This resource is using open-source code maintained in github (see the quick-start-guide section) and available for download from NGC

Bidirectional Encoder Representations from Transformers (BERT) is a new method of pre-training language representations that obtains state-of-the-art results on a wide array of Natural Language Processing (NLP) tasks. This model is based on the BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding paper.

This repository contains scripts to interactively launch data download, training, benchmarking, and inference routines in a Docker container for pre-training and fine-tuning tasks such as question answering. The major differences between the original implementation of the paper and this version of BERT are as follows:

Scripts to download the Wikipedia dataset
Scripts to preprocess downloaded data into inputs and targets for pre-training in a modular fashion
LAMB optimizer to support training with larger batches
Adam optimizer for fine-tuning tasks
Automatic mixed precision (AMP) training support

Other publicly available implementations of BERT include:

This model trains with mixed precision Tensor Cores on NVIDIA Ampere and provides a push-button solution to pre-training on a corpus of choice. As a result, researchers can get results 4x faster than training without Tensor Cores. This model is tested against each NGC monthly container release to ensure consistent accuracy and performance over time.

Model architecture

The BERT model uses the same architecture as the encoder of the Transformer. Input sequences are projected into an embedding space before being fed into the encoder structure. Additionally, positional and segment encodings are added to the embeddings to preserve positional information. The encoder structure is simply a stack of Transformer blocks, which consist of a multi-head attention layer followed by successive stages of feed-forward networks and layer normalization. The multi-head attention layer accomplishes self-attention on multiple input representations.

An illustration of the architecture taken from the Transformer paper is shown below.

$\"BERT\"$

Default configuration

The architecture of the BERT model is almost identical to the Transformer model that was first introduced in the Attention Is All You Need paper. The main innovation of BERT lies in the pre-training step, where the model is trained on two unsupervised prediction tasks using a large text corpus. Training on these unsupervised tasks produces a generic language model, which can then be quickly fine-tuned to achieve state-of-the-art performance on language processing tasks such as question answering.

The BERT paper reports the results for two configurations of BERT, each corresponding to a unique model size. This implementation provides the same default configurations, which are described in the table below.

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n

tr:nth-child(odd)]:bg-surface-sunken\">

Model	Hidden layers	Hidden unit size	Attention heads	Feedforward filter size	Max sequence length	Parameters
BERTBASE	12 encoder	768	12	4 x 768	512	110M
BERTLARGE	24 encoder	1024	16	4 x 1024	512	330M

Feature support matrix

The following features are supported by this model.

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n

tr:nth-child(odd)]:bg-surface-sunken\">

Feature	BERT
Paddle AMP	Yes
Paddle Fleet	Yes
LAMB	Yes
LDDL	Yes
Multi-node	Yes

Features

Fleet is a unified API for distributed training of PaddlePaddle.

LAMB stands for Layerwise Adaptive Moments based optimizer, which is a large batch optimization technique that helps accelerate the training of deep neural networks using large minibatches. It allows using a global batch size of 65536 and 32768 on sequence lengths 128 and 512, respectively, compared to a batch size of 256 for Adam. The optimized implementation accumulates 1024 gradient batches in phase 1 and 4096 steps in phase 2 before updating weights once. This results in a 15% training speedup. On multi-node systems, LAMB allows scaling up to 1024 GPUs resulting in training speedups of up to 72x in comparison to Adam. Adam has limitations on the learning rate that can be used since it is applied globally on all parameters, whereas LAMB follows a layerwise learning rate strategy.

LDDL is a library that enables scalable data preprocessing and loading. LDDL is used by this PaddlePaddle BERT example.

Mixed precision training

Mixed precision is the combined use of different numerical precisions in a computational method. Mixed precision training offers significant computational speedup by performing operations in half-precision format while storing minimal information in single-precision to retain as much information as possible in critical parts of the network. Since the introduction of Tensor Cores in NVIDIA Volta, and following with both the NVIDIA Turing and NVIDIA Ampere architectures, significant training speedups are experienced by switching to mixed precision -- up to 3x overall speedup on the most arithmetically intense model architectures. Using mixed precision training requires two steps:

Porting the model to use the FP16 data type where appropriate.
Adding loss scaling to preserve small gradient values.

The ability to train deep learning networks with lower precision was introduced in the Pascal architecture and first supported in CUDA 8 in the NVIDIA Deep Learning SDK.

For information about:

How to train using mixed precision in PaddlePaddle, refer to the Mixed Precision Training paper and Automatic Mixed Precision Training documentation.
Techniques used for mixed precision training, refer to the Mixed-Precision Training of Deep Neural Networks blog.

Enabling mixed precision

Mixed precision is enabled in Paddle by using the Automatic Mixed Precision (AMP) while storing variables in single-precision format. Furthermore, to preserve small gradient magnitudes in backpropagation, a loss scaling step must be included when applying gradients.\nIn PaddlePaddle, loss scaling can be easily applied by passing in arguments to GradScaler(). The scaling value to be used can be dynamic or fixed.

For an in-depth walkthrough on AMP, check out sample usage here. Paddle AMP is a PaddlePaddle built-in module that provides functions to construct AMP workflow. The details can be found in Automatic Mixed Precision (AMP), which requires minimal network code changes to leverage Tensor Cores performance.

Code example to enable mixed precision for static graph:

Use paddle.static.amp.decorate to wrap optimizer

import paddle.static.amp as amp\nmp_optimizer = amp.decorate(optimizer=optimizer, init_loss_scaling=8.0)\n

Minimize loss , and get scaled_loss, which is useful when you need customized loss.

ops, param_grads = mp_optimizer.minimize(loss)\nscaled_loss = mp_optimizer.get_scaled_loss()\n

For distributed training, it is recommended to use Fleet to enable amp, which is a unified API for distributed training of PaddlePaddle. For more information, refer to Fleet

import paddle.distributed.fleet as fleet\nstrategy = fleet.DistributedStrategy()\nstrategy.amp = True # by default this is false\noptimizer = fleet.distributed_optimizer(optimizer, strategy=dist_strategy)\n

Enabling TF32

TensorFloat-32 (TF32) is the new math mode in NVIDIA A100 GPUs for handling the matrix math, also called tensor operations. TF32 running on Tensor Cores in A100 GPUs can provide up to 10x speedups compared to single-precision floating-point math (FP32) on NVIDIA Volta GPUs.

TF32 Tensor Cores can speed up networks using FP32, typically with no loss of accuracy. It is more robust than FP16 for models, which require a high dynamic range for weights or activations.

For more information, refer to the TensorFloat-32 in the A100 GPU Accelerates AI Training, HPC up to 20x blog post.

TF32 is supported in the NVIDIA Ampere GPU architecture and is enabled by default.

Glossary

Fine-tuning
\nTraining an already pre-trained model further using a task-specific dataset for subject-specific refinements by adding task-specific layers on top if required.

Language Model
\nAssigns a probability distribution over a sequence of words. Given a sequence of words, it assigns a probability to the whole sequence.

Pre-training
\nTraining a model on vast amounts of data on the same (or different) task to build general understandings.

Transformer
\nThe paper Attention Is All You Need introduces a novel architecture called Transformer that uses an attention mechanism and transforms one sequence into another.

Phase 1
\nPre-training on samples of sequence length 128 and 20 masked predictions per sequence.

Phase 2
\nPre-training on samples of sequence length 512 and 80 masked predictions per sequence.

