### Installation Output Log

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/sharding/sharding.ipynb

Example output showing successful installation of TorchRec dependencies.

```text
Looking in indexes: https://download.pytorch.org/whl/cu121
Requirement already satisfied: torch in /usr/local/lib/python3.10/dist-packages (2.5.1+cu121)
Requirement already satisfied: sympy==1.13.1 in /usr/local/lib/python3.10/dist-packages (from torch) (1.13.1)
Requirement already satisfied: networkx in /usr/local/lib/python3.10/dist-packages (from torch) (3.3)
Requirement already satisfied: nvidia-cuda-cupti-cu12==12.1.105 in /usr/local/lib/python3.10/dist-packages (from torch) (12.1.105)
Requirement already satisfied: typing-extensions>=4.8.0 in /usr/local/lib/python3.10/dist-packages (from torch) (4.13.2)
Requirement already satisfied: fsspec in /usr/local/lib/python3.10/dist-packages (from torch) (2024.6.1)
Requirement already satisfied: jinja2 in /usr/local/lib/python3.10/dist-packages (from torch) (3.1.6)
Requirement already satisfied: nvidia-cuda-nvrtc-cu12==12.1.105 in /usr/local/lib/python3.10/dist-packages (from torch) (12.1.105)
Requirement already satisfied: nvidia-cuda-runtime-cu12==12.1.105 in /usr/local/lib/python3.10/dist-packages (from torch) (12.1.105)
Requirement already satisfied: nvidia-cublas-cu12==12.1.3.1 in /usr/local/lib/python3.10/dist-packages (from torch) (12.1.3.1)
Requirement already satisfied: nvidia-nvtx-cu12==12.1.105 in /usr/local/lib/python3.10/dist-packages (from torch) (12.1.105)
Requirement already satisfied: nvidia-cudnn-cu12==9.1.0.70 in /usr/local/lib/python3.10/dist-packages (from torch) (9.1.0.70)
Requirement already satisfied: triton==3.1.0 in /usr/local/lib/python3.10/dist-packages (from torch) (3.1.0)
Requirement already satisfied: nvidia-nccl-cu12==2.21.5 in /usr/local/lib/python3.10/dist-packages (from torch) (2.21.5)
Requirement already satisfied: nvidia-cufft-cu12==11.0.2.54 in /usr/local/lib/python3.10/dist-packages (from torch) (11.0.2.54)
Requirement already satisfied: nvidia-cusparse-cu12==12.1.0.106 in /usr/local/lib/python3.10/dist-packages (from torch) (12.1.0.106)
Requirement already satisfied: filelock in /usr/local/lib/python3.10/dist-packages (from torch) (3.13.1)
Requirement already satisfied: nvidia-curand-cu12==10.3.2.106 in /usr/local/lib/python3.10/dist-packages (from torch) (10.3.2.106)
Requirement already satisfied: nvidia-cusolver-cu12==11.4.5.107 in /usr/local/lib/python3.10/dist-packages (from torch) (11.4.5.107)
Requirement already satisfied: nvidia-nvjitlink-cu12 in /usr/local/lib/python3.10/dist-packages (from nvidia-cusolver-cu12==11.4.5.107->torch) (12.1.105)
Requirement already satisfied: mpmath<1.4,>=1.1.0 in /usr/local/lib/python3.10/dist-packages (from sympy==1.13.1->torch) (1.3.0)
Requirement already satisfied: MarkupSafe>=2.0 in /usr/local/lib/python3.10/dist-packages (from jinja2->torch) (3.0.2)
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv
Looking in indexes: https://download.pytorch.org/whl/cu121
Requirement already satisfied: fbgemm_gpu in /usr/local/lib/python3.10/dist-packages (1.0.0+cu121)
Requirement already satisfied: numpy in /usr/local/lib/python3.10/dist-packages (from fbgemm_gpu) (2.2.5)
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv
Requirement already satisfied: torchmetrics in /usr/local/lib/python3.10/dist-packages (1.0.3)
Requirement already satisfied: lightning-utilities>=0.7.0 in /usr/local/lib/python3.10/dist-packages (from torchmetrics) (0.14.3)
Requirement already satisfied: torch>=1.8.1 in /usr/local/lib/python3.10/dist-packages (from torchmetrics) (2.5.1+cu121)
Requirement already satisfied: packaging in /usr/local/lib/python3.10/dist-packages (from torchmetrics) (25.0)
Requirement already satisfied: numpy>1.20.0 in /usr/local/lib/python3.10/dist-packages (from torchmetrics) (2.2.5)
Requirement already satisfied: typing_extensions in /usr/local/lib/python3.10/dist-packages (from lightning-utilities>=0.7.0->torchmetrics) (4.13.2)
Requirement already satisfied: setuptools in /usr/lib/python3/dist-packages (from lightning-utilities>=0.7.0->torchmetrics) (59.6.0)
Requirement already satisfied: filelock in /usr/local/lib/python3.10/dist-packages (from torch>=1.8.1->torchmetrics) (3.13.1)
```

