Quick Start Guide To Large Language Models Second Edition Sinan Ozdemir

Quick Start Guide To Large Language Models
Second Edition Sinan Ozdemir download
https://ebookbell.com/product/quick-start-guide-to-large-
language-models-second-edition-sinan-ozdemir-57362590
Explore and download more ebooks at ebookbell.com

Here are some recommended products that we believe you will be
interested in. You can click the link to download.
Quick Start Guide To Large Language Models Strategies And Best
Practices For Using Hatgpt And Other Llms Sinan Ozdemir
https://ebookbell.com/product/quick-start-guide-to-large-language-
models-strategies-and-best-practices-for-using-hatgpt-and-other-llms-
sinan-ozdemir-58738084
Quick Start Guide To Large Language Models Sinan Ozdemir
models-sinan-ozdemir-58738142
Quick Start Guide To Large Language Models Strategies And Best
Practices For Using Chatgpt And Other Llms Sinan Ozdemir
models-strategies-and-best-practices-for-using-chatgpt-and-other-llms-
sinan-ozdemir-231944618
Early Release Quick Start Guide To Large Language Models Strategies
And Best Practices For Using Chatgpt And Other Llms Sinan Ozdemir
https://ebookbell.com/product/early-release-quick-start-guide-to-
large-language-models-strategies-and-best-practices-for-using-chatgpt-
and-other-llms-sinan-ozdemir-49607190

Quick Start Guide To Ffmpeg Learn To Use The Open Source
Multimediaprocessing Tool Like A Pro 1st Edition V Subhash
https://ebookbell.com/product/quick-start-guide-to-ffmpeg-learn-to-
use-the-open-source-multimediaprocessing-tool-like-a-pro-1st-edition-
v-subhash-47649426
Quick Start Guide To Ffmpeg Learn To Use The Open Source
Multimediaprocessing Tool Like A Pro 1st Edition Vsubhash
https://ebookbell.com/product/quick-start-guide-to-ffmpeg-learn-to-
use-the-open-source-multimediaprocessing-tool-like-a-pro-1st-edition-
vsubhash-47682238
Quick Start Guide To Dart Programming Create Highperformance
Applications For The Web And Mobile 1st Edition Sanjib Sinha
https://ebookbell.com/product/quick-start-guide-to-dart-programming-
create-highperformance-applications-for-the-web-and-mobile-1st-
edition-sanjib-sinha-50195018
Quick Start Guide To Azure Data Factory Azure Data Lake Server And
Azure Data Warehouse Mark Beckner
https://ebookbell.com/product/quick-start-guide-to-azure-data-factory-
azure-data-lake-server-and-azure-data-warehouse-mark-beckner-51027562
Quick Start Guide To Verilog 2nd Edition 2nd Brock J Lameres
https://ebookbell.com/product/quick-start-guide-to-verilog-2nd-
edition-2nd-brock-j-lameres-52959080

Quick Start Guide to Large
Language Models, Second
Edition
Sinan Ozdemir
A NOTE FOR EARLY RELEASE READERS
With Early Release eBooks, you get books in their earliest
form—the author’s raw and unedited content as they write—
so you can take advantage of these technologies long
before the official release of these titles.
If you have comments about how we might improve the
content and/or examples in this book, or if you notice
missing material within this title, please reach out to
Pearson at PearsonITAcademics@pearson.com

Contents
Preface
Acknowledgments
About the Author
Part I: Introduction to Large Language Models
Chapter 1: Introduction to Large Language Models
Chapter 2: Semantic Search with LLMs
Chapter 3: First Steps with Prompt Engineering
Chapter 4: The LLM/AI Ecosystem--RAG + Agent
Case Study
Part II: Getting the Most Out of LLMs
Chapter 5: Optimizing LLMs with Customized Fine-
Tuning
Chapter 6: Advanced Prompt Engineering
Chapter 7: Customizing Embeddings and Model
Architectures
Chapter 8: Alignment First Principles
Part III: Advanced LLM Usage
Chapter 9: Moving Beyond Foundation Models
Chapter 10: Advanced Open-Source LLM Fine
Tuning

Chapter 11: Moving LLMs into Production
Chapter 12: Evaluating LLMs/LLMOps

Table of Contents
Preface
Acknowledgments
About the Author
Part I: Introduction to Large Language Models
1. Overview of Large Language Models
Introduction
What Are Large Language Models?
Popular Modern LLMs
Domain-Specific LLMs
Applications of LLMs
Summary
2. Semantic Search with LLMs
Introduction
The Task
Solution Overview
The Components
Putting It All Together
The Cost of Closed-Source Components
Summary
3. First Steps with Prompt Engineering
Introduction
Prompt Engineering

Working with Prompts Across Models
Summary
4. The AI Ecosystem—Putting the Pieces Together
Introduction
The Ever-Shifting Performance of Closed-
Source AI
AI Reasoning versus Thinking
Case Study 1: Retrieval Augmented
Generation (RAG)
Case Study 2: Automated AI Agents
Conclusion
Part II: Getting the Most Out of LLMs
5. Optimizing LLMs with Customized Fine-Tuning
6. Advanced Prompt Engineering
7. Customizing Embeddings and Model
Architectures
8. AI Alignment: First Principles
Introduction
Aligned to Whom and to What End?
Alignment as a Bias Mitigator
The Pillars of Alignment
Constitutional AI—A Step Toward Self-
Alignment
Conclusion
Part III: Advanced LLM Usage
9. Moving Beyond Foundation Models
10. Advanced Open-Source LLM Fine Tuning

11. Moving LLMs into Production
12. Evaluating LLMs
Introduction
Evaluating Generative Tasks
Evaluating Understanding Tasks
Conclusion
Keep Going!

Preface
This content is currently in development.

Acknowledgments

About the Author

Part I
Introduction to Large
Language Models

1. Overview of Large
Language Models
Introduction
In 2017, a team at Google Brain introduced an advanced
artificial intelligence (AI) deep learning model called the
Transformer. Since then, the Transformer has become the
standard for tackling various natural language processing
(NLP) tasks in academia and industry. It is likely that you
have interacted with the Transformer model in recent years
without even realizing it, as Google uses BERT to enhance
its search engine by better understanding users’ search
queries. The GPT family of models from OpenAI have also
received attention for their ability to generate human-like
text and images.
Note
We cannot fit all of the ever-shifting code for this book
within these pages so to get the always free and update
to code, check out our github repo at
https://github.com/sinanuozdemir/quick-start-guide-to-
llms.

These Transformers now power applications such as
GitHub’s Copilot (developed by OpenAI in collaboration with
Microsoft), which can convert comments and snippets of
code into fully functioning source code that can even call
upon other large language models (LLMs) (as in Listing 1.1)
to perform NLP tasks.
Listing 1.1 Using the Copilot LLM to get an output
from Facebook’s BART LLM
from transformers import pipeline
def classify_text(email):
“””
Use Facebook’s BART model to classify an email
Args:
email (str): The email to classify
Returns:
str: The classification of the email
“””
# COPILOT START. EVERYTHING BEFORE THIS COMMENT
classifier = pipeline(
‘zero-shot-classification’, model=’facebook/bar
labels = [‘spam’, ‘not spam’]
hypothesis_template = ‘This email is {}.’
results = classifier(
email, labels, hypothesis_template=hypothesis_t
return results[‘labels’][0]
# COPILOT END

In Listing 1.1, I used Copilot to take in only a Python function
definition and some comments I wrote, and I wrote all of the
code to make the function do what I wrote. There’s no
cherry-picking here, just a fully working Python function that
I can call like this:
classify_text(‘hi I am spam’) # spam
It appears we are surrounded by LLMs, but just what are
they doing under the hood? Let’s find out!
What Are Large Language Models?
Large language models (LLMs) are AI models that are
usually (but not necessarily) derived from the Transformer
architecture and are designed to understand and generate
human language, code, and much more. These models are
trained on vast amounts of text data, allowing them to
capture the complexities and nuances of human language.
LLMs can perform a wide range of language-related tasks,
from simple text classification to text generation, with high
accuracy, fluency, and style.
In the healthcare industry, LLMs are being used for
electronic medical record (EMR) processing, clinical trial
matching, and drug discovery. In finance, they are being
utilized for fraud detection, sentiment analysis of financial
news, and even trading strategies. LLMs are also used for
customer service automation via chatbots and virtual
assistants. Owing to their versatility and highly performant
natures, Transformer-based LLMs are becoming an
increasingly valuable asset in a variety of industries and
applications.
Note

I will use the term understand a fair amount in this text.
In this context, I am usually referring to “natural
language understanding” (NLU)—a research branch of
NLP that focuses on developing algorithms and models
that can accurately interpret human language. As we
will see, NLU models excel at tasks such as
classification, sentiment analysis, and named entity
recognition. However, it is important to note that while
these models can perform complex language tasks, they
do not possess true understanding in the same way that
humans do.
The success of LLMs and Transformers is due to the
combination of several ideas. Most of these ideas had been
around for years but were also being actively researched
around the same time. Mechanisms such as attention,
transfer learning, and scaling up neural networks, which
provide the scaffolding for Transformers, were seeing
breakthroughs right around the same time. Figure 1.1
outlines some of the biggest advancements in NLP in the
last few decades, all leading up to the invention of the
Transformer.

Figure 1.1 A brief history of modern NLP highlights the
use of deep learning to tackle language modeling,
advancements in large-scale semantic token
embeddings (Word2vec), sequence-to-sequence models
with attention (something we will see in more depth
later in this chapter), and finally the Transformer in
2017.
The Transformer architecture itself is quite impressive. It can
be highly parallelized and scaled in ways that previous
state-of-the-art NLP models could not be, allowing it to scale
to much larger datasets and training times than was
possible with previous NLP models. The Transformer uses a
special kind of attention calculation called self-attention to

allow each word in a sequence to “attend to” (look to for
context) all other words in the sequence, enabling it to
capture long-range dependencies and contextual
relationships between words. Of course, no architecture is
perfect. Transformers are still limited to an input context
window, which represents the maximum length of text they
can process at any given moment.
Since the advent of the Transformer architecture in 2017,
the ecosystem around using and deploying Transformers
has exploded. The aptly named “Transformers” library and
its supporting packages have enabled practitioners to use,
train, and share models, greatly accelerating this model’s
adoption, to the point that it is now being used by
thousands of organizations (and counting). Popular LLM
repositories such as Hugging Face have popped up,
providing access to powerful open-source models to the
masses. In short, using and productionizing a Transformer
has never been easier.
That’s where this book comes in.
My goal is to guide you on how to use, train, and optimize
all kinds of LLMs for practical applications while giving you
just enough insight into the inner workings of the model to
know how to make optimal decisions about model choice,
data format, fine-tuning parameters, and so much more.
My aim is to make use of Transformers accessible for
software developers, data scientists, analysts, and hobbyists
alike. To do that, we should start on a level playing field and
learn a bit more about LLMs.
Definition of LLMs

To back up only slightly, we should talk first about the
specific NLP task that LLMs and Transformers are being used
to solve, which provides the foundation layer for their ability
to solve a multitude of tasks. Language modeling is a
subfield of NLP that involves the creation of statistical/deep
learning models for predicting the likelihood of a sequence
of tokens in a specified vocabulary (a limited and known
set of tokens). There are generally two kinds of language
modeling tasks out there: autoencoding tasks and
autoregressive tasks (Figure 1.2).

