Month: December 2022

The human hand is one of the most remarkable outcomes of millions of years of evolution. The ability to pick up all sorts of objects and use them as tools is a…

The human hand is one of the most remarkable outcomes of millions of years of evolution. The ability to pick up all sorts of objects and use them as tools is a crucial differentiator enabling us to shape our world.

For robots to work in the everyday human world, the ability to deftly interact with our tools and the environment around them is critical. Without that capability, they will continue to be useful only in specialized domains such as factories or warehouses.

While it has been possible to teach robots with legs how to walk for some time, robots with hands have generally proven to be much more tricky to control. A hand with fingers has more joints that must move in specific coordinated ways to accomplish a given task. Traditional robotics control methods with precise grasps and motions are incapable of the kind of generalized fine motor control skills that humans take for granted.

One approach to these problems has been the application of deep reinforcement learning (deep RL) techniques that train a neural network to control the robot’s joints. With deep RL, a robot learns from trial and error and is rewarded for the successful completion of the assigned task. Unfortunately, this technique can require millions or even billions of samples to learn from, making it almost impossible to apply directly to real robots.

Applying simulation

Enter the NVIDIA Isaac robotics simulator, which enables robots to be trained inside a simulated universe that can run more than 10,000x faster than the real world and yet obeys the laws of physics.

Using NVIDIA Isaac Gym, an RL training robotics simulator, NVIDIA researchers on the DeXtreme project taught this robot hand how to manipulate a cube to match a provided target position and orientation or pose. The neural network brain learned to do this entirely in simulation before being transplanted to control a robot in the real world.

Similar work has only been shown one time before, by researchers at OpenAI. Their work required a far more sophisticated and expensive robot hand, a cube tricked out with precise motion control sensors, and a supercomputing cluster of hundreds of computers to train.

Democratizing dexterity

The hardware used by the DeXtreme project was chosen to be as simple and inexpensive as possible to enable researchers worldwide to replicate our experiments.

The robot itself is an Allegro Hand, which costs as little as 1/10th the cost of some alternatives, has four fingers instead of five, and has no moving wrist. We can use three off-the-shelf RGB cameras to track the 3D cube with vision, which can be repositioned easily as needed without requiring special hardware. The cube is 3D-printed with stickers affixed to each face.

DeXtreme is trained using Isaac Gym, which provides an end-to-end GPU-accelerated simulation environment for reinforcement learning. NVIDIA PhysX simulates the world on the GPU, and results stay in GPU memory during the training of the deep learning control policy network.

As a result, training can happen on a single Omniverse OVX server. Training a good policy takes about 32 hours on this system, equivalent to 42 years of a single robot’s experience in the real world.

Not needing a separate CPU cluster for simulation means a 10–200x reduction in computing costs for training at current cloud rental rates. Because we can use Isaac Gym to train the model, training time and cost can be dramatically reduced.

Perception and synthetic data

For the robot to know the current position and orientation of the cube that it’s holding, it needs a perception system. To keep costs low and leave open the potential for manipulation of other objects in the future, DeXtreme uses three off-the-shelf cameras and another neural network that can interpret the cube pose.

This network is trained using about 5 million frames of synthetic data generated using Omniverse Replicator and no real images whatsoever. The network learns how to perform the task under challenging circumstances in the real world. To make the training more robust, we use a technique called domain randomization to change lighting and camera positions, plus data augmentation to apply random crops, rotation, and backgrounds.

The DeXtreme pose estimation system is reliable and can perceive accurate poses even when the object in question is partly occluded from view, or when the image has significant motion blur.

Real robots are still challenging

One of the key reasons to use simulations is that training robots directly in the real world are riddled with various challenges. For example, robot hardware is prone to breaking after excessive usage. Experiment iteration cycles and turnaround time can also be slow.

During our experiments, we often found ourselves repairing the hand after prolonged usage, for example, tightening the loose screws, replacing the ribbon cables, and resting the hand to cool down after running 10-15 trials. Simulations enabled us to sidestep many of these issues by training on a robot that doesn’t wear out but also provides the large diversity of data needed to learn challenging tasks. At the same time, because simulations can run much faster than in real time, the iteration cycle is massively improved.

When training in simulation, the most significant challenge is bridging the gaps between the simulations and the real world. To address this, DeXtreme uses domain randomization of the physics properties set in the simulator: changing object masses, friction levels, and other attributes at scale across over a hundred thousand simulated environments at one time.

One interesting upshot of these randomizations is that we train the AI with all kinds of unusual combinations of scenarios, which translates to robustness when performing the task in the real world. For instance, most of our experiments on the real robot took place with a slightly malfunctioning thumb due to a loose connection on the circuit board. We were positively surprised that the policies transferred from simulation to the real world reliably, regardless.

Video 5. After over 32 hours of training, the DeXtreme robot was capable of repeated success at the task of rotating a cube to match a specific target

Sim-to-real

Future breakthroughs in robotic manipulation will enable a new wave of robotics applications beyond traditional industrial uses.