--------------------------------

### Build and Serve Docs Locally

Source: https://github.com/meta-pytorch/torchrec/blob/main/docs/README.md

Run these commands to build the documentation and start a local HTTP server for preview. Ensure you are in the 'docs' directory and have installed the necessary requirements.

```bash
cd docs
pip3 install -r requirements.txt
make html
python3 -m http.server 8082 --bind ::
```

```bash
make html
python3 -m http.server 8082 --bind ::
```

--------------------------------

### Complete Recommendation Pipeline Example

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/INTEGRATION_GUIDE.md

A full example demonstrating a recommendation pipeline that combines retrieval and ranking using TorchRec. This script requires PyTorch and Faiss to be installed.

```python
#!/usr/bin/env python3
"""
Complete recommendation pipeline combining retrieval and ranking.
"""

import torch
import faiss
from torchrec.modules.embedding_configs import EmbeddingBagConfig
from torchrec.modules.embedding_modules import EmbeddingBagCollection
from torchrec.sparse.jagged_tensor import KeyedJaggedTensor

```

--------------------------------

### Install Cloud CLI Tools

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/cloud_deployment/README.md

Install command-line interface tools for AWS, Azure, and GCP.

```bash
# AWS
pip install awscli boto3

# Azure
pip install azure-cli

# GCP
pip install google-cloud-sdk
```

--------------------------------

### Run Simple TorchRec Example

Source: https://github.com/meta-pytorch/torchrec/blob/main/docs/source/setup-torchrec.md

A basic Python script demonstrating the creation and forward pass of an EmbeddingBagCollection with KeyedJaggedTensor. This example verifies a successful TorchRec installation.

```python
import torch

import torchrec
from torchrec.sparse.jagged_tensor import JaggedTensor, KeyedJaggedTensor

ebc = torchrec.EmbeddingBagCollection(
    device="cpu",
    tables=[
        torchrec.EmbeddingBagConfig(
            name="product_table",
            embedding_dim=16,
            num_embeddings=4096,
            feature_names=["product"],
            pooling=torchrec.PoolingType.SUM,
        ),
        torchrec.EmbeddingBagConfig(
            name="user_table",
            embedding_dim=16,
            num_embeddings=4096,
            feature_names=["user"],
            pooling=torchrec.PoolingType.SUM,
        )
    ]
)

product_jt = JaggedTensor(
    values=torch.tensor([1, 2, 1, 5]), lengths=torch.tensor([3, 1])
)
user_jt = JaggedTensor(values=torch.tensor([2, 3, 4, 1]), lengths=torch.tensor([2, 2]))

# Q1: How many batches are there, and which values are in the first batch for product_jt and user_jt?
kjt = KeyedJaggedTensor.from_jt_dict({"product": product_jt, "user": user_jt})

print("Call EmbeddingBagCollection Forward: ", ebc(kjt))
```