Figure 1.2 Both the autoencoding and autoregressive
language modeling tasks involve filling in a missing
token, but only the autoencoding task allows for context
to be seen on both sides of the missing token.
Note
A token is the smallest unit of semantic meaning, which
is created by breaking down a sentence or piece of text
into smaller units; it is the basic input for an LLM. Tokens
can be words but also can be “sub-words,” as we will
see in more depth throughout this book. Some readers
may be familiar with the term “n-gram,” which refers to
a sequence of n consecutive tokens.
Autoregressive language models are trained to predict the
next token in a sentence, based on only the previous tokens
in the phrase. These models correspond to the decoder part
of the Transformer model, with a mask being applied to the
full sentence so that the attention heads can see only the
tokens that came before. Autoregressive models are ideal
for text generation. A good example of this type of model is
GPT.
Autoencoding language models are trained to reconstruct
the original sentence from a corrupted version of the input.
These models correspond to the encoder part of the
Transformer model and have access to the full input without
any mask. Autoencoding models create a bidirectional
representation of the whole sentence. They can be fine-
tuned for a variety of tasks such as text generation, but
their main application is sentence classification or token
classification. A typical example of this type of model is
BERT.

To summarize, LLMs are language models may be either
autoregressive, autoencoding, or a combination of the two.
Modern LLMs are usually based on the Transformer
architecture (which we will use in this book), but can also be
based on another architecture. The defining features of
LLMs are their large size and large training datasets, which
enable them to perform complex language tasks, such as
text generation and classification, with high accuracy and
with little to no fine-tuning.
For now, let’s look at some of the popular LLMs we’ll be
using throughout this book.
Popular Modern LLMs
BERT, GPT, T5, and Llama are four popular LLMs developed
by Google, OpenAI, Google, and Meta respectively. These
models differ quite dramatically in terms of their
architecture, even though they all share the Transformer as
a common ancestor. Other widely used variants of LLMs in
the Transformer family include RoBERTa, BART (which we
saw earlier performing some text classification), and
ELECTRA.
BERT
BERT (Figure 1.3) is an autoencoding model that uses
attention to build a bidirectional representation of a
sentence. This approach makes it ideal for sentence
classification and token classification tasks.

Figure 1.3 BERT was one of the first LLMs and
continues to be popular for many NLP tasks that involve
fast processing of large amounts of text.
BERT uses the encoder of the Transformer and ignores the
decoder to become exceedingly good at
processing/understanding massive amounts of text very
quickly relative to other, slower LLMs that focus on
generating text one token at a time. BERT-derived
architectures, therefore, are best for working with and
analyzing large corpora quickly when we don’t need to write
free-text.
BERT itself doesn’t classify text or summarize documents,
but it is often used as a pre-trained model for downstream
NLP tasks. BERT has become a widely used and highly
regarded LLM in the NLP community, paving the way for the
development of even more advanced language models.
The GPT Family and ChatGPT

GPT (Figure 1.4), in contrast to BERT, is an autoregressive
model that uses attention to predict the next token in a
sequence based on the previous tokens. The GPT family of
algorithms (which include ChatGPT and GPT-4) is primarily
used for text generation and has been known for its ability
to generate natural-sounding, human-like text.
Figure 1.4 The GPT family of models excels at
generating free-text aligned with the user’s intent.
GPT relies on the decoder portion of the Transformer and
ignores the encoder, so it is exceptionally good at
generating text one token at a time. GPT-based models are
best for generating text given a rather large context
window. They can also be used to process/understand text,
as we will see later in this book. GPT-derived architectures
are ideal for applications that require the ability to freely
write text.
T5

T5 is a pure encoder/decoder Transformer model that was
designed to perform several NLP tasks, from text
classification to text summarization and generation, right off
the shelf. It is one of the first popular models to be able to
boast of such a feat, in fact. Before T5, LLMs like BERT and
GPT-2 generally had to be fine-tuned using labeled data
before they could be relied on to perform such specific
tasks.
T5 uses both the encoder and the decoder of the
Transformer, so it is highly versatile in both processing and
generating text. T5-based models can perform a wide range
of NLP tasks, from text classification to text generation, due
to their ability to build representations of the input text
using the encoder and generate text using the decoder
(Figure 1.5). T5-derived architectures are ideal for
applications that “require both the ability to process and
understand text and the ability to generate text freely.”
Figure 1.5 T5 was one of the first LLMs to show
promise in solving multiple tasks at once without any
fine-tuning.

T5’s ability to perform multiple tasks with no fine-tuning
spurred the development of other versatile LLMs that can
perform multiple tasks with efficiency and accuracy with
little or no fine-tuning. GPT-3, released around the same
time as T5, also boasted this ability but was closed source
and under OpenAI’s control.
More modern open source LLMs like Llama (seen in Figure
1.6) pop up seemingly by the day and represent a wonderful
and massive shift towards a more open and transparent
community of AI. This shift is not without speedbumps,
however. Even Llama – considered one of the most powerful
open-source family of auto-regressive models – is not 100%
open. To download the parameter weights you must agree
to a relatively strict license, and we do not have access to
the training data nor the code they used to make the model.
Figure 1.6 The Llama family of models is considered
one of the more powerful (mostly) open-source families
of LLMs, trained on trillions of tokens and ready to be
fine-tuned for specific tasks.
Nearly all LLMs are highly versatile and are used for various
NLP tasks, such as text classification, text generation,
machine translation, and sentiment analysis, among others.

These LLMs, along with flavors (variants) of them, will be the
main focus of this book and our applications.
Table 1.1 shows the disk size, memory usage, number of
parameters – the internal numbers that make up the
matrices of the deep learning architecture itself, and
approximate size of the pre-training data for several popular
LLMs. Note that these sizes are approximate and may vary
depending on the specific implementation and hardware
used.
Table 1.1 Comparison of Popular Large Language
Models

But size isn’t everything. Let’s look at some of the key
characteristics of LLMs and then dive into how they learn to
read and write.
Key Characteristics of LLMs
The original Transformer architecture, as devised in 2017,
was a sequence-to-sequence model, which means it had
two main components:
An encoder, which is tasked with taking in raw text,
splitting it up into its core components (more on this
later), converting those components into vectors
(similar to the Word2vec process), and using attention
to understand the context of the text
A decoder, which excels at generating text by using a
modified type of attention to predict the next best token
As shown in Figure 1.7, the Transformer has many other
subcomponents (which we won’t get into) that promote
faster training, generalizability, and better performance.
Today’s LLMs are, for the most part, variants of the original
Transformer. Models like BERT and GPT dissect the
Transformer into only an encoder and a decoder
(respectively) so as to build models that excel in
understanding and generating (also respectively).

Figure 1.7 The original Transformer has two main
components: an encoder, which is great at
understanding text, and a decoder, which is great at
generating text. Putting them together makes the entire
model a “sequence-to-sequence” model.
As mentioned earlier, in general, LLMs can be categorized
into three main buckets:
Autoregressive models, such as GPT, which predict
the next token in a sentence based on the previous
tokens. These LLMs are effective at generating coherent
free-text following a given context.
Autoencoding models, such as BERT, which build a
bidirectional representation of a sentence by masking
some of the input tokens and trying to predict them
from the remaining ones. These LLMs are adept at
capturing contextual relationships between tokens
quickly and at scale, which makes them great
candidates for text classification tasks, for example.
Combinations of autoregressive and autoencoding,
such as T5, which can use the encoder and decoder to
be more versatile and flexible in generating text. Such
combination models can generate more diverse and
creative text in different contexts compared to pure
decoder-based autoregressive models due to their
ability to capture additional context using the encoder.
Figure 1.8 shows the breakdown of the key characteristics of
LLMs based on these three buckets.

Figure 1.8 A breakdown of the key characteristics of
LLMs based on how they are derived from the original
Transformer architecture.
More Context, Please
No matter how the LLM is constructed and which parts of
the Transformer it is using, they all care about context
(Figure 1.9). The goal is to understand each token as it
relates to the other tokens in the input text. Since the
introduction of Word2vec around 2013, NLP practitioners
and researchers have been curious about the best ways of
combining semantic meaning (basically, word definitions)
and context (with the surrounding tokens) to create the
most meaningful token embeddings possible. The
Transformer relies on the attention calculation to make this
combination a reality.

Figure 1.9 LLMs are great at understanding context.
The word “Python” can have different meanings
depending on the context. We could be talking about a
snake or a pretty cool coding language.
Choosing what kind of Transformer you want isn’t enough.
Just choosing the encoder doesn’t mean your Transformer
magically becomes good at understanding text. Let’s look at
how these LLMs actually learn to read and write.
How LLMs Work
How an LLM is pre-trained and fine-tuned makes all the
difference between an okay-performing model and a state-
of-the-art, highly accurate LLM. We’ll need to take a quick
look into how LLMs are pre-trained to understand what they
are good at, what they are bad at, and whether we would
need to update them with our own custom data.

Pre-training
Every LLM on the market has been pre-trained on a large
corpus of text data and on specific language modeling-
related tasks. During pre-training, the LLM tries to learn and
understand general language and relationships between
words. Every LLM is trained on different corpora and on
different tasks.
BERT, for example, was originally pre-trained on two publicly
available text corpora (Figure 1.10):
English Wikipedia: a collection of articles from the
English version of Wikipedia, a free online encyclopedia.
It contains a range of topics and writing styles, making it
a diverse and representative sample of English
language text (at the time, 2.5 billion words).
The BookCorpus: a large collection of fiction and
nonfiction books. It was created by scraping book text
from the web and includes a range of genres, from
romance and mystery to science fiction and history. The
books in the corpus were selected to have a minimum
length of 2000 words and to be written in English by
authors with verified identities (approximately 800
million words in total).