At the heart of the DeXtreme project is the message that simulation can be an incredibly effective tool for training complex robotic systems. This is true even for the systems that must handle environments with objects in continual contact with the robot. We hope that by demonstrating this using relatively low-cost hardware, we can inspire others to use our simulation tools and build on this work.

For more information, see DeXtreme: Transfer of Agile In-hand Manipulation from Simulation to Reality and visit DeXtreme.

For a further dive into simulators and how they can affect your projects, see How GPUs Can Democratize Deep Reinforcement Learning for Robotics Development. You can also download the latest version of NVIDIA Omniverse Isaac Sim and learn about training your own reinforcement learning policies.

NVIDIA Modulus is now available on NVIDIA LaunchPad. Sign-up for a free, hands-on lab that will teach you how to develop physics-informed machine-learning…

NVIDIA Modulus is now available on NVIDIA LaunchPad. Sign-up for a free, hands-on lab that will teach you how to develop physics-informed machine-learning solutions.

HDBSCAN is a state-of-the-art, density-based clustering algorithm that has become popular in domains as varied as topic modeling, genomics, and geospatial…

HDBSCAN is a state-of-the-art, density-based clustering algorithm that has become popular in domains as varied as topic modeling, genomics, and geospatial analytics.

RAPIDS cuML has provided accelerated HDBSCAN since the 21.10 release in October 2021, as detailed in GPU-Accelerated Hierarchical DBSCAN with RAPIDS cuML – Let’s Get Back To The Future. However, support for soft clustering (also known as fuzzy clustering) was not included. With soft clustering, a vector of values (rather than a single cluster label) is created for each point representing the probability that the point is a member of each cluster. 

Performing HDBSCAN soft clustering on CPUs has been slow. It can take hours or even days on medium-sized datasets due to the heavy computational burden. Now, in the 22.10 RAPIDS release, cuML provides accelerated soft clustering for HDBSCAN, enabling the use of this technique on large datasets.

This post highlights the importance of using soft clustering to better capture nuance in downstream analysis and the performance gains possible with RAPIDS. In a document clustering example, soft clustering that takes hours on a CPU can be completed in seconds with cuML on a GPU.

Soft clustering code

In the CPU-based scikit-learn-contrib/hdbscan library, soft clustering is available through the all_points_membership_vectors top-level module function. 

cuML matches this API:

import cuml

blobs, labels = cuml.datasets.make_blobs(n_samples=2000, n_features=10, random_state=12)

clusterer = cuml.cluster.hdbscan.HDBSCAN(prediction_data=True)
clusterer.fit(blobs)
cuml.cluster.hdbscan.all_points_membership_vectors(clusterer)

Why soft clustering

In many clustering algorithms, each record in a dataset is either assigned to a single cluster or considered noise and not assigned to any clusters. However, in the real world, many records do not perfectly fit into a single cluster. HDBSCAN acknowledges this reality by providing a mechanism for soft clustering by representing the degree to which each point belongs to each cluster. This is similar to other algorithms such as mixture models and fuzzy c-means.

Imagine a news article about a sports-themed musical. If you wanted to assign the article to a cluster, would it belong to a sports cluster or a musical cluster? Perhaps a smaller cluster specifically for sports-themed musicals? Or should it have some level of membership in both clusters?

By forcing the choice of a single label, this nuance is lost, which can be significant when the results of clustering are used for downstream analysis or actions. If only a sports label is assigned, would a recommendation system surface the article to readers also interested in musicals? What about the inverse (a reader interested in one but not the other)? This article should be potentially in the queue for readers interested in either or both of these topics.

Soft clustering solves this problem, enabling actions based on thresholds for each category and creating applications that provide a better experience.

Example of document clustering

You can use document clustering to measure the potential real-world performance benefits and impact of cuML’s accelerated soft clustering. 

Many modern document clustering workflows are composed of the following steps:

• Convert each document into a numeric representation (often using neural network embeddings)
• Reduce the dimensionality of the numeric-transformed documents
• Perform clustering on the dimension-reduced dataset
• Take actions based on the results

If you would like to run the workflow on your system, a Jupyter Notebook is available by visiting hdbscan-blog.ipynb on GitHub.

Preparing the dataset

This example uses the A Million News Headlines dataset from Kaggle, which contains over 1 million news article headlines from the Australian Broadcasting Corporation. 

After downloading the dataset, convert each headline into an embedding vector. To do this, use the all-MiniLM-L6-v2 neural network from the Sentence Transformers library. Note that a few other libraries imported will be used later.

import numpy as np
import pandas as pd
import cuml
from sentence_transformers import SentenceTransformer

N_HEADLINES = 10000

model = SentenceTransformer('all-MiniLM-L6-v2')

df = pd.read_csv("/path/to/million-headlines.zip")
embeddings = model.encode(df.headline_text[:N_HEADLINES])

Reducing dimensionality

With the embeddings ready, reduce the dimensionality down to five features using cuML’s UMAP. This example is only using 10,000 records, as it is intended as a guide.