--------------------------------

### Install eksctl

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/cloud_deployment/aws/README.md

Install the eksctl command-line tool, a simple CLI for provisioning and managing EKS clusters.

```bash
curl --silent --location "https://github.com/weaveworks/eksctl/releases/latest/download/eksctl_$(uname -s)_amd64.tar.gz" | tar xz -C /tmp
sudo mv /tmp/eksctl /usr/local/bin
```

--------------------------------

### Run TorchRec Quickstart

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/cloud_deployment/README.md

Commands to build and run the TorchRec quickstart locally using Docker or deploy it to a Kubernetes cluster.

```bash
# Build and run locally (single GPU)
cd quickstart && docker build -t torchrec-quickstart . && docker run --gpus all torchrec-quickstart train_torchrec_quickstart.py

# Or deploy to Kubernetes (multi-GPU, multi-node)
kubectl apply -f quickstart/kubernetes_job.yaml
```

--------------------------------

### Run TorchRec Example Script

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/prediction/README.md

Execute the main example script for TorchRec inference.

```bash
python3 predict_using_torchrec.py
```

--------------------------------

### Install PyTorch

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/prediction/README.md

Install the PyTorch library, a fundamental dependency for TorchRec.

```bash
pip install torch torchvision
```

--------------------------------

### Execute TorchRec Example Script

Source: https://github.com/meta-pytorch/torchrec/blob/main/docs/source/setup-torchrec.md

Command to run the saved TorchRec Python example script from the terminal. This confirms the installation and basic functionality.

```bash
python torchrec_example.py
```

--------------------------------

### Activate Virtual Environment and Run Example

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/prediction/README.md

Activate a Python virtual environment before running the TorchRec example script.

```bash
source your_venv/bin/activate
python predict_using_torchrec.py
```

--------------------------------

### Install TorchRec

Source: https://github.com/meta-pytorch/torchrec/blob/main/torchrec/distributed/planner/Planner_Introduction.ipynb

Installs the nightly build of TorchRec.

```python
# install torchrec
!pip3 install torchrec-nightly
```

--------------------------------

### Run Profiling Example with Sorted ZCH

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/zch/Readme.md

Execute the main script to run the profiling example using the sorted ZCH technique.

```bash
python main.py
```

--------------------------------

### Install ZCH Benchmarking Dependencies

Source: https://github.com/meta-pytorch/torchrec/blob/main/torchrec/distributed/benchmark/benchmark_zch/Readme.md

Commands to install PyTorch, FBGEMM-GPU, TorchRec, and the generative recommenders library.

```bash
# Install torch and fbgemm_gpu following instructions in https://docs.pytorch.org/FBGEMM/fbgemm_gpu/development/InstallationInstructions.html
pip install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu126/
pip install --pre fbgemm-gpu --index-url https://download.pytorch.org/whl/nightly/cu126/
# Install torchrec
pip install torchrec --index-url https://download.pytorch.org/whl/nightly/cu126
# Install generative recommenders
git clone https://github.com/meta-recsys/generative-recommenders.git
cd generative-recommenders
pip install -e .
```

--------------------------------

### Install TorchRec

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/prediction/README.md

Install the TorchRec library, which provides specialized modules for recommendation systems.

```bash
pip install torchrec
```

--------------------------------

### Install TorchRec and Dependencies

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/ray/README.md

Installs the required packages for TorchRec, TorchX, and Ray.

```bash
pip install --pre torchrec -f https://download.pytorch.org/whl/torchrec/index.html
pip install torchx-nightly
pip install "ray[default]" -qqq
```

