Checkpoint support is in the works: https://
github.com/Lightning-AI/l
it-gpt/issues/464
…
LLMS
-
Checkpoint Support Development in Progress for Lit-GPT
By
–
-
Benchmarks comparison with GPT-4 and ChatGPT-3.5
By
–
The benchmarks include GPT-4 and ChatGPT-3.5 (table on the right)
-
LLM Agents Learning Continuously: Advancing Autonomous AI Systems
By
–
Our work on LLM agents than learn continually
-
Fine-tuning with 16k tokens but capable of handling 100k
By
–
When I understood correctly, they used 16k for finetuning, but it can handle up to 100k even?
-
Why no 70B model for 100k context windows
By
–
Also why no 70B model?
1) Because of the 4x smaller dataset (compared to Llama 2 pretraining)?
2) to make 100k context windows work hardware-wise? -
RAG Dominates AI Discourse While Fine-tuning Costs Rise
By
–
Did a deep dive into this yesterday! RAG is far and away the most-talked about subject in AI right now. Also important re: the costs for GPT-3.5 Turbo fine-tune which is… not cheap. https://
supervised.news/p/a-clever-way
-to-sidestep-some-of
… -
CodeLlama-Python 34B finetuned on unnatural instructions
By
–
It's CodeLlama-Python 34B finetuned on 15,000 unnatural instructions. What's an unnatural instruction? Would like to see a few examples 😛
-
Evaluating RAG Pipelines with RAGAS and LangSmith
By
–
Blog: https://
blog.langchain.dev/evaluating-rag
-pipelines-with-ragas-langsmith/
… Webinar: https://
crowdcast.io/c/bnx91nz59cqq Hope to see you there! -
RAGAS Framework for RAG Systems Evaluation with LangSmith
By
–
RAGAS is an awesome open-source framework for evaluating RAG systems We recently integrated LangSmith RAGAS for optimal ease of evaluation We're also doing a webinar on RAG evaluation with them in 5 minutes!!! Link to webinar and blog
-
Scale Fine-Tunes GPT-3.5 Outperforming Base Model 66 Percent
By
–
We executed a pilot with @brexHQ on fine-tuning the models and where the fine-tuned GPT-3.5 model outperformed base GPT-3.5 66% of the time. We see strong evidence that you can fine-tune GPT-3.5 to provide dramatically better performance. https://
scale.com/blog/open-ai-s
cale-partnership-gpt-3-5-fine-tuning
…