For reproducibility, set a random_state. The results in this workflow may differ slightly when run on your machine, depending on the UMAP results.

umap = cuml.manifold.UMAP(n_components=5, n_neighbors=15, min_dist=0.0, random_state=12)
reduced_data = umap.fit_transform(embeddings)

Soft clustering

Next, fit the HDBSCAN model on the dataset and enable using all_points_membership_vectors for soft clustering by setting prediction_data=True. Also set min_cluster_size=50, which means that groupings of fewer than 50 headlines will be considered as part of another cluster (or noise), rather than as a separate cluster.

clusterer = cuml.cluster.hdbscan.HDBSCAN(min_cluster_size=50, metric='euclidean', prediction_data=True)
clusterer.fit(reduced_data)
soft_clusters = cuml.cluster.hdbscan.all_points_membership_vectors(clusterer)
soft_clusters[0]
array([0.01466704, 0.01035215, 0.0220814 , 0.01829496, 0.0127591 ,
       0.02333117, 0.01993877, 0.02453639, 0.03369896, 0.02514531,
       0.05555269, 0.04149485, 0.05131698, 0.02297594, 0.03559102,
       0.02765776, 0.04131499, 0.06404213, 0.01866449, 0.01557038,
       0.01696391], dtype=float32)

Inspecting a couple of clusters

Before going further, look at the assigned labels (-1 represents noise). It is evident that there are quite a few clusters in this data:

pd.Series(clusterer.labels_).value_counts()
-1     3943
 3     1988
 5     1022
 20     682
 13     367
 7      210
 0      199
 2      192
 8      190
 16     177
 12     161
 17     143
 15     107
 19      86
 9       83
 11      70
 4       68
 18      66
 6       65
 10      61
 1       61
 14      59
dtype: int64

Select two clusters at random, clusters 7 and 10, for example, and examine a few points from each.

df[:N_HEADLINES].loc[clusterer.labels_ == 7].headline_text.head()
572    man accused of selling bali bomb chemicals goe...
671      timor sea treaty will be ratfied shortly martin
678     us asks indonesia to improve human rights record
797           shop owner on trial accused over bali bomb
874    gatecrashers blamed for violence at bali thank...
Name: headline_text, dtype: object

df[:N_HEADLINES].loc[clusterer.labels_ == 10].headline_text.head()
40     direct anger at govt not soldiers crean urges
94                   mayor warns landfill protesters
334                    more anti war rallies planned
362     pm criticism of protesters disgraceful crean
363      pm defends criticism of anti war protesters
Name: headline_text, dtype: object

Then, use the soft clustering membership scores to find some points, which might belong to both of the clusters. From the soft cluster scores, identify the top two clusters for each point. Exclude outliers by filtering for soft memberships to clusters 7 and 10, which are both proportionately larger than their memberships to other clusters:

df2 = pd.DataFrame(soft_clusters.argsort()[:,::-1][:,:2])
df2["sum"] = soft_clusters.sum(axis=1)
df2["cluster_7_ratio"] = soft_clusters[:,7] / df2["sum"]
df2["cluster_10_ratio"] = soft_clusters[:,10] / df2["sum"]
df2[(df2[0] == 7) & (df2[1] == 10) & (df2["cluster_7_ratio"] > 0.1) & (df2["cluster_10_ratio"] > 0.1)]

3824  7  10  0.630313         0.170000         0.160453
6405  7  10  0.695286         0.260162         0.119036

Inspect the headlines for these points, and notice they are both about Indonesia, which is also in the cluster 7 headline 678 (above). Also notice that both of these headlines are about antiwar and peace, which are topics included in several of the cluster 10 headlines.

df[:N_HEADLINES].iloc[3824]
publish_date                                              20030309
headline_text    indonesians stage mass prayer against war in iraq
Name: 3824, dtype: object

df[:N_HEADLINES].iloc[6405]
publish_date                           20030322
headline_text    anti war fury sweeps indonesia
Name: 6405, dtype: object

Why soft clustering matters

How confident should you be that these results belong to their assigned cluster, rather than another cluster? As previously observed, some clusters contain headlines that have a meaningful probability of being in a different cluster.

The degree of confidence can be partially quantified by calculating the difference between the membership probabilities of each point’s top two clusters. This example excludes noise points, to drive home how this does not just happen with “noisy” points but also those assigned cluster labels:

soft_non_noise = soft_clusters[clusterer.labels_ != -1]
probs_top2_non_noise = np.take_along_axis(soft_non_noise, soft_non_noise.argsort(), axis=1)[:, -2:]
diffs = np.diff(probs_top2_non_noise).ravel()

Plotting a histogram and the empirical cumulative distribution function of these differences shows that many points were close to being assigned a different cluster label. In fact, about 30% of points had top-two cluster membership probabilities within 0.2 of one another (Figure 1). 

Using these soft clustering probabilities enables you to incorporate this uncertainty to build much more robust machine learning pipelines and applications.

Performance benchmark results

Next, run the preceding core workload on different numbers of news article headlines, varying from 25,000 to 400,000 rows. Note that HDBSCAN parameters can be tweaked for larger datasets. 

For this example, CPU benchmarks were run on an x86 Intel Xeon Gold 6128 CPU at 3.40 GHz. GPU benchmarks were recorded on an NVIDIA Quadro RTX 8000 with 48 GB of memory.