Figure 1.10 BERT was originally pre-trained on English
Wikipedia and the BookCorpus. More modern LLMs are
trained on datasets thousands of times larger.
BERT was also pre-trained on two specific language
modeling tasks (Figure 1.11):

Masked Language Modeling (MLM) task (autoencoding
task): helps BERT recognize token interactions within a
single sentence.
Next Sentence Prediction (NSP) task: helps BERT
understand how tokens interact with each other
between sentences.
Figure 1.11 BERT was pre-trained on two tasks: the
autoencoding language modeling task (referred to as
the “masked language modeling” task) to teach it
individual word embeddings and the “next sentence
prediction” task to help it learn to embed entire
sequences of text.
Pre-training on these corpora allowed BERT (mainly via the
self-attention mechanism) to learn a rich set of language
features and contextual relationships. The use of large,
diverse corpora like these has become a common practice
in NLP research, as it has been shown to improve the
performance of models on downstream tasks.
Note

The pre-training process for an LLM can evolve over
time as researchers find better ways of training LLMs
and phase out methods that don’t help as much. For
example, within a year of the original Google BERT
release that used the NSP pre-training task, a BERT
variant called RoBERTa (yes, most of these LLM names
will be fun) by Facebook AI was shown to not require the
NSP task to match and even beat the original BERT
model’s performance in several areas.
BERT, as we now know, is an auto-encoding model so it’s
pretraining will be different than how, say Llama-3 is
pretrained. Instead of MLM and NSP, autoregressive
models are pretrained simply on the auto-regressive
language modeling task over a predefined corpus of data.
Put another way, pretraining models like Llama-3 just mean
that they read vast amounts of unstructured text mostly
from the internet and trained to emulate the language as
closely as possible.
Depending on which LLM you decide to use, it will likely be
pre-trained differently from the rest. This is what sets LLMs
apart from each other. Some LLMs are trained on proprietary
data sources, including OpenAI’s GPT family of models, to
give their parent companies an edge over their competitors.
We won’t revisit the idea of pre-training often in this book
because it’s not exactly the “quick” part of a “quick start
guide.” Nevertheless, it can be worth knowing how these
models were pre-trained because this pre-training enables
us to apply transfer learning, which lets us achieve the
state-of-the-art results we want—which is a big deal!
Transfer Learning

Transfer learning is a technique used in machine learning to
leverage the knowledge gained from one task to improve
performance on another related task. Transfer learning for
LLMs involves taking an LLM that has been pre-trained on
one corpus of text data and then fine-tuning it for a specific
“downstream” task, such as text classification or text
generation, by updating the model’s parameters with task-
specific data.
The idea behind transfer learning is that the pre-trained
model has already learned a lot of information about the
language and relationships between words, and this
information can be used as a starting point to improve
performance on a new task. Transfer learning allows LLMs to
be fine-tuned for specific tasks with much smaller amounts
of task-specific data than would be required if the model
were trained from scratch. This greatly reduces the amount
of time and resources needed to train LLMs. Figure 1.12
provides a visual representation of this relationship.

Figure 1.12 The general transfer learning loop involves
pre-training a model on a generic dataset on some
generic self-supervised task and then fine-tuning the
model on a task-specific dataset.
Fine-Tuning
Once an LLM has been pre-trained, it can be fine-tuned for
specific tasks. Fine-tuning involves training the LLM on a
smaller, task-specific dataset to adjust its parameters for
the specific task at hand. This allows the LLM to leverage its
pre-trained knowledge of the language to improve its
accuracy for the specific task. Fine-tuning has been shown
to drastically improve performance on domain-specific and
task-specific tasks and lets LLMs adapt quickly to a wide
variety of NLP applications.
Figure 1.13 shows the basic fine-tuning loop that we will use
for our models in later chapters. Whether they are open-
source or closed-source, the loop is more or less the same:
1. We define the model we want to fine-tune as well as
any fine-tuning parameters (e.g., learning rate).
2. We aggregate some training data (the format and
other characteristics depend on the model we are
updating).
3. We compute losses (a measure of error) and gradients
(information about how to change the model to
minimize error).
4. We update the model through backpropagation—a
mechanism to update model parameters to minimize
errors.

Figure 1.13 The Transformers package from Hugging
Face provides a neat and clean interface for training and
fine-tuning LLMs.
If some of that went over your head, not to worry: We will
rely on prebuilt tools from Hugging Face’s Transformers
package (Figure 1.9) and OpenAI’s Fine-Tuning API to
abstract away a lot of this so we can really focus on our data
and our models.
Note
You will not need a Hugging Face account or key to
follow along and use any of the code in this book, apart
from the very specific advanced exercises where I will
call it out.
Attention
The title of the original paper that introduced the
Transformer was “Attention Is All You Need.” Attention is a
mechanism used in deep learning models (not just
Transformers) that assigns different weights to different
parts of the input, allowing the model to prioritize and
emphasize the most important information while performing
tasks like translation or summarization. Essentially,
attention allows a model to “focus” on different parts of the
input dynamically, leading to improved performance and
more accurate results. Before the popularization of
attention, most neural networks processed all inputs equally
and the models relied on a fixed representation of the input
to make predictions. Modern LLMs that rely on attention can
dynamically focus on different parts of input sequences,
allowing them to weigh the importance of each part in
making predictions.

To recap, LLMs are pre-trained on large corpora and
sometimes fine-tuned on smaller datasets for specific tasks.
Recall that one of the factors behind the Transformer’s
effectiveness as a language model is that it is highly
parallelizable, allowing for faster training and efficient
processing of text. What really sets the Transformer apart
from other deep learning architectures is its ability to
capture long-range dependencies and relationships between
tokens using attention. In other words, attention is a crucial
component of Transformer-based LLMs, and it enables them
to effectively retain information between training loops and
tasks (i.e., transfer learning), while being able to process
lengthy swatches of text with ease.
Attention is considered the aspect most responsible for
helping LLMs learn (or at least recognize) internal world
models and human-identifiable rules. A Stanford University
study conducted in 2019 showed that certain attention
calculations in BERT corresponded to linguistic notions of
syntax and grammar rules. For example, the researchers
noticed that BERT was able to notice direct objects of verbs,
determiners of nouns, and objects of prepositions with
remarkably high accuracy from only its pre-training. These
relationships are presented visually in Figure 1.14.

Figure 1.14 Research has probed into LLMs and
revealed that they seem to be recognizing grammatical
rules even when they were never explicitly told these
rules.
Other research has explored which other kinds of “rules”
LLMs are able to learn simply by pre-training and fine-
tuning. One example is a series of experiments led by
researchers at Harvard University that explored an LLM’s
ability to learn a set of rules for a synthetic task like the
game of Othello (Figure 1.15). They found evidence that an
LLM was able to understand the rules of the game simply by
training on historical move data.

Figure 1.15 LLMs may be able to learn all kinds of
things about the world, whether it be the rules and
strategy of a game or the rules of human language.
For any LLM to learn any kind of rule, however, it has to
convert what we perceive as text into something machine
readable. This is done through the process of embedding.
Embeddings
Embeddings are the mathematical representations of
words, phrases, or tokens in a large-dimensional space. In

NLP, embeddings are used to represent the words, phrases,
or tokens in a way that captures their semantic meaning
and relationships with other words. Several types of
embeddings are possible, including position embeddings,
which encode the position of a token in a sentence, and
token embeddings, which encode the semantic meaning of
a token (Figure 1.16).
Figure 1.16 An example of how BERT uses three layers
of embedding for a given piece of text. Once the text is
tokenized, each token is given an embedding and then
the values are added up, so each token ends up with an
initial embedding before any attention is calculated. We
won’t focus too much on the individual layers of LLM
embeddings in this text unless they serve a more

practical purpose, but it is good to know about some of
these parts and how they look under the hood.
LLMs learn different embeddings for tokens based on their
pre-training and can further update these embeddings
during fine-tuning.
Tokenization
Tokenization, as mentioned previously, involves breaking
text down into the smallest unit of understanding—tokens.
These tokens are the pieces of information that are
embedded into semantic meaning and act as inputs to the
attention calculations, which leads to . . . well, the LLM
actually learning and working. Tokens make up an LLM’s
static vocabulary and don’t always represent entire words.
For example, tokens can represent punctuation, individual
characters, or even a sub-word if a word is not known to the
LLM. Nearly all LLMs also have special tokens that have
specific meaning to the model. For example, the BERT
model has the special [CLS] token, which BERT
automatically injects as the first token of every input and is
meant to represent an encoded semantic meaning for the
entire input sequence.
Readers may be familiar with techniques like stop-words
removal, stemming, and truncation that are used in
traditional NLP. These techniques are not used, nor are they
necessary, for LLMs. LLMs are designed to handle the
inherent complexity and variability of human language,
including the usage of stop words like “the” and “an,” and
variations in word forms like tenses and misspellings.
Altering the input text to an LLM using these techniques
could potentially harm the model’s performance by reducing
the contextual information and altering the original meaning
of the text.

