Skip to content

Commit 68ff6a1

Browse files
ericstjericstj
authored
Add a doc with information about components and dependencies (#7537)
* Add a doc with information about components and dependencies * Formatting fixups and edits * Add a diagram of components * Improve diagram readability * Update mermaid style
1 parent 3abbb17 commit 68ff6a1

File tree

1 file changed

+269
-0
lines changed

1 file changed

+269
-0
lines changed

‎docs/project-docs/components-and-dependencies.md‎

Lines changed: 269 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,269 @@
1+
ML.NET is a modular set of libraries that enables building a pipeline from data loaders, trainers/estimators (in case of training), transformers (in case of inferencing), and various data structures to facilitate the building of pipelines and representing data. The core of ML.NET – the Microsoft.ML package has no external dependencies. It's largely managed code.
2+
3+
Microsoft.ML does have a helper native math library CPUMath - which is only used on .NETFramework. .NET 6.0 and later have a managed implementation using intrinsics/TensorPrimitives and do not require the native build of CPUMath.
4+
5+
Microsoft.ML contains one other native library, LDANative, which is used by the LatentDirichletAllocationEstimator/Transform to support the LightLDA algorithm. If this component is used it will require the LightLDA native library. The native library is built for linux-arm, linux-arm64, linux-x64, osx-arm64 (M1), osx-x64, win-arm64, win-x64, win-x86. This library has only platform/CRT dependencies.
6+
7+
Some components that represent an algorithm or binding to another framework are factored into separate packages to allow opt-in to using those and their dependencies.
8+
9+
ML.NET redistributes Intel MKL as Microsoft.ML.MKL.Redist in which is a minimized MKL library linked with just exports needed by ML.NET. This component follows the support matrix of Intel MKL and is only supported on x86 and x64 architectures: linux-x64, osx-x64 (no longer supported by Intel), win-x64, and win-x86. Similarly some components have light-up to use an Intel OneDAL implementation which is only supported on x64.
10+
11+
| NuGet Package | Entry-Point Components | Native Dependencies | Status | Notes |
12+
|-------------------------------------|----------------------------------------------------------------|------------------------------------------------|----------|--------------------------------------------------------------------------------------------|
13+
| `Microsoft.ML` | `MLContext`, core transforms, trainers | None | Stable | |
14+
| `Microsoft.Extensions.ML` | `PredictionEnginePool` | None | Stable | |
15+
| `Microsoft.ML.AutoML` | `AutoCatalog` for AutoML | *As required by other components* | Preview | Support varies based on components used |
16+
| `Microsoft.ML.CodeGenerator` | | None | Preview | Part of AutoML |
17+
| `Microsoft.ML.CpuMath` | | Optional native | Stable | Internal implementation; only used on .NET Framework |
18+
| `Microsoft.ML.DataView` | `IDataView` | None | Stable | |
19+
| `Microsoft.ML.DnnImageFeaturizer.*` | | None | Preview | Data-only |
20+
| `Microsoft.ML.Ensemble` | | None | Preview | Supports ML.NET component catalog |
21+
| `Microsoft.ML.EntryPoints` | | None | Preview | Supports ML.NET component catalog |
22+
| `Microsoft.ML.Experimental` | | None | Preview | Experimental API |
23+
| `Microsoft.ML.FairLearn` | `FairlearnCatalog` | None | Preview | |
24+
