XGBoost is an optimized library for distributed gradient boosting. It has become the leading machine learning library for solving regression and classification problems.
+::: + +:::{grid-item-card} +:padding: 1 +:img-top: ../images/finance-7.png +:class-img-top: small-sd-card-img-top +:class-body: small-sd-card +:class: small-sd-card ++++ + +Financial institutions deal with massive datasets and complex models that demand high computational power. +With AMD ROCm™ and GPU-accelerated libraries like LightGBM and ThunderGBM, we can significantly reduce training time while improving model performance.
+::: + +:::: diff --git a/docs/finance/index.md b/docs/finance/index.md new file mode 100644 index 0000000..64f0a13 --- /dev/null +++ b/docs/finance/index.md @@ -0,0 +1,103 @@ +# ROCm-Finance: ROCm toolkit for finance + +ROCm-Finance pulls the trajectory of tomorrow into today: an open toolkit on the [ROCm](https://rocm.docs.amd.com/) stack that +delivers GPU-native gradient-boosting stacks that the industry already trusts. XGBoost, +LightGBM, and ThunderGBM, tuned for [AMD Instinct](https://www.amd.com/en/products/accelerators/instinct.html) +accelerators, so training, scoring, and scenario work land closer to real time than the CPU-era +baselines could achieve. + +ROCm-Finance collapses the distance between signal and decision. The same workloads that once +queued overnight now run in minutes. Risk, fraud, forecasting, and simulation pipelines step +into the high-bandwidth, multi-GPU computing ROCm was built to serve. ROCm-Finance provides production-oriented kernels, memory paths, and scaling behavior so your boosting jobs feel like +they arrived from the next generation, even on this week's cluster. + +For more information on ROCm-Finance, including comparisons, prerequisites, installation, and deep API +reference, see the [ROCm-Finance documentation](https://rocm.docs.amd.com/projects/rocm-finance/en/latest/index.html). + +:::::{grid} 2 2 2 2 + +::::{grid-item-card} +:padding: 1 +:img-top: ../images/finance-1.png +:class-img-top: small-sd-card-img-top +:class-body: small-sd-card +:class: small-sd-card ++++ + +General-purpose GPU gradient boosting. Start here for many finance tabular workflows. +
+:::: + +::::{grid-item-card} +:padding: 1 +:img-top: ../images/finance-2.png +:class-img-top: small-sd-card-img-top +:class-body: small-sd-card +:class: small-sd-card ++++ + +Leaf-wise training. Strong fit when dataset size drives the bottleneck. +
+:::: + +::::{grid-item-card} +:padding: 1 +:img-top: ../images/finance-3.png +:class-img-top: small-sd-card-img-top +:class-body: small-sd-card +:class: small-sd-card ++++ + +GPU-oriented boosting for highly parallel, GPU-intensive training and simulation-style runs. +
+:::: + +::::{grid-item-card} +:padding: 1 +:img-top: ../images/finance-4.png +:class-img-top: small-sd-card-img-top +:class-body: small-sd-card +:class: small-sd-card ++++ + +Source for ROCm-Finance and related packaging on GitHub. +
+:::: + +::::{grid-item-card} +:padding: 1 +:img-top: ../images/finance-5.png +:class-img-top: small-sd-card-img-top +:class-body: small-sd-card +:class: small-sd-card ++++ + +Runnable examples on GitHub to explore the code. +
+:::: + +::::{grid-item-card} +:padding: 1 +:img-top: ../images/finance-6.png +:class-img-top: small-sd-card-img-top +:class-body: small-sd-card +:class: small-sd-card ++++ + +Browse blogs detailing how to accelerate your finance boost workloads on AMD Instinct GPUs. +
+:::: + +::::: diff --git a/docs/finance/lightgbm.md b/docs/finance/lightgbm.md new file mode 100644 index 0000000..fb24769 --- /dev/null +++ b/docs/finance/lightgbm.md @@ -0,0 +1,41 @@ +# LightGBM (ROCm-Finance) + +LightGBM is how ROCm-Finance answers scale: leaf-wise training that shines when dataset size—wide +feature stores, long histories, dense microstructure matrices—would otherwise push +decisions into the next shift. On Instinct, that wall between overnight queues and minutes +thins out; the same boosting idiom, routed through ROCm's high-bandwidth, multi-GPU environment. + +Reach for LightGBM when volume is the bottleneck, and you still want gradient boosting semantics +with GPU-native backing. + +::::{grid} 2 2 2 2 + +:::{grid-item-card} +:padding: 1 +:img-top: ../images/finance-6.png +:class-img-top: small-sd-card-img-top +:class-body: small-sd-card +:class: small-sd-card ++++ + +Installation instructions, how-to guides, and API reference material are on the ROCm Documentation site. +