On the CPU (indicated by the hdbscan backend), soft clustering performance scales loosely linearly as the number of documents is doubled. 50,000 documents took 50 seconds; 100,000 documents took 500 seconds; 200,000 documents took 5,500 seconds (1.5 hours); and 400,000 documents took over 60,000 seconds (17 hours). See Table 1 for more details.

Using the GPU-accelerated soft clustering in cuML, soft clusters for 400,000 documents can be calculated in less than 2 seconds, rather than 17 hours.

If you would like to run this benchmark on your system, use the benchmark-membership-vectors.py GitHub gist. Note that performance will vary depending on the CPU and GPU used.

Key takeaways

We are excited to report these performance results. Soft clustering can meaningfully improve workflows powered by machine learning. Until now, using a clustering technique like HDBSCAN has been computationally challenging for even a few hundred thousand records.

With the addition of HDBSCAN soft clustering, RAPIDS and cuML continue to break through barriers and make state-of-the-art computational techniques more accessible at scale. 

To get started with cuML, visit the RAPIDS Getting Started page, where conda packages, pip packages, and Docker containers are available. cuML is also available in the NVIDIA optimized PyTorch and Tensorflow Docker containers on NVIDIA NGC, making this end-to-end workflow even easier.

Acknowledgments

This post describes cuML functionality contributed by Tarang Jain during his internship at NVIDIA under the mentorship of Corey Nolet.

X-ray-powered research is aiming to target sneaky hazardous materials making their way through airport security. The study, recently published in Scientific…

X-ray-powered research is aiming to target sneaky hazardous materials making their way through airport security. The study, recently published in Scientific Reports, proposes a new design for a fast and powerful X-ray diffraction (XRD) technology able to identify potential threats. The work could be a notable step toward more accurate luggage scanning in airports. 

“The main goal of my project was to speed up this new X-ray imaging modality so it can be economically viable for airport security. Ultimately such a scanner could help find even the most creatively hidden explosives, drugs, and contraband, without excessive costs for the operator or delays for passengers,” said study author Airidas Korolkovas. He conducted part of the study while working as an X-ray physicist and imaging scientist at iTomography Corporation.

An awardee of the NVIDIA Academic Hardware Grant Program, Korolkovas was granted an NVIDIA TITAN V during his postdoctoral fellowship at Uppsala University in Sweden. Applicants must demonstrate how access to world-class computing resources could boost their research. 

In this case, Korolkovas designed and implemented a GPU-accelerated tomographic reconstruction algorithm for XRD to supplement existing computed tomography (CT) X-ray scans in airport security.  

Currently, airports rely on X-ray transmission alone to reveal luggage contents in 3D. X-ray beams can penetrate through, absorb, or scatter depending on the composition and spatial arrangement of atoms within each material. By measuring changes in the beam at various angles, sophisticated algorithms and computer vision technology can reconstruct 3D images of the bag contents. 

This gives airport security a peek into luggage without having to touch it. However, transmission-based CT scans have limitations. 

“Standard CT is sensitive to the average density and composition of materials. It is not sensitive to the internal arrangement of atoms, which makes all the difference when identifying a benign piece of plastic from a plastic explosive, or between sugar and cocaine,” Korolkovas said. 

According to the study, XRD could be a powerful new addition when scanning luggage, because it is sensitive to the internal arrangement of atoms.

This makes XRD especially well suited for identifying crystals as the repetitive arrangement of molecules in crystalline materials results in concentrated X-ray scattering along precise angles unique to each material. Access to this data could help airport security determine if objects in a bag contain threats like cocaine, crystal methamphetamine, or even explosives that are naturally crystalline, semi-crystalline, or crystalline powders.

Unfortunately, XRD scans of whole passenger luggage are quite slow, making them unusable in commercial aviation, which demands real-time results. 

To shorten the time, Korolkovas employed a novel scanner design aimed at high-intensity rather than high-resolution X-ray beams, which is traditionally a no-go as it degrades the XRD signal quality beyond recognition. 

Through a multistep approach, involving CT image segmentation and complex algebraic reconstructions, he was able to recover the XRD resolution, despite the limitations of beam intensity.

Korolkovas used the NVIDIA TITAN V to calculate the probabilities of all possible diffraction pathways.

“In this study, I was able to maintain acceptable resolution by combining data from transmission, diffraction, all the viewing angles, and the full spectrum of X-ray energies,” he said.

This can easily run into a quintillion mathematical operations for every slice of the bag.

According to Korolkovas, coding the reconstruction algorithm on a GPU was very helpful in keeping the computation time manageable. Changing from a CPU to a GPU, he was able to speed up the runtime from 10 hours to less than 1 hour. By further improving the algorithm and using multiple GPUs or cloud computing, he envisions eventually running the scan in real time. 

“X-rays can penetrate and scatter within the bag in every possible direction. CUDA texture mapping has turned out to be an efficient way to access the photon survival probabilities along any such pathway. The calculations of various pathways are partially independent of each other, and benefit from parallel computing afforded by CUDA,” he said.

