René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"Roadmap","articleBody":"# Roadmap\r\n\r\n## Functionality\r\n- [x] #36\r\n- [x] #39 \r\n- [ ] Distill from any model without access to the original training data\r\n- [ ] Batched inference\r\n- [ ] Fine-grained KV cache management\r\n\r\n## Integration\r\n### Local Deployment\r\n- [ ] #33\r\n- [ ] #32\r\n- [ ] #35\r\n### Serving\r\n- [ ] [vllm](https://github.com/vllm-project/vllm)\r\n- [ ] [TGI](https://github.com/huggingface/text-generation-inference)\r\n- [ ] [lightllm](https://github.com/ModelTC/lightllm)\r\n\r\n## Research\r\n- [x] #34\r\n- [ ] Optimize the tree-based attention to reduce additional computation\r\n- [ ] Improve the acceptance scheme to generate more diverse sequences","author":{"url":"https://github.com/ctlllll","@type":"Person","name":"ctlllll"},"datePublished":"2023-09-12T01:49:06.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":15},"url":"https://github.com/3/Medusa/issues/3"}