Please enable JavaScript.
Coggle requires JavaScript to display documents.
Machine Learning Journey - Coggle Diagram
Machine Learning Journey
Tools
Hugging Face
Models
Pipelines :check:
task
model
AutoClasses :fire:
Trainer
Pre-processing
Tokenizer
Datasets
Pytorch
Tensor
Example
[[1, 2, 3], [4, 5, 6]]
properties
shape
data type
device
operations
addition (+) / subtraction (-)
element-wise multiplication (*)
transposition
concatenation
matrix multiplication (@)
Layer
types
Linear
Sequential
Stack of layers
Activation function
types
Sigmoid
Zoom In
Softmax
Zoom In
LLM
Architecture
Transformers
Tokenizers 🗿
Normalization
"I am Stéve Jôbs" :arrow_right: "i am steve jobs"
Pre-Tokenization
"i am steve jobs" :arrow_right: ["i", "am", "steve", "jobs"]
Tokenization
(Tokens)
["i", "am", "steve", 'jobs"] :arrow_right: [2022, 484, 5055, 788]
Decode
:arrow_left:
Embedding layer (output dimension = 5)
Embedding :bed:
[2022, 484, 5055, 788]
:arrow_right:
[[0.230, -0.433, 0.000, 0.055, 0.409],
[-0.850, -0.473, 0.010, -0.125, -0.444],
[-0.703, -0.883, -0.010, 0.525, -0.949],
[-0.663, -0.443, 0.220, -0.055, 0.410]]
1 more item...
Encode
:arrow_right:
Zoom In
Use cases
Text 🔤
Sentiment Analysis
Text Entailment (QNLI)
Check Grammar
Generate Responses
Text summarization
Image 🖼️
Stock photograph
Agriculture monitoring
Wildlife monitoring
Q&A
Audio 🔉
Speech Recognition
Speaker identification
Language identification
Environment sounds
Search 🔎
Semantic Search :bed: (≠ Keyword Search)
Pre processing
Images
Cropping & Resizing
Text
Tokenization
Audio
Sampling
16kHz
Resampling
Filtering
by length
Fine-tuning
LLM
🤖
+
Data
📊
Training
Models Evaluation
Audio
WER
Deep Learning
Architecture
Neural Network
Input, Hidden and Output layers
Data
Large amount of data