| `Microsoft.ML.FastTree` | `FastTreeRankingTrainer` | Optional native acceleration | Stable | Native library used on x86/x64; managed fallback |
25+
| `Microsoft.ML.ImageAnalytics` | `MLImage` (image exchange type) | `libSkiaSharp` | Stable | Wrapper over SkiaSharp / Google Skia; supported where dependency is supported |
26+
| `Microsoft.ML.LightGBM` | `LightGbm\*Trainer` | `LightGBM` | Stable | Wrapper over LightGBM; supported where dependency is supported |
27+
| `Microsoft.ML.MKL.Components` | `SymbolicSgdLogisticRegressionBinaryTrainer` | Intel MKL | Stable | Only works where Intel MKL works |
28+
| `Microsoft.ML.MKL.Redist` | Internal native Intel MKL | `libomp` | Stable | Not for direct reference; win-x86/x64 only |
29+
| `Microsoft.ML.OneDal` | Internal native Intel OneDal | Intel OneDAL | Preview | Not for direct reference; x64 only |
30+
| `Microsoft.ML.OnnxConverter` | Adds ONNX export support | `Microsoft.ML.OnnxRuntime` | Stable | Wrapper over ONNX Runtime; supports "bring your own" runtime |
31+
| `Microsoft.ML.OnnxTransformer` | `OnnxCatalog` | `Microsoft.ML.OnnxRuntime` | Stable | Wrapper over ONNX Runtime; supports "bring your own" runtime |
32+
| `Microsoft.ML.Parquet` | `ParquetLoader` | None | Preview | Uses managed Parquet.Net (port of Apache Parquet) |
33+
| `Microsoft.ML.Recommender` | `MatrixFactorizationTrainer` | LIBMF (bundled) | Stable | Includes libmf built for all runtimes supported by ML.NET |
34+
| `Microsoft.ML.TensorFlow` | `TensorFlowModel`, `Transformer`, `Estimator` | TensorFlow via `TensorFlow.NET` | Stable | Wrapper over TensorFlow; supports "bring your own" runtime |
35+
| `Microsoft.ML.TimeSeries` | `ForecastingCatalog` | Intel MKL, `libomp` | Stable | Only works where Intel MKL works |
36+
| `Microsoft.ML.TorchSharp` | `QATrainer`, `TextClassificationTrainer`, `SentenceSimilarityTrainer` | libTorch via `TorchSharp` | Preview | Wrapper over libTorch; supported where TorchSharp is supported |
37+
| `Microsoft.ML.Vision` | `ImageClassificationTrainer` | TensorFlow | Stable | Depends on `Microsoft.ML.TensorFlow` for implementation |
38+
39+
40+
Other packages:
41+
| NuGet Package | Entry-Point Components | Native Dependencies | Status | Notes |
42+
|----------------------------------|--------------------------------------------------|-----------------------------------------------|------------|---------|
43+
| `Microsoft.Data.Analysis` | `DataFrame` | `Apache.Arrow` | Preview | |
44+
| `Microsoft.ML.GenAI.*` | | | Preview | |
45+
| `Microsoft.ML.Tokenizers.*` | `Tokenizer` | | Stable | |
46+
| `Microsoft.ML.SampleUtils` | | | Preview | |
47+
48+
49+
## Package Dependencies Diagram
50+
51+
The following diagram shows the relationships between ML.NET packages and their external dependencies:
52+
53+
```mermaid
54+
%%{init: {'theme':'base', 'themeVariables': {'fontSize':'18px'}, 'flowchart': {'useMaxWidth': false, 'htmlLabels': true, 'curve': 'linear'}, 'securityLevel': 'loose'}}%%
55+
graph TD
56+
%% Core packages - arranged vertically at top
57+
subgraph CorePackages["🔧 Core ML.NET Packages"]
58+
direction TB
59+
DataView["Microsoft.ML.DataView"]
60+
Core["Microsoft.ML"]
61+
Extensions["Microsoft.Extensions.ML"]
62+
CpuMath["Microsoft.ML.CpuMath"]
63+
end
64+
65+
%% AutoML packages
66+
subgraph AutoMLPackages["🤖 AutoML Packages"]
67+
direction TB
68+
AutoML["Microsoft.ML.AutoML"]
69+
CodeGen["Microsoft.ML.CodeGenerator"]
70+
Ensemble["Microsoft.ML.Ensemble"]
71+
EntryPoints["Microsoft.ML.EntryPoints"]
72+
FairLearn["Microsoft.ML.FairLearn"]
73+
end
74+
75+
%% Algorithm packages
76+
subgraph AlgorithmPackages["⚙️ ML Algorithm Packages"]