Testing the approach on a simulated bag containing both benign and threat materials, he found that the XRD reconstruction adds material-specific information, improving what CT alone captures and threat detection. 

“An XRD imaging add-on to existing CT scanners is feasible and is well positioned to provide unique, material-specific information, at a low cost of installing an extra detector or two and developing suitable reconstruction software,” Korolkovas writes in the study. 

The next steps in the research include building an experimental prototype and testing the algorithm on real-world data. 

“The study has received encouraging feedback from Rapiscan Systems, a major manufacturer of X-ray scanners. Now that air travel is returning to pre-pandemic levels, there is renewed interest in advanced X-ray imaging and I hope to contribute to this endeavor,” Korolkovas said. 

He also plans on using machine learning to train neural networks that fingerprint the reconstructed diffraction patterns against a broad range of materials found in suitcases. This will improve the robustness of flagging threat materials, even with limited data that can be acquired in real time.

Read the study, Fast X-ray diffraction (XRD) tomography for enhanced identification of materials.
Access research code on the XRD_Tomography GitHub page.
Learn more about NVIDIA Higher Education and Research Developer Resources.

Funding for this research includes a grant from the U.S. Department of Homeland Security, Science, and Technology Directorate and a Titan V donated by NVIDIA.

The NVIDIA Optical Flow SDK 4.0 is now available, enabling you to fully harness the new NVIDIA Optical Flow Accelerator on the NVIDIA Ada architecture with…

The NVIDIA Optical Flow SDK 4.0 is now available, enabling you to fully harness the new NVIDIA Optical Flow Accelerator on the NVIDIA Ada architecture with NvOFFRUC.

Optical flow on the NVIDIA Ada Lovelace architecture

Starting from the NVIDIA Turing architecture, NVIDIA GPUs have dedicated hardware for optical flow computation between a pair of frames. NVIDIA has continued to invest in improving the optical flow hardware engine in the NVIDIA Ampere architecture and NVIDIA Ada Lovelace architecture generations, thanks to the continued feedback from application developers and researchers.

Significant performance improvements

The Optical Flow algorithm requires certain pre– and post-processing steps to improve the quality of the flow vectors.

In the NVIDIA Turing and NVIDIA Ampere architecture generation GPUs, most of these algorithms use a compute engine to perform the required tasks. As a result, when the compute engine workload is high, the performance of the NVIDIA Optical Flow Accelerator (NVOFA) could be affected.

On NVIDIA Ada-generation GPUs, most of these algorithms are moved to dedicated hardware within the NVOFA, reducing the dependency on the compute engine significantly.

In addition, NVIDIA Ada-generation GPUs bring several other optimizations related to reducing the overhead of interaction between driver and hardware. This increases the overall performance and context switches between various hardware engines on the GPU.

With these changes, the speed of the NVIDIA Ada Lovelace architecture NVOFA is improved ~2x compared to the NVIDIA Ampere architecture NVOFA.

Quality improvements

Based on the feedback from earlier generations of NVOFA, there are several quality improvements incorporated in the hardware. Using the same preset, you can see a 10-15% improvement in quality (tested on the KITTI2015 data set) compared to NVIDIA Ampere architecture GPUs.

For more information, see 1.4 NVOFA Quality and Performance.

Optical Flow SDK 4.0

The NVIDIA Optical Flow SDK enables you to access NVOFA functionality. The NVIDIA Optical Flow SDK is a set of Optical Flow C APIs, reusable C++ wrapper classes, and a set of sample applications. These APIs and C++ wrapper classes facilitate the programming of the NVOFA for the efficient computation of the optical flow between a pair of images.

Optical Flow SDK 4.0 comes with the following enhancements and features:

•  External hint support
•  NVIDIA Optical Flow-assisted Frame-Rate Up-Conversion (NvOFFRUC)

External hint support

When hints are generated with low evolution images or are available from other sources such as a game engine, NVOFA can refine the hints further to improve the quality of the flow vectors.

Though external hint support is already available through C-API, support was missing in earlier versions of SDK C++ wrapper classes.

Optical Flow SDK 4.0 adds necessary support in the C++ classes and the use of external hints is demonstrated in the sample application AppOFCuda. The hint format is the same as the output flow vector format: an array of NV_OF_FLOW_VECTOR structures. Each array element represents a motion vector for the corresponding block in raster scan order.

AppOFCuda accepts hints in Middlebury flo format but converts them into the required format (an array of NV_OF_FLOW_VECTOR structures) before passing it to the NVOF API. NVOFA prioritizes external hints when they are provided; you are expected to provide reasonable quality hints.

Frame-rate up-conversion (FRUC) is a technique that generates higher frame-rate video from lower frame-rate video by inserting interpolated frames into it. Such high frame-rate video shows smooth continuity of motion across frames, improving the perceived visual quality of the video.

The NvOFFRUC library exposes APIs that take two consecutive frames and generate an interpolated frame in between. The interpolation is instant and does not have to be exactly in the middle of the two frames: it can be specified arbitrarily. For more information, see the NVOFA FRUC Programming Guide.

