DBRX: Creating an LLM from Scratch Using Databricks
摘要
In Chapter X, we discussed the model that Databricks trained, DBRX. Let’s recap what it is. DBRX is a state-of-the-art large language model (LLM) trained from scratch on the Databricks and MosaicML platforms. At the time of model release, it outperformed established open-source models on language understanding (MMLU), programming (HumanEval), and math (GSM8K), as shown in Figure 13-1.