N Neurarch Architectures Models Checks Data Docs Open the app

Comparisons / DLRM vs Wide & Deep

DLRM vs Wide & Deep

Two production ranking models with different interaction layers.

Wide & Deep has 61M fewer parameters than DLRM: 3 layers added, 4 removed, 6 changed.

Baseline

DLRM

Layers
11
Parameters
64M
Input
13
Output
1
Forward-passes
yes
Est. train cost
$0.063
Compared

Wide & Deep

Layers
10
Parameters
3.7M
Input
10000
Output
1
Forward-passes
yes
Est. train cost
$0.043

The deltas

Every number is Wide & Deep relative to DLRM.

Parameters
-61M (-94.3%)
Layers
-1
Added
3
Removed
4
Changed
6
Unchanged
4

Which GPUs each one fits

Each side is measured at its own declared input (13 against 10000). Both columns are right about their own model; the difference between them is not a fact about the designs.

GPUDLRMWide & Deep
T4 16GBfitsfits
A100 40GBfitsfits
H100 80GBfitsfits

Layer by layer

Aligned in topological order. 4 of 17 rows are the same layer with the same parameters.

Hide all 17 rows
DLRMWide & Deep
LayerParamsOutputLayerParamsOutput
1changed
shape
Dense Features
Input
13Wide Input (cross feats)
Input
10000
2changed
shape
Sparse Features
Input
26Deep Input (sparse cat)
Input
50
3removedBottom MLP 1
Linear
89664
4removedEmbeddingBag
Embedding Bag
64M26 × 64
5removedReLU
Relu
64
6changed
inFeatures, outFeatures
Bottom MLP 2
Linear
4.2K64Wide Linear
Linear
10K1
7removedFeature Interaction
Feature Interaction
415
8addedEmbeddings
Embedding
3.2M50 × 32
9addedFlatten
Flatten
1600
10changed
inFeatures, outFeatures
Top MLP 1
Linear
213K512Deep FC 1
Linear
410K256
11sameReLU
Relu
512ReLU 1
Relu
256
12changed
inFeatures, outFeatures
Top MLP 2
Linear
131K256Deep FC 2
Linear
33K128
13sameReLU
Relu
256ReLU 2
Relu
128
14changed
inFeatures
CTR Head
Linear
2571Deep Out
Linear
1291
15addedWide + Deep
Add
1
16sameSigmoid
Sigmoid
1Sigmoid CTR
Sigmoid
1
17sameP(click)
Output
1P(click)
Output
1

What this is not

Take it further

Open either graph in the editor, change it, and check it again: DLRM · Wide & Deep

Compare any two models of your own, including anything on Hugging Face: the comparison tool.

Machine-readable: this page as markdown · the pair index · POST https://www.neurarch.com/api/v1/plan for a graph of your own.

Related comparisons

Wide & Deep vs Neural Collaborative Filtering
The memorisation-plus-generalisation recommender against a purely neural one.