77+
direction TB
78+
FastTree["Microsoft.ML.FastTree"]
79+
LightGBM["Microsoft.ML.LightGBM"]
80+
Recommender["Microsoft.ML.Recommender"]
81+
TimeSeries["Microsoft.ML.TimeSeries"]
82+
TorchSharp["Microsoft.ML.TorchSharp"]
83+
end
84+
85+
%% Image and vision packages
86+
subgraph ImagePackages["🖼️ Image & Vision Packages"]
87+
direction TB
88+
ImageAnalytics["Microsoft.ML.ImageAnalytics"]
89+
Vision["Microsoft.ML.Vision"]
90+
DnnFeaturizerAlexNet["Microsoft.ML.DnnImageFeaturizer.AlexNet"]
91+
DnnFeaturizerResNet18["Microsoft.ML.DnnImageFeaturizer.ResNet18"]
92+
DnnFeaturizerResNet50["Microsoft.ML.DnnImageFeaturizer.ResNet50"]
93+
DnnFeaturizerResNet101["Microsoft.ML.DnnImageFeaturizer.ResNet101"]
94+
DnnFeaturizerModelRedist["Microsoft.ML.DnnImageFeaturizer.ModelRedist"]
95+
end
96+
97+
%% ONNX and TensorFlow packages
98+
subgraph FrameworkPackages["🔗 Framework Integration Packages"]
99+
direction TB
100+
OnnxConverter["Microsoft.ML.OnnxConverter"]
101+
OnnxTransformer["Microsoft.ML.OnnxTransformer"]
102+
TensorFlow["Microsoft.ML.TensorFlow"]
103+
end
104+
105+
%% Intel MKL packages
106+
subgraph IntelPackages["⚡ Intel MKL Packages"]
107+
direction TB
108+
MKLComponents["Microsoft.ML.MKL.Components"]
109+
MKLRedist["Microsoft.ML.MKL.Redist"]
110+
OneDal["Microsoft.ML.OneDal"]
111+
end
112+
113+
%% AI/GenAI packages
114+
subgraph AIPackages["🧠 AI & GenAI Packages"]
115+
direction TB
116+
GenAICore["Microsoft.ML.GenAI.Core"]
117+
GenAILLaMA["Microsoft.ML.GenAI.LLaMA"]
118+
GenAIMistral["Microsoft.ML.GenAI.Mistral"]
119+
GenAIPhi["Microsoft.ML.GenAI.Phi"]
120+
AutoGenCore["AutoGen.Core"]
121+
MSExtensionsAI["Microsoft.Extensions.AI.Abstractions"]
122+
SemanticKernel["Microsoft.SemanticKernel.Abstractions"]
123+
end
124+
125+
%% Tokenizer packages
126+
subgraph TokenizerPackages["📝 Tokenizer Packages"]
127+
direction TB
128+
Tokenizers["Microsoft.ML.Tokenizers"]
129+
TokenizersGpt2["Microsoft.ML.Tokenizers.Data.Gpt2"]
130+
TokenizersR50k["Microsoft.ML.Tokenizers.Data.R50kBase"]
131+
TokenizersP50k["Microsoft.ML.Tokenizers.Data.P50kBase"]
132+
TokenizersO200k["Microsoft.ML.Tokenizers.Data.O200kBase"]
133+
TokenizersCl100k["Microsoft.ML.Tokenizers.Data.Cl100kBase"]
134+
end
135+
136+
%% Data packages
137+
subgraph DataPackages["📊 Data Packages"]
138+
direction TB
139+
Parquet["Microsoft.ML.Parquet"]
140+
DataAnalysis["Microsoft.Data.Analysis"]
141+
end
142+
143+
%% Other packages
144+
subgraph OtherPackages["🔧 Other Packages"]
145+
direction TB
146+
Experimental["Microsoft.ML.Experimental"]
147+
SampleUtils["Microsoft.ML.SampleUtils"]
148+
end
149+
150+
%% External dependencies - arranged vertically at bottom
151+
subgraph ExternalDeps["🌐 External Dependencies"]
152+
direction TB
153+
SkiaSharp["SkiaSharp"]
154+
LightGBMNative["LightGBM"]
155+
OnnxRuntime["Microsoft.ML.OnnxRuntime"]
156+
TensorFlowNET["TensorFlow.NET"]
157+
TorchSharpLib["TorchSharp"]
158+
ApacheArrow["Apache.Arrow"]
159+
ParquetNet["Parquet.Net"]
160+
GoogleProtobuf["Google.Protobuf"]
161+
end
162+
163+
%% Core dependencies
164+
Core --> DataView
165+
Core --> CpuMath
166+
Extensions --> Core
167+
168+
%% AutoML dependencies
169+
AutoML --> Core
170+
AutoML --> CpuMath
171+
AutoML --> DnnFeaturizerAlexNet
172+
AutoML --> DnnFeaturizerResNet18
173+
AutoML --> DnnFeaturizerResNet50
174+
AutoML --> DnnFeaturizerResNet101
175+
AutoML --> OnnxTransformer
176+
AutoML --> TimeSeries
177+
AutoML --> TorchSharp
178+
AutoML --> Vision
179+
AutoML --> ImageAnalytics
180+
AutoML --> LightGBM
181+
AutoML --> MKLComponents
182+