","resourceVersions",[234,235,236],"resourceVersionsTotal","hasMoreVersions","versionMetadata",{"_213":29,"_214":-5,"_53":118,"_115":-5,"_119":215,"_121":-5,"_216":132,"_126":18,"_123":35,"_217":-5,"_122":35,"_218":219,"_220":90,"_221":222,"_223":90,"_224":225,"_226":90,"_227":228,"_229":90,"_230":231,"_232":90,"_124":233},"versionNotFound","isLoggedIn","latestVersionPath","/orgs/nvidia/dle/resources/bert_paddle/-","imageReference","cliDownloadCommand","ngc registry resource download-version nvidia/dle/bert_paddle:22.12.0","helmFetchCommand","viewAllVersionsPath","/orgs/nvidia/dle/resources/bert_paddle/22.12.0/version-history","accessCheckResult",{"_212":90},"hasRelatedCollections","context",{"_208":209,"_210":211},1787604173175,[],[],[202,203,204,205,206,207],{"_49":63,"_51":73,"_53":65,"_55":66,"_57":74,"_59":60},{"_49":63,"_51":70,"_53":65,"_55":66,"_57":67,"_59":60},{"_49":63,"_51":68,"_53":65,"_55":66,"_57":69,"_59":60},{"_49":63,"_51":64,"_53":65,"_55":66,"_57":67,"_59":60},{"_49":61,"_51":62,"_53":151,"_55":-5,"_57":-5,"_59":60},{"_49":50,"_51":52,"_53":54,"_55":56,"_57":58,"_59":60},"orgName","nvidia","teamName","dle","allowed","versionId","description",["D",1675810830256],"totalSizeInBytes","malwareScanStatus","releaseNotes","### Changelog\n\nJanuary 2023\n- [Pre-training using Language Datasets and Data Loaders (LDDL)](https://github.com/NVIDIA/LDDL)\n- Binned pretraining for phase2 with LDDL using a bin size of 64\n\nAugust 2022\n- Pre-training support with LAMB optimizer.\n- Updated Data download and Preprocessing.\n- Integrate DLLogger.\n- Pre-training with AMP, Fleet.\n- Pre-training support with Lamb optimizer.\n- SQuAD finetune support with AdamW optimizer.\n- Updated accuracy and performance tables tested on A100.\n- Initial release.\n \n### Known issues\n \nThere are no known issues with this model.","hasReleaseNotes","setup","The following section lists the requirements you need to meet to start training the BERT model. \n \n### Requirements\n \nThis repository contains a Dockerfile that extends the CUDA NGC container and encapsulates some dependencies. Aside from these dependencies, ensure you have the following components:\n \n* [NVIDIA Docker](https://github.com/NVIDIA/nvidia-docker)\n* [PaddlePaddle 22.12-py3 NGC container](https://catalog.ngc.nvidia.com/orgs/nvidia/containers/paddlepaddle) or newer\n* Supported GPUs:\n * [NVIDIA Ampere architecture](https://www.nvidia.com/en-us/data-center/nvidia-ampere-gpu-architecture/)\n\nFor more information about how to get started with NGC containers, refer to the\nfollowing sections from the NVIDIA GPU Cloud Documentation and the Deep Learning\nDGX Documentation:\n* [Getting Started Using NVIDIA GPU Cloud](https://docs.nvidia.com/ngc/ngc-getting-started-guide/index.html)\n* [Accessing And Pulling From The NGC Container Registry](https://docs.nvidia.com/deeplearning/dgx/user-guide/index.html#accessing_registry)\n\nFor those unable to use the PaddlePaddle NGC container, to set up the required environment or create your own container, refer to the versioned [NVIDIA Container Support Matrix](https://docs.nvidia.com/deeplearning/dgx/support-matrix/index.html).\n\nFor multi-node, the sample provided in this repository requires [Enroot](https://github.com/NVIDIA/enroot) and [Pyxis](https://github.com/NVIDIA/pyxis) set up on a [SLURM](https://slurm.schedmd.com) cluster.\n \nMore information on how to set up and launch can be found in the [Multi-node Documentation](https://docs.nvidia.com/ngc/multi-node-bert-user-guide).","hasSetup","quickStartGuide","To train your model using mixed or TF32 precision with Tensor Cores, perform the following steps using the default parameters of the BERT model. Training configurations to run on 8 x A100 cards and examples of usage are provided at the end of this section. For the specifics concerning training and inference, refer to the [Advanced](#advanced) section.\n \n1. Clone the repository.\n```\ngit clone https://github.com/NVIDIA/DeepLearningExamples.git\ncd DeepLearningExamples/PaddlePaddle/LanguageModeling/BERT\n```\n \n2. Download the NVIDIA pre-trained checkpoint.\nIf you want to use a pre-trained checkpoint, visit [NGC](https://catalog.ngc.nvidia.com/orgs/nvidia/teams/dle/models/bert_large_paddle_ckpt_mode-pretrain/files). This pre-trained checkpoint is used to fine-tune on SQuAD. Ensure you unzip the downloaded file and place the checkpoint in the `checkpoints/` folder. For a checkpoint already fine-tuned for QA on SQuAD v1.1 visit [NGC](https://catalog.ngc.nvidia.com/orgs/nvidia/teams/dle/models/bert_large_paddle_ckpt_mode-qa_ds-squad11/files).\n\n3. Build BERT on top of the NGC container.\n```\nbash scripts/docker/build.sh\n```\n \n4. Start an interactive session in the NGC container to run training/inference.\n```\nbash scripts/docker/launch.sh\n```\n\nBy default: \n- Checkpoints of pretraining and fine-tuning routines are stored in the `results/` folder.\n- Paddle native logs are stored in the `log/` folder.\n- DLLogger's outputs are stored in the `results/` folder. \n\n5. Download the dataset.\n\nThis repository provides scripts to download, verify, and extract the following datasets:\n \n- [SQuAD](https://rajpurkar.github.io/SQuAD-explorer/) (fine-tuning for question answering)\n- Wikipedia (pre-training)\n\nTo download, verify, extract the datasets, run: \n```shell\nbash data/create_datasets_from_start.sh\n```\n\nNote: For fine-tuning only, downloading the Wikipedia dataset can be skipped by commenting it out.\n\nNote: Ensure a complete Wikipedia download. But if the download failed in LDDL,\nremove the output directory `data/wikipedia/` and start over again. \n\n6. Start pre-training.\n \nTo run on a single node 8 x A100 80G cards from within the container, you can use the following script to run pre-training.\n```\nbash scripts/run_pretraining.sh\n```\n \nThe default hyperparameters are set to run on 8x A100 80G cards.\n\nTo run on multiple nodes, refer to the [Multi-node](#multi-node) section. \n\n7. Start fine-tuning with the SQuAD dataset.\n \nThe above pre-trained BERT representations can be fine-tuned with just one additional output layer for a state-of-the-art question answering system. Running the following script launches fine-tuning for question answering with the SQuAD dataset.\n```\nbash scripts/run_squad.sh /workspace/bert/checkpoints/\n```\n\n8. Start validation/evaluation.\n \nFor SQuAD, validation can be performed with the `bash scripts/run_squad.sh /workspace/bert/checkpoints/`, setting `mode` to `eval` in `scripts/run_squad.sh` as follows:\n\n```\nmode=${12:-\"eval\"}\n```\n\n9. Start inference/predictions.\nInference can be performed with the `bash scripts/run_squad.sh /workspace/bert/checkpoints/`, setting `mode` to `prediction` in `scripts/run_squad.sh` as follows:\n\n```\nmode=${12:-\"prediction\"}\n```\n\nNote:\n- Both in `prediction` and `eval` mode, the inference process will be performed and the prediction results will be saved into `/bert-large-uncased/squad/predictions.json`, set in `scripts/run_squad.sh` as follows:\n\n```\nOUT_DIR=${11:-\"/results\"} # For SQuAD.\n```\n- In `eval` mode, after the inference process is completed, the script will further evalute the `predictions.json` on the test dateset and output two metrics: the average `exact match` score and the average `F1` score.\n\n- `predictions.json` only contains a dict of mapping the questions' id to their predicted answers. For example:\n```json\n{\n \"56be4db0acb8001400a502ec\": \"Denver Broncos\",\n \"56be4db0acb8001400a502ed\": \"Carolina Panthers\",\n}\n```\n- All the reference (such as contexts, questions, answers) can be located in test dataset (`dev-v1.1.json`) according to the unique `id`. For example:\n```json\n{\n \"answers\": [{\"answer_start\": 177, \"text\": \"Denver Broncos\"}, {\"answer_start\": 177, \"text\": \"Denver Broncos\"}, {\"answer_start\": 177, \"text\": \"Denver Broncos\"}], \n \"question\": \"Which NFL team represented the AFC at Super Bowl 50?\", \n \"id\": \"56be4db0acb8001400a502ec\"\n}\n```\n\nThis repository contains some predefined configurations to run the pre-training and SQuAD on NVIDIA DGX A100 nodes in `scripts/configs/pretrain_config.sh` and `scripts/configs/squad_config.sh`. For example, to use the default DGX A100 8 gpu config, run:\n```\nbash scripts/run_pretraining.sh $(source scripts/configs/pretrain_config.sh && dgxa100-80g_8gpu_amp) # For pre-training with mixed precision.\nbash scripts/run_pretraining.sh $(source scripts/configs/pretrain_config.sh && dgxa100-80g_8gpu_tf32) # For pre-training with TF32 precision.\nbash scripts/run_squad.sh $(source scripts/configs/squad_config.sh && dgxa100-80g_8gpu_amp) # For the SQuAD v1.1 dataset with mixed precision.\nbash scripts/run_squad.sh $(source scripts/configs/squad_config.sh && dgxa100-80g_8gpu_tf32) # For the SQuAD v1.1 dataset with TF32 precision.\n```","hasQuickStartGuide","advanced","The following sections provide greater details of the dataset, running training and inference, and the training results.\n \n### Scripts and sample code\n\nDescriptions of the key scripts and folders are provided below.\n \n- `data/` - Contains scripts for downloading and preparing individual datasets and will contain downloaded and processed datasets.\n- `scripts/` - Contains shell scripts to launch data download, pre-training, and fine-tuning.\n- `run_squad.sh` - Interface for launching question answering fine-tuning with `run_squad.py`.\n- `run_pretraining.sh` - Interface for launching BERT pre-training with `run_pretraining.py`.\n- `create_pretraining_data.py` - Creates `.hdf5` files from shared text files in the final step of dataset creation.\n- `modeling.py` - Implements the BERT pre-training and fine-tuning model architectures with PaddlePaddle.\n- `optimizer.py` - Implements the LAMB optimizer with PaddlePaddle.\n- `tokenizer.py` - Implements the BERT tokenizer.\n- `loss.py` - Implement the loss function for BERT pre-training and fine-tuning.\n- `run_squad.py` - Implements fine-tuning training and evaluation for question answering on the [SQuAD](https://rajpurkar.github.io/SQuAD-explorer/) dataset.\n- `run_pretraining.py` - Implements BERT pre-training.\n\n### Parameters\n \n#### Pre-training parameters\n\nBERT is designed to pre-train deep bidirectional networks for language representations. The following scripts replicate pre-training on Wikipedia from this [paper](https://arxiv.org/pdf/1810.04805.pdf). These scripts are general and can be used for pre-training language representations on any corpus of choice.\n \nThe complete list of the available parameters for the `run_pretraining.py` script is :\n```\nGlobal:\n --input-dir INPUT_DIR\n The input data directory. Should be specified by users and contain .hdf5 files for the task. (default: None)\n --vocab-file VOCAB_FILE\n Vocabulary mapping/file BERT was pretrainined on. (default: None)\n --output-dir OUTPUT_DIR\n The output directory where the model checkpoints will be written. Should be specified by users. (default: None)\n --bert-model {bert-base-uncased,bert-base-cased,bert-large-uncased,bert-large-cased,custom}\n Specifies the type of BERT model to use. If it is set as custom, the path to the config file must be given by specifying --config-file (default: bert-large-uncased)\n --config-file CONFIG_FILE\n The BERT model config. If set to None, `<--bert-model>.json` in folder `bert_configs` will be used. (default: None)\n --max-steps MAX_STEPS\n Total number of training steps to perform. (default: None)\n --log-freq LOG_FREQ Frequency of logging loss. (default: 10)\n --num-steps-per-checkpoint NUM_STEPS_PER_CHECKPOINT\n Number of update steps until a model checkpoint is saved to disk. (default: 100)\n --from-pretrained-params FROM_PRETRAINED_PARAMS\n Path to pretrained parameters. If set to None, no pretrained parameter will be used. (default: None)\n --from-checkpoint FROM_CHECKPOINT\n A checkpoint path to resume training. If set to None, no checkpoint will be used. If not None, --from-pretrained-params will be ignored. (default: None)\n --last-step-of-checkpoint LAST_STEP_OF_CHECKPOINT\n The step id of the checkpoint given by --from-checkpoint. It should be None, auto, or integer > 0. If it is set as None, then training will start from the 1-th epoch. If it is set as auto,\n then it will search the largest integer-convertible folder --from-checkpoint, which contains the required checkpoint. (default: None)\n --from-phase1-final-params FROM_PHASE1_FINAL_PARAMS\n Path to final checkpoint of phase1, which will be used to initialize the parameter in the first step of phase2, and ignored in the rest steps of phase2. (default: None)\n --seed SEED Random seed. (default: 42)\n --report-file REPORT_FILE\n A file in which to store JSON experiment reports. (default: ./report.json)\n --model-prefix MODEL_PREFIX\n The prefix name of model files to save/load. (default: bert_paddle)\n --show-config SHOW_CONFIG\n To show arguments. (default: True)\n --enable-cpu-affinity ENABLE_CPU_AFFINITY\n To enable in-built GPU-CPU affinity. (default: True)\n --benchmark To enable benchmark mode. (default: False)\n --benchmark-steps BENCHMARK_STEPS\n Steps for a benchmark run, only applied when --benchmark is set. (default: 20)\n --benchmark-warmup-steps BENCHMARK_WARMUP_STEPS\n Warmup steps for a benchmark run, only applied when --benchmark is set. (default: 20)\n\nTraining:\n --optimizer OPTIMIZER\n The name of optimizer. It should be one of {Lamb, AdamW}. (default: Lamb)\n --gradient-merge-steps GRADIENT_MERGE_STEPS\n Number of update steps to accumulate before performing a backward/update pass. (default: 1)\n --learning-rate LEARNING_RATE\n The initial learning rate. (default: 0.0001)\n --warmup-start-lr WARMUP_START_LR\n The initial learning rate for warm up. (default: 0.0)\n --warmup-proportion WARMUP_PROPORTION\n Proportion of training to perform linear learning rate warmup for. For example, 0.1 = 10% of training. (default: 0.01)\n --beta1 BETA1 The exponential decay rate for the 1st moment estimates. (default: 0.9)\n --beta2 BETA2 The exponential decay rate for the 2st moment estimates. (default: 0.999)\n --epsilon EPSILON A small float value for numerical stability. (default: 1e-06)\n --weight-decay WEIGHT_DECAY\n The weight decay coefficient. (default: 0.01)\n --max-seq-length MAX_SEQ_LENGTH\n The maximum total input sequence length after WordPiece tokenization. Sequences longer than this will be truncated, and sequences shorter than this will be padded. (default: 512)\n --batch-size BATCH_SIZE\n The batch size for training (default: 32)\n --phase1 The phase of BERT pretraining. It should not be set with --phase2 at the same time. (default: False)\n --phase2 The phase of BERT pretraining. It should not be set with --phase1 at the same time. (default: False)\n --max-predictions-per-seq MAX_PREDICTIONS_PER_SEQ\n The maximum total of masked tokens in the input sequence (default: 80)\n\nAdvanced Training:\n --amp Enable automatic mixed precision training (AMP). (default: False)\n --scale-loss SCALE_LOSS\n The loss scalar for AMP training, only applied when --amp is set. (default: 1.0)\n --use-dynamic-loss-scaling\n Enable dynamic loss scaling in AMP training, only applied when --amp is set. (default: False)\n --use-pure-fp16 Enable pure FP16 training, only applied when --amp is set. (default: False)\n```\n\n#### Fine tuning parameters\n\n* SQuAD \n\nDefault arguments are listed below in the order `scripts/run_squad.sh` expects:\n \n- Initial checkpoint - The default is `checkpoints/squad`.\n- Number of training Epochs - The default is `2`.\n- Batch size - The default is `32`.\n- Learning rate - The default is `4.6e-5`.\n- Warmup proportion - The default is `0.2`.\n- Precision (either `amp` or `fp32`) - The default is `amp`.\n- Number of GPUs - The default is `8`.\n- Seed - The default is `1`.\n- SQuAD directory - The default is `data/download/squad/v1.1`.\n- Vocabulary file (token to ID mapping) - The default is `vocab/bert-large-uncased-vocab.txt`.\n- Output directory for results - The default is `/results`.\n- Mode (`train`, `eval`, `prediction`, `train_eval`) - The default is `train_eval`.\n- Config file for the BERT model (It should be the same as the pre-trained model) - The default is `None`, which means `bert_configs/<--bert-model>.json` will be used.\n- Max steps - The default is -1.\n- Enable benchmark - The default is `false`.\n- Benchmark steps - The default is `100`.\n- Benchmark warmup steps - The default is `100`.\n\nThe script saves the final checkpoint to the `/results/bert-large-uncased/squad` folder.\n\nNote:\n- For SQuAD fine-tuning, `<--max-steps>` is not required since it's usually trained for two or three epochs. If `<--max-steps>` is not set or set to -1, it will be trained for `<--epochs>` epochs. If `<--max-steps>` is set to a positive number, the total training steps is calculated by: `total_steps = min(max_steps, epochs * steps_per_epoch)`.\n- For pre-training, `<--max-steps>` is required and `<--epochs>` is deprecated. Because We typically train for a specified number of steps rather than epochs.\n\n#### Multi-node\nMulti-node runs can be launched on a pyxis/enroot Slurm cluster (refer to [Requirements](#requirements)) with the `run.sub` script with the following command for a 4-node DGX-A100 example for both phase 1 and phase 2:\n \n```\nTRAIN_BATCH_SIZE=256 GRADIENT_ACCUMULATION_STEPS=8 PHASE=1 sbatch -N4 run.sub\nTRAIN_BATCH_SIZE=32 GRADIENT_ACCUMULATION_STEPS=32 PHASE=2 sbatch -N4 run.sub\n```\n \nCheckpoints after phase 1 will be saved in `checkpointdir` specified in `run.sub`. The checkpoint will be automatically picked up to resume training on phase 2. Note that phase 2 should be run after phase 1.\n \nThe batch variables `BATCHSIZE`, `GRADIENT_STEPS`,`PHASE` refer to the Python arguments `--batch-size`, `--gradient-merge-steps`, `--phase1/--phase2` respectively.\n \nNote that the `run.sub` script is a starting point that has to be adapted depending on the environment. In particular, variables such as `datadir` handle the location of the files for each phase. \n \nRefer to the file's contents to find the full list of variables to adjust for your system.\n\n### Command-line options\n \nTo view the full list of available options and their descriptions, use the `-h` or `--help` command-line option, for example:\n \n`python run_pretraining.py --help`\n \n`python run_squad.py --help`\n\nDetailed descriptions of command-line options can be found in the [Parameters](#parameters) section.\n \n### Getting the data\n\nFor pre-training BERT, we use the Wikipedia (2500M words) dataset. We extract \nonly the text passages and ignore headers, lists, and tables. BERT requires that\ndatasets are structured as a document level corpus rather than a shuffled \nsentence-level corpus because it is critical to extract long contiguous \nsentences. `data/create_datasets_from_start.sh` uses the LDDL downloader to \ndownload the Wikipedia dataset, and `scripts/run_pretraining.sh` uses the LDDL \npreprocessor and load balancer to preprocess the Wikipedia dataset into Parquet\nshards which are then streamed during the pre-training by the LDDL data loader.\nRefer to [LDDL's README](https://github.com/NVIDIA/LDDL/blob/main/README.md) for more \ninformation on how to use LDDL. Depending on the speed of your internet \nconnection, downloading and extracting the Wikipedia dataset takes a few hours,\nand running the LDDL preprocessor and load balancer takes half an hour on a \nsingle DGXA100 node.\n\nFor fine-tuning a pre-trained BERT model for specific tasks, by default, this repository prepares the following dataset:\n \n- [SQuAD](https://rajpurkar.github.io/SQuAD-explorer/): for question answering\n\n#### Dataset guidelines\n \nThe procedure to prepare a text corpus for pre-training is described in the above section. This section provides additional insight into how exactly raw text is processed so that it is ready for pre-training.\n \nFirst, raw text is tokenized using [WordPiece tokenization](https://arxiv.org/pdf/1609.08144.pdf). A [CLS] token is inserted at the start of every sequence, and the two sentences in the sequence are separated by a [SEP] token.\n \nNote: BERT pre-training looks at pairs of sentences at a time. A sentence embedding token [A] is added to the first sentence and token [B] to the next.\n \nBERT pre-training optimizes for two unsupervised classification tasks. The first is Masked Language Modeling (Masked LM). One training instance of Masked LM is a single modified sentence. Each token in the sentence has a 15% chance of being replaced by a [MASK] token. The chosen token is replaced with [MASK] 80% of the time, 10% with a random token and the remaining 10% the token is retained. The task is then to predict the original token.\n \nThe second task is next sentence prediction. One training instance of BERT pre-training is two sentences (a sentence pair). A sentence pair may be constructed by simply taking two adjacent sentences from a single document or by pairing up two random sentences with equal probability. The goal of this task is to predict whether or not the second sentence followed the first in the original document.\n\n### Training process\n \nThe training process consists of two steps: pre-training and fine-tuning.\n \n#### Pre-training\n\nPre-training is performed using the `run_pretraining.py` script along with parameters defined in the `scripts/run_pretraining.sh`.\n \nThe `run_pretraining.sh` script runs a job on a single node that trains the BERT-large model from scratch using Wikipedia datasets as training data using the LAMB optimizer. By default, the training script runs two phases of training with a hyperparameter recipe specific to 8x A100 80G cards:\n\nPhase 1: (Maximum sequence length of 128)\n- Runs on 8 GPUs with a training batch size of 256 per GPU.\n- Uses a learning rate of 6e-3.\n- Has AMP enabled.\n- Runs for 7038 steps, where the first 28.43% (2000) are warm-up steps\n- Saves a checkpoint every 200 iterations (keeps only the latest three checkpoints) and at the end of training. All checkpoints and training logs are saved to the `results/` directory (in the container, which can be mounted to a local directory).\n- Creates a log file containing all the output\n\nPhase 2: (Maximum sequence length of 512)\n- Runs on 8 GPUs with training batch size of 32 per GPU\n- Uses a learning rate of 4e-3\n- Has AMP enabled\n- Runs for 1563 steps, where the first 12.8% are warm-up steps\n- Saves a checkpoint every 200 iterations (keeps only the latest 3 checkpoints) and at the end of training. All checkpoints, and training logs are saved to the `results/` directory (in the container which can be mounted to a local directory).\n- Creates a log file containing all the output\n \nThe default hyperparameters in `run_pretraining.sh` will train BERT-large model using FP16 arithmetic on the Wikipedia dataset to state-of-the-art accuracy on a DGXA100 with 8x80GB A100 cards. 90% of the training steps are done with sequence length 128 (phase 1 of training) and 10% of the training steps are done with sequence length 512 (phase 2 of training).\n \n```shell\nbash run_pretraining.sh \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \n```\n\nWhere:\n- `` is the per-GPU batch size used for training. Larger batch sizes run more efficiently but require more memory.\n- `` is the base learning rate for training\n- `` is the type of math in your model, which can be either `amp` or `fp32`. The options mean:\n - AMP: Mixed precision 16 and 32-bit floats.\n - FP32: 32-bit IEEE single precision floats.\n- `` is the number of GPUs to use for training. Must be equal to or smaller than the number of GPUs attached to your node.\n- `` is the percentage of training steps used for warm-up at the start of training.\n- `` is the total number of training steps.\n- `` controls how often checkpoints are saved.\n- `` a flag indicating if output should be written to a log file or not (acceptable values are `true` or `false`. `true` indicates output should be saved to a log file.)\n- `` an integer indicating the number of steps to accumulate gradients over.\n- `` random seed for the run.\n- `` is per-GPU batch size used for training in phase 2. Larger batch sizes run more efficiently, but require more memory.\n- `` is the base learning rate for training phase 2.\n- `` is the percentage of training steps used for warm-up at the start of training.\n- `` is the total number of training steps for phase 2, to be continued in addition to phase 1.\n- `` is an integer indicating the number of steps to accumulate gradients in phase 2.\n- `` is the path to dataset of phase 1. It should be a path to the folder containing `.hdf` files.\n- `` is the root path to bert code.\n- `` is the path to the checkpoint to start the pretraining routine on (Usually a BERT pre-trained checkpoint).\n- `wikipedia_source` is the path to the 'source' subdirectory for the Wikipedia corpus.\n- `num_dask_workers` is the number of dask workers to preprocess the bert dataset.\n- `num_shards_per_workers` is the number of the output parquet/txt shards per worker.\n- `num_workers` is the number of workers for dataloading.\n- `sample_ratio` is the ratio of how many articles/documents are sampled from each corpus.\n- `phase2_bin_size` is the stride of the sequence length for each binbin size for phase2.\n- `masking` LDDL supports both static and dynamic masking. Refer to [LDDL's README](https://github.com/NVIDIA/LDDL/blob/main/README.md) for more information.\n- `` is the path to the bert config file.\n- `` a flag to enable benchmark. The train process will warmup for `` and then measure the throughput of the following ``.\n\nNote that: \n- If users follow [Quick Start Guide](#quick-start-guide) to set up container and dataset, there is no need to set any parameters. For example:\n```shell\nbash scripts/run_pretraining.sh\n```\n- It's common for users to mount their own datasets or customize the hyperparameters. If so, it's better to specify all parameters in `scripts/run_pretraining.sh` manually and ensure that all paths are correct. For example:\n```shell\nbash scripts/run_pretraining.sh \\\n 256 6e-3 amp 8 0.2843 7038 200 false 32 0 bert_pretraining \\\n 32 4e-3 0.128 1563 128 \\\n /path/to/dataset/phase1 \\\n /path/to/dataset/phase2 \\\n /workspace/bert \\\n None \\\n /path/to/wikipedia/source \\\n 32 128 4 0.9 64 static \\\n None false\n```\n \nTo run the pre-training routine on an initial checkpoint, point the `from-checkpoint` variable to the location of the checkpoint folder in `scripts/run_pretraining.sh`.\n\nBoth `scripts/run_pretraining.sh` and `scripts/run_squad.sh` rely on positional arguments, which means that if you want to change just one variable from the default value, you need to explicitly specify the values of all variables before it in the argument list, which can be annoying sometimes. Thus, it's also recommended for users to modify the predefined configurations or create their own configuration in `scripts/configs/` folder and run scripts like: `bash scripts/run_pretraining.sh $(source scripts/configs/.sh && )`. \n\nIf users want to move away from bash scripts and launch python scripts directly. It's also recommended to run `run_pretraining.py` and `run_squad.py` with named arguments. For example:\n```shell\npython3 -m paddle.distributed.launch \\\n --gpus=\"0,1,2,3,4,5,6,7\" \\\n ./run_pretraining.py \\\n --input-dir=/path/to/dataset/phase1 \\\n --vocab-file=vocab/bert-large-uncased-vocab.txt \\\n --output-dir=./results \\\n --bert-model=bert-large-uncased \\\n --from-checkpoint=./results/bert-large-uncased/phase1 \\\n --last-step-of-checkpoint=auto \\\n --batch-size=256 \\\n --max-steps=7038 \\\n --num-steps-per-checkpoint=200 \\\n --log-freq=1 \\\n --max-seq-length=128 \\\n --max-predictions-per-seq=20 \\\n --gradient-merge-steps=32 \\\n --amp \\\n --use-dynamic-loss-scaling \\\n --optimizer=Lamb \\\n --phase1 \\\n --scale-loss=1048576 \\\n --learning-rate=6e-3 \\\n --warmup-proportion=0.2843 \\\n --report-file=./results/dllogger_p1.json\n```\nWe provide examples in `scripts/run_pretraining_p1.sh` for phase 1 pre-training and `scripts/run_pretraining_p2.sh` for phase 2 pre-training. For more imformation about each named argument, run `python run_pretraining.py -h` or `python run_squad.py -h`.\n\nNote that:\n- All `` arguments denote the size of a batch computed simultaneously on each GPU. For example, to run phase 1 pretraining on 8xA100, the `` is set to 256 and the `` is set to 32 by default. Each GPU will compute 32 batches before performing a backward/update pass, which means the effective batch size per GPU is `256 * 32 = 8192` and the global batch size is `8192 * #GPUs = 65536`.\n- Checkpoints will be stored in `<--output-dir>/<--bert-model>/<--phase>/` folders.\n\nFor example, to run bert-large-uncased pre-training phase1 for 1000 steps and save checkpoints into `results/bert-large-uncased/phase1` for every 200 steps, set the following args:\n```\n--output-dir results\n--bert-model bert-large-uncased\n--phase1\n--num-steps-per-checkpoint 200\n--max_steps 1000\n```\nThe lastest 3 checkpoints will be saved into `results/bert-large-uncased/phase1/600`, `results/bert-large-uncased/phase1/800` and `results/bert-large-uncased/phase1/1000`.\n\nEach checkpoint includes four files:\n- `<--model-prefix>.pdparams`: A file contains all the trainable tensors.\n- `<--model-prefix>.pdopts`: A file contains all the tensors used by the optimizer. For Adam optimizer, it contains beta1, beta2, momentum, and so on. (If the optimizer has no Tensor need to save (like SGD), no file will be generated).\n- `<--model-prefix>.pdmodel`: A file to keep the description of the program. It's only used for deployment.\n- `<--model-prefix>_progress.json>`: The record of training progress, including file_id, step_id, phase_id, etc.\n\nMake sure:\n- Resume from checkpoints with `--from-checkpoint`: Both `<--model-prefix>.pdopts` and `<--model-prefix>.pdparams` must be in the given path.\n- Start from pre-trained weights with `--from-pretrained-params`: `<--model-prefix>.pdparams` must be in the given path.\n- Don't set `--from-checkpoint` and `--from-pretrained-params` at the same time.\n\nThe difference between those two is that `--from-pretrained-params` contain only model weights, and `--from-checkpoint`, apart from model weights, contain the optimizer state and LR scheduler state.\n\n`--from-checkpoint` is suitable for dividing the training into parts, for example, in order to divide the training job into shorter stages, or restart training after infrastructure faults.\n\n`--from-pretrained-params` can be used to set the initial parameter of phase2 or as a base for fine-tuning the model to a different dataset.\n\nAssume the training process aborts during the 700th step due to infrastructure faults in the previous example, which means the latest checkpoint is saved in `results/bert-large-uncased/phase1/600`. To resume training from the checkpoint, specify `--from-checkpoint` and `--last-step-of-checkpoint` with following these steps: \n- Set `--from-checkpoint` to `results/bert-large-uncased/phase1/600`.\n- Set `--last-step-of-checkpoint` to `600`, which can also be omitted because `results/bert-large-uncased/phase1/600/bert_paddle_progress.json` records the last step id.\nThen rerun the training to resume training from the 601th step to the 1000th step.\n\nWe also provide automatic searching for the checkpoint from the last step. It can be enabled by setting `--last-step-of-checkpoint` as `auto`. Noted that if enable automatic searching, `--from-checkpoint` should be a folder containing checkpoint files or `/`. In the previous example, it should be `results/bert-large-uncased/phase1`.\n\n`--from-phase1-final-params` actually does the same thing as `--from-pretrained-params`. The difference is that the former only works in the first step of phase2, while the latter always works as long as it is set.\n\nAssume you want to train bert-large from scratch however the train progress needs to be restarted many times due to the cluster time limit. To avoid changing `--from-checkpoint` at each run, set the following args:\n- Phase1\n```\n--from-checkpoint results/bert-large-uncased/phase1\n--last-step-of-checkpoint auto\n```\n- Phase2\n```\n--from-checkpoint results/bert-large-uncased/phase2\n--last-step-of-checkpoint auto\n--from-phase1-final-params results/bert-large-uncased/phase1/7038\n```\nAt each rerun, the program will automatically find the checkpoint without having to specify it manually.\n\n#### Fine-tuning\n \nFine-tuning is provided for a variety of tasks. The following tasks are included with this repository through the following scripts (only support SQuAD for now): \n \n- Question Answering (`scripts/run_squad.sh`)\n \nBy default, each Python script implements fine-tuning a pre-trained BERT model for a specified number of training epochs as well as evaluation of the fine-tuned model. Each shell script invokes the associated Python script with the following default parameters:\n \n- Uses 8 GPUs\n- Has FP16 precision enabled\n- Saves a checkpoint at the end of training to the `./results/checkpoints/squad` folder\n \nFine-tuning Python scripts implement support for mixed precision and multi-GPU training. For a full list of parameters and associated explanations, refer to the [Parameters](#parameters) section.\n\nThe fine-tuning shell scripts have positional arguments outlined below:\n \n```shell\nbash scripts/run_squad.sh \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \\\n \n```\n \nBy default, the `mode` argument is set to `train eval`. Refer to the [Quick Start Guide](#quick-start-guide) for explanations of each positional argument.\n \nNote: The first positional argument (the path to the checkpoint to load) is required.\n \nThe fine-tuning script assumes that the corresponding dataset files exist in the `data/` directory or a separate path can be a command-line input to `run_squad.sh`.\n\n### Inference process\n \nFine-tuning inference can be run to obtain predictions on fine-tuning tasks, for example, Q&A on SQuAD. Evaluation fine-tuning is enabled by the same scripts as training:\n \n- Question Answering (`run_squad.py`)\n \nThe `mode` argument of the shell script is used to run in evaluation mode. The fine-tuned BERT model will be run on the evaluation dataset, and the evaluation loss and accuracy will be displayed.\n \nEach inference shell script expects dataset files to exist in the same locations as the corresponding training scripts. The inference scripts can be run with default settings. By setting the `mode` variable in the script to either `eval` or `prediction` flag, you can choose between running predictions and evaluating them on a given dataset or just obtain the model predictions.\n\n`bash scripts/run_squad.sh `","hasAdvanced","performance","### Benchmarking\n \nThe following section shows the steps to run benchmarks measuring the model performance in training and inference modes.\n \n#### Training performance benchmark\n \nTraining performance benchmarks for pre-training can be obtained by running `scripts/run_pretraining.sh`, and fine-tuning can be obtained by running `scripts/run_squad.sh` for SQuAD, respectively. The required parameters can be passed through the command-line as described in [Training process](#training-process).\n\nTo benchmark the training performance on a specific batch size for pre-training, refer to [Pre-training](#pre-training) and turn on the `` flags. An example call to run pretraining for 20 steps (10 steps for warmup and 10 steps to measure, both in phase1 and phase2) and generate throughput numbers:\n```shell\nbash scripts/run_pretraining.sh \\\n 256 6e-3 amp 8 0.2843 7038 200 false \\\n 32 0 bert_pretraining 32 4e-3 0.128 1563 128 \\\n /path/to/dataset/phase1 \\\n /path/to/dataset/phase2 \\\n /workspace/bert \\\n None \\\n /path/to/wikipedia/source \\\n 32 128 4 0.9 64 static \\\n None true 10 10\n```\n\nTo benchmark the training performance on a specific batch size for SQuAD, refer to [Fine-tuning](#fine-tuning) and turn on the `` flags. An example call to run training for 200 steps (100 steps for warmup and 100 steps to measure), and generate throughput numbers:\n```shell\nbash scripts/run_squad.sh \\\n /path/to/pretrained/model \\\n 2 32 4.6e-5 0.2 amp 8 42 \\\n /path/to/squad/v1.1 \\\n vocab/bert-large-uncased-vocab.txt \\\n results/checkpoints \\\n train \\\n bert_configs/bert-large-uncased.json \\\n -1 true 100 100\n```\n \n#### Inference performance benchmark\nInference performance benchmark for fine-tuning can be obtained by running `scripts/run_squad.sh`. The required parameters can be passed through the command-line as described in [Inference process](#inference-process).\n \nTo benchmark the inference performance on a specific batch size for SQuAD, run:\n```shell\nbash scripts/run_squad.sh \\\n \\\n \\\n \\\n \\\n \\\n eval \\\n \\\n \n```\n \nAn example call to run inference and generate throughput numbers:\n```\nbash scripts/run_squad.sh \\\n /path/to/pretrained/model \\\n 2 32 4.6e-5 0.2 amp 8 42 \\\n /path/to/squad/v1.1 \\\n vocab/bert-large-uncased-vocab.txt \\\n results/checkpoints \\\n eval \\\n bert_configs/bert-large-uncased.json \\\n -1 true 100 100\n```\n\n### Results\n \nThe following sections provide details on how we achieved our performance and accuracy in training and inference. \n \n#### Training accuracy results\n \nOur results were obtained by running the `scripts/run_squad.sh` and `scripts/run_pretraining.sh` training scripts in the paddle NGC container unless otherwise specified.\n \n##### Pre-training loss results: NVIDIA DGX A100 (8x A100 80GB)\n\n| DGX System | GPUs / Node | Precision | Accumulated Batch size / GPU (Phase 1 and Phase 2) | Accumulation steps (Phase 1 and Phase 2) | Final Loss | Time to train(hours) | Time to train speedup (TF32 to mixed precision) |\n|--------------------|-------------|-----------|----------------------------------------------------|------------------------------------------|-------------------|----------------------|-------------------------------------------------|\n| 32 x DGX A100 80GB | 8 | AMP | 256 and 128 | 1 and 4 | 1.409 | ~ 1.2 hours | 1.72 |\n| 32 x DGX A100 80GB | 8 | TF32 | 128 and 16b | 2 and 8 | 1.421 | ~ 2.5 hours | 1 |\n\n##### Pre-training loss curves\n![Pre-training Loss Curves](https://raw.githubusercontent.com/NVIDIA/DeepLearningExamples/master/PaddlePaddle/LanguageModeling/BERT//images/bert_loss.png)\n\n\n##### Fine-tuning accuracy results: NVIDIA DGX A100 (8x A100 80GB)\n* SQuAD\n\n| GPUs | Batch size / GPU (TF32 and FP16) | Accuracy - TF32(% F1) | Accuracy - mixed precision(% F1) | Time to train(hours) - TF32 | Time to train(hours) - mixed precision | Time to train speedup (TF32 to mixed precision) |\n|------|----------------------------------|-----------------------|----------------------------------|-----------------------------|----------------------------------------|-------------------------------------------------|\n| 8 | 32 | 91.13 | 91.11 | 0.078 | 0.056 | 1.39 |\n\n##### Training stability test\n \n###### Pre-training stability test\n \n| Accuracy Metric | Seed 0 | Seed 1 | Seed 2 | Seed 3 | Seed 4 | Mean | Standard Deviation |\n|-----------------|--------|--------|--------|--------|--------|-------|--------------------|\n| Final Loss | 1.409 | 1.367 | 1.528 | 1.434 | 1.470 | 1.442 | 0.049 |\n \n###### Fine-tuning stability test\n* SQuAD\n \nTraining stability with 8 GPUs, FP16 computations, batch size of 32:\n \n| Accuracy Metric | Seed 0 | Seed 1 | Seed 2 | Seed 3 | Seed 4 | Seed 5 | Seed 6 | Seed 7 | Seed 8 | Seed 9 | Mean | Standard Deviation |\n|-----------------|--------|--------|--------|--------|--------|--------|--------|--------|--------|--------|-------|--------------------|\n| Exact Match % | 84.07 | 84.39 | 83.94 | 83.78 | 83.85 | 84.47 | 84.13 | 84.20 | 84.03 | 83.80 | 84.07 | 0.225 |\n| f1 % | 90.86 | 91.00 | 90.82 | 90.56 | 90.76 | 91.11 | 90.77 | 90.90 | 90.65 | 90.54 | 90.80 | 0.173 |\n \n#### Training performance results\n\n##### Training performance: NVIDIA DGX A100 (8x A100 80GB)\n\nOur results were obtained by running the script `run_pretraining.sh` in the PaddlePaddle:22.12-py3 NGC container on NVIDIA DGX A100 (8x A100 80GB) GPUs. Performance numbers (in sequences per second) were averaged over a few training iterations.\n\n###### Pre-training NVIDIA DGX A100 (8x A100 80GB)\n\n| GPUs | Batch size / GPU (TF32 and FP16) | Accumulation steps (TF32 and FP16) | Sequence length | Throughput - TF32(sequences/sec) | Throughput - mixed precision(sequences/sec) | Throughput speedup (TF32 - mixed precision) | Weak scaling - TF32 | Weak scaling - mixed precision |\n|------|----------------------------------|------------------------------------|-----------------|----------------------------------|---------------------------------------------|---------------------------------------------|---------------------|--------------------------------|\n| 1 | 8192 and 8192 | 64 and 32 | 128 | 307 | 633 | 2.06 | 1.00 | 1.00 |\n| 8 | 8192 and 8192 | 64 and 32 | 128 | 2428 | 4990 | 2.06 | 7.91 | 7.88 |\n| 1 | 4096 and 4096 | 256 and 128 | 512 | 107 | 219 | 2.05 | 1.00 | 1.00 |\n| 8 | 4096 and 4096 | 256 and 128 | 512 | 851 | 1724 | 2.26 | 7.95 | 7.87 |\n\n###### Pre-training NVIDIA DGX A100 (8x A100 80GB) Multi-node Scaling\n\n| Nodes | GPUs / node | Batch size / GPU (TF32 and FP16) | Accumulated Batch size / GPU (TF32 and FP16) | Accumulation steps (TF32 and FP16) | Sequence length | Mixed Precision Throughput | Mixed Precision Strong Scaling | TF32 Throughput | TF32 Strong Scaling | Speedup (Mixed Precision to TF32) |\n|-------|-------------|----------------------------------|------------------------------------|-----------------|----------------------------|--------------------------------|-----------------|---------------------|-----------------------------------|-----|\n| 1 | 8 | 126 and 256 | 8192 and 8192 | 64 and 32 | 128 | 4990 | 1 | 2428 | 1 | 2.06 |\n| 2 | 8 | 126 and 256 | 4096 and 4096 | 32 and 16 | 128 | 9581 | 1.92 | 4638 | 1.91 | 2.07 |\n| 4 | 8 | 126 and 256 | 2048 and 2048 | 16 and 8 | 128 | 19262 | 3.86 | 9445 | 3.89 | 2.04 |\n| 8 | 8 | 126 and 256 | 1024 and 1024 | 8 and 4 | 128 | 37526 | 7.52 | 18335 | 7.55 | 2.05 |\n| 16 | 8 | 126 and 256 | 512 and 512 | 4 and 2 | 128 | 71156 | 14.26 | 35526 | 14.63 | 2.00 |\n| 32 | 8 | 126 and 256 | 256 and 256 | 2 and 1 | 128 | 142087 | 28.47 | 69701 | 28.71 | 2.04 |\n| 1 | 8 | 16 and 32 | 4096 and 4096 | 256 and 128 | 512 | 1724 | 1 | 851 | 1 | 2.03 |\n| 2 | 8 | 16 and 32 | 2048 and 2048 | 128 and 64 | 512 | 3305 | 1.92 | 1601 | 1.88 | 2.06 |\n| 4 | 8 | 16 and 32 | 1024 and 1024 | 64 and 32 | 512 | 6492 | 3.77 | 3240 | 3.81 | 2.00 |\n| 8 | 8 | 16 and 32 | 512 and 512 | 32 and 16 | 512 | 12884 | 7.47 | 6329 | 7.44 | 2.04 |\n| 16 | 8 | 16 and 32 | 256 and 256 | 16 and 8 | 512 | 25493 | 14.79 | 12273 | 14.42 | 2.08 |\n| 32 | 8 | 16 and 32 | 128 and 128 | 8 and 4 | 512 | 49307 | 28.60 | 24047 | 28.26 | 2.05 |\n\n###### Fine-tuning NVIDIA DGX A100 (8x A100 80GB)\n\n* SQuAD\n \n| GPUs | Batch size / GPU (TF32 and FP16) | Throughput - TF32(sequences/sec) | Throughput - mixed precision(sequences/sec) | Throughput speedup (TF32 - mixed precision) | Weak scaling - TF32 | Weak scaling - mixed precision |\n|------|----------------------------------|----------------------------------|---------------------------------------------|---------------------------------------------|---------------------|--------------------------------|\n| 1 | 32 and 32 | 83 | 120 | 1.45 | 1.00 | 1.00 |\n| 8 | 32 and 32 | 629 | 876 | 1.39 | 7.59 | 7.30 |\n \n#### Inference performance results\n\n##### Inference performance: NVIDIA DGX A100 (1x A100 80GB) \n \nOur results were obtained by running `scripts/run_squad.sh` in the PaddlePaddle:22.08-py3 NGC container on NVIDIA DGX A100 with (1x A100 80G) GPUs.\n \n###### Fine-tuning inference on NVIDIA DGX A100 (1x A100 80GB)\n\n* SQuAD\n \n| GPUs | Batch Size (TF32/FP16) | Sequence Length | Throughput \\- TF32\$sequences/sec\$ | Throughput \\- Mixed Precision\$sequences/sec\$ |\n|------|---------------------------|-----------------|-------------------------------------|------------------------------------------------|\n| 1 | 32/32 | 384 | 131 | 158 |\n\nTo achieve these same results, follow the steps in the [Quick Start Guide](#quick-start-guide).\n \nThe inference performance metrics used were items/second.","hasPerformance",[],{"_85":239,"_114":240,"_115":-5,"_123":35,"_122":35,"_124":238,"_220":90},{"_85":29,"_114":29,"_115":-5,"_123":35,"_122":35,"_124":238,"_220":90},{"_85":113,"_114":113,"_115":-5,"_123":35,"_122":35,"_124":237,"_220":90},[],[],"-","Latest (22.12.0)","nvidia/dle/bert_paddle","key","nvidia/dle/resources/bert_paddle","resourceId","bert_paddle","shortDescription","BERT is a method of pre-training language representations which obtains state-of-the-art results on a wide array of NLP tasks.","This resource is using open-source code maintained in github (see the quick-start-guide section) and available for download from [NGC](https://catalog.ngc.nvidia.com/orgs/nvidia/teams/dle/resources/bert-paddle)\n\n \nBidirectional Encoder Representations from Transformers (BERT) is a new method of pre-training language representations that obtains state-of-the-art results on a wide array of Natural Language Processing (NLP) tasks. This model is based on the [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://arxiv.org/abs/1810.04805) paper.\n \nThis repository contains scripts to interactively launch data download, training, benchmarking, and inference routines in a Docker container for pre-training and fine-tuning tasks such as question answering. The major differences between the original implementation of the paper and this version of BERT are as follows:\n \n- Scripts to download the Wikipedia dataset\n- Scripts to preprocess downloaded data into inputs and targets for pre-training in a modular fashion\n- [LAMB](https://arxiv.org/pdf/1904.00962.pdf) optimizer to support training with larger batches\n- Adam optimizer for fine-tuning tasks\n- Automatic mixed precision (AMP) training support\n \nOther publicly available implementations of BERT include:\n1. [NVIDIA PyTorch](https://github.com/NVIDIA/DeepLearningExamples/tree/master/PyTorch/LanguageModeling/BERT)\n2. [NVIDIA TensorFlow](https://github.com/NVIDIA/DeepLearningExamples/tree/master/TensorFlow/LanguageModeling/BERT)\n3. [Hugging Face](https://github.com/huggingface/pytorch-pretrained-BERT)\n4. [codertimo](https://github.com/codertimo/BERT-pytorch)\n5. [gluon-nlp](https://github.com/dmlc/gluon-nlp/tree/v0.10.x/scripts/bert)\n6. [Google's implementation](https://github.com/google-research/bert)\n \nThis model trains with mixed precision Tensor Cores on NVIDIA Ampere and provides a push-button solution to pre-training on a corpus of choice. As a result, researchers can get results 4x faster than training without Tensor Cores. This model is tested against each NGC monthly container release to ensure consistent accuracy and performance over time.\n \n### Model architecture\n \nThe BERT model uses the same architecture as the encoder of the Transformer. Input sequences are projected into an embedding space before being fed into the encoder structure. Additionally, positional and segment encodings are added to the embeddings to preserve positional information. The encoder structure is simply a stack of Transformer blocks, which consist of a multi-head attention layer followed by successive stages of feed-forward networks and layer normalization. The multi-head attention layer accomplishes self-attention on multiple input representations.\n \nAn illustration of the architecture taken from the [Transformer paper](https://arxiv.org/pdf/1706.03762.pdf) is shown below.\n \n ![BERT](https://raw.githubusercontent.com/NVIDIA/DeepLearningExamples/master/PaddlePaddle/LanguageModeling/BERT//images/model.png)\n\n \n### Default configuration\n \nThe architecture of the BERT model is almost identical to the Transformer model that was first introduced in the [Attention Is All You Need paper](https://arxiv.org/pdf/1706.03762.pdf). The main innovation of BERT lies in the pre-training step, where the model is trained on two unsupervised prediction tasks using a large text corpus. Training on these unsupervised tasks produces a generic language model, which can then be quickly fine-tuned to achieve state-of-the-art performance on language processing tasks such as question answering.\n \nThe BERT paper reports the results for two configurations of BERT, each corresponding to a unique model size. This implementation provides the same default configurations, which are described in the table below.\n \n| **Model** | **Hidden layers** | **Hidden unit size** | **Attention heads** | **Feedforward filter size** | **Max sequence length** | **Parameters** |\n|:---------:|:-----------------:|:--------------------:|:-------------------:|:---------------------------:|:-----------------------:|:--------------:|\n| BERTBASE | 12 encoder | 768 | 12 | 4 x 768 | 512 | 110M |\n| BERTLARGE | 24 encoder | 1024 | 16 | 4 x 1024 | 512 | 330M |\n \n### Feature support matrix\n \nThe following features are supported by this model. \n \n| **Feature** | **BERT** |\n|:-----------:|:--------:|\n| [Paddle AMP](https://www.paddlepaddle.org.cn/documentation/docs/en/guides/performance_improving/amp_en.html) | Yes |\n| [Paddle Fleet](https://www.paddlepaddle.org.cn/documentation/docs/en/api/paddle/distributed/fleet/Fleet_en.html#fleet) | Yes |\n| [LAMB](https://www.paddlepaddle.org.cn/documentation/docs/en/api/paddle/optimizer/Lamb_en.html) | Yes |\n| [LDDL](https://github.com/NVIDIA/LDDL) | Yes |\n| Multi-node | Yes |\n \n#### Features\n \n[Fleet](https://www.paddlepaddle.org.cn/documentation/docs/en/api/paddle/distributed/fleet/Fleet_en.html#fleet) is a unified API for distributed training of PaddlePaddle.\n \n[LAMB](https://arxiv.org/pdf/1904.00962.pdf) stands for Layerwise Adaptive Moments based optimizer, which is a large batch optimization technique that helps accelerate the training of deep neural networks using large minibatches. It allows using a global batch size of 65536 and 32768 on sequence lengths 128 and 512, respectively, compared to a batch size of 256 for [Adam](https://arxiv.org/pdf/1412.6980.pdf). The optimized implementation accumulates 1024 gradient batches in phase 1 and 4096 steps in phase 2 before updating weights once. This results in a 15% training speedup. On multi-node systems, LAMB allows scaling up to 1024 GPUs resulting in training speedups of up to 72x in comparison to Adam. Adam has limitations on the learning rate that can be used since it is applied globally on all parameters, whereas LAMB follows a layerwise learning rate strategy.\n\n[LDDL](https://github.com/NVIDIA/LDDL) is a library that enables scalable data preprocessing and loading. LDDL is used by this PaddlePaddle BERT example.\n\n### Mixed precision training\n\nMixed precision is the combined use of different numerical precisions in a computational method. [Mixed precision](https://arxiv.org/abs/1710.03740) training offers significant computational speedup by performing operations in half-precision format while storing minimal information in single-precision to retain as much information as possible in critical parts of the network. Since the introduction of [Tensor Cores](https://developer.nvidia.com/tensor-cores) in NVIDIA Volta, and following with both the NVIDIA Turing and NVIDIA Ampere architectures, significant training speedups are experienced by switching to mixed precision -- up to 3x overall speedup on the most arithmetically intense model architectures. Using mixed precision training requires two steps:\n1. Porting the model to use the FP16 data type where appropriate.\n2. Adding loss scaling to preserve small gradient values.\n\nThe ability to train deep learning networks with lower precision was introduced in the Pascal architecture and first supported in CUDA 8 in the NVIDIA Deep Learning SDK.\n\nFor information about:\n- How to train using mixed precision in PaddlePaddle, refer to the [Mixed Precision Training](https://arxiv.org/abs/1710.03740) paper and [Automatic Mixed Precision Training](https://www.paddlepaddle.org.cn/documentation/docs/en/guides/01_paddle2.0_introduction/basic_concept/amp_en.html#automatic-mixed-precision-training) documentation.\n- Techniques used for mixed precision training, refer to the [Mixed-Precision Training of Deep Neural Networks](https://devblogs.nvidia.com/mixed-precision-training-deep-neural-networks/) blog.\n\n#### Enabling mixed precision\n\nMixed precision is enabled in Paddle by using the Automatic Mixed Precision (AMP) while storing variables in single-precision format. Furthermore, to preserve small gradient magnitudes in backpropagation, a [loss scaling](https://docs.nvidia.com/deeplearning/sdk/mixed-precision-training/index.html#lossscaling) step must be included when applying gradients.\nIn PaddlePaddle, loss scaling can be easily applied by passing in arguments to [GradScaler()](https://www.paddlepaddle.org.cn/documentation/docs/en/api/paddle/amp/GradScaler_en.html). The scaling value to be used can be dynamic or fixed.\n\nFor an in-depth walkthrough on AMP, check out sample usage [here](https://www.paddlepaddle.org.cn/documentation/docs/en/guides/01_paddle2.0_introduction/basic_concept/amp_en.html). Paddle AMP is a PaddlePaddle built-in module that provides functions to construct AMP workflow. The details can be found in [Automatic Mixed Precision (AMP)](https://www.paddlepaddle.org.cn/documentation/docs/en/guides/01_paddle2.0_introduction/basic_concept/amp_en.html#automatic-mixed-precision-training), which requires minimal network code changes to leverage Tensor Cores performance.\n\nCode example to enable mixed precision for static graph:\n- Use `paddle.static.amp.decorate` to wrap optimizer\n ```python\n import paddle.static.amp as amp\n mp_optimizer = amp.decorate(optimizer=optimizer, init_loss_scaling=8.0)\n ```\n- Minimize `loss` , and get `scaled_loss`, which is useful when you need customized loss.\n ```python\n ops, param_grads = mp_optimizer.minimize(loss)\n scaled_loss = mp_optimizer.get_scaled_loss()\n ```\n- For distributed training, it is recommended to use Fleet to enable amp, which is a unified API for distributed training of PaddlePaddle. For more information, refer to [Fleet](https://www.paddlepaddle.org.cn/documentation/docs/en/api/paddle/distributed/fleet/Fleet_en.html#fleet)\n\n ```python\n import paddle.distributed.fleet as fleet\n strategy = fleet.DistributedStrategy()\n strategy.amp = True # by default this is false\n optimizer = fleet.distributed_optimizer(optimizer, strategy=dist_strategy)\n ```\n\n#### Enabling TF32\n\nTensorFloat-32 (TF32) is the new math mode in [NVIDIA A100](https://www.nvidia.com/en-us/data-center/a100/) GPUs for handling the matrix math, also called tensor operations. TF32 running on Tensor Cores in A100 GPUs can provide up to 10x speedups compared to single-precision floating-point math (FP32) on NVIDIA Volta GPUs.\n\nTF32 Tensor Cores can speed up networks using FP32, typically with no loss of accuracy. It is more robust than FP16 for models, which require a high dynamic range for weights or activations.\n\nFor more information, refer to the [TensorFloat-32 in the A100 GPU Accelerates AI Training, HPC up to 20x](https://blogs.nvidia.com/blog/2020/05/14/tensorfloat-32-precision-format/) blog post.\n\nTF32 is supported in the NVIDIA Ampere GPU architecture and is enabled by default.\n\n### Glossary\n \n**Fine-tuning** \nTraining an already pre-trained model further using a task-specific dataset for subject-specific refinements by adding task-specific layers on top if required.\n \n**Language Model** \nAssigns a probability distribution over a sequence of words. Given a sequence of words, it assigns a probability to the whole sequence.\n \n**Pre-training** \nTraining a model on vast amounts of data on the same (or different) task to build general understandings.\n \n**Transformer** \nThe paper [Attention Is All You Need](https://arxiv.org/abs/1706.03762) introduces a novel architecture called Transformer that uses an attention mechanism and transforms one sequence into another.\n \n**Phase 1** \nPre-training on samples of sequence length 128 and 20 masked predictions per sequence.\n \n**Phase 2** \nPre-training on samples of sequence length 512 and 80 masked predictions per sequence.","displayName","BERT for PaddlePaddle",["D",1680632378086],"publisher","NVIDIA Deep Learning Examples","latestVersion","collectionCounts","imageUrl","/api/image/nvidia/dle/resources/bert_paddle","rawImageUrl","https://assets.nvidiagrid.net/ngc/logos/Deep-Learning-Examples.png","canGuestDownload","latestVersionSizeInBytes","licenseTerms","isMultinodeEnabled","isMultiArchitecture","bias","explainability","privacy","safetyAndSecurity","featureLabels",[],"displayLabels",[280],"policyLabels",[],"systemLabels",[],"products",[],"hasNimMetadata",{"_249":281,"_282":283,"_284":285,"_286":87},"Natural Language Understanding","value","uscs_natural_language_understanding","filter",{"_287":288,"_282":283},"category","field","application",{"_156":1268,"_158":1269},"ngcApps",[1221],"csaApiBaseUrl","https://www.nvidia.com/ai-agents/csa-agent/dcsm","datadogRum",{"_1202":1203,"_1204":1205,"_1206":1207,"_1208":1209,"_1210":300,"_109":326,"_1211":1212,"_1213":60,"_1214":90,"_1215":90,"_1216":1217,"_1218":90,"_1219":1220},"amplitude",{"_1199":1200,"_840":-7,"_1201":35},"disableJs","environment","prod","featureFlagsBlob","JTdCJTIydGVzdEZsYWclMjIlM0FmYWxzZSUyQyUyMnNob3dTZWFyY2hMYWJlbFNldEZpbHRlcnMlMjIlM0FmYWxzZSUyQyUyMmVuYWJsZUNzYVdpZGdldCUyMiUzQWZhbHNlJTJDJTIyZW5hYmxlUHJvZHVjdEdyb3VwaW5nQnlTdWJzY3JpcHRpb24lMjIlM0FmYWxzZSUyQyUyMmNvbnRhaW5lclNjYW5uaW5nUmVkZXNpZ25NMiUyMiUzQWZhbHNlJTdE","featureFlagsGuarded","isDevLikeEnv","isPlaywright","isMobileDevice","categorizedOrgs",{"_1193":1194,"_1195":1196,"_1197":1198},"currentOrgContext","currentContextRoles",[],"enterpriseOrgs",[],"manageableExternalUserGroups",[],"ngcApiOrigin","ngcHelmOrigin","origin","https://catalog.ngc.nvidia.com","usesMocks","theme","year",2026,"now",1787604173118,"1.12.0","defaultSearchSuggestions",{"_1080":407,"_1081":1082,"_1083":1084,"_1085":1086},"announcements",[],"consentState",{"_364":1070,"_1071":90,"_1072":35,"_1073":35,"_1074":1075},"consentConfig",{"_346":347,"_348":349,"_350":351,"_352":353,"_354":355,"_356":357,"_358":359,"_360":361,"_362":363,"_364":365},"showContactAdmin",1787604173119,[],[],[340,341,342,343,344,345],{"_49":63,"_51":73,"_53":151,"_55":-5,"_57":-5,"_59":60},{"_49":63,"_51":70,"_53":151,"_55":-5,"_57":-5,"_59":60},{"_49":63,"_51":68,"_53":151,"_55":-5,"_57":-5,"_59":60},{"_49":63,"_51":64,"_53":151,"_55":-5,"_57":-5,"_59":60},{"_49":61,"_51":62,"_53":151,"_55":-5,"_57":-5,"_59":60},{"_49":50,"_51":52,"_53":151,"_55":-5,"_57":-5,"_59":60},"mainInfoText","We and our third-party partners (including social media, advertising, and analytics partners) use cookies and other tracking technologies to collect, store, monitor, and process certain information about you when you visit our website. The information collected might relate to you, your preferences, or your device. We use that information to make the site work, analyze performance and traffic on our website, provide a more personalized web experience, and assist in our marketing efforts.

