Posts by Anshu Raina

Enabling DeepSeek-V4-Flash Training on AMD Instinct MI355X GPUs with Primus

DeepSeek-AI released the DeepSeek-V4 series on April 24, 2026: a preview pair of MIT-licensed Mixture-of-Experts models, with DeepSeek-V4-Flash at 284B total parameters (13B activated) and a one-million-token context window. Flash pushes sparse attention further than any open-weight model before it: its 43 transformer layers interleave three different attention types, each layer sits inside a hyper-connection block rather than a plain residual, and every token is routed through 256 experts. Each of those choices breaks an assumption baked into stock Megatron-LM training code.

Read more ...


Primus Tuning Agent: Closing the Configuration-Search Loop

Error parsing meta tag attribute “keywords”: No content.

Read more ...


TraceLens: Democratizing AI Performance Analysis

Profiling modern AI workloads produces huge traces that are hard to interpret. Framework profilers record thousands of operations, kernels, and communication events, and engineers often end up staring at tools like Perfetto UI doing manual calculations. TraceLens speeds this up: it consumes existing framework traces and turns them into structured summaries and comparisons, allowing you to move on to the actual diagnosis and optimization.

Read more ...


Primus Projection: Estimate Memory and Performance Before You Train

Error parsing meta tag attribute “keywords”: No content.

Read more ...


Primus-Pipeline: A More Flexible and Scalable Pipeline Parallelism Implementation

Error parsing meta tag attribute “keywords”: No content.

Read more ...