+::: + +:::{grid-item-card} +:padding: 1 +:img-top: ../images/finance-2.png +:class-img-top: small-sd-card-img-top +:class-body: small-sd-card +:class: small-sd-card ++++ + +View the LightGBM source code on Github. +
+::: + +:::: diff --git a/docs/finance/thundergbm.md b/docs/finance/thundergbm.md new file mode 100644 index 0000000..e499ae6 --- /dev/null +++ b/docs/finance/thundergbm.md @@ -0,0 +1,39 @@ +# ThunderGBM (ROCm-Finance) + +Use ThunderGBM when parallelism and raw throughput dominate the story: +massively parallel trees, simulation-scale batches, and scenario grids that want the accelerator +to do the heavy lifting today—not in some speculative later hardware generation. ThunderGBM collapses the +signal-to-decision distance, optimized for highly parallel, +GPU-intensive training runs on [AMD Instinct](https://www.amd.com/en/products/accelerators/instinct.html) silicon. + +::::{grid} 2 2 2 2 + +:::{grid-item-card} +:padding: 1 +:img-top: ../images/finance-6.png +:class-img-top: small-sd-card-img-top +:class-body: small-sd-card +:class: small-sd-card ++++ + +ThunderGBM on ROCm—component documentation on the ROCm Documentation site. +
+::: + +:::{grid-item-card} +:padding: 1 +:img-top: ../images/finance-3.png +:class-img-top: small-sd-card-img-top +:class-body: small-sd-card +:class: small-sd-card ++++ + +View the ThunderGBM source code on Github. +
+::: + +:::: diff --git a/docs/finance/xgboost.md b/docs/finance/xgboost.md new file mode 100644 index 0000000..338c8c6 --- /dev/null +++ b/docs/finance/xgboost.md @@ -0,0 +1,37 @@ +# XGBoost (ROCm-Finance) + +XGBoost is the general-purpose engine in ROCm-Finance: tabular risk, fraud, pricing-side features, and trading-adjacent workloads. +Use it when you want a familiar level-wise boosting path with broad finance coverage and a +straightforward on-ramp from yesterday's pipelines to tomorrow's throughput. + +::::{grid} 2 2 2 2 + +:::{grid-item-card} +:padding: 1 +:img-top: ../images/finance-6.png +:class-img-top: small-sd-card-img-top +:class-body: small-sd-card +:class: small-sd-card ++++ + +XGBoost on ROCm—installation, tuning, and API reference on the ROCm Documentation site. +
+::: + +:::{grid-item-card} +:padding: 1 +:img-top: ../images/finance-1.png +:class-img-top: small-sd-card-img-top +:class-body: small-sd-card +:class: small-sd-card ++++ + +View the XGBoost source code on Github. +
+::: + +:::: diff --git a/docs/images/finance-1.png b/docs/images/finance-1.png new file mode 100644 index 0000000..feb6562 Binary files /dev/null and b/docs/images/finance-1.png differ diff --git a/docs/images/finance-10.png b/docs/images/finance-10.png new file mode 100644 index 0000000..fb347c3 Binary files /dev/null and b/docs/images/finance-10.png differ diff --git a/docs/images/finance-2.png b/docs/images/finance-2.png new file mode 100644 index 0000000..0f4cc0c Binary files /dev/null and b/docs/images/finance-2.png differ diff --git a/docs/images/finance-3.png b/docs/images/finance-3.png new file mode 100644 index 0000000..577bbc8 Binary files /dev/null and b/docs/images/finance-3.png differ diff --git a/docs/images/finance-4.png b/docs/images/finance-4.png new file mode 100644 index 0000000..701ad6d Binary files /dev/null and b/docs/images/finance-4.png differ diff --git a/docs/images/finance-5.png b/docs/images/finance-5.png new file mode 100644 index 0000000..e0bb3f9 Binary files /dev/null and b/docs/images/finance-5.png differ diff --git a/docs/images/finance-6.png b/docs/images/finance-6.png new file mode 100644 index 0000000..06f2d35 Binary files /dev/null and b/docs/images/finance-6.png differ diff --git a/docs/images/finance-7.png b/docs/images/finance-7.png new file mode 100644 index 0000000..3a36614 Binary files /dev/null and