These APIs can be used for up-conversion of any video content. Internally, the library uses the NVOFA hardware engine and CUDA compute cores. As a result, frame interpolation using the NvOFFRUC library is much faster compared to software-only methods.

For more information, see AV1 Encoding and FRUC: Video Performance Boosts and Higher Fidelity on the NVIDIA Ada Lovelace Architecture.

Optical Flow SDK 4.0 is available now.

The NVIDIA Base Command Platform enables an intuitive, fully featured development experience for AI applications. It was built to serve the needs of…

The NVIDIA Base Command Platform enables an intuitive, fully featured development experience for AI applications. It was built to serve the needs of the internal NVIDIA research and product development teams. Now, it has become an essential method for accessing on-demand compute resources to train neural network models and execute other accelerated computing experiments. 

Base Command Platform simplifies AI experimentation workflows by providing a cohesive service that integrates users, jobs, and data. It provides easy access to a private registry to host custom containers as well as the breadth of software from the NGC Catalog. It offers all these features without sacrificing reliable NVIDIA performance, flexibility, and scalability. You can use Base Command Platform for experiments requiring a single GPU or a data center’s worth of them.

Base Command Platform interface and features

Base Command Platform supports a CLI, API, and web interface, all built into the NGC portal. The integrated web interface makes software discovery in the NGC Catalog and subsequent use in Base Command Platform smooth. You don’t have to transition between tools not designed to be used together.

In addition to providing access to the public NGC Catalog, you also gain access to a private registry dedicated to the Base Command Platform environment. The private registry is useful for keeping containers, models, and software private and secure, as dictated by developer requirements.

Base Command Platform provides a rich set of user management controls. When you are invited to use a Base Command Platform environment (called an organization), the administrator can restrict your ability to upload and interact with content on the platform through a set of role-based access controls. These controls can apply to the root organization and also to the concept of a team. 

A team can differ minimally or significantly from the root organization, depending on how an admin configures that team. For example, a team may only be provided access to a subset of private registry containers or resources. When onboarded to a team, you could be disallowed from uploading your own containers to the private registry. 

These capabilities can be mixed and matched to provide the right level of functionality for a given user or group by the org administrator. Administrators can also set hardware usage quotas in the organization for specific users, both GPU and storage capacity.

The Base Command Platform web interface places the key user interaction points front and (left of) center: 

• Jobs: A list of containers running on NVIDIA Base Command Platform compute resources.
• Datasets: Read-only data inputs that can be mounted into jobs.
• Workspaces: Read/write persistent storage that can also be mounted into jobs.

Simple yet powerful hardware abstraction

In Base Command Platform, the managed hardware resources are presented to the user through two concepts: accelerated computing environments (ACEs) and instances within an ACE. 

An ACE is a composition of a set of hardware resources: compute, network, and storage. An instance selects the CPU, RAM, and GPU resource quantities that a job requires from a system within an ACE. 

ACEs can support a variety of instance types depending on their underlying hardware composition. Administrators can restrict the use of these resources through a quota for GPU hours, as well as completely restricting instance type availability for specific users in the org. 

Base Command Platform resources are connected through industry-leading technology provided by the underlying infrastructure. NVIDIA NVLink, NVIDIA InfiniBand, and high-performance Ethernet connectivity are integrated as part of a Base Command Platform environment’s design to maximize the value of the managed hardware resources. 

The scheduler in Base Command Platform is designed to take advantage of topology awareness to provide optimal resource use for jobs as they are submitted.

Datasets and workspaces

Data management is core to Base Command Platform’s capabilities. Datasets, models, and source code must be made available to compute resources during experimentation. 

The dataset and workspace concepts are how Base Command Platform solves this problem. A dataset is a read-only storage construct after creation, with all the same sharing capabilities as private registry contents. They can be private to a specific user, shared with any number of teams in an org, or shared with the entire org. 

Workspaces are more flexible. They are readable and writable but can be marked read-only when used in a job if desired. Workspace-sharing capabilities are identical to what datasets support.

Datasets are created through either the web interface or the CLI at upload or conversion time. We cover conversion when jobs are discussed later in this post. A workspace is created first, then populated with data as part of a job, or through direct upload (similar to datasets).

So, why would you use one over the other? 

Datasets are a great fit for immutable data that must be widely shared as-is. Frequently, that is a dataset that no longer requires modification but could also include a license file or API key intended for shared use. They can be shared widely because there is no chance that they will be modified in place. 

Workspaces are a great fit as a landing place for data that is a work-in-progress: datasets, source code maintained outside a container, or even a collection of models under development. Workspaces can be shared widely but given that they are writable by default, wide sharing may require additional coordination and awareness between users.

The aggregate dataset and workspace capacity available for a given user in Base Command Platform is controlled by the user’s storage quota, set by the org administrator. You can see your storage quota along with your current storage usage on the Base Command Platform dashboard. 

There is an additional storage type, result, that factors into this capacity, which we discuss later in the context of jobs. When your quota is exceeded, you can request additional capacity if enabled by your environment administrator.

Bringing it all together in a job

Base Command Platform operationalizes data, compute resources, the NGC Catalog, and private registry contents with jobs. 

