Aligned, a number one innovator in AI infrastructure and scalable machine studying options, introduced a strategic collaboration with AMD and the College of Southern California’s Info Sciences Institute (USC ISI) to speed up the event and coaching of MEGALODON, a cutting-edge giant language mannequin (LLM). The collaboration facilities on harnessing the highly effective AMD Intuition™ MI300 GPUs and Aligned’s experience in AI operations to coach the mannequin effectively on non-NVIDIA architectures.
MEGALODON is USC ISI’s flagship effort in large-scale language modeling, exploring a novel Transferring Common Geared up Gated Consideration (MEGA) structure designed to reinforce long-context retention whereas lowering computational complexity. With mannequin coaching requiring large GPU sources and complicated system orchestration, the collaboration between AMD and Aligned goals to push efficiency boundaries past conventional CUDA-based environments.
“As researchers push the boundaries of mannequin measurement and functionality, MEGALODON represents a leap ahead in structure and scalability,” stated Jonathan Could, analysis affiliate professor on the Viterbi Faculty of Engineering. “This collaboration brings the computational energy and infrastructure experience wanted to comprehend its full potential.”
Additionally Learn: AiThority Interview with Tom Findling, CEO of Conifers.ai
Tackling AI at Scale with AMD {Hardware} and ROCmTM Open Software program
Aligned will play a vital position in optimizing MEGALODON’s coaching on AMD’s Intuition™ MI325X GPUs, leveraging AMD’s ROCm™ platform as a CUDA various. Aligned is offering the instruments and engineering help to assist allow USC ISI’s researchers to faucet into the newest AMD GPU compute infrastructure.
“We’re excited to work alongside AMD and USC ISI to assist innovate on large-scale AI coaching,” stated Chris Ensey, CEO of Aligned. “Our mission is to allow highly effective AI on any {hardware}—and this partnership exemplifies that imaginative and prescient.”
World-Class Infrastructure Meets World-Class Analysis
The challenge additionally intersects with ISI’s broader ecosystem of private and non-private partnerships—together with DARPA, NSF, Lockheed Martin, and Chevron—demonstrating the real-world impression of foundational AI analysis. MEGALODON is being developed and examined utilizing high-performance compute clusters and superior storage applied sciences, with future plans to scale with extra superior AMD GPUs such because the MI355x.
Additionally Learn: The Way forward for Resolution Intelligence: Combining GenAI and Agentic AI
Key Collaboration Highlights:
- Novel LLM Structure: MEGALODON integrates the MEGA consideration mechanism, providing improved scalability and long-context efficiency.
- Cross-Platform AI Enablement: First main effort to coach a large-scale mannequin on AMD MI325X through ROCmTM open software program ecosystem.
- Superior Infrastructure: Joint entry to supercomputing clusters, high-throughput storage, and customized chip prototyping labs like MOSIS.
- Funding & Institutional Backing: Supported by NSF grants and USC’s dedication to advancing open, interoperable AI analysis.
[To share your insights with us as part of editorial or sponsored content, please write to psen@itechseries.com]
