Collective Communication Enabled Transformer Acceleration on Heterogeneous Clusters
Bibliographic record
Abstract
Field Programmable Gate Arrays (FPGAs) are becoming increasingly popular in data centers because of their ability to accelerate High-Performance Computing (HPC) and Machine Learning (ML) workloads. There has been extensive research on providing efficient tools such as Message Passing Interface (MPI) and ML model libraries that target FPGA platforms to developers and customers, due to the power efficiency, low latency, and flexibility provided by FPGAs. However, in the context of data centers, there could be hundreds or thousands of FPGAs deployed to accelerate one large-scale application, and the scalability of FPGAs is not well-researched in the academic field. In this thesis, we focus on extending the scalability of the Galapagos framework and propose a multi-cluster Galapagos project. We develop a tool, called Cluster Builder, for partitioning large-scale applications into multiple Galapagos clusters. To further increase the efficiency of the communication portion of the workloads, we propose to add one extra layer on top of the current Galapagos Stack, called the Galapagos Messaging Interface (GMI), which provides the necessary collective communication capabilities to the multi-cluster Galapagos framework. We also design an efficient multi-FPGA transformer model and use the Cluster Builder to deploy it on the multi-cluster Galapagos framework. Lastly, we propose a modified Galapagos framework that supports the AMD Versal Adaptive Compute Acceleration Platforms (ACAPs) [1] architecture. Based on the modified Galapagos framework, we estimate the performance of the transformer.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.003 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".