b/docs/images/finance-7.png differ diff --git a/docs/images/finance-8.png b/docs/images/finance-8.png new file mode 100644 index 0000000..4bd5cf2 Binary files /dev/null and b/docs/images/finance-8.png differ diff --git a/docs/images/finance-9.png b/docs/images/finance-9.png new file mode 100644 index 0000000..25c75c9 Binary files /dev/null and b/docs/images/finance-9.png differ diff --git a/docs/index.md b/docs/index.md index 428b822..5215fef 100644 --- a/docs/index.md +++ b/docs/index.md @@ -3,7 +3,7 @@ html_theme.sidebar_secondary.remove: true --- # AMD Instinct Data Center GPU Documentation -The AMD Instinct Documentation site provides comprehensive guides and technical documentation for system administrators and technical users deploying AMD Instinct Data Center GPUs in enterprise environments. This site focuses on large-scale deployment, cluster management, monitoring, and operational best practices for both HPC and AI workloads. For API documentation and core software stack details, please visit the [ROCm documentation](https://rocm.docs.amd.com). +The AMD Instinct Documentation site provides comprehensive guides and technical documentation for system administrators and technical users deploying AMD Instinct Data Center GPUs in enterprise environments. This site focuses on large-scale deployment, cluster management, monitoring, and operational best practices for both HPC and AI workloads. For API documentation and core software stack details, visit the [ROCm documentation](https://rocm.docs.amd.com). ::::::::::{dropdown} Industries/Verticals :open: @@ -166,6 +166,112 @@ The AMD Instinct Documentation site provides comprehensive guides and technical ::::::: +:::::::{tab-item} Finance + +::::{grid} 2 2 3 4 + +:::{grid-item-card} +:padding: 1 +:img-top: ./images/finance-6.png +:class-img-top: small-sd-card-img-top +:class-body: small-sd-card +:class: small-sd-card ++++ + +Boost your financial workloads with the ROCm Toolkit for Finance (ROCm-Finance). +
+::: + +:::{grid-item-card} +:padding: 1 +:img-top: ./images/finance-5.png +:class-img-top: small-sd-card-img-top +:class-body: small-sd-card +:class: small-sd-card ++++ + +General-purpose GPU gradient boosting. Start here for many finance tabular workflows. +
+::: + +:::{grid-item-card} +:padding: 1 +:img-top: ./images/finance-10.png +:class-img-top: small-sd-card-img-top +:class-body: small-sd-card +:class: small-sd-card ++++ + +Leaf-wise training is the way to go. It is a strong fit when dataset size drives the bottleneck. +
+::: + +:::{grid-item-card} +:padding: 1 +:img-top: ./images/finance-8.png +:class-img-top: small-sd-card-img-top +:class-body: small-sd-card +:class: small-sd-card ++++ + +GPU-oriented boosting for highly parallel, GPU-intensive training and simulation-style runs. +
+::: + +:::{grid-item-card} +:padding: 1 +:img-top: ./images/finance-4.png +:class-img-top: small-sd-card-img-top +:class-body: small-sd-card +:class: small-sd-card ++++ + +Source for ROCm-Finance and related packaging on GitHub. +
+::: + +:::{grid-item-card} +:padding: 1 +:img-top: ./images/finance-9.png +:class-img-top: small-sd-card-img-top +:class-body: small-sd-card +:class: small-sd-card ++++ + +Runnable examples on GitHub to explore the code. +
+::: + +:::{grid-item-card} +:padding: 1 +:img-top: ./images/finance-7.png +:class-img-top: small-sd-card-img-top +:class-body: small-sd-card +:class: small-sd-card ++++ + +Browse blogs detailing how to accelerate your finance boost workloads on AMD Instinct GPUs. +
+::: + +:::: + +::::::: + :::::::{tab-item} Life Science ::::{grid} 2 2 3 4 diff --git a/docs/sphinx/_toc.yml.in b/docs/sphinx/_toc.yml.in index ecb258a..1563cc5 100644 --- a/docs/sphinx/_toc.yml.in +++ b/docs/sphinx/_toc.yml.in @@ -37,6 +37,16 @@ subtrees: - entries: - file: life-science/MONAI.md title: MONAI + - file: finance/index.md + title: Finance (ROCm-Finance) + subtrees: + - entries: + - file: finance/xgboost.md + title: XGBoost + - file: finance/lightgbm.md + title: LightGBM + - file: finance/thundergbm.md + title: ThunderGBM - file: isv-apps/index.md title: Simulation & Modeling Apps subtrees: