AMD-AGI
Popular repositories Loading
Repositories
- AgentKernelArena Public
AgentKernelArena provides an end-to-end siloed-benchmarking environment where different LLM-powered agents—such as Cursor Agent, Claude Code, Codex, SWE-agent, and GEAK—can be evaluated side-by-side on the same GPU kernel tasks, using objective and reproducible metrics.
AMD-AGI/AgentKernelArena’s past year of commit activity - Primus-Turbo Public
A high-performance acceleration library dedicated to large-scale model training on AMD GPUs
AMD-AGI/Primus-Turbo’s past year of commit activity - Apex Public
Agents, and RL environment, for optimizing GPU kernels on AMD ROCm using LLM agents. Benchmarks LLM serving workloads end-to-end, profiles bottleneck kernels, optimizes them via Claude Code or Codex, and scores on compilation, correctness, and speedup.
AMD-AGI/Apex’s past year of commit activity - Primus Public
A flexible and high-performance training framework designed for large-scale foundation model training on AMD GPUs
AMD-AGI/Primus’s past year of commit activity - Nitro-1 Public
This repo provides a working re-implementation of Latent Adversarial Diffusion Distillation by AMD
AMD-AGI/Nitro-1’s past year of commit activity - DUET-VLM Public
DUET-VLM: Dual stage Unified Efficient Token reduction for VLM Training and Inference
AMD-AGI/DUET-VLM’s past year of commit activity
Top languages
Loading…
Most used topics
Loading…