1. Data Parallelism Training steps: The master device loads the model and copies the model parameters to every worker device The master device splits the training data along the batch dimension and passes each batch to every worker device Each worker device performs training The master device aggregates the gradients from each worker device and updates the model parameters The master device broadcasts the model parameters to every worker device, preparing for the next batch of training Core idea:

1. Compiling Models with TensorRT-LLM 1.1 Introduction to TensorRT-LLM When using TensorRT, you usually need to convert the model to ONNX format, then convert the ONNX to TensorRT format, and finally run inference in TensorRT or Triton Server. But this conversion process is not simple, and you often run into

Paid integration service available at 200 RMB/year; technical support is also available at 200 RMB per session, capped at 1 hour. Follow the official account to get contact details. 1. Information You Need to Provide Go to https://mp.weixin.qq.com/ and open 【Settings and

1. Problems with Using LLMs Directly Unstable output One characteristic of generative AI is the diversity of its output. Ask an LLM the same question several times and you may get different answers. This uncertainty in output is a pleasant surprise for users in conversation and creative scenarios. But in

Two months ago, I gave a talk to a business team about chaos engineering in practice. This post mainly organizes the content of that talk. Click to view the presentation 1. The Origin of Chaos 1.1 The Birth of the Discipline of Chaos Before discussing chaos, let us first think about the relationship between chaos, chaos engineering, and our production services.