Job creation starts with resource selection. You are presented with available ACEs for use by your org and team. You may have access to more than one ACE, but a job must execute within a single ACE. 

After an ACE is selected, you can select from the available instance types in that ACE. For multi-node jobs, the only instances available are those that leverage the maximum CPU, RAM, and GPU resources available for a given system type within the selected ACE.

Choose an ACE, an instance type, and multi-node launch options, if necessary. Next, you can specify datasets and workspaces that are a part of the chosen ACE to be mounted into the target job, along with the desired mount point for each of them. Here, a workspace can be marked as read-only. 

The job’s result storage path must be specified as well. A result is a job-specific read/write storage pool intended to hold artifacts that you’d like to preserve from the job upon completion, along with stderr and stdout. As we mentioned previously, the capacity consumed by results counts against your quota.

Then, you must select a Container object and a valid container Tag. You can choose containers from the NGC Catalog as well as the private registry containers that you have permission to access. If you select a multi-node job, only containers marked in the NGC Catalog or private registry as supporting multi-node functionality are presented as options.

You now specify one or more commands, or even a service (such as JupyterLab, Visual Studio Code Server, or NVIDIA Triton Inference Server) to run inside the selected container when the job is active. If an external port is needed to expose a web interface or some other endpoint, one or more ports must be added with the Add a Container Port option.

Now that the job is specified, there are several more options available to configure: 

• Job priority level
• New job name
• How a job is capable of behaving if preempted
• Maximum runtime of the job
• Time slice interval for telemetry collection
• Custom labels

Interacting with running and finished jobs

After a job has been launched, you are redirected to a page specific to that job, where the launch details are on the Overview tab. In fact, an equivalent CLI version of the launch form is available under the Command section. 

Jobs are presented in a way that makes them easy to reproduce: either by copying a CLI representation or by cloning the job through the web interface. If ports were added to the job when launched, a URL endpoint is also available (Video 2).

Several additional tabs are present for a scheduled, running, or completed job:

• Telemetry: Key performance metrics for a job over time.
• Status History: A list of states that the job and associated replicas have been in.
• Results: A way to view the files that have been saved to a job’s results directory.
• Log: Searchable, live access to a job’s stdout output.

After a job completes, the job-specific page is still accessible and can be used as a reference for future jobs, or further debugging if something didn’t work out as intended. The results directory and log files from the job can be easily retrieved. Depending on how the job was written, the desired job artifacts could be in a workspace instead of the results directory. 

Base Command Platform provides CLI support for downloading data from a workspace. The resulting artifacts from a job can then be uploaded into Base Command Platform, either made public or kept in the private registry for the org. It provides the starting point for further experimentation in Base Command Platform or a critical component for a model deployed elsewhere.

For more information, see Managing Jobs.

MLOps integration using the NGC API

You can further augment and extend Base Command Platform capabilities with external software integration through the documented NGC API. 

The NGC API can be used for workflow integration or dashboards outside of Base Command Platform, such as third-party MLOps platforms and tools. MLOps software and service providers that have integrated their unique offerings with Base Command Platform include Weights & Biases and Rescale.

As Base Command Platform features evolve and expand, the NGC API enables new and existing software ecosystems to integrate its strengths into other purpose-built solutions.

Conclusion

Base Command Platform is one of the key NVIDIA tools for making AI infrastructure accessible to developers. To get a hands-on sense of how Base Command Platform works, NVIDIA offers a series of labs through NVIDIA LaunchPad. Some labs cover specific use cases around natural language processing and medical imaging, and others are tailored toward gaining experience with Base Command Platform capabilities. 

Learn how to use NVIDIA Base Command Platform to accelerate your containerized AI training workloads. 

Real-time natural language understanding will transform how we interact with intelligent machines and applications.

Real-time natural language understanding will transform how we interact with intelligent machines and applications.

The role of artificial intelligence (AI) in boosting performance and energy efficiency in cellular network operations is rapidly becoming clear. This is…

The role of artificial intelligence (AI) in boosting performance and energy efficiency in cellular network operations is rapidly becoming clear. This is especially the case for radio access networks (RANs), which account for over 60% of industry costs. 

This post explains how AI is transforming the 5G RAN, improving energy and cost efficiency while supporting better use of RAN computing infrastructure.

5G is in full motion 

It is now over 4 years since the first 5G networks were launched. A GSMA study projects that there will be over 1.4 billion 5G connections by 2025. The deployment of standalone 5G networks globally is also beginning to increase. To learn more, see Operators Keep Pushing Forward on 5G Standalone Networks.

In the consumer market, 5G is the default upgrade for an ubiquitous cellular communications service. For the enterprise market, 5G has the optimal combination of high performance, mobility, flexibility, and security to provide the connectivity fabric for enterprise use cases (Figure 1). 

NVIDIA is driving innovation in cellular networks with a fully programmable NVIDIA Aerial SDK for building and deploying GPU-accelerated 5G virtual radio access networks (vRANs). This is providing the building blocks for a standard public 5G network for a telco or a private 5G network implementation with AI-on-5G. 