Under certain privacy laws, you have the right to direct us not to \"sell\" or \"share\" your personal information for targeted advertising. To opt-out of the \"sale\" and \"sharing\" of personal information through cookies, you must opt-out of optional cookies using the toggles below. To opt out of the \"sale\" and \"sharing\" of data collected by other means (e.g., online forms) you must also update your data sharing preferences through the NVIDIA Preference Center.

Click on the different category headings below to find out more and change the settings according to your preference. You cannot opt out of Required Cookies as they are deployed to ensure the proper functioning of our website (such as prompting the cookie banner and remembering your settings, etc.). By clicking \"Save and Accept\" or \"Decline All\" at the bottom, you consent to the use of cookies and other tools as described in our Cookie Policy in accordance with your settings and accept our Terms of Service (which contains important waivers). For more information about our privacy practices, please see our Privacy Policy.","mainText","Cookie Settings","alertNoticeText",{"_1062":1063,"_1064":1065,"_1066":1067,"_1068":1069},"bannerAcceptLabel","Accept All","bannerRejectLabel","Reject Optional","bannerSettingsLabel","Manage Settings","preferenceSaveLabel","Save Choices","preferenceRejectLabel","Decline All","alwaysActiveText","Always Active","groups",[366,367,368,369,370],{"_371":752,"_114":753,"_214":754,"_375":90,"_376":18,"_378":755,"_380":756},{"_371":655,"_114":656,"_214":657,"_375":35,"_376":58,"_378":658,"_380":659},{"_371":562,"_114":563,"_214":564,"_375":35,"_376":565,"_378":566,"_380":567},{"_371":382,"_114":383,"_214":384,"_375":35,"_376":385,"_378":386,"_380":387},{"_371":372,"_114":373,"_214":374,"_375":35,"_376":377,"_378":379,"_380":381},"optanonGroupId","C0005","Social Media Cookies","These cookies are set by a range of social media services that we have added to the site to enable you to share our content with your friends and networks. They are capable of tracking your browser across other sites and building up a profile of your interests. This may impact the content and messages you see on other websites you visit. If you do not allow these cookies you may not be able to use or see these sharing tools.","alwaysActive","order",11,"cookies",[],"hosts",[],"C0004","Advertising Cookies","These cookies record your visit to our websites, the pages you have visited and the links you have followed to influence the advertisements that you see on other websites. These cookies and the information they collect may be managed by other companies, including our advertising partners, and may be used to build a profile of your interests and show you relevant advertising on other sites. We and our advertising partners will use this information to make our websites and the advertising displayed on it, more relevant to your interests.",5,[504,505,506,507,508,509,510,511,512,513,514,515,516,517,518,519,520,521,522,523,524,525],[388,389,390,391,392,393,394,395,396,397,398,399,400,401,402,403,404],{"_405":499,"_249":499,"_214":407,"_378":500},{"_405":495,"_249":495,"_214":407,"_378":496},{"_405":490,"_249":490,"_214":407,"_378":491},{"_405":486,"_249":486,"_214":407,"_378":487},{"_405":481,"_249":481,"_214":407,"_378":482},{"_405":476,"_249":476,"_214":407,"_378":477},{"_405":468,"_249":468,"_214":407,"_378":469},{"_405":460,"_249":460,"_214":407,"_378":461},{"_405":455,"_249":455,"_214":407,"_378":456},{"_405":450,"_249":450,"_214":407,"_378":451},{"_405":445,"_249":445,"_214":407,"_378":446},{"_405":440,"_249":440,"_214":407,"_378":441},{"_405":435,"_249":435,"_214":407,"_378":436},{"_405":428,"_249":428,"_214":407,"_378":429},{"_405":423,"_249":423,"_214":407,"_378":424},{"_405":416,"_249":416,"_214":407,"_378":417},{"_405":406,"_249":406,"_214":407,"_378":408},"hostName","amazon-adsystem.com","",[409,410],{"_114":415,"_412":406,"_413":35,"_214":414},{"_114":411,"_412":406,"_413":35,"_214":414},"ad-privacy","host","isSession","This domain is owned by online retailer Amazon and is used as part of its affiliate marketing programme.","ad-id","doubleclick.net",[418,419],{"_114":422,"_412":416,"_413":35,"_214":421},{"_114":420,"_412":416,"_413":35,"_214":421},"IDE","This domain is owned by Doubleclick (Google). The main business activity is: Doubleclick is Googles real time bidding advertising exchange","test_cookie","bing.com",[425],{"_114":426,"_412":423,"_413":35,"_214":427},"MUID","This domain is owned by Mircosoft - it is the site for the search engine Bing.","assets.adobedtm.com",[430,431],{"_114":434,"_412":428,"_413":35,"_214":433},{"_114":432,"_412":428,"_413":35,"_214":433},"kndctr_F207D74D549850760A4C98C6_AdobeOrg_*","If you know anything about this host and its uses please add your contributions here. ","kndctr_F207D74D549850760A4C98C6_AdobeOrg_consent","tiktok.com",[437],{"_114":438,"_412":435,"_413":35,"_214":439},"_ttp","TikTok is a social media platform owned by ByteDance, known for short-form video sharing. The service is used to create and share videos, and it utilizes cookies for various purposes including advertising and analytics.","t.co",[442],{"_114":443,"_412":440,"_413":35,"_214":444},"muc_ads","This domain is owned by Twitter. It is used as a URL shortening service, which allows Twitter to collect data on how links are shared.","snowplow.com",[447],{"_114":448,"_412":445,"_413":35,"_214":449},"sp","Testing","google.com",[452],{"_114":453,"_412":450,"_413":35,"_214":454},"NID","This domain is owned by Google Inc. Although Google is primarily known as a search engine, the company provides a diverse range of products and services. Its main source of revenue however is advertising. Google tracks users extensively both through its own products and sites, and the numerous technologies embedded into many millions of websites around the world. It uses the data gathered from most of these services to profile the interests of web users and sell advertising space to organisations based on such interest profiles as well as aligning adverts to the content on the pages where its customer's adverts appear.","linkedin.com",[457],{"_114":458,"_412":455,"_413":35,"_214":459},"bcookie","This domain is owned by LinkedIn, the business networking platform. It typically acts as a third party host where website owners have placed one of its content sharing buttons in their pages, although its content and services can be embedded in other ways. Although such buttons add functionality to the website they are on, cookies are set regardless of whether or not the visitor has an active Linkedin profile, or agreed to their terms and conditions. For this reason it is classified as a primarily tracking/targeting domain.","twitter.com",[462,463],{"_114":466,"_412":460,"_413":35,"_214":467},{"_114":464,"_412":460,"_413":35,"_214":465},"personalization_id","This domain is owned by Twitter. The main business activity is: Social Networking Services. Where twitter acts as a third party host, it collects data through a range of plug-ins and integrations, that is primarily used for tracking and targeting.","guest_*","This domain is owned by X (formerly Twitter). The main business activity is: Social Networking Services. Where X (formerly Twitter) acts as a third party host, it collects data through a range of plug-ins and integrations, that is primarily used for tracking and targeting.","youtube.com",[470,471],{"_114":474,"_412":468,"_413":35,"_214":475},{"_114":472,"_412":468,"_413":35,"_214":473},"VISITOR_INFO1_LIVE","This cookie is used as a unique identifier to track viewing of videos","__Secure-*","YouTube is a Google owned platform for hosting and sharing videos. YouTube collects user data through videos embedded in websites, which is aggregated with profile data from other Google services in order to display targeted advertising to web visitors across a broad range of their own and other websites.","demdex.net",[478],{"_114":479,"_412":476,"_413":35,"_214":480},"demdex","This cookie helps Adobe Audience Manger perform basic functions such as visitor identification, ID synchronization, segmentation, modeling, reporting, etc.","im-apps.net",[483],{"_114":484,"_412":481,"_413":35,"_214":485},"receive-cookie-deprecation","This domain is owned by im-apps.net which provides information and resources about various instant messaging (IM) applications to help users make informed decisions based on their communication needs.","influ2.com",[488],{"_114":489,"_412":486,"_413":35,"_214":449},"R","6sc.co",[492],{"_114":493,"_412":490,"_413":35,"_214":494},"6suuid","This domain is associated with the company 6Sense Insights, Inc., which provides predictive analytics and account engagement services. It is primarily used for tracking and targeting in marketing automation and advertising.","hushly.com",[497],{"_114":498,"_412":495,"_413":35,"_214":449},"_hly_gvid","t.contentsquare.net",[501],{"_114":502,"_412":499,"_413":90,"_214":503},"_cs_*","This is a session cookie. This is a pattern type cookie where a random 13-digit number is appended to the prefix _cs_.",{"_114":560,"_412":527,"_413":35,"_214":561},{"_114":558,"_412":527,"_413":35,"_214":559},{"_114":556,"_412":527,"_413":35,"_214":557},{"_114":438,"_412":527,"_413":35,"_214":555},{"_114":532,"_412":527,"_413":90,"_214":534},{"_114":554,"_412":527,"_413":35,"_214":407},{"_114":553,"_412":530,"_413":35,"_214":407},{"_114":551,"_412":527,"_413":35,"_214":552},{"_114":549,"_412":527,"_413":35,"_214":550},{"_114":548,"_412":527,"_413":35,"_214":407},{"_114":546,"_412":527,"_413":35,"_214":547},{"_114":532,"_412":545,"_413":90,"_214":534},{"_114":543,"_412":530,"_413":35,"_214":544},{"_114":543,"_412":533,"_413":35,"_214":544},{"_114":502,"_412":527,"_413":35,"_214":503},{"_114":541,"_412":527,"_413":35,"_214":542},{"_114":448,"_412":539,"_413":35,"_214":540},{"_114":537,"_412":527,"_413":35,"_214":538},{"_114":535,"_412":527,"_413":35,"_214":536},{"_114":532,"_412":533,"_413":90,"_214":534},{"_114":529,"_412":530,"_413":35,"_214":531},{"_114":526,"_412":527,"_413":90,"_214":528},"AMCVS_xxxxxAdobeOrg","nvidia.com","This is a pattern type cookie name associated with Adobe Marketing Cloud. It stores a unique visitor identifier, and uses an organisation identifier to allow a company to track users across their domains and services.This is a pattern type cookie name associated with Adobe Marketing Cloud. It stores a unique visitor identifier, and uses an organisation identifier.","_rdt_pn","www.nvidia.com","This cookie name is associated with www.mind.org.uk which is used for marketing purposes.","AMP_*","build.nvidia.com","Supports Amplitude product analytics.","_gcl_au","Used by Google AdSense for experimenting with advertisement efficiency across websites using their services","_mkto_trk","This cookie is associated with an email marketing service provided by Marketo. This tracking cookie allows a website to link visitor behaviour to the recipient of an email marketing campaign, to measure campaign effectiveness.","investor.nvidia.com","This cookie is associated with Eventbrite and is used to deliver content tailored to the end user's interests and improve content creation. This cookie is also used for event-booking purposes.","_gat_gtag_UA_140825383_2","Google Analytics","_rdt_uuid","This cookie is set by Reddit and is used for tracking the views of embedded videos and performance of the advertisements.","catalog.ngc.nvidia.com","_uetsid","This cookie is used by Bing to determine what ads should be shown that may be relevant to the end user perusing the site. ","kndctr_F207D74D549850760A4C98C6_AdobeOrg_bc_session_id","ttcsid_XXXXXXXXXXXXXXXXXXXXCategory","This cookie name is associated with TikTok which is used to track your activity on the platform, including how you interact with their embedded services on other websites.This cookie is associated with TikTok. It helps track your activity on the platform, including how you interact with their embedded services on other websites.","_fbp","Used by Facebook to deliver a series of advertisement products such as real time bidding from third party advertisers","_zitok","_twpid","This cookie is typically associated with the use of Taptap Digital, a company specializing in digital advertising. It is used for tracking and ad personalization purposes.","_tt_enable_cookie","This cookie name is associated with TikTok which is used to collect data about behavior and purchases on the website and to measure the effect of advertising. It is also used to track website activity to help optimize advertising efforts.","AMCV_xxxxxAdobeOrg","This cookie is used to identify a unique visitor for tracking purposes by Adobe Experience Cloud.This is a pattern type cookie name associated with Adobe Marketing Cloud. It stores a unique visitor identifier, and uses an organisation identifier to allow a company to track users across their domains and services.","ttcsid","This cookie name is associated with TikTok which is used to enable TikTok Ads session tracking by assigning a unique session ID to users who click on TikTok ads, allowing for the attribution of user behavior and conversions within the same session.This cookie is associated with TikTok. It helps track your activity on the platform, including how you interact with their embedded services on other websites.","C0003","Personalization Cookies","These cookies collect data about how you have interacted with our website to help us improve your web experience, such as which pages you have visited. These cookies may store a unique ID so that our system will remember you when you return. They may be set by us or by third party providers whose services we have added to our pages. These cookies enable us to provide enhanced website functionality and personalization as well as make the marketing messages we send to you more relevant to your interests. If you do not allow these cookies, then some or all of these services may not function properly.",4,[597,598,599,600,601,602,603,604,605,606,607,608,609,610,611,612,613,614,615,616,617,618,619,620,621],[568,569,570,571,572],{"_405":481,"_249":481,"_214":407,"_378":592},{"_405":455,"_249":455,"_214":407,"_378":587},{"_405":582,"_249":582,"_214":407,"_378":583},{"_405":577,"_249":577,"_214":407,"_378":578},{"_405":573,"_249":573,"_214":407,"_378":574},"nvidia.partners",[575],{"_114":576,"_412":573,"_413":35,"_214":449},"edge","app.hushly.com",[579],{"_114":580,"_412":577,"_413":35,"_214":581},"*CORS","This domain is associated with Hushly, a platform that provides lead conversion solutions for content marketing by using personalization and content engagement strategies.","github.com",[584],{"_114":585,"_412":582,"_413":35,"_214":586},"_octo","This domain is owned by Github. The main business activity is: Source code repository and collaboration platform",[588,589],{"_114":591,"_412":455,"_413":35,"_214":459},{"_114":590,"_412":455,"_413":35,"_214":459},"lidc","li_gc",[593,594],{"_114":596,"_412":481,"_413":35,"_214":485},{"_114":595,"_412":481,"_413":35,"_214":485},"imid_*","imid",{"_114":622,"_412":530,"_413":35,"_214":624},{"_114":653,"_412":527,"_413":35,"_214":654},{"_114":622,"_412":652,"_413":35,"_214":624},{"_114":650,"_412":527,"_413":90,"_214":651},{"_114":648,"_412":527,"_413":35,"_214":649},{"_114":647,"_412":630,"_413":90,"_214":407},{"_114":622,"_412":626,"_413":35,"_214":624},{"_114":645,"_412":527,"_413":35,"_214":646},{"_114":622,"_412":533,"_413":35,"_214":624},{"_114":633,"_412":530,"_413":35,"_214":407},{"_114":622,"_412":644,"_413":35,"_214":624},{"_114":642,"_412":527,"_413":35,"_214":643},{"_114":639,"_412":640,"_413":35,"_214":641},{"_114":637,"_412":638,"_413":35,"_214":407},{"_114":622,"_412":634,"_413":35,"_214":624},{"_114":636,"_412":527,"_413":35,"_214":407},{"_114":622,"_412":635,"_413":35,"_214":624},{"_114":633,"_412":634,"_413":35,"_214":407},{"_114":632,"_412":630,"_413":90,"_214":407},{"_114":622,"_412":631,"_413":35,"_214":624},{"_114":622,"_412":630,"_413":35,"_214":624},{"_114":622,"_412":629,"_413":35,"_214":624},{"_114":628,"_412":527,"_413":90,"_214":624},{"_114":625,"_412":626,"_413":90,"_214":627},{"_114":622,"_412":623,"_413":35,"_214":624},"_gd_*","info.nvidia.com","This cookie contains no identifying information and is used for diagnostic purposes by Impact Radius, a service we use to track signup commissions for our affiliate program. We have categorized it under Functionality because it is necessary for the operation of our affiliate program.","nv_country_code","developer.nvidia.com","This cookie is likely used to store the country code of the visitor to provide localized content and possibly for legal compliance with regional data privacy laws.","_gdxxxxxxxxxxxxx","nvidianews.nvidia.com","docs.api.nvidia.com","research.nvidia.com","readme_library","wcmmode","ai-podcast.nvidia.com","resources.nvidia.com","geo_locator","openedx-language-preference","learn.nvidia.com","doxygen_width","raytracing-docs.nvidia.com","This cookie is likely used to remember the width preferences of a sidebar or interface element within a documentation platform like Doxygen.","_sfid_eb78","This cookie is categorized as Functionality Cookie.","crowdsource.nvidia.com","mbox","Cookie set by Adobe SiteCatalyst analytics software - used for measuring the performance of page content using A/B split testing","readme_language","vid","This cookie name 'vid' is used for tracking the search request made by the user on the 'Find a practitioner' page.","at_check","Used by Adobe Target to check if cookies are enabled/supported on the browser","blogs.nvidia.com","nvHeadless","This cookie is likely used to track if a user is using the site in a 'headless' non-GUI mode, which could be for accessibility reasons or for automated processes.","C0002","Performance Cookies","These cookies are used to provide quantitative measures of our website visitors, such as the number of times you visit, time on page, your mouse movements, scrolling, clicks and keystroke activity on the websites; other browsing, search, or product research behavior; and what brought you to our site. These cookies may store a unique ID so that our system will remember you when you return. Information collected with these cookies is used to measure and find ways to improve website performance.",[680,681,682,683,684,685,686,687,688,689,690,691,692,693,694,695,696,697,698,699,700,701,702,703,704,705,706,707,708],[660,661,662,663],{"_405":675,"_249":675,"_214":407,"_378":676},{"_405":670,"_249":670,"_214":407,"_378":671},{"_405":468,"_249":468,"_214":407,"_378":667},{"_405":582,"_249":582,"_214":407,"_378":664},[665],{"_114":666,"_412":582,"_413":90,"_214":586},"_gh_sess",[668],{"_114":669,"_412":468,"_413":90,"_214":475},"YSC","bizible.com",[672],{"_114":673,"_412":670,"_413":35,"_214":674},"_BUID","Bizible is a marketing attribution software that tracks the performance of marketing efforts and connects them to revenue, primarily used for B2B environments.","a.usbrowserspeed.com",[677],{"_114":678,"_412":675,"_413":35,"_214":679},"tuid","This domain is owned by Ookla which provides internet speed testing services through its Speedtest application.",{"_114":750,"_412":527,"_413":35,"_214":751},{"_114":735,"_412":749,"_413":90,"_214":736},{"_114":718,"_412":530,"_413":35,"_214":643},{"_114":718,"_412":631,"_413":35,"_214":643},{"_114":747,"_412":527,"_413":35,"_214":748},{"_114":735,"_412":533,"_413":90,"_214":736},{"_114":745,"_412":527,"_413":35,"_214":746},{"_114":735,"_412":530,"_413":90,"_214":736},{"_114":743,"_412":630,"_413":35,"_214":744},{"_114":735,"_412":652,"_413":90,"_214":736},{"_114":715,"_412":530,"_413":35,"_214":717},{"_114":741,"_412":527,"_413":90,"_214":742},{"_114":739,"_412":527,"_413":35,"_214":740},{"_114":737,"_412":527,"_413":90,"_214":738},{"_114":735,"_412":626,"_413":90,"_214":736},{"_114":733,"_412":527,"_413":35,"_214":734},{"_114":731,"_412":527,"_413":35,"_214":732},{"_114":729,"_412":527,"_413":90,"_214":730},{"_114":727,"_412":635,"_413":35,"_214":728},{"_114":715,"_412":626,"_413":35,"_214":717},{"_114":725,"_412":527,"_413":35,"_214":726},{"_114":723,"_412":527,"_413":90,"_214":724},{"_114":721,"_412":527,"_413":35,"_214":722},{"_114":719,"_412":527,"_413":35,"_214":720},{"_114":718,"_412":527,"_413":35,"_214":643},{"_114":715,"_412":716,"_413":35,"_214":717},{"_114":713,"_412":527,"_413":35,"_214":714},{"_114":711,"_412":527,"_413":90,"_214":712},{"_114":709,"_412":527,"_413":35,"_214":710},"_ga_xxxxxxxxxx","Used by Google Analytics to identify and track an individual session with your device.","s_ppvl","This cookie is set by Adobe Analytics to store information on the percentage of the page displayed to the end user.","coveo_visitorId","This cookie name is associated with Coveo which is used to store a unique user ID for marketing purposes and to track the number of distinct users who submitted a query or clicked an item, counting them only once even if they did the same query or clicked the same item over the course of many visits.","_dd_s","marketplace.nvidia.com","Datadog Cookie used to group all events generated from a unique user session across multiple pages. It contains the current session ID, whether the session is excluded due to sampling, and the expiration date of the session. The cookie is extended for an extra 15 minutes every time the user interacts with the website, up to the maximum user session duration (4 hours).","nvweb_A","_ga","This cookie name is associated with Google Universal Analytics - which is a significant update to Google's more commonly used analytics service. This cookie is used to distinguish unique users by assigning a randomly generated number as a client identifier. It is included in each page request in a site and used to calculate visitor, session and campaign data for the sites analytics reports. By default it is set to expire after 2 years, although this is customisable by website owners.","_uetvid","This is a cookie utilised by Microsoft Bing Ads and is a tracking cookie. It allows us to engage with a user that has previously visited our website.","tp","This cookie is associated with Adobe Analytics site tracking. It registers if the website's scroll-depth detection is active. This function remembers how far the user has scrolled on the current session across the website's sub-pages.","_hly_vid","This cookie _hly_vid is used to record website and server performance with each visit to the site. The data is aggregated and analyzed; therefore, it remains anonymous upon use.","_session_id","This cookie creates an interim session ID used as an in-session user ID.