Tokenization can also involve preprocessing steps like
casing, which refers to the capitalization of the tokens. Two
types of casing are distinguished: uncased and cased. In
uncased tokenization, all the tokens are lowercase, and
usually accents are stripped from letters. In cased
tokenization, the capitalization of the tokens is preserved.
The choice of casing can impact the model’s performance,
as capitalization can provide important information about
the meaning of a token. Figure 1.17 provides an example.
Figure 1.17 The choice of uncased versus cased
tokenization depends on the task. Simple tasks like text
classification usually prefer uncased tokenization,
whereas tasks that derive meaning from case, such as
named entity recognition, prefer a cased tokenization.
Note
Even the concept of casing carries some bias,
depending on the model. To uncase a text—that is, to
implement lowercasing and stripping of accents—is
generally a Western-style preprocessing step. I speak
Turkish, so I know that the umlaut (e.g., the “Ö” in my
last name) matters and can actually help the LLM

understand the word being said in Turkish. Any language
model that has not been sufficiently trained on diverse
corpora may have trouble parsing and utilizing these
bits of context.
Figure 1.18 shows an example of tokenization—namely, how
LLMs tend to handle out-of-vocabulary (OOV) phrases. OOV
phrases are simply phrases/words that the LLM doesn’t
recognize as a token and has to split up into smaller sub-
words. For example, my name (Sinan) is not a token in most
LLMs (the story of my life), so in BERT, the tokenization
scheme will split my name up into two tokens (assuming
uncased tokenization):
Sin: the first part of my name
##an: a special sub-word token that is different from
the word “an” and is used only as a means to split up
unknown words
Figure 1.18 Every LLM has to deal with words it has
never seen before. How an LLM tokenizes text can

matter if we care about the token limit of an LLM. In the
case of BERT, “sub-words” are denoted with a preceding
“##”, indicating they are part of a single word and not
the beginning of a new word. Here the token “##an” is
an entirely different token than the word “an”.
Some LLMs limit the number of tokens we can input at any
one time. How the LLM tokenizes text can matter if we are
trying to be mindful about this limit.
So far, we have talked a lot about language modeling—
predicting missing/next tokens in a phrase. However,
modern LLMs can also borrow from other fields of AI to make
their models more performant and, more importantly, more
aligned—meaning that the AI is performing in accordance
with a human’s expectation. Put another way, an aligned
LLM has an objective that matches a human’s objective.
Beyond Language Modeling: Alignment + RLHF
Alignment in language models refers to how well the
model can respond to input prompts that match the user’s
expectations. Standard language models predict the next
word based on the preceding context, but this can limit their
usefulness for specific instructions or prompts. Researchers
are coming up with scalable and performant ways of
aligning language models to a user’s intent. One such broad
method of aligning language models is through the
incorporation of reinforcement learning (RL) into the training
loop. Modern models are even being released in their pre-
alignment and post-alignment form. Figure 1.19 shows
Llama-2’s non aligned and aligned version answering the
same question. The difference is quite stark

Random documents with unrelated
content Scribd suggests to you:

in two places,—here are the marks. I bled a good deal. The other
lodgers didn’t like to hit him for it, on account of his having no legs,
but they kicked him out of the house, and would not let him back
any more. They all wanted me to lock him up, but I wouldn’t, as he
was an old pal. Two or three silk handkerchiefs was tied round my
leg, and the next day I was took to St. Thomas’s Hospital, where I
remained for about nine days. When I left the head-nurse gave me
ten shillings on account of being so destitute—for I was without a
ha’penny to call my own. As soon as I got out of the hospital I went
down to Billingsgate, and bought some bread and pickled whelks at
a stall, but when I pulled out my money to pay for ’em some
costermongering chaps knocked me down, and robbed me of 5s. I
was completely stunned by the blow. The police came up to see
what was the matter, and took me to the station-house, where I
stopped till the next morning, when the inspector made me tell
where my father lived, and I was taken home to him. For about a
month my father kept me under lock and key, and after I had been
with him about three months more I ‘stept it’ again, and as I could
always whistle very well, I thought I’d try it for a living; so I made a
‘pitch’ in New-street, Covent Garden, and began by whistling ‘Will
you love me then as now?’ but there wasn’t many in the world as
loved me. I did very well though that day, for I got about 3s. 6d. or
4s., so I thought I’d practise it and stick to it. I worked all about
town till I got well known. I used, sometimes, to go into public-
houses and whistle upon a piece of ’bacco pipe, blowing into the
bowl, and moving my fingers as if I was playing a flute, and nobody
could tell the difference if they had not seen me. Sometimes I used
to be asked to stand outside hotels, taverns, and even club-houses,
and give ’em a tune: I often had sixpences, shillings, and half-
crowns thrown me. I only wish I had sich luck now, for the world’s
topsy-turvy, and I can’t get hardly anything. I used then to earn 3s.
or 4s. a-day, and now it don’t amount to more than 1s. 6d.
“After I’d worked London pretty well, I sometimes would start off a
few miles out to the towns and villages; but, generally, it wasn’t
much account. The country chaps like sich tunes as ‘The Barley

Stack,’ or ‘The Little House under the Hill.’ I often used to whistle to
them while they danced. They liked jigs mostly, and always paid me
a penny a dance each.
“I recollect once when I was whistling before a gentleman’s house
down at Hounslow, he sent his servant and called me in. I was taken
into a fine large room, full of looking-glasses, and time-pieces, and
pictures. I was never in sich a room before, all my life. The
gentleman was there with his family,—about six on ’em,—and he
told me if I’d whistle, and learn his birds to sing, he’d give me a
sovereign. He had three fine brass-wire cages, with a bird in each,
slung all of a row from the ceiling. I set to work ‘like a brick,’ and the
birds begun to sing directly, and I amused ’em very much. I stopped
about an hour and a half, and let ’em have all sorts of tunes, and
then he gave me a sovereign, and told me to call again when I come
that way; but before I left he said the servants was to give me
something to eat and drink, so I had dinner in the kitchen with the
servants, and a jolly good dinner it was.
“From Hounslow I walked to Maidenhead, and took a lodging for the
night at the Turk’s Head. In the evening some countrymen come into
the tap-room and kicked up a row with the missus because she
couldn’t lodge ’em. She run in to turn then away, when three of ’em
pitched into her right and left; and if it hadn’t been for me and
another chap she’d have got killed. When they got her down I
jumped upon the table and snatched up the only weapon I could
find, a brass candlestick, and knocked one of ’em down senseless,
and the other fellow got hold of a broomstick and give it ’em as hard
as he could, till we beat ’em right out of the place. There happened
to be some police outside, drilling, who came over and took three of
them to the stocks, where they was locked in for twenty-four hours.
The next day the magistrate sentenced ’em to three months’
imprisonment each, and I started for London and never whistled a
tune till I reached it, which was three days afterwards. I kept on at
the old game, earning about 2s. 6d. a-day, till the militia was being
called out, and then I joined them, for I thought it would be the best
thing I could do. I was sworn in by Colonel Scrivens at Eton Mews.

We was taken into a stable, where there was three horses. Four of
us laid hold of a book altogether; and then, after asking us if we had
any complaints, or were lame, or any way unfit for service, or was
married, or had any children; and when we had said No, he asked us
if we was free, able, and willing to serve in her Majesty’s militia, in
either England, Ireland, Scotland, or Wales, for the term of five
years, if so long her Majesty required our services; and when we
said we was, we took the oath and kissed the book.
“The same day, which was the 11th of June, 1854, we was packed
off from the Waterloo Station for Portsmouth. After being drilled for
three weeks I was returned for duty, and went on guard. The first
guard I mounted was at Detached Dock at Portsmouth—it’s where
the convicts are. I didn’t do any whistling there, I can tell yer; I’d
different sorts of work, for part of our duty was to bury the poor
fellows that died after coming home invalided from the Crimea. The
people through that used to call us the ‘garrison undertakers.’ I was
there thirteen months, and never, the whole time, had more than
two nights’ bed a-week; and some part of the time the weather was
very frosty, and we was often over our ankles in snow. I belonged to
the 4th Middlesex, and no corps ever did so much duty, or went
through so much hardships, as ours. From Portsmouth I was
ordered, with my regiment, 950 strong, to Buttervant, county Cork,
Ireland. When we reached the Irish Channel a storm arose, and we
was all fastened under hatches, and not suffered to come upon deck
for four days, by which time we reached the Cove of Cork: the
Colonel’s horse had to be thrown overboard, and they, more than
once, had serious thoughts of throwing all the luggage into the sea
as well. I was ten months in Ireland. I didn’t do any whistling there;
and then the regiment was ordered home again on account of the
peace. But before we left we had a day’s sport, consisting of greasy-
pole climbing, jumping in sacks, racing after a pig with a greasy tail,
and all them sort of things; and at night the officers had a grand
ball. We landed at Portsmouth on a Monday morning at four o’clock,
and marched through to the station, and reached Hounslow about
four o’clock the same afternoon. A month after we were

disembodied, and I came at once to London. I had about 1l. 5s. in
my pocket, and I resolved in my own mind never to go whistling any
more. I went to my father, but he refused to help me in any way. I
tried for work, but couldn’t get any, for the people said, they didn’t
like a militia man; so, after having spent all my money, I found that I
must either starve or whistle, and so, you see, I’m once more on the
streets.
“While I was in Ireland I absented myself from the barracks for
twenty-one days, but fearing that a picket would get hold of me, I
walked in one morning at six o’clock. I was instantly placed under
arrest in the guard-room, where I remained four days, when I was
taken before the Colonel, and to my great surprise I saw, sitting
aside of him, the very gentleman who had given me the pound to
whistle to his birds; his name was Colonel Bagot, as I found out
afterwards, and he was deputy-magistrate for Middlesex. He asked
me if I was not the chap as had been to his house; I told him I was,
so he got me off with a good reprimand, and saved me being tried
by a court-martial. When I first took to sleeping at lodging-houses
they was very different to what they are now. I’ve seen as many as
eighteen people in one cellar sleeping upon loose straw, covered
with sheets or blankets, and as many as three in one bed; but now
they won’t take in any little boys like as I was, unless they are with
their parents; and there’s very few beds in a room, and never more
than one in a bed. Married people have a place always parted off for
themselves. The inspector comes in all times—often in the middle of
the night—to see that the regulations ain’t broken.
“I used, one time, to meet another man whistling, but like old Dick,
who was the first at the profession, he’s gone dead, and so I’m the
only one at it now anywhere. It’s very tiring work, and makes you
precious hungry when you keep at it for two or three hours; and I
only wish I could get something else to do, and you’d see how soon
I’d drop it.
“The tunes that are liked best in the streets is sich as ‘Ben Bolt’ and
‘Will you love me then as now?’ but a year or two ago, nothin’ went

down like the ‘Low-back Car.’ I was always being asked for it. I soon
gets hold of the new tunes that comes up. I don’t think whistling
hurts me, because I don’t blow so hard as ‘old Dick’ used. A
gentleman come up to me once in the street that was a doctor, and
asked me whether I drunk much, and whether I drawed my breath
in or blowed it out. I told him I couldn’t get much to drink, and he
said I ought at least to have three half-pints of beer a-day, or else I
should go into a consumption; and when I said I mostly blowed out
when I whistled, he said that was the best, because it didn’t strain
the lungs so much.”
Whistling and Dancing Boy.
At the present time there is only one English boy going about the
streets of London dancing, and at the same time playing his own
musical accompaniment on a tin whistle. There are two or three
Italian boys who dance whilst they perform on either the flute or the
hurdy-gurdy, but the lad who gave me the following statement
assured me that he was the only Englishman who had made street
whistling and dancing “his profession.”
He was a red-headed lad, of that peculiar white complexion which
accompanies hair of that colour. His forehead was covered with
freckles, so thick, that they looked as if a quantity of cayenne pepper
had been sprinkled over it; and when he frowned, his hair moved
backwards and forwards like the twitching of a horse shaking off
flies.
“I’ve put some ile on my hair, to make me look tidy,” he said. The
grease had turned his locks to a fiery crimson colour, and as he
passed his hands through it, and tossed it backwards, it positively
glittered with the fat upon it.
The lad soon grew communicative enough, and proceeded to show
me a blue jacket which he had bought that morning for a shilling,

