GREAT MODEL
#1
by doberst - opened
Congratulations on DeciLM-7B ... Great Model. We just fine-tuned the model for RAG and posted at: llmware/dragon-deci-7b-v0 - check out the RAG benchmark scores, which compare favorably with Mistral, Llama-2, and other leading open source 7B foundation models.
Excellent accuracy with especially strong performance in key analytical areas such as not-found classification, Boolean questions, common-sense math and logic, multiple-choice questions, and table reading.
Also, the inference speed looks to be the fastest of the 7B models .... (we ran inference tests on Nvidia A6000)