--------------------------------

### Torchx Job Output Example

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/bert4rec/README.MD

Example log output from a Torchx job, showing scheduling, progress, and warnings.

```text
torchx 2022-03-23 05:38:40 INFO     Log files located in: /tmp/torchx_fby9w0w9/torchx/bert4rec_main-txq11m6736w20c/bert4rec_main/0
local_cwd://torchx/bert4rec_main-txq11m6736w20c
torchx 2022-03-23 05:38:40 INFO     Waiting for the app to finish...
bert4rec_main/0 WARNING:
bert4rec_main/0 *****************************************
bert4rec_main/0 Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed.
bert4rec_main/0 *****************************************
bert4rec_main/0 [2]:Filtering triplets
bert4rec_main/0 [7]:Filtering triplets
bert4rec_main/0 [4]:Filtering triplets
bert4rec_main/0 [6]:Filtering triplets
bert4rec_main/0 [3]:Filtering triplets
bert4rec_main/0 [5]:Filtering triplets
bert4rec_main/0 [2]:Densifying index
bert4rec_main/0 [7]:Densifying index
bert4rec_main/0 [4]:Densifying index
bert4rec_main/0 [6]:Densifying index
bert4rec_main/0 [3]:Densifying index
bert4rec_main/0 [5]:Densifying index
bert4rec_main/0 [1]:Filtering triplets
bert4rec_main/0 [0]:Filtering triplets
bert4rec_main/0 [1]:Densifying index
bert4rec_main/0 [0]:Densifying index
bert4rec_main/0 [2]:Splitting
bert4rec_main/0 [2]:<pandas.core.groupby.generic.DataFrameGroupBy object at 0x7f52d986dcd0>
bert4rec_main/0 [6]:Splitting
bert4rec_main/0 [6]:<pandas.core.groupby.generic.DataFrameGroupBy object at 0x7f22b398bbe0>
bert4rec_main/0 [7]:Splitting
bert4rec_main/0 [7]:<pandas.core.groupby.generic.DataFrameGroupBy object at 0x7f90fe1b3be0>
bert4rec_main/0 [4]:Splitting
bert4rec_main/0 [4]:<pandas.core.groupby.generic.DataFrameGroupBy object at 0x7f6f53e3ccd0>
bert4rec_main/0 [3]:Splitting
bert4rec_main/0 [3]:<pandas.core.groupby.generic.DataFrameGroupBy object at 0x7f3babf7ecd0>
bert4rec_main/0 [5]:Splitting
bert4rec_main/0 [5]:<pandas.core.groupby.generic.DataFrameGroupBy object at 0x7f675bedbcd0>
bert4rec_main/0 [1]:Splitting
bert4rec_main/0 [1]:<pandas.core.groupby.generic.DataFrameGroupBy object at 0x7f84351f7cd0>
```

--------------------------------

### Run Profiling Example with Custom Parameters

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/zch/Readme.md

Run the MPZCH profiling example with custom batch size, number of iterations, and embeddings per table.

```bash
python main.py --use_mpzch --batch_size 8 --num_iters 100 --num_embeddings_per_table 1000
```

--------------------------------

### Run Profiling Example with MPZCH

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/zch/Readme.md

Execute the main script with the --use_mpzch flag to run the profiling example using the MPZCH technique.

```bash
python main.py --use_mpzch
```

--------------------------------

### Install TorchRec and Dependencies

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/sharding/sharding.ipynb

Use these commands to install the required packages from the PyTorch index.

```bash
!pip3 install --pre torch --index-url https://download.pytorch.org/whl/cu121 -U
!pip3 install fbgemm_gpu --index-url https://download.pytorch.org/whl/cu121
!pip3 install torchmetrics
!pip3 install torchrec --index-url https://download.pytorch.org/whl/cu121
```

--------------------------------

### Install NVIDIA GPU Drivers

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/cloud_deployment/gcp/README.md