AI is shaping the current state—and evolution—of 5G 

The role of AI in cellular network operations is growing at a fast pace. AI delivers value through the terabytes of data collected every day—from network elements to customer interactions—and through the resulting insights. These insights are related to managing increased network demand, combating cyberthreats, optimizing services, and improving the customer experience.  

AI is currently applied across different domains in cellular networks such as RAN, core network, operations support systems (OSS), business support systems (BSS), and cloud infrastructure. These AI-enabled functionalities emerged in 4G, are becoming entrenched in 5G, and will become native in 6G. 

While AI will permeate the entire value chain of the industry, its impact on the RAN will be the most profound, particularly given the disproportionate share of industry capex and opex the RAN accounts for. Accordingly, both O-RAN and 3GPP have identified and are working on AI initiatives that improve the performance, flexibility, scalability, and efficiency of the RAN. To learn more, see Embracing AI in 5G-Advanced Towards 6G: A Joint 3GPP and O-RAN Perspective.

AI is transforming the RAN in four key ways: energy savings, mobility management and optimization, load balancing, and Cloud RAN. Read on for more details about each.

Energy savings

The rapid growth of 5G deployment has coincided with a rapid increase in energy costs globally, leading to concerns about high operational costs and carbon emissions. There is the additional concern that some 5G deployments may face hard limits on how much power can be supplied, even if the owners are willing to pay. These concerns create a powerful incentive to increase operational efficiency to achieve higher power efficiencies from current and future network deployments. See Take the Green Train: NVIDIA BlueField DPUs Drive Data Center Efficiency for more details. 

The industry has been using reactive and inflexible rule-based techniques to conserve energy, such as switching on/off cells based on different thresholds of cell load. However, AI offers a proactive and adaptive approach, enabling telcos to predict energy efficiency and load in future states. AI also provides better integration between the RAN and virtualized core network functions (with User Plane Function, for example) through offloading networking, security, and RAN tasks to NVIDIA GPUs and DPUs. 

Mobility management and optimization

Mobile communications systems have the distinct ability to support handovers of devices from one access point to another. This provides service continuity, supports mobility, and optimizes performance. Expectedly, disruptions, delays, and frequency of handovers add up to inefficiencies in network performance. 

Using AI to optimize paging and predict the next cell for handovers offers a significant opportunity to improve performance for advanced features. Such features include sophisticated dual connectivity options, conditional handover, and dual active protocol stack (DAPS) handover. 

AI prediction will rely on insights about the possible movement of the device. To achieve this, the Network Data Analytics Function (NWDAF) from 3GPP SA2 provides data from the core network, applications, and the OSS to improve handover performance, predict device location and performance, and steer traffic to achieve quality network performance.

NVIDIA continues to innovate around the NVIDIA Aerial SDK to support these new expectations for data collection and the use of AI for network management. 

Load balancing 

Handovers enable mobile networks to steer traffic to balance the load across different cell sites and to improve the use of spectrum, RAN, transport, and core infrastructure. This load-balancing decision is achieved by optimizing handover parameters and decisions using current or historical load information. 

However, this task is becoming more challenging due to the use of multiple frequency bands and interworking with different RANs. Current rules will increasingly struggle to cope with fast time-varying scenarios with high mobility, and dynamic traffic patterns with a large number of connections. 

AI models perform better and can predict load to optimize critical tasks using the collection of RAN data. This will improve network performance and user experience. This is a key driver in the development of proprietary AI tools and the current push for some industry standardization to unlock this opportunity at scale. 

Cloud RAN: Colocating RAN and AI in the cloud

While Cloud RAN is not a core application of AI in itself, it is the logical extension of the preceding load-balancing discussion. In this case, a softwarized and cloud-enabled RAN can be colocated on the same cloud infrastructure with AI workloads. Boosting RAN use beyond the typical average 25% of many sites, this will support telcos to extract efficiency gains from an asset that gulps at least 60% of industry capex. 

The suitability to share the same computational resources with other AI workloads, the availability of such orthogonal AI workloads, and the ability to use AI to switch dynamically between the different workloads are key to unlocking this opportunity. 

This Cloud RAN vision will begin in the 5G era and then mature in the 6G era, as the RAN as a workload in the cloud becomes the ultimate destination. By pooling baseband computing resources into a cloud-native environment, the Cloud RAN solution delivers significant improvements in asset use for both Cloud Service Providers (CSPs) and telcos. 

CSPs can run the RAN as a workload alongside their AI workloads within their existing data center architecture while telcos can increase RAN operational efficiency by more than 2x for an estimated >25% impact on EBITDA. 

NVIDIA is shaping this evolution with the Cloud RAN solution based on NVIDIA Spectrum switch, NVIDIA H100 CNX converged accelerator, and NVIDIA Aerial SDK. To learn more, see Unlocking New Opportunities with AI Cloud Infrastructure for 5G vRAN.

The role of AI in the RAN is only part of the overall role of AI in telecom network operations. In addition to the RAN, NVIDIA is working with partners on AI-powered operations to use data insights from telco data to create new revenues and improve operational efficiency. Visit the NVIDIA Telecommunications page to learn more.