\n","s_cc","This cookie is associated with the Adobe Site Catalyst. It determines whether cookies are enabled in the web browser.","s_ecid","Adobe Analytics cookie containing the the Adobe Experience Cloud 1st party visitor id if the Analytics AMCV visitor id is not found.","_hjSessionUser_xxxxxx","Hotjar cookie that is set when a user first lands on a page with the Hotjar script. It is used to persist the Hotjar User ID, unique to that site on the browser. This ensures that behavior in subsequent visits to the same site will be attributed to the same user ID.","_hjHasCachedUserAttributes","This cookie name is associated with Hotjar which is used to know whether the data set in _hjUserAttributes Local Storage item is up to date or not.","s_ppv","This cookie is associated with Adobe Analytics site tracking and is used to measure a user's scroll activity to see how much of a page they view before moving to another page.","_gid","This cookie name is associated with Google Universal Analytics. This appears to be a new cookie and as of Spring 2017 no information is available from Google. It appears to store and update a unique value for each page visited.","_hjTLDTest","When the Hotjar script executes we try to determine the most generic cookie path we should use, instead of the page hostname. This is done so that cookies can be shared across subdomains (where applicable). To determine this, we try to store the _hjTLDTest cookie for different URL substring alternatives until it fails. After this check, the cookie is removed.","ekfls-2025-03-27","This cookie name is associated with readme.io which is used to help maintain data for the API reference support page.","s_getNewRepeat","This cookie is used to determine whether a visitor is a new visitor or a repeat visitor to the site, and is used for analytical purposes.","_hly_sid","This cookie is likely associated with website analytics, as the prefix '_hly_' suggests a connection to Holistics, a business intelligence and analytics platform.","docs.nvidia.com","_hjSession_xxxxxx","A cookie that holds the current session data. This ensues that subsequent requests within the session window will be attributed to the same Hotjar session.","C0001","Required Cookies","These cookies enable core functionality such as security, network management, and accessibility. These cookies are required for the site to function and cannot be turned off.",[897,898,899,900,901,902,903,904,905,906,907,908,909,910,911,912,913,914,915,916,917,918,919,920,921,922,923,924,925,926,927,928,929,930,931,932,933,934,935,936,937,938,939,940,941,942,943,944,945,946,947,948,949,950,951,952,953,954,955,956,957,958,959,960,961,962,963,964,965,966,967,968,969],[757,758,759,760,761,762,763,764,765,766,767,768,769,770,771,772,773,774,775,776,777,778,779,780,781],{"_405":894,"_249":894,"_214":407,"_378":895},{"_405":888,"_249":888,"_214":407,"_378":889},{"_405":885,"_249":885,"_214":407,"_378":886},{"_405":873,"_249":873,"_214":407,"_378":874},{"_405":870,"_249":870,"_214":407,"_378":871},{"_405":867,"_249":867,"_214":407,"_378":868},{"_405":862,"_249":862,"_214":407,"_378":863},{"_405":468,"_249":468,"_214":407,"_378":859},{"_405":856,"_249":856,"_214":407,"_378":857},{"_405":460,"_249":460,"_214":407,"_378":854},{"_405":455,"_249":455,"_214":407,"_378":852},{"_405":847,"_249":847,"_214":407,"_378":848},{"_405":842,"_249":842,"_214":407,"_378":843},{"_405":837,"_249":837,"_214":407,"_378":838},{"_405":834,"_249":834,"_214":407,"_378":835},{"_405":582,"_249":582,"_214":407,"_378":831},{"_405":440,"_249":440,"_214":407,"_378":828},{"_405":577,"_249":577,"_214":407,"_378":821},{"_405":573,"_249":573,"_214":407,"_378":815},{"_405":809,"_249":809,"_214":407,"_378":810},{"_405":804,"_249":804,"_214":407,"_378":805},{"_405":799,"_249":799,"_214":407,"_378":800},{"_405":795,"_249":795,"_214":407,"_378":796},{"_405":790,"_249":790,"_214":407,"_378":791},{"_405":782,"_249":782,"_214":407,"_378":783},"polyfill.io",[784,785],{"_114":788,"_412":782,"_413":35,"_214":789},{"_114":786,"_412":782,"_413":35,"_214":787},"cf_ob_info","This cookie is associated with sites using CloudFlare. It is used to improve page load times and to override any security restrictions based on the visitors IP address. It does not contain any user identification information.","cf_use_ob","This cookie is associated with sites using CloudFlare. It is used to improve page load times and to override any security restrictions based on the visitors IP address. It does not contain any user identification information. ","www.recaptcha.net",[792],{"_114":793,"_412":790,"_413":35,"_214":794},"_GRECAPTCHA","This domain is used by Google reCAPTCHA, which provides spam protection services to websites by detecting if a user is a human or a bot.","api2.hcaptcha.com",[797],{"_114":798,"_412":795,"_413":35,"_214":449},"__cflb","instagram.com",[801],{"_114":802,"_412":799,"_413":35,"_214":803},"csrftoken","Instagram is a social networking service owned by Facebook, Inc. It's designed for sharing photos and videos from a smartphone. The service also uses cookies for various purposes, including analytics, personalization, and advertising.","readme.io",[806],{"_114":807,"_412":804,"_413":35,"_214":808},"__cf_bm","The __cf_bm cookie is a cookie necessary to support Cloudflare Bot Management, currently in private beta. As part of our bot management service, this cookie helps manage incoming traffic that matches criteria associated with bots.","nvidiacorp.us-5.evergage.com",[811,812],{"_114":814,"_412":809,"_413":35,"_214":449},{"_114":813,"_412":809,"_413":35,"_214":449},"AWSALBTG","AWSALBTGCORS",[816,817],{"_114":819,"_412":573,"_413":35,"_214":820},{"_114":818,"_412":573,"_413":35,"_214":449},"ak_bmsc","bm_mi","This domain is likely associated with NVIDIA's partner program, which could involve marketing, sales, and support services for partners.",[822,823,824],{"_114":827,"_412":577,"_413":35,"_214":449},{"_114":826,"_412":577,"_413":90,"_214":449},{"_114":825,"_412":577,"_413":35,"_214":449},"AWSALBCORS","JSESSIONID","AWSALB",[829],{"_114":807,"_412":440,"_413":35,"_214":830},"This is a CloudFoundry cookie",[832],{"_114":833,"_412":582,"_413":35,"_214":586},"logged_in","hcaptcha.com",[836],{"_114":807,"_412":834,"_413":35,"_214":830},"tracking.intentsify.io",[839],{"_114":840,"_412":837,"_413":35,"_214":841},"userId","Used to track usage of the embedded video player.","static.reo.dev",[844],{"_114":845,"_412":842,"_413":35,"_214":846},"cookietest","Common cookie name could have a number of different origins. Where this is first party and a session cookie, its most likely to do with checking to see if the browser is set to block or allow cookies.","www.datadoghq-browser-agent.com",[849],{"_114":850,"_412":847,"_413":35,"_214":851},"dd_cookie_test_","dd_cookie_test",[853],{"_114":807,"_412":455,"_413":35,"_214":830},[855],{"_114":807,"_412":460,"_413":35,"_214":830},"supabase.co",[858],{"_114":807,"_412":856,"_413":35,"_214":830},[860],{"_114":861,"_412":468,"_413":35,"_214":475},"VISITOR_PRIVACY_METADATA","dosa-prd.azurewebsites.net",[864],{"_114":865,"_412":862,"_413":35,"_214":866},"access_token","This cookie is used to obtain the permissions for users of the portal to allow for API access in order to retrieve information required to perform portal functions.","w.hcaptcha.com",[869],{"_114":807,"_412":867,"_413":35,"_214":830},"cdn.readme.io",[872],{"_114":807,"_412":870,"_413":35,"_214":830},"queue-it.net",[875,876,877,878,879],{"_114":884,"_412":873,"_413":35,"_214":407},{"_114":883,"_412":873,"_413":35,"_214":407},{"_114":882,"_412":873,"_413":35,"_214":407},{"_114":881,"_412":873,"_413":35,"_214":407},{"_114":880,"_412":873,"_413":35,"_214":407},"Queue-it-{queueId}","Queue-it-{dmonvidia }________{eventId}","Queue-it-token","queueitsoftblock_{queueId}","Queue-it-{nvidia}________{eventId}","x.com",[887],{"_114":807,"_412":885,"_413":35,"_214":808},"api.hcaptcha.com",[890,891],{"_114":893,"_412":888,"_413":35,"_214":449},{"_114":798,"_412":888,"_413":35,"_214":892},"This domain is associated with hCaptcha, a service that provides CAPTCHA and reCAPTCHA alternatives for user verification and bot protection.","hmt_id","vimeo.com",[896],{"_114":807,"_412":894,"_413":35,"_214":830},{"_114":1030,"_412":527,"_413":35,"_214":1061},{"_114":1059,"_412":623,"_413":90,"_214":1060},{"_114":1052,"_412":1021,"_413":90,"_214":1053},{"_114":1058,"_412":716,"_413":90,"_214":407},{"_114":1056,"_412":527,"_413":35,"_214":1057},{"_114":814,"_412":1042,"_413":35,"_214":981},{"_114":989,"_412":1051,"_413":90,"_214":991},{"_114":813,"_412":980,"_413":35,"_214":981},{"_114":992,"_412":1051,"_413":35,"_214":993},{"_114":1054,"_412":527,"_413":90,"_214":1055},{"_114":1052,"_412":1028,"_413":90,"_214":1053},{"_114":986,"_412":1051,"_413":35,"_214":988},{"_114":1049,"_412":527,"_413":35,"_214":1050},{"_114":813,"_412":972,"_413":35,"_214":981},{"_114":977,"_412":1048,"_413":35,"_214":979},{"_114":825,"_412":978,"_413":35,"_214":994},{"_114":827,"_412":978,"_413":35,"_214":1045},{"_114":992,"_412":987,"_413":35,"_214":993},{"_114":1046,"_412":626,"_413":35,"_214":1047},{"_114":827,"_412":652,"_413":35,"_214":1045},{"_114":1043,"_412":527,"_413":35,"_214":1044},{"_114":819,"_412":652,"_413":35,"_214":1038},{"_114":1015,"_412":631,"_413":90,"_214":407},{"_114":813,"_412":1042,"_413":35,"_214":981},{"_114":1040,"_412":980,"_413":90,"_214":1041},{"_114":1004,"_412":527,"_413":35,"_214":995},{"_114":977,"_412":527,"_413":35,"_214":979},{"_114":1039,"_412":539,"_413":35,"_214":407},{"_114":819,"_412":527,"_413":35,"_214":1038},{"_114":827,"_412":972,"_413":35,"_214":1037},{"_114":1035,"_412":635,"_413":90,"_214":1036},{"_114":1033,"_412":527,"_413":90,"_214":1034},{"_114":1032,"_412":527,"_413":35,"_214":407},{"_114":1030,"_412":978,"_413":35,"_214":1031},{"_114":1029,"_412":539,"_413":35,"_214":407},{"_114":1020,"_412":1028,"_413":90,"_214":1022},{"_114":1026,"_412":631,"_413":35,"_214":1027},{"_114":1024,"_412":716,"_413":35,"_214":1025},{"_114":1023,"_412":533,"_413":35,"_214":407},{"_114":1020,"_412":1021,"_413":90,"_214":1022},{"_114":1018,"_412":987,"_413":90,"_214":1019},{"_114":1016,"_412":527,"_413":35,"_214":1017},{"_114":1015,"_412":530,"_413":90,"_214":407},{"_114":986,"_412":990,"_413":35,"_214":988},{"_114":818,"_412":1005,"_413":35,"_214":995},{"_114":1013,"_412":530,"_413":35,"_214":1014},{"_114":1011,"_412":527,"_413":35,"_214":1012},{"_114":1009,"_412":527,"_413":35,"_214":1010},{"_114":1008,"_412":527,"_413":35,"_214":407},{"_114":850,"_412":530,"_413":35,"_214":1007},{"_114":1004,"_412":652,"_413":35,"_214":995},{"_114":807,"_412":623,"_413":35,"_214":1006},{"_114":1004,"_412":1005,"_413":35,"_214":995},{"_114":1003,"_412":539,"_413":35,"_214":407},{"_114":1002,"_412":527,"_413":35,"_214":975},{"_114":1000,"_412":987,"_413":35,"_214":1001},{"_114":813,"_412":998,"_413":35,"_214":999},{"_114":997,"_412":539,"_413":35,"_214":407},{"_114":996,"_412":527,"_413":35,"_214":407},{"_114":989,"_412":987,"_413":90,"_214":991},{"_114":818,"_412":652,"_413":35,"_214":995},{"_114":825,"_412":652,"_413":35,"_214":994},{"_114":992,"_412":990,"_413":35,"_214":993},{"_114":989,"_412":990,"_413":90,"_214":991},{"_114":986,"_412":987,"_413":35,"_214":988},{"_114":984,"_412":980,"_413":90,"_214":985},{"_114":982,"_412":638,"_413":35,"_214":983},{"_114":814,"_412":980,"_413":35,"_214":981},{"_114":977,"_412":978,"_413":35,"_214":979},{"_114":807,"_412":630,"_413":35,"_214":976},{"_114":974,"_412":527,"_413":35,"_214":975},{"_114":825,"_412":972,"_413":35,"_214":973},{"_114":970,"_412":527,"_413":35,"_214":971},"LoginHints","Keeps users logged in across NVIDIA domains.","api.ngc.nvidia.com","This cookie is managed by AWS and is used for load balancing.","bm_so","This cookie name is used to load the site as quickly as possible and to recognize bots using Akamai.","The __cf_bm cookie is a cookie necessary to support Cloudflare Bot Management, currently in private beta. As part of our bot management service, this cookie helps manage incoming traffic that matches criteria associated with bots.\nThis is a CloudFoundry cookie","OptanonConsent","nvid.nvidia.com","This cookie is set by the cookie compliance solution from OneTrust. It stores information about the categories of cookies the site uses and whether visitors have given or withdrawn consent for the use of each category. This enables site owners to prevent cookies in each category from being set in the users browser, when consent is not given. The cookie has a normal lifespan of one year, so that returning visitors to the site will have their preferences remembered. It contains no information that can identify the site visitor.","events.nvidia.com","This cookie is associated with the Amazon Web Services Elastic Load Balancing functionality for routing client request on the server.Used to attribute commission to affiliates when you arrive at the website from an affiliate referral link. It is set when you click on one of our links and used to let the advertiser and us know the website from which you came.","sessionid","This is a very generic cookie name that may have different purposes on different sites, but generally it will be some kind of anonymous session identifier.","_csrf","This cookie is used to prevent Cross-Site Request Forgery (CSRF) attacks, ensuring that each request made to the server is from the user's browser and is intentional.","LSKey-c$CookieConsentPolicy","enterprise-support.nvidia.com","Used to apply end-user cookie consent preferences set by our client-side utility.Custom- Short Duration. Unique ID of an Enquiry- used to validate access to case, and load exisiting data.\n","renderCtx","academicgrants.nvidia.com","Used to store site parameters in the session for reuse across requests by a single client for functionality and performance reasons.","CookieConsentPolicy","Used to apply end-user cookie consent preferences set by our client-side utility.","This cookie is managed by AWS and is used for load balancing.Used to attribute commission to affiliates when you arrive at the website from an affiliate referral link. It is set when you click on one of our links and used to let the advertiser and us know the website from which you came.","This cookie is associated with Akamai and is used to differentiate between traffic from humans and bots.","nv_co","__uxq412__ses.88c1","learn.learn.nvidia.com","This cookie is associated with the Amazon Web Services Elastic Load Balancing functionality for routing client request on the server.","pctrk","Used to count page views by unauthenticated users against license usage.","bm_s","bpazaws52gukakzc__ctrl0_ctl36_uccaptcha","bm_sv","store.nvidia.com","The __cf_bm cookie is a cookie necessary to support Cloudflare Bot Management, currently in private beta. As part of our bot management service, this cookie helps manage incoming traffic that matches criteria associated with bots.\n","This cookie name is used by DataDog to determine if cookies are supported on the website.dd_cookie_test","AMP_TLDTEST_xxxxxxxx","SESSlsd","This cookie is likely to be a session identifier used to maintain user session state across page requests.","mboxEdgeCluster","This cookie name is associated with Adobe Target which is used to communicate the correct edge cluster to at.js so that Target profiles can stay in sync as users navigate across a site.","_rdt_em","This cookie name is associated with Reddit which is used to manage session state and user interactions.","megaNavCollapsed","_abck","The _abck cookie is associated with Akamai's Bot Manager, which is used to prevent fraud and abuse by safeguarding online interactions and detecting irregularities that could suggest malicious activity, such as bot-driven attacks or hacking attempts.","idccsrf","This cookie name is associated with Salesforce which is used to track CrossSiteRequestForgery validation for certain single sign-on (SSO) flows.","ApplicationGatewayAffinity","partners.nvidia.com","This cookie is used to maintain a user session by the server in a load balanced environment, to ensure the user has a consistent experience.","AMP_TEST_xxxxxxxx","bm_lso","This cookie name is associated with Akamai which is used to load the site as quickly as possible and to recognize bots.","SSESS###########","This cookie is likely to be a session cookie used to maintain an anonymized user session by the server in a secure environment, often seen in sites built with the Drupal content management system.","nvdeveloper.nvidia.com","__uxq412__id.88c1","OptanonAlertBoxClosed","This cookie is set by websites using certain versions of the cookie law compliance solution from OneTrust. It is set after visitors have seen a cookie information notice and in some cases only when they actively close the notice down. It enables the website not to show the message more than once to a user. The cookie has a normal lifespan of one year and contains no personal information.","QueueITAccepted-SDFrts345E-V3_{eventId}","TEST_AMCV_COOKIE_WRITE","This cookie is used to test whether cookies are enabled in the browser. It is a temporary cookie used to determine if the browser accepts cookies.","XSRF-TOKEN","This cookie is written to help with site security in preventing Cross-Site Request Forgery attacks.This cookie helps with site security in preventing Cross-Site Request Forgery attacks.","AWS ELB application load balancer","This cookie is associated with Akamai. It is used to analyse traffic to determine if it is automated traffic generated by bots or a human user.","bpazaws52gukakzc__ctrl0_ctl78_list_ctl09_systemcaptcha","PHPSESSID","PHP session cookie associated with embedded content from this domain.\n","base.manager.learn.nvidia.com","bm_ss","This cookie is used to help protect websites from automated bot traffic.","AWS ELB application load balancerUsed to attribute commission to affiliates when you arrive at the website from an affiliate referral link. It is set when you click on one of our links and used to let the advertiser and us know the website from which you came.","_devzone_session","This cookie is likely used to maintain a user session on the website by the server. It is common for session identifiers to be used to keep users logged in and track their interaction with the site during a session.","docs.ngc.nvidia.com","bm_sz","This cookie is used by Akamai to optimize site performance and security. It is typically associated with the delivery of web content and services.","vcalliance.nvidia.com","ASP.NET_SessionId","General purpose platform session cookie, used by sites written with Microsoft .NET based technologies. Usually used to maintain an anonymised user session by the server.","com.adobe.alloy.getTld","This cookie name is associated with the Adobe WebSDK which is used to determine the top-most domain that can accept cookies.","SESSls","This cookie is likely to be a session identifier, used to maintain a user's session state across page requests.","c_code","BIGipServerx","This cookie name is associated with the BIG-IP product suite from company F5. Usually associated with managing sessions on load balanced servers, to ensure user requests are routed consistently to the correct server. The common root is BIGipServer most commonly followed by a domain name, usually the one that it is hosted on, but not always.","This cookie is set by websites using certain versions of the cookie law compliance solution from OneTrust. It is set after visitors have seen a cookie information notice and in some cases only when they actively close the notice down. It enables the website not to show the message more than once to a user. The cookie has a one year lifespan and contains no personal information.","default","