Deploy the NVIDIA driver installer daemonset to the GKE cluster nodes.

```bash
# Install NVIDIA GPU device drivers
kubectl apply -f https://raw.githubusercontent.com/GoogleCloudPlatform/container-engine-accelerators/master/nvidia-driver-installer/cos/daemonset-preloaded.yaml
```

--------------------------------

### Install TorchX with Kubernetes Support

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/cloud_deployment/gcp/README.md

Install TorchX with the necessary Kubernetes integration. This command ensures all dependencies for running on Kubernetes are included.

```bash
pip install torchx[kubernetes]
```

--------------------------------

### Install TorchRec and dependencies

Source: https://github.com/meta-pytorch/torchrec/blob/main/Torchrec_Introduction.ipynb

Use these commands to install the necessary packages, including PyTorch, FBGEMM, and TorchRec, from the nightly wheels.

```bash
!pip3 install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu121 -U
!pip3 install fbgemm_gpu --index-url https://download.pytorch.org/whl/nightly/cu121
!pip3 install torchmetrics==1.0.3
!pip3 install torchrec --index-url https://download.pytorch.org/whl/nightly/cu121
```

--------------------------------

### Activate Conda Environment and Run Example

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/prediction/README.md

Activate a specific conda environment before running the TorchRec example script.

```bash
conda activate your_environment_name
python predict_using_torchrec.py
```

--------------------------------

### E2E Example for RecMetricModule

Source: https://github.com/meta-pytorch/torchrec/blob/main/torchrec/metrics/README.md

Demonstrates the end-to-end usage of RecMetricModule with NE and Throughput metrics. Call update() for each batch and compute() to get the final metrics.

```python
task = RecTaskInfo(
    name="task_name",
    label_name="label",
    prediction_name="prediction",
    weight_name="weight",
)
metrics_config = MetricsConfig(
    rec_tasks=[task],
    rec_metrics={
        RecMetricEnum.NE: RecMetricDef(
            rec_tasks=[task], window_size=10_000_000
        ),
    },
    throughput_metric=ThroughputDef(),
    state_metrics=[],
)
batch_size = 128
metric_module = generate_metric_module(
    RecMetricModule,
    metrics_config=metrics_config,
    batch_size=batch_size,
    world_size=64,
    my_rank=0,
    state_metrics_mapping={},
    device=torch.device("cpu"),
)
n_batches = 10
for _ in range(n_batches):
    model_output = gen_test_batch(
        label_name=task.label_name,
        prediction_name=task.prediction_name,
        weight_name=task.weight_name,
        batch_size=batch_size,
    )
    metric_module.update(model_output)
metrics_result = metric_module.compute()
```

```python
{'ne-task_name|lifetime_ne': tensor([0.8873], dtype=torch.float64), 'ne-task_name|window_ne': tensor([0.6875], dtype=torch.float64), 'throughput-throughput|total_examples': tensor(1400)}
```

--------------------------------

### Create Filestore Instance

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/cloud_deployment/gcp/README.md

Provisions a High Scale SSD Filestore instance for high-performance data access.

```bash
# Create Filestore instance (High Scale tier for best performance)
gcloud filestore instances create torchrec-filestore \
  --zone us-central1-a \
  --tier HIGH_SCALE_SSD \
  --file-share name=torchrec_data,capacity=10TB \
  --network name=default
```

--------------------------------

### Install PyTorch and TorchRec

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/prediction/README.md

Install specific versions of PyTorch and TorchRec to ensure compatibility. This is useful if you encounter installation issues.

```bash
pip install torch==2.0.0 torchvision==0.15.0
pip install torchrec==0.5.0
```

--------------------------------

### Configure Spot Instances in eksctl

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/cloud_deployment/aws/README.md

Example configuration for enabling spot instances in EKS using eksctl. This allows for significant cost savings on GPU compute nodes.

```yaml
# In eksctl config, add spot instances
managedNodeGroups:
  - name: gpu-spot-nodes
    instanceTypes: ["p4d.24xlarge"]
    spot: true
    minSize: 0
    maxSize: 4
```