and explained to me at the same time how artful he had been over
the bargain, for the boy had asked eighteenpence.
I remarked that his shoes seemed in a bad state, for they were
really as white as a baker’s slippers from want of blacking, and the
toe of one gaped like the opening to a tortoise-shell. He explained to
me that he wore all his boots out dancing, doing the double shuffle.
“Now these ’ere shoes,” he said, “cost me a shilling in Petticoat-lane
not a week since, and looked as good as new then, and even now,
with a little mending, they’ll make a tidy pair of crab-shells again.”
To give force to this remark, he lifted his leg up, but, despite his
explanation, I could not see how the leather could possibly be
repaired.
He went through his dances for me, at the same time accompanying
himself on his penny whistle. He took his shoes off and did a
hornpipe, thumping his feet upon the floor the while, like palms on a
panel, so that I felt nervous lest there should be a pin in the carpet
and he be lamed by it.
The boy seemed to have no notion of his age, for although he
accounted for twenty-two years of existence, yet he insisted he was
only seventeen “come two months.” I was sorry to find, moreover,
that he was in the habit of drinking, seldom going home after his
night’s work without being intoxicated; and, indeed, his thin body
and pinched face bore evidence of his excess in this respect, though,
but for his assertion that “he was never hungry, and food was no
good to him,” I should have imagined, at the first glance, that he
was pining with want.
He seems to be among the more fortunate of those who earn their
living in the streets, for although I questioned and cross-questioned
him in every possible way, he still clung to his assertion that he
made 2l. per week. His clothes, however, bore no evidence of his
prosperity, for his outer garment was a washed-out linen blouse,
such as glaziers wear, whilst his trousers were of coarse canvas, and
as black on the thighs as the centre of a drum-head.

He brought with him a penny whistle to show me his musical talents,
and, certainly, his execution of the tin instrument was rapid and
certain.
The following is the statement he gave me:—
“Whistling Billy.
That’s my name, and I’m known all round about in the Borough as
‘Whistling Billy,’ though some, to be sure, calls me ‘Whistling Bill,’ but
in general I’m ‘Billy.’ I’m not looking very respectable now, but you
should see me when I’m going to the play; I looks so uncommon
respectable, nobody knows me again. I shall go to the theatre next
week, and I should just like you to see me. It’s surprising.
“I ain’t a very fat chap, am I? but I’m just meaty enough for my
perfession, which is whistling and dancing in public-houses, where I
gives ’em the hornpipe and the bandy jig, that’s dancing with my
toes turned in.
“My father was a barber. He only charged a penny for shaving, but
he wouldn’t cut your hair under twopence, and he used to do well—
very well sometimes; I don’t know whether he’s alive now, for I ain’t
seen him these ten years, nor asked him for a halfpenny. Mother
was alive when I left, and so was my two brothers. I don’t know
whether they’re alive now. No, I don’t want to go and see him, for I
can get my own living. He used to keep a shop near Fitzroy-square.
“I was always fond of dancing, and I runned away from home for to
follow it. I don’t know my own age exactly: I was as tall then as I
am now. I was twelve when I left home, and it must be ten years
ago, but I ain’t twenty-two: oh, dear no! Why, I ain’t got no whiskers
nor things. I drink such a lot of beer and stuff, that I can’t grow no
taller; gentlemen at the public-houses gives it me. Why, this morning
I was near tipsy, dancing to some coalheavers, who gave me drink.

“I used, when I was at father’s, to go to a ball, and that’s where I
learned to dance. It was a shilling ball in the New-road, where there
was ladies, regular nice ones, beautifully dressed. They used to see
me dancing, and say, when I growed up I should make a beautiful
dancer; and so I do, for I’d dance against anybody, and play the
whistle all the time. The ladies at these balls would give me money
then for dancing before them. Ah! I’d get my entrance shilling back,
and four or five into the bargain. I’d generally take it home to
mother, after buying a little sweet-stuff, or such-like, and I think
that’s why mother would let me go, ’cos I picked up a good bit of
money.
“It was another boy that put me up to running away from home. He
axed me to go along with him, and I went. I dare say it troubled
father a bit when he found I’d gone. I ain’t troubled him for ten
years now. If I was to go back to him, he’d only send me to work,
and I make a better living by myself. I don’t like work, and, to tell
you the truth, I never did work, for it’s like amusement to me to
dance; and it must be an amusement, ’cos it amuses the people,
and that’s why I gets on so well.
“When I hooked it with that chap, we went to Croydon, in Surrey.
We went to a lodging-house where there was men and women, and
boys and chaps, and all like that; we all slept in one room. I had no
money with me, only my clothes; there was a very nice velvet cap;
and I looked very different then to what I do now. This young chap
had some tin, and he kept me. I don’t know how he earned his tin,
for he’d only go out at night time, and then he’d come home and
bring in money, and meat and bread, and such-like. He said to me,
before I went pals with him, that he’d keep me, and that he’d make
plenty of money. He told me he wanted a chum to mate with, so I
went with him right off. I can’t say what he was. He was about
thirteen or fourteen, and I never seed him do no work. He might
have been a prig for all I knows.
“After I’d been in the lodging-house, this chap bought a stock of
combs and cheap jewels, and then we went out together, and he’d

knock at the houses and offer the things for sale, and I’d stand by.
There’s a lot of gentlemen’s houses, if you recollect, sir, round
Croydon, on the London road. Sometimes the servants would give us
grub instead of money. We had plenty to eat. Now you comes to
speak of it, I do remember he used to bring back some old silver
with him, such as old tablespoons or ladles, broke up into bits, and
he’d make a deal selling them. I think he must have been a prig. At
night we used to go to the public-houses and dance. He never
danced, but sit down and looked on. He said he was my relation,
and I always shared my drink with him, and the people would say,
‘Feed me, feed my dog,’ seeing me going halves with him.
“I kept along with him for three years, he working in the day, and I
at night, dancing. We parted at Plymouth, and I took up with
another mate, and worked on to Exeter. I think my new mate was a
regular prig, for it was through his putting me up to prigging that I
got into trouble there. This chap put me on to taking a brass cock
from a foundry. It was in a big wooden butt, with 150 gallons of
water in it. I got over a gate and pulled it out, and set all the
foundry afloat. We cut away, but two hours afterwards the
policeman come to the lodging-house, and though there was a lot of
boys and girls, he picked me out, and I had two months for it, and
all my hair was cut off, and I only had dry bread and gruel every
day, and soup twice a-week. I was jolly sorry for that cock business
when I was caught, and I made up my mind never to take nothing
more. It’s going to the lodging-houses puts fellows up to prigging.
The chaps brings in legs of beef, and puddens, and clothes, and
then they sells ’em cheap. You can sometimes buy a pair of
breeches, worth ten shillings, for two bob, for the chaps don’t like to
take ’em to sell at the shops, and would sooner sell ’em for a’most
nothing rather than be found out.
“When I came out of quod I had a shilling give me, and I went and
bought a penny whistle. I was always fond of music and dancing,
and I know’d a little of playing the whistle. Mother and father was
both uncommon fond of dancing and music, and used to go out
dancing and to concerts, near every night pretty well, after they’d

locked the shop up. I made about eleven bob the first week I was
out, for I was doing very well of a night, though I had no hair on my
head. I didn’t do no dancing, but I knew about six tunes, such as
‘Rory O’More,’ and ‘The Girl I left behind me,’ two hornpipes, (the
Fishers’ and the Sailors’) ‘St. Patrick’s Day,’ and ‘The Shells of the
Ocean,’ a new song as had just come up. I can play fifty tunes now.
Whistles weren’t so common then, they weren’t out a quarter so
much as now. Swinden had the making of them then, but he weren’t
the first maker of them. Clarke is the largest manufactory of them
now, and he followed Swinden. People was astonished at seeing a
tune played on a tin whistle, and gave pretty liberal. I believe I was
the first as ever got a living on a tin whistle. Now there’s more. It
was at that time as I took to selling whistles. I carried ’em on a tin
tray before me, and a lid used to shut on it, fixed. I’d pitch before a
hotel amongst the gentlemen, and I’d get 2d. a-piece for the
whistles, and some would give me sixpence or a shilling, just
according. The young gents was them as bought most, and then
they’d begin playing on them, and afterwards give them to the
young ladies passing. They was very pleased with me, for I was so
little, and I done well. The first two months I made about 17s. or
18s. a-week, but after that they got rather dull, so I gived up selling
of them and took to dancing. It didn’t pay me so well as the
whistles, for it was pretty near all profit on them—they only cost me
3d. a-dozen. I travelled all round Devonshire, and down to Land’s
End, in Cornwall—320 miles from London, and kept on playing the
whistle on the road. I knew all about them parts. I generally pitched
before the hotels and the spirit-shops, and began whistling and
dancing; but sometimes I’d give the cottagers a turn, and they’d
generally hand over a ha’penny a-piece and some bread.
“I stopped travelling about the south of England, and playing and
dancing, for a little better than four years and a half. I didn’t do so
well in winter as in summer. Harvest time was my best time. I’d go
to the fields where they was working, and play and dance.
Sometimes the master would hollar out, ‘Here, you get out of this!’
but the men would speak up for me and say, ‘Let him stop, master.’

