-tuned our model to strengthen its assistant-like chat ability as well
|
|
|
|
|
|
|
|
|
|
| 25 |
|
| 26 |
-
|
|
|
|
|
|
|
|
|
|
| 27 |
|
| 28 |
**Approach:**
|
| 29 |
|
|
|
|
| 13 |
|
| 14 |
Join our custom agent and long context (262k-1M+) waitlist: https://forms.gle/L6TDY7dozx8TuoUv7
|
| 15 |
|
| 16 |
+
Gradient incorporates your data to deploy autonomous assistants that power critical operations across your business. If you're looking to build custom AI models or agents, email us a message [email protected]. For more info see our [end-to-end development service for custom LLMs and AI systems](https://gradient.ai/development-lab)
|
|
|
|
|
|
|
| 17 |
|
| 18 |
[Join our Discord](https://discord.com/invite/2QVy2qt2mf)
|
| 19 |
|
| 20 |
This model extends LLama-3 8B's context length from 8k to > 1040K, developed by Gradient, sponsored by compute from [Crusoe Energy](https://huggingface.co/crusoeai). It demonstrates that SOTA LLMs can learn to operate on long context with minimal training by appropriately adjusting RoPE theta. We trained on 830M tokens for this stage, and 1.4B tokens total for all stages, which is < 0.01% of Llama-3's original pre-training data.
|
| 21 |
|
| 22 |
+
**Update (5/3): We further fine-tuned our model to strengthen its assistant-like chat ability as well.**
|
| 23 |
+
|
| 24 |
+
Updated NIAH result:
|
| 25 |
+
<img src="https://cdn-uploads.huggingface.co/production/uploads/6585dc9be92bc5f258156bd6/-qaI__83ksClzoJzlqZjq.png" width="900" />
|
| 26 |
|
| 27 |
+
RULER evals:
|
| 28 |
+
- Our model is behind only GPT-4 and Yi in the retrieval and Q&A tasks
|
| 29 |
+
- It’s the smallest parameter model to rank in the top 7 overall
|
| 30 |
+
<img src="https://cdn-uploads.huggingface.co/production/uploads/655bb613e8a8971e89944f3e/0mLjl0Latrjc8gOrdtbc6.png" width="900" />
|
| 31 |
|
| 32 |
**Approach:**
|
| 33 |
|