--------------------------------

### Install Multiprocess Library

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/sharding/sharding.ipynb

Installs the multiprocess package using pip.

```bash
!pip3 install multiprocess
```

--------------------------------

### Execute Recommendation Inference

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/INTEGRATION_GUIDE.md

Example usage showing how to instantiate the pipeline and generate recommendations for a specific user.

```python
if __name__ == "__main__":
    # Create pipeline
    pipeline = create_pipeline(
        retrieval_checkpoint_dir="./retrieval/checkpoints",
        ranking_checkpoint_path="./golden_training/checkpoints/model.pt",
    )

    # Generate recommendations for a user
    user_features = KeyedJaggedTensor(
        keys=["user_id"],
        values=torch.tensor([12345]),
        lengths=torch.tensor([1]),
    )
    user_dense = torch.randn(1, 13)  # 13 dense features

    recommendations = pipeline.recommend(user_features, user_dense)
    print(f"Recommended items: {recommendations}")
```

--------------------------------

### Install TorchRec Dependencies

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/cloud_deployment/README.md

Install the required PyTorch, TorchRec, and FBGEMM-GPU packages.

```bash
pip install torch torchrec fbgemm-gpu
```

--------------------------------

### Upload Training Data to Azure Blob Storage

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/cloud_deployment/azure/README.md

Sets up a storage account and container to host the training dataset.

```bash
# Create storage account
az storage account create \
  --name torcrecstorage \
  --resource-group torchrec-rg \
  --location eastus2 \
  --sku Standard_LRS

# Create container
az storage container create --name criteo-data --account-name torcrecstorage

# Upload dataset
az storage blob upload-batch \
  --destination criteo-data \
  --source ./criteo_data \
  --account-name torcrecstorage
```

--------------------------------

### Provision Cloud Infrastructure for Training

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/cloud_deployment/quickstart/README.md

Commands to create GPU-enabled clusters on major cloud providers.

```bash
eksctl create cluster --name torchrec-cluster \
    --node-type p4d.24xlarge \
    --nodes 2 \
    --region us-east-1
```

```bash
az aks create --resource-group mygroup \
    --name torchrec-cluster \
    --node-vm-size Standard_ND96asr_v4 \
    --node-count 2
```

```bash
gcloud container clusters create torchrec-cluster \
    --machine-type a2-highgpu-8g \
    --num-nodes 2 \
    --accelerator type=nvidia-tesla-a100,count=8
```

--------------------------------

### Install TorchRec Dependencies

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/retrieval/README.md

Install the necessary packages including PyTorch, FAISS, and TorchRec.

```bash
# Install dependencies
conda install -c conda-forge pytorch faiss-gpu
pip install torchx torchrec
```

--------------------------------

### Install TorchRec on Colab

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/sharding/uvm.ipynb

Installs Miniconda, PyTorch with CUDA support, and the TorchRec library.

```bash
# install conda to make installying pytorch with cudatoolkit 11.3 easier. 
!wget https://repo.anaconda.com/miniconda/Miniconda3-py37_4.9.2-Linux-x86_64.sh
!chmod +x Miniconda3-py37_4.9.2-Linux-x86_64.sh
!bash ./Miniconda3-py37_4.9.2-Linux-x86_64.sh -b -f -p /usr/local

# install pytorch with cudatoolkit 11.3
!conda install pytorch cudatoolkit=11.3 -c pytorch-nightly -y

# install torchrec
!pip3 install torchrec-nightly
```

--------------------------------

### Test TorchRec Installation

Source: https://github.com/meta-pytorch/torchrec/blob/main/README.MD

Commands to verify the installation using TorchX for distributed execution.

```bash
GPU mode

torchx run -s local_cwd dist.ddp -j 1x2 --gpu 2 --script test_installation.py

CPU Mode

torchx run -s local_cwd dist.ddp -j 1x2 --script test_installation.py -- --cpu_only
```