AutoML --> Recommender
183+
CodeGen --> AutoML
184+
185+
%% Algorithm dependencies
186+
FastTree --> Core
187+
LightGBM --> Core
188+
LightGBM --> FastTree
189+
LightGBM --> LightGBMNative
190+
Recommender --> Core
191+
TimeSeries --> Core
192+
TimeSeries --> MKLRedist
193+
TorchSharp --> Core
194+
TorchSharp --> ImageAnalytics
195+
TorchSharp --> Tokenizers
196+
TorchSharp --> TorchSharpLib
197+
198+
%% Image and vision dependencies
199+
ImageAnalytics --> Core
200+
ImageAnalytics --> SkiaSharp
201+
Vision --> Core
202+
Vision --> TensorFlow
203+
204+
%% Framework dependencies
205+
OnnxConverter --> Core
206+
OnnxTransformer --> Core
207+
OnnxTransformer --> OnnxRuntime
208+
OnnxTransformer --> GoogleProtobuf
209+
TensorFlow --> Core
210+
TensorFlow --> ImageAnalytics
211+
TensorFlow --> TensorFlowNET
212+
213+
%% Intel MKL dependencies
214+
MKLComponents --> Core
215+
MKLComponents --> MKLRedist
216+
MKLComponents --> OneDal
217+
218+
%% Other package dependencies
219+
Ensemble --> Core
220+
EntryPoints --> Core
221+
Experimental --> Core
222+
FairLearn --> Core
223+
FairLearn --> DataAnalysis
224+
FairLearn --> AutoML
225+
Parquet --> Core
226+
Parquet --> ParquetNet
227+
DataAnalysis --> ApacheArrow
228+
229+
%% GenAI dependencies
230+
GenAICore --> TorchSharpLib
231+
GenAICore --> AutoGenCore
232+
GenAICore --> MSExtensionsAI
233+
GenAICore --> SemanticKernel
234+
GenAILLaMA --> GenAICore
235+
GenAILLaMA --> TorchSharpLib
236+
GenAIMistral --> GenAICore
237+
GenAIPhi --> GenAICore
238+
239+
%% DNN Image Featurizer dependencies
240+
DnnFeaturizerAlexNet --> OnnxTransformer
241+
DnnFeaturizerAlexNet --> DnnFeaturizerModelRedist
242+
DnnFeaturizerResNet18 --> OnnxTransformer
243+
DnnFeaturizerResNet18 --> DnnFeaturizerModelRedist
244+
DnnFeaturizerResNet50 --> OnnxTransformer
245+
DnnFeaturizerResNet50 --> DnnFeaturizerModelRedist
246+
DnnFeaturizerResNet101 --> OnnxTransformer
247+
DnnFeaturizerResNet101 --> DnnFeaturizerModelRedist
248+
249+
%% Tokenizer dependencies
250+
Tokenizers --> GoogleProtobuf
251+
TokenizersGpt2 --> Tokenizers
252+
TokenizersR50k --> Tokenizers
253+
TokenizersP50k --> Tokenizers
254+
TokenizersO200k --> Tokenizers
255+
TokenizersCl100k --> Tokenizers
256+
257+
%% Styling for readability and larger text
258+
classDef external fill:#ffebcd,stroke:#d2691e,stroke-width:4px,font-size:18px,font-weight:bold
259+
classDef core fill:#e6f3ff,stroke:#0066cc,stroke-width:4px,font-size:18px,font-weight:bold
260+
classDef algorithm fill:#f0f8e6,stroke:#228b22,stroke-width:4px,font-size:18px,font-weight:bold
261+
classDef bundled fill:#ffefd5,stroke:#ff8c00,stroke-width:4px,font-size:18px,font-weight:bold
262+
classDef subgraphStyle fill:#f9f9f9,stroke:#333,stroke-width:3px,font-size:20px,font-weight:bold
263+
264+
class SkiaSharp,LightGBMNative,OnnxRuntime,TensorFlowNET,TorchSharpLib,ApacheArrow,ParquetNet,GoogleProtobuf,AutoGenCore,MSExtensionsAI,SemanticKernel external
265+
class DataView,Core,Extensions core
266+
class AutoML,CodeGen,FastTree,LightGBM,Recommender,TimeSeries,TorchSharp,ImageAnalytics,DnnFeaturizerAlexNet,DnnFeaturizerResNet18,DnnFeaturizerResNet50,DnnFeaturizerResNet101,DnnFeaturizerModelRedist,Vision,OnnxConverter,OnnxTransformer,TensorFlow,MKLComponents,Ensemble,EntryPoints,Experimental,FairLearn,Parquet,DataAnalysis,GenAICore,GenAILLaMA,GenAIMistral,GenAIPhi,Tokenizers,TokenizersGpt2,TokenizersR50k,TokenizersP50k,TokenizersO200k,TokenizersCl100k,SampleUtils algorithm
267+
class CpuMath,MKLRedist,OneDal bundled
268+
```
269+

0 commit comments

Comments
 (0)