Many a chap’s got the sack through me, by leaving off his work and
beginning to dance. Sometimes, when the last load of hay was going
home (you see, that’s always considered the jolliest part of the
work), they’d make me get up to the top of the load, and then
whistle to them. They was all merry—as merry as could be, and
would follow after dancing about, men, women, and boys. I
generally played at the harvest suppers, and the farmer himself
would give me 4s. 6d. or 5s. the night, besides my quart of ale.
Then I’d pick up my 6s. or 7s. in ha’pence among the men. I’ve had
as many as two harvest suppers a week for three weeks or a month
following, for I used to ax the people round what time they was
going to have a supper, and where, and set off, walking nine or ten
miles to reach the farm, and after that we find another spot.
“It’s very jolly among farm people. They give you plenty of cider and
ale. I’ve drunk the cider hot, whilst they was brewing it—new cider,
you know. You never want food neither, for there’s more than you
can eat, generally bread and cheese, or maybe a little cold biled
pork. At night, the men and women used to sleep in a kind of barn,
among the clean straw; and after the beer-shops had closed—they
are all little beer-shops, 3d. a quart in your own jugs, and like that—
they’d say to me, ‘Come up to the doss and give us a tune,’ and
they’d come outside and dance in the open air, for they wouldn’t let
them have no candles nor matches. Then they’d make theirselves
happy, and I’d play to ’em, and they’d club up and give me money,
sometimes as much as 7s., but I’ve never had no higher than that,
but never no less than 3s. One man used to take all the money for
me, and I’d give him a pot o’ ale in the morning. It was a penny a
dance for each of ’em as danced, and each stand-up took a quarter
of a hour, and there was generally two hours of it; that makes about
seven dances, allowing for resting. I’ve had as many as forty
dancing at a time, and sometimes there was only nine of ’em. I’ve
seen all the men get up together and dance a hornpipe, and the
women look on. They always did a hornpipe or a country dance. You
see, some of ’em would sit down and drink during the dance, but it
amounted to almost three dances each person, and generally there

was about fifty present. Usually the men would pay for the women,
but if they was hard up and been free with their money, the girls
would pay for them. They was mostly Irish, and I had to do jigs for
them, instead of a hornpipe. My country dance was to the tune ‘Oh
don’t you tease me, pretty little dear.’ Any fiddler knows that air. It’s
always played in the country for country dances. First they dances to
each other, and then it’s hands across, and then down the middle,
and then it’s back again and turn. That’s the country dance, sir. I
used to be regular tired after two hours. They’d stick me up on a
box, or a tub, or else they’d make a pile of straw, and stick me a-top
of it; or if there was any carts standing by loaded with hay, and the
horses out, I was told to mount that. There was very little drinking
all this time, because the beer-shops was shut up. Perhaps there
might be such a thing as a pint of beer between a man and his
partner, which he’d brought in a can along with him. They only
danced when it was moonlight. It never cost me nothing for lodgings
all the harvest times, for they would make me stop in the barn along
with them; and they was very good company, and took especial care
of me. You mustn’t think this dancing took place every night, but
only three or four nights a-week. I find ’em out travelling along the
road. Sometimes they’ve sent a man from one farm-house to
bespeak me whilst I was playing at another. There was a man as
played on the clarionet as used to be a favourite among haymakers,
but they prefer the penny tin whistle, because it makes more noise,
and is shriller, and is easier heard; besides, I’m very rapid with my
fingers, and makes ’em keep on dancing till they are tired out.
Please God, I’ll be down among them again this summer. I goes
down regular. Last year and the year before, and ever since I can
recollect.
“When I’m in London I make a good living at dancing and playing,
for I’m the only one that plays the whistle and dances at the same
time. I’m reckoned the best hand at it of any man in town or
country. I’ve often been backed by the company to dance and play
against another man, and I generally win. I’ve been in hotels, and
danced to gentlemen, and made plenty of money at it. I do all

manner of tricks, just to make ’em laugh—capering, or ‘hanky-
panky,’ as I term it. I once had half-a-sovereign given to me, but I
think it was a mistake, for he says, ‘Take that, and go on.’ I went
home to clean myself, and had my trousers washed, and my shoes
blacked, and went half-price to the theatre—the ‘Wic,’ I think it was
—and paid my shilling, and went in as tidy as a gentleman.
“When I first go into a public-house I go into the tap-room, and say,
‘Would you like to hear a tune, gentlemen, or see a dance, or a little
bit of amusement?’ If they say ‘No,’ I stand still, and begin a talking,
to make ’em laugh. I’m not to be choked off easy. I say, ‘Come,
gentlemen, can’t you help a poor fellow as is the best dancer in
England? I must have some pudden for breakfast, because I ain’t
had nothing for three weeks.’ Then some say, ‘Well, I will see the
best dancer in England; I’ve got a mag.’ Then after dancing I go to
the gentleman who has given me most, and ask him six or seven
times ‘to give me a copper,’ declaring he’s the only one as has given
me nothing, and that makes the others laugh. I also ask the landlord
to give a half-pint of beer to grease my feet, and that makes ’em
merry. I generally gets good nobbings (that’s a collection, you
know). They likes the dancing better than music; but it’s doing them
together that takes. I ax them if they’ll have the hornpipe or the
Irish jig, and if they says the jig I do it with my toes turned in, like
as if I was bandy; and that’s very popular. I have been to as many
as forty public-houses in a evening, and dance inside; or if they
won’t let me come in, they’ll say, ‘Dance outside as much as you
like,’ and that’s very near as good for me. If I gets inside, I’ll mop up
1s. if it’s good company, or perhaps 3d. or 4d., and always plenty to
drink—more than I can take, for I’m generally drunk before I can get
home. They never gives me nothing to eat, but it don’t matter, for
I’m seldom hungry; but ‘I like a drop of good beer,’ as the song says.
“I’ve been engaged at concert-rooms to dance. I have pumps put
on, and light trousers, and a Guernsey, dressed up as a sailor. That
was in the country, at Canterbury, and I had 7s. and plenty to eat
and drink. I’ve never appeared at a London concert-room, though

I’ve been axed to come in and amuse the company; but I wasn’t tidy
enough, and didn’t like.
“When I dance in a public-house I take my shoes off and say, ‘Now,
gentlemen, watch my steps.’ For the hornpipe I begin with walking
round, or ‘twisting’ as the term is; then I stands up, and does a
double-shuffle—or the ‘straight fives’ as we calls it; then I walk
round again before doing the back-snatches, another kind of double-
shuffle. Then I does the rocks of Scilly, that’s when you twists your
feet and bends sideways; next comes the double steps and rattles,
that is, when the heels makes a rattle coming down; and I finishes
with the square step. My next step is to walk round and collect the
money. The Irish like to see me do the jig better than the hornpipe.
Them two are the only dances I know.
“I make regular 2l. a-week. Yesterday I made 7s. 3d., and it was
rainy, so I couldn’t get out till late. At Brighton Regatta I and my
mate made 5l. 10s. between us, and at Dover Regatta we made 8l.
between us. We squandered 2l. 10s. at the lodging-house in one
night, betting and tossing, and playing at cards. We always follows
up the regatta. We made only 2l. 10s. at Hastings Regatta. You see
we pick up on a Saturday night our 11s. a-piece, and on other days
perhaps 5s. or 8s., according to the day.
“I used to go about with a mate who had a wooden leg. He was a
beautiful dancer, for he made ’em all laugh. He’s a little chap, and
only does the hornpipe, and he’s uncommon active, and knocks his
leg against the railings, and makes the people grin. He was very
successful at Brighton, because he was pitied.
“I’ve also been about with a school of tumblers. I used to do the
dancing between the posturing and likes of that. I’ve learnt
tumbling, and I was cricked for the purpose, to teach me. I couldn’t
walk for three days. They put my legs round my neck, and then
couldn’t get ’em back again. I was in that state, regular doubled up,
for two hours, and thought I was done for. Some of my mates said,
‘There, you’ve been and spoiled that chap.’ It’s dreadful painful
learning tumbling. When I was out with the posturers I used to play

the drum and mouth-pipes; I had a old hat and coat on. Then when
my turn come, I’d appear in my professional costume, and a young
chap who was a fluter—not a whistler, like me,—would give a tune,
and I’d go on the carpet and give the Irish jig or the hornpipe.
“There was four of us in the school, and we’d share a pound a-week
each. We were down at Dover there, and put up at the Jolly Sailors.
I left them there, and went alone on to the camp where the German
Legion was—at Shorncliffe, that’s the place. I stopped there for three
weeks, and did very well, taking my 7s. or 8s. a-day.
“After that I got tired of dancing, and thought I’d like a change, so I
went out on a fishing-boat. They didn’t give me nothing a-week,
only 4s. when we come home after two months, and your clothes,
and victuals aboard. We first went fishing for plaice, and soles, and
turbots, and we’d land them at Yarmouth, and they’d send them on
to Lowestoft, and from there on to London. Then we went codding
off the coast of Holland, for cod and haddock. It was just drawing on
winter, and very cold. They set me with a line and I had to keep
sawing it backwards and forwards till I felt a fish bite, then to hawl it
up. One night I was a near froze, and suddenly I had two cods bite
at once, and they nearly pulled me over, for they dart about like
mad, and tug awful; so I said to the master, ‘I don’t like this work.’
But he answers, ‘You must like it the time you stops here.’ So I made
up my mind to bolt the first time I got to shore. I only did it as a
change, to see if I liked it. You’re right there, there ain’t no drinking
on board.
“When you hawl up a cod they bound about the deck, and they’re as
strong as a Scotch terrier dog. When we hold ’em down, we prick
them under the fin, to let the wind out of them. It would choke them
if we didn’t let it out, for it hisses as it comes off. It’s from dragging
them up so quick out of fifteen-fathom water that gives ’em the
wind. When they were pricked, we chucked them into the well in the
hold, and let them swim about. We killed them when we got to
Gravesend by hitting them on the head with tom-boys—the sticks we

hauls the line through. After three or four blows they’re stunned,
and the blood comes, and they’re killed.
“When I goes into the public-houses, part of my performance is to
play the whistle up my nose. I don’t do it in the streets, because if I
did there’d be thousands looking at me, and then the police would
make a row. Last night I did it. I only pitched at one place, and did
my night’s work right off. I took 4s. 3½d. and lots of beer in an hour,
from the cabbies and the people and all. At last the police told me to
move on. When I plays the whistle up my nose, I puts the end of it
in my nostril, and blows down it. I can do that just as easy as with
my mouth, only not as loud. I do it as a variety, first in my mouth,
then in my nose, and then back again in my mouth. It makes the
people laugh. I’ve got a cold now, so I can’t do it so well as at times,
but I’ll let you see what it is like.”
He then inserted the wooden tongue of the whistle into his nostril,
and blowing down it, began a hornpipe, which, although not so shrill
as when he played it with the mouth, was still loud enough to be
heard all over the house.

IV.—STREET ARTISTS.
I now come to the Street Artists. These include the artists in
coloured chalks on the pavements, the black profile-cutters, and
others.
Street Photography.
Within the last few years photographic portraits have gradually been
diminishing in price, until at the present time they have become a
regular article of street commerce. Those living at the west-end of
London have but little idea of the number of persons who gain a
livelihood by street photography.
There may be one or two “galleries” in the New-road, or in
Tottenham-court-road, but these supply mostly shilling portraits. In
the eastern and southern districts of London, however, such as in
Bermondsey, the New-cut, and the Whitechapel-road, one cannot
walk fifty yards without passing some photographic establishment,
where for sixpence persons can have their portrait taken, and
framed and glazed as well.
It was in Bermondsey that I met with the first instance of what may
be called pure street photography. Here a Mr. F——l was taking
sixpenny portraits in a booth built up out of old canvas, and erected
on a piece of spare ground in a furniture-broker’s yard.
Mr. F——l had been a travelling showman, but finding that
photography was attracting more attention than giants and dwarfs,
he relinquished the wonders of Nature for those of Science.