--------------------------------

### Training Progress Visualization

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/prediction/README.md

Example output showing training progress over epochs, including loss values and a visual progress bar.

```text
Epoch 1:  Loss: 2.5432  ████░░░░░░░░░░░░░░░░░░░░░░░░░░                        
Epoch 2:  Loss: 1.8721  ████████░░░░░░░░░░░░░░░░░░░░░░                        
Epoch 3:  Loss: 1.2345  ████████████░░░░░░░░░░░░░░░░░░                        
Epoch 4:  Loss: 0.8234  ████████████████░░░░░░░░░░░░░░                        
Epoch 5:  Loss: 0.5123  ████████████████████░░░░░░░░░░                        
...                                                                            
Epoch 10: Loss: 0.1234  ████████████████████████████████  ✓                   
```

--------------------------------

### Install FBGEMM-GPU

Source: https://github.com/meta-pytorch/torchrec/blob/main/README.MD

Commands to install the FBGEMM-GPU library required for optimized RecSys kernels.

```bash
CUDA 12.6

pip install fbgemm-gpu --index-url https://download.pytorch.org/whl/nightly/cu126

CUDA 12.8

pip install fbgemm-gpu --index-url https://download.pytorch.org/whl/nightly/cu128

CUDA 12.9

pip install fbgemm-gpu --index-url https://download.pytorch.org/whl/nightly/cu129

CPU

pip install fbgemm-gpu --index-url https://download.pytorch.org/whl/nightly/cpu
```

--------------------------------

### Build TorchRec Docker container

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/cloud_deployment/quickstart/README.md

Build a Docker image for the TorchRec quickstart environment.

```bash
# Build the container
docker build -t torchrec-quickstart .
```

--------------------------------

### Install NumPy

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/prediction/README.md

Install the NumPy library, often used for numerical operations in conjunction with PyTorch.

```bash
pip install numpy
```

--------------------------------

### Initialize Distributed Communication Backend

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/sharding/sharding.ipynb

Sets up environment variables for distributed communication. Ensure MASTER_ADDR and MASTER_PORT are correctly configured for your environment.

```python
import os
import copy
import torch
import torchrec
import multiprocess
from torchrec.distributed.types import ShardingEnv

os.environ["MASTER_ADDR"] = "localhost"
os.environ["MASTER_PORT"] = "10000"
```

--------------------------------

### Initialize Sharding Topology and Planner

Source: https://github.com/meta-pytorch/torchrec/blob/main/torchrec/distributed/planner/Planner_Introduction.ipynb

Sets up a 2-GPU topology and generates an initial sharding plan for a model using the EmbeddingShardingPlanner.

```python
topology_2gpu = Topology(
    world_size=2, 
    compute_device="cuda"
)

planner = EmbeddingShardingPlanner(
    topology=topology_2gpu,
    debug=True,
)

sharding_plan = planner.plan(small_toy_model, sharders=[EmbeddingBagCollectionSharder()])
```

--------------------------------

### Install Requirements and TorchRec

Source: https://github.com/meta-pytorch/torchrec/blob/main/README.MD

Commands to install Python dependencies and the TorchRec package in development mode.

```bash
pip install -r requirements.txt
```

```bash
python setup.py install develop
```

--------------------------------

### Implement Real-time Serving with Caching

Source: https://github.com/meta-pytorch/torchrec/blob/main/examples/INTEGRATION_GUIDE.md

Cache retrieval results in Redis to optimize re-ranking performance.

```python
# Cache retrieval results, re-rank on the fly
cached_candidates = redis.get(f"candidates:{user_id}")
if cached_candidates is None:
    cached_candidates = retrieval_model(user_kjt)
    redis.set(f"candidates:{user_id}", cached_candidates, ttl=3600)
scores = ranking_model(dense, sparse)
```