NVIDIA uses cookies to improve your experience on our web site. We and our third-party partners also use cookies and other tools to collect and record information you provide as well as information about your interactions with our websites for performance improvement, analytics, and to assist in marketing efforts. By clicking \"Accept All\", you consent to our use of cookies and other tools as described in our Cookie Policy. You can manage your cookie settings by clicking on \"Manage Settings.\" By continuing to use this site or by clicking one of the buttons below, you agree to our Terms of Service (which contains important waivers). Please see our Privacy Policy for more information on our privacy practices.

","gpc","

We have detected the Global Privacy Control (GPC) signal and recorded your rejection of all optional cookies on this site for this browser. You can manage your cookie settings by clicking on \"Manage Settings\". Please see our Cookie Policy for more information. To opt out of non-cookie personal information \"sales\" / \"sharing\" for targeted advertising purposes, please visit the NVIDIA Preference Center. Please see our Privacy Policy for more information on our privacy practices.

","gpcLoggedIn","

We have detected the Global Privacy Control Signal (GPC) and have opted you out of all optional cookies on this browser. You can manage your cookie settings by clicking on \"Manage Settings\". Please see our Cookie Policy for more information. We have also opted you out of \"sharing\"/\"sales\" of personal information outside of cookies. You can manage these settings in the NVIDIA NVIDIA Preference Center. Please see our Privacy Policy for more information.

","gpcLoggedInOverridden","

We have detected the Global Privacy Control Signal (GPC) and have opted you out of all optional cookies on this browser. You can manage your cookie settings by clicking on \"Manage Settings\". Please see our Cookie Policy for more information. We have also opted you out of \"sharing\"/\"sales\" of personal information outside of cookies which overrides at least one of your previous settings. You can manage them in the NVIDIA Preference Center. Please see our Privacy Policy for more information.

",{"_752":18,"_655":60,"_562":60,"_382":60,"_372":60},"hasConsented","gpcApplied","showBanner","geolocation",{"_1076":1077,"_1078":1079},"country","US","state","OR","query","categories",[1099,1100,1101,1102,1103],"optionUrls",[1087,1088,1089,1090,1091,1092,1093,1094,1095,1096,1097,1098],"total",12,"/orgs/nim/deepseek-ai/containers/deepseek-r1/-","/orgs/nim/nvidia/containers/llama-3.1-nemotron-70b-instruct/-","/orgs/nvidia/-/containers/pytorch/-","/orgs/nvidia/omniverse/collections/kit/-","/orgs/nvidia/-/collections/deepstream_sdk/-","/orgs/nvidia/omniverse/collections/kit-appstreaming-collection/-","/orgs/nvidia/research/models/stylegan3/-","/orgs/nvidia/tao/models/peoplenet/-","/orgs/nvidia/tao/models/trafficcamnet/-","/orgs/nvidia/riva/resources/riva_quickstart/-","/orgs/nvidia/-/helm-charts/gpu-operator/-","/orgs/nvidia/-/helm-charts/k8s-nim-operator/-",{"_21":1177,"_1105":1178,"_1107":60},{"_21":1142,"_1105":1143,"_1107":58},{"_21":1125,"_1105":1126,"_1107":1127},{"_21":22,"_1105":1118,"_1107":1119},{"_21":1104,"_1105":1106,"_1107":20},"helm-charts","items",[1108,1109],"startIndex",{"_21":1104,"_249":1115,"_208":209,"_210":-7,"_244":1116,"_252":1112,"_277":1117,"_1114":1097},{"_21":1104,"_249":1110,"_208":209,"_210":-7,"_244":1111,"_252":1112,"_277":1113,"_1114":1098},"NVIDIA NIM Operator","k8s-nim-operator","NVIDIA",[],"url","GPU Operator","gpu-operator",[],[1120],9,{"_21":22,"_249":1121,"_208":209,"_210":1122,"_244":1123,"_252":1112,"_277":1124,"_1114":1096},"Riva Skills Quick Start","riva","riva_quickstart",[],"models",[1128,1129,1130],6,{"_21":1125,"_249":1138,"_208":209,"_210":1139,"_244":1140,"_252":1112,"_277":1141,"_1114":1093},{"_21":1125,"_249":1135,"_208":209,"_210":1132,"_244":1136,"_252":1112,"_277":1137,"_1114":1094},{"_21":1125,"_249":1131,"_208":209,"_210":1132,"_244":1133,"_252":1112,"_277":1134,"_1114":1095},"TrafficCamNet","tao","trafficcamnet",[],"PeopleNet","peoplenet",[],"StyleGAN3 pretrained models","research","stylegan3",[],"collections",[1144,1145,1146],{"_21":1142,"_249":1174,"_208":209,"_210":1148,"_244":1175,"_252":1112,"_277":1176,"_1114":1090},{"_21":1142,"_249":1171,"_208":209,"_210":-7,"_244":1172,"_252":1112,"_277":1173,"_1114":1091},{"_21":1142,"_249":1147,"_208":209,"_210":1148,"_244":1149,"_252":1112,"_277":1150,"_1114":1092},"Omniverse Kit App Streaming","omniverse","kit-appstreaming-collection",[1151,1152,1153],{"_114":1165,"_249":1166,"_246":1167,"_256":1168,"_1158":1169},{"_114":1148,"_249":1161,"_246":1162,"_256":1157,"_1158":1163},{"_114":1154,"_249":1155,"_246":1156,"_256":1157,"_1158":1159},"omniverse-dev","NVIDIA Developer Program","Join the Developer Program for access to free tools, support, and tech resources.","https://store-images.s-microsoft.com/image/apps.33936.6f1bcf42-55d0-45d1-b4b6-f08c68a50bbe.e91ca707-db64-49d6-9761-4781590a9685.7ba64fc5-7c44-4864-a5e6-7bde05126935","supportedFeatures",[1160],"UNIFIED_CATALOG","NVIDIA Omniverse","NVIDIA Omniverse is a platform of APIs, services, and software development kits (SDKs) that enable developers to build generative AI-enabled tools and applications.",[1160,1164],"SHAREABLE_TO_USER_GROUPS","nv-ai-enterprise","NVIDIA AI Enterprise","Accelerate your AI agent development","https://assets.ngc.nvidia.com/products/catalog/images/product-NVAIE.jpg",[1160,1164,1170],"NGC_CONSOLE","DeepStream SDK","deepstream_sdk",[],"Omniverse Kit (FB)","kit",[1151,1152,1153],"containers",[1179,1180,1181],{"_21":1177,"_249":1189,"_208":1186,"_210":1190,"_244":1191,"_252":1112,"_277":1192,"_1114":1087},{"_21":1177,"_249":1185,"_208":1186,"_210":209,"_244":1187,"_252":1112,"_277":1188,"_1114":1088},{"_21":1177,"_249":1182,"_208":209,"_210":-7,"_244":1183,"_252":1112,"_277":1184,"_1114":1089},"PyTorch","pytorch",[],"Llama-3.1-Nemotron-70B-Instruct","nim","llama-3.1-nemotron-70b-instruct",[],"DeepSeek-R1","deepseek-ai","deepseek-r1",[],"ownerOrgs",[],"memberOrgs",[],"externalOrgs",[],"apiKey","2b44f86fea50e90ffd2db29ee2028419","consentGranted","applicationId","4c857e4f-5ffa-45fa-b75e-af773c22f0ad","clientToken","pubb6927db33dae60ba5031a186cde0242a","site","datadoghq.com","service","catalog-ui","env","sessionSampleRate",100,"sessionReplaySampleRate","compressIntakeRequests","storeContextsAcrossPages","defaultPrivacyLevel","mask-user-input","trackViewsManually","trackingConsent","not-granted",{"_1222":1223,"_1224":319,"_1225":35,"_1226":1227,"_1105":1228,"_1229":1230,"_1231":1223,"_1232":35,"_1233":35,"_1234":35,"_1235":1236},"appName","ngc-catalog","appOrigin","hideOnPrivateOrg","icon","ObjectsModules2",[1237,1238,1239,1240,1241,1242,1243],"menuIcon","editor-swatches","productName","requiresAllScopes","requiresLogin","requiresOrgLevelScopesOnly","title","Catalog",{"_1235":1265,"_51":1266,"_1229":1267},{"_1258":1259,"_1235":1260,"_51":1261,"_1231":1165},{"_1235":1256,"_51":1257,"_1229":1230},{"_1235":1253,"_51":1254,"_1229":1255},{"_1235":1250,"_51":1251,"_1229":1252},{"_1235":1247,"_51":1248,"_1229":1249},{"_1235":1244,"_51":1245,"_1229":1246},"Resources","/resources","misc-library","Models","/models","misc-neural-network","Helm Charts","/helm-charts","misc-ship-wheel","Containers","/containers","3d-cube","Collections","/collections","access",[1262],"Enterprise Catalog","/enterprise",[1263,1264],"NVAIE","READ","Explore Catalog","/","common-home",{"_161":162,"_163":164,"_165":166,"_167":168,"_169":170,"_171":172},{"_160":35}]

	Files	Compressed Size		Download
22.12.0Selected	1	62.75 KB	February 7, 2023 UTC
22.08.0	1	57.68 KB	November 22, 2022 UTC