Into this yard he had driven his yellow caravan, where it stood like
an enormous Noah’s ark, and in front of the caravan (by means of
clothes-horses and posts, over which were spread out the large sail-
like paintings (show-cloths), which were used at fairs to decorate the
fronts of booths), he had erected his operating-room, which is about
as long and as broad as a knife-house, and only just tall enough to
allow a not particularly tall customer to stand up with his hat off:
whilst by means of two window-sashes a glazed roof had been
arranged for letting light into this little tent.
On the day of my visit Mr. F——l was, despite the cloudy state of the
atmosphere, doing a large business. A crowd in front of his tent was
admiring the photographic specimens, which, of all sizes and in all
kinds of frames, were stuck up against the canvas-wall, as irregularly
as if a bill-sticker had placed them there. Others were gazing up at
the chalky-looking paintings over the door-way, and on which a lady
was represented photographing an officer, in the full costume of the
11th Hussars.
Inside the operating room we found a crowd of women and children
was assembled, all of them waiting their turn to be taken. Mr. F——l
remarked, as I entered, that ‘It was wonderful the sight of children
that had been took;’ and he added, ‘when one girl comes for her
portrait, there’s a dozen comes along with her to see it took.’
The portraits I discovered were taken by Mrs. F——l, who, with the
sleeves of her dress tucked up to the elbows, was engaged at the
moment of my visit in pointing the camera at a lady and her little
boy, who, from his wild nervous expression, seemed to have an idea
that the operatress was taking her aim previous to shooting him. Mr.
F——l explained to me the reason why his wife officiated. “You see,”
said he, “people prefers more to be took by a woman than by a
man. Many’s a time a lady tells us to send that man away, and let
the missis come. It’s quite natural,” he continued; “for a lady don’t
mind taking her bonnet off and tucking up her hair, or sticking a pin
in here and there before one of her own sect, which before a man
proves objectionable.”

After the portrait had been taken I found that the little square piece
of glass on which it was impressed was scarcely larger than a visiting
card, and this being handed over to a youth, was carried into the
caravan at the back, where the process was completed. I was invited
to follow the lad to the dwelling on wheels.
The outside of the caravan was very remarkable, and of that peculiar
class of architecture which is a mixture of coach-and-ship building.
In the centre of the front of the show were little folding-doors with
miniature brass knockers, and glass let into the upper panels. On
each side of the door were long windows, almost big enough for a
shop-front, whilst the white curtains, festooned at their sides, gave
them a pleasant appearance. The entire erection was coloured
yellow, and the numerous little wooden joists and tie-beams, which
framed and strengthened the vehicle, conferred upon it a singular
plaid-like appearance.
I mounted the broad step-ladder and entered. The room reminded
me of a ship’s cabin, for it was panelled and had cross-beams to the
arched roof, whilst the bolts and fastenings were of bright brass. If
the windows had not been so large, or the roof so high, it would
have resembled the fore-cabin of a Gravesend steamer. There were
tables and chairs, as in an ordinary cottage room. At one end was
the family bed, concealed during the day by chintz curtains, which
hung down like a drop-scene before a miniature theatre; and
between the openings of these curtains I could catch sight of some
gaudily attired wax figures stowed away there for want of room, but
standing there like a group of actors behind the scenes.
Along one of the beams a blunderbuss and a pistol rested on hooks,
and the showman’s speaking trumpet (as large as the funnel to a
grocer’s coffee-mill) hung against the wall, whilst in one corner was
a kind of cabin stove of polished brass, before which a boy was
drying some of the portraits that had been recently taken.
“So you’ve took him at last,” said the proprietor, who accompanied
us as he snatched the portrait from the boy’s hand. “Well, the eyes
ain’t no great things, but as it’s the third attempt it must do.”

On inspecting the portrait I found it to be one of those drab-looking
portraits with a light back-ground, where the figure rises from the
bottom of the plate as straight as a post, and is in the cramped,
nervous attitude of a patient in a dentist’s chair.
After a time I left Mr. F——l’s, and went to another establishment
close by, which had originally formed part of a shop in the penny-
ice-and-bull’s-eye line—for the name-board over “Photographic
Depôt” was still the property of the confectioner—so that the
portraits displayed in the window were surmounted by an
announcement of “Ginger beer 1d. and 2d.”
A touter at the door was crying out “Hi! hi!—walk inside! walk inside!
and have your c’rect likeness took, frame and glass complete, and
only 6d.!—time of sitting only four seconds!”
A rough-looking, red-faced tanner, who had been staring at some
coloured French lithographs which decorated the upper panes, and
who, no doubt, imagined that they had been taken by the
photographic process, entered, saying, “Let me have my likeness
took.”
The touter instantly called out, “Here, a shilling likeness for this here
gent.”
The tanner observed that he wanted only a sixpenny.
“Ah, very good, sir!” and raising his voice, the touter shouted louder
than before—“A sixpenny one first, and a shilling one afterwards.”
“I tell yer I don’t want only sixpennorth,” angrily returned the
customer, as he entered.
At this establishment the portraits were taken in a little alley
adjoining the premises, where the light was so insufficient, that even
the blanket hung up at the end of it looked black from the deep
shadows cast by the walls.
When the tanner’s portrait was completed it was nearly black; and,
indeed, the only thing visible was a slight light on one side of the
face, and which, doubtlessly, accounted for the short speech which

the operator thought fit to make as he presented the likeness to his
customer.
“There,” he said, “there is your likeness, if you like! look at it
yourself; and only eightpence”—“Only sixpence,” observed the man.
—“Ah!” continued the proprietor, “but you’ve got a patent American
preserver, and that’s twopence more.”
Then followed a discussion, in which the artist insisted that he lost
by every sixpenny portrait he took, and the tanner as strongly
protesting that he couldn’t believe that, for they must get some
profit any how. “You don’t tumble to the rig,” said the artist; “it’s the
half-guinea ones, you see, that pays us.”
The touter, finding that this discussion was likely to continue,
entered and joined the argument. “Why, it’s cheap as dirt,” he
exclaimed indignantly; “the fact is, our governor’s a friend of the
people, and don’t mind losing a little money. He’s determined that
everybody shall have a portrait, from the highest to the lowest.
Indeed, next Sunday, he do talk of taking them for threepence-
ha’penny, and if that ain’t philandery, what is?”
After the touter’s oration the tanner seemed somewhat contented,
and paying his eightpence left the shop, looking at his picture in all
lights, and repeatedly polishing it up with the cuff of his coat-sleeve,
as if he were trying to brighten it into something like distinctness.
Whilst I was in this establishment a customer was induced to pay
twopence for having the theory of photography explained to him.
The lecture was to the effect, that the brass tube of the “camerer”
was filled with clock-work, which carried the image from the lens to
the ground glass at the back. To give what the lecturer called
“hockeylar proof” of this, the camera was carried to the shop-door,
and a boy who was passing by ordered to stand still for a minute.
“Now, then,” continued the lecturer to the knowledge-seeker, “look
behind here; there’s the himage, you see;” and then addressing the
boy, he added, “Just open your mouth, youngster;” and when the
lad did so, the student was asked, “Are you looking down the young

un’s throat?” and on his nodding assent, he was informed, “Well,
that’s the way portraits is took.”
Statement of a Photographic Man.
“I’ve been on and off at photographic-portrait taking since its
commencement—that is to say, since they were taken cheap—two
years this summer. I lodged in a room in Lambeth, and I used to
take them in the back-yard—a kind of garden; I used to take a
blanket off the bed, and used to tack it on a clothes-horse, and my
mate used to hold it, if the wind was high, whilst I took the portrait.
“The reason why I took to photographing was, that I thought I
should like it better than what I was at. I was out busking and drag-
pitching with a banjo then. Busking is going into public-houses and
playing, and singing, and dancing; and drag-pitching is going out in
the day down the little courts—tidy places, little terraces, no
thoroughfares, we call drags. I’m a very determined chap, and when
I take a hidea into my head I always do it somehow or other. I didn’t
know anything about photographs then, not a mite, but I saved up
my money; sometimes a 1s.; if I had a good day, 1s. 6d.; and my
wife she went to work at day boot-binding, and at night dancing at a
exhibition, or such-like (she’s a tolerable good dancer—a penny
exhibition or a parade dancer at fairs; that is, outside a show);
sometimes she is Mademoiselle, or Madame, or what it may be. I
got a loan of 3l. (and had to pay 4l. 3s. for it), and with what I’d
saved, I managed to get together 5l. 5s., and I went to Gilbert
Flemming’s, in Oxford-street, and bought a complete apparatus for
taking pictures; 6½ by 4¾, for 5l. 5s. Then I took it home, and
opened the next day to take portraits for what we could get—1s.
and over. I never knew anything about taking portraits then, though
they showed me when I bought the apparatus (but that was as good
as nothing, for it takes months to learn). But I had cards ready
printed to put in the window before I bought the apparatus. The
very next day I had the camera, I had a customer before I had even

tried it, so I tried it on him, and I gave him a black picture (for I
didn’t know how to make the portrait, and it was all black when I
took the glass out), and told him that it would come out bright as it
dried, and he went away quite delighted. I took the first Sunday
after we had opened 1l. 5s. 6d., and everybody was quite pleased
with their spotted and black pictures, for we still told them they
would come out as they dried. But the next week they brought them
back to be changed, and I could do them better, and they had
middling pictures—for I picked it up very quick.
“I had one fellow for a half-guinea portrait, and he was from
Woolwich, and I made him come three times, like a lamb, and he
stood pipes and ’bacca, and it was a thundering bad one after all. He
was delighted, and he swears now it’s the best he ever had took, for
it don’t fade, but will stop black to the end of the world; though he
remarks that I deceived him in one thing, for it don’t come out
bright.
“You see, when first photography come up I had my eye on it, for I
could see it would turn me in something some time. I went and
worked as a regular labourer, carrying pails and so on, so as to try
and learn something about chemistry; for I always had a hankling
after science. Me and Jim was out at Stratford, pitching with the
banjo, and I saw some men coming out of a chemical works, and we
went to ‘nob’ them (that’s get some halfpence out of them). Jim was
tambo beating, and we was both black, and they called us lazy
beggars, and said we ought to work as they did. So we told them we
couldn’t get work, we had no characters. As we went home I and
Jim got talking, and he says, ‘What a fine thing if we could get into
the berth, for you’d soon learn about them portraits if you get
among the chemicals;’ so I agreed to go and try for the situation,
and told him that if I got the berth I’d ‘nanti panka his nabs snide;’
that means, I wouldn’t turn him up, or act nasty to him, but would
share money the same as if we were pitching again. That slang is
mummers’ slang, used by strolling professionals.

“I stopped there for near twelve months, on and off. I had 10s. at
first, but I got up to 16s.; and if I’d stopped I’ve no doubt I should
have been foreman of one of the departments, for I got at last to
almost the management of the oxalic acid. They used to make
sulphate of iron—ferri sulp is the word for it—and carbonate of iron,
too, and I used to be like the red man of Agar then, all over red, and
a’most thought of cutting that to go for a soldier, for I shouldn’t have
wanted a uniform. Then I got to charging the retorts to make
carbonate of ammonia, and from that I went to oxalic acid.
“At night me and Jim used to go out with the banjo and tamborine,
and we could manage to make up our shares to from 18s. to a
guinea a-week each; that is, sharing my wages and all; for when we
chum together we always panka each other bona (that is, share).
We always made our ponta (that is, a pound) a-week, for we could
average our duey bionk peroon a darkey, or two shillings each, in
the night.
“That’s how I got an idea of chemicals, and when I went to
photography many of the very things I used to manufacture was the
very same as we used to take portraits, such as the hyposulphate of
soda, and the nitrate of silver, and the sulphate of iron.

PHOTOGRAPHIC SALOON, EAST END OF LONDON.
[From a Sketch.]
“One of the reasons why I couldn’t take portraits was, that when I
bought my camera at Flemming’s he took a portrait of me with it to
show me how to use it, and as it was a dull afternoon he took 90
seconds to produce the picture. So, you see, when I went to work I
thought I ought to let my pictures go the same time; and hang me if
I didn’t, whether the sun was shining or not. I let my plate stop 90
seconds, and of course they used to come out overdone and quite
white, and as the evening grew darker they came better. When I got
a good one I was surprised, and that picture went miles to be shown
about. Then I formed an idea that I had made a miscalculation as to
my time, and by referring to the sixpenny book of instructions I saw
my mistake, and by the Sunday—that was five days after—I was
very much improved, and by a month I could take a very tidy
picture.
“I was getting on so well I got some of my portraits, when they was
good ones, put in a chandler’s shop; and to be sure I got first-rate

specimens. I used to go to the different shilling portrait galleries and
have a likeness of myself or friends done, to exhibit in my own
window. That’s the way I got my samples to begin with, and I
believe it’s done all over London.
“I kept at this all the winter, and all the time I suppose I earned 30s.
a-week. When summer come again I took a place with a garden in
the Old Kent-road, and there I done middling, but I lost the majority
of my business by not opening on a Sunday, for it was a religious
neighbourhood, and I could have earned my 5l. a-week comfortable,
for as it was I cleared my 2l. regular. Then I had a regular tent built
up out of clothes-horses. I stopped there till I had an offer of a good
situation, and I accepted of it, at 2l. a-week.
“My new place was in Whitechapel, and we lowered the price from a
shilling to sixpence. We did well there, that is the governor did, you
know, for I’ve taken on the average from 60 to 100 a-day, varying in
price from sixpence to half-a-guinea, and the majority was shilling
ones. The greatest quantity I ever took was 146 in one day, and 124
was taken away as they was done. The governor used to take 20l. a-
week, and of that 8l. clear profit, after paying me 2l., the men at the
door 24s., a man and woman 29s., and rent 2l. My governor had, to
my knowledge, 11 other shops, and I don’t know all of his
establishments; I managed my concern for him, and he never come
near us sometimes for a month.
“I left on my own accord after four months, and I joined two others
on equal shares, and opened a place of my own in Southwark.
Unfortunately, I begun too late in the season, or I should have done
well there; but at first we realised about 2l. a-week each, and up to
last week we have shared our 25s. a-head.
“Sunday is the best day for shilling portraits; in fact, the majority is
shilling ones, because then, you see, people have got their wages,
and don’t mind spending. Nobody knows about men’s ways better
than we do. Sunday and Monday is the Derby-day like, and then
after that they are about cracked up and done. The largest amount
I’ve taken at Southwark on a Sunday is 80—over 4l. worth, but then

in the week-days it’s different; Sunday’s 15s. we think that very tidy,
some days only 3s. or 4s.
“You see we are obliged to resort to all sort of dodges to make
sixpenny portraits pay. It’s a very neat little picture our sixpenny
ones is; with a little brass rim round them, and a neat metal inside,
and a front glass; so how can that pay if you do the legitimate
business? The glass will cost you 2d. a-dozen—this small size—and
you give two with every picture; then the chemicals will cost quite a
halfpenny, and varnish, and frame, and fittings, about 2d. We reckon
3d. out of each portrait. And then you see there’s house-rent and a
man at the door, and boy at the table, and the operator, all to pay
their wages out of this 6d.; so you may guess where the profit is.
“One of our dodges is what we term ‘An American Air-Preserver;’
which is nothing more than a card,—old benefit tickets, or, if we are
hard up, even brown paper, or anythink,—soap wrappings, just
varnished on one side. Between our private residence and our shop,
no piece of card or old paper escapes us. Supposing a party come
in, and says ‘I should like a portrait;’ then I inquire which they’ll
have, a shilling or a sixpenny one. If they prefer a sixpenny one, I
then make them one up, and I show them one of the air-preservers,
—which we keep ready made up,—and I tell them that they are all
chemicalized, and come from America, and that without them their
picture will fade. I also tell them that I make nothing out of them,
for that they are only 2d. and cost all the money; and that makes
’em buy one directly. They always bite at them; and we’ve actually
had people come to us to have our preservers put upon other
persons’ portraits, saying they’ve been everywhere for them and
can’t get them. I charge 3d. if it’s not one of our pictures. I’m the
original inventor of the ‘Patent American Air-Preserver.’ We first
called them the ‘London Air-Preservers;’ but they didn’t go so well as
since they’ve been the Americans.
“Another dodge is, I always take the portrait on a shilling size; and
after they are done, I show them what they can have for a shilling,
—the full size, with the knees; and table and a vase on it,—and let

them understand that for sixpence they have all the back-ground
and legs cut off; so as many take the shilling portraits as sixpenny
ones.
“Talking of them preservers, it is astonishing how they go. We’ve
actually had photographers themselves come to us to buy our
‘American Air-Preservers.’ We tells them it’s a secret, and we
manufacture them ourselves. People won’t use their eyes. Why, I’ve
actually cut up an old band-box afore the people’s eyes, and
varnished it and dried it on the hob before their eyes, and yet they
still fancy they come from America! Why, we picks up the old paper
from the shop-sweeping, and they make first-rate ‘Patent American
Air-Preservers.’ Actually, when we’ve been short, I’ve torn off a bit of
old sugar-paper, and stuck it on without any varnish at all, and the
party has gone away quite happy and contented. But you must
remember it is really a useful thing, for it does do good and do
preserve the picture.
“Another of our dodges,—and it is a splendid dodge, though it wants
a nerve to do it,—is the brightening solution, which is nothing more
than aqua distilled, or pure water. When we take a portrait, Jim, my
mate, who stops in the room, hollows to me, ‘Is it bona?’ That is,—Is
it good? If it is, I say, ‘Say.’ That is,—Yes. If not, I say ‘Nanti.’ If it is a
good one he takes care to publicly expose that one, that all may see
it, as a recommendation to others. If I say ‘Nanti,’ then Jim takes it
and finishes it up, drying it and putting it up in its frame. Then he
wraps it up in a large piece of paper, so that it will take sometime to
unroll it, at the same time crying out ‘Take sixpence from this lady, if
you please.’ Sometimes she says, ‘O let me see it first;’ but he
always answers, ‘Money first, if you please ma’am; pay for it first,
and then you can do what you like with it. Here, take sixpence from
this lady.’ When she sees it, if it is a black one, she’ll say, ‘Why this
ain’t like me; there’s no picture at all.’ Then Jim says, ‘It will become
better as it dries, and come to your natural complexion.’ If she still
grumbles, he tells her that if she likes to have it passed through the
brightening solution, it will come out lighter in an hour or two. They
in general have it brightened; and then, before their face, we dip it

into some water. We then dry it off and replace it in the frame, wrap
it up carefully, and tell them not to expose it to the air, but put it in
their bosom, and in an hour or two it will be all right. This is only
done when the portrait come out black, as it doesn’t pay to take two
for sixpence. Sometimes they brings them back the next day, and
says, ‘It’s not dried out as you told us;’ and then we take another
portrait, and charge them 3d. more.
“We also do what we call the ‘bathing,’—another dodge. Now to-day
a party came in during a shower of rain, when it was so dark it was
impossible to take a portrait; or they will come in, sometimes, just as
we are shutting up, and when the gas is lighted, to have their
portraits taken; then we do this. We never turn business away, and
yet it’s impossible to take a portrait; so we ask them to sit down,
and then we go through the whole process of taking a portrait, only
we don’t put any plate in the camera. We always make ’em sit a long
time, to make ’em think it’s all right,—I’ve had them for two-and-a-
half minutes, till their eyes run down with water. We then tell them
that we’ve taken the portrait, but that we shall have to keep it all
night in the chemical bath to bring it out, because the weather’s so
bad. We always take the money as a deposit, and give them a
written paper as an order for the picture. If in the morning they
come themselves we get them to sit again, and then we do really
take a portrait of them; but if they send anybody, we either say that
the bath was too strong and eat the picture out, or that it was too
weak and didn’t bring it out; or else I blow up Jim, and pretend he
has upset the bath and broke the picture. We have had as many as
ten pictures to bathe in one afternoon.
“If the eyes in a portrait are not seen, and they complain, we take a
pin and dot them; and that brings the eye out, and they like it. If the
hair, too, is not visible we takes the pin again, and soon puts in a
beautiful head of hair. It requires a deal of nerve to do it; but if they
still grumble I say, ‘It’s a beautiful picture, and worth half-a-crown,
at the least;’ and in the end they generally go off contented and
happy.

“When we are not busy, we always fill up the time taking specimens
for the window. Anybody who’ll sit we take him; or we do one
another, and the young woman in the shop who colours. Specimens
are very useful things to us, for this reason,—if anybody comes in a
hurry, and won’t give us time to do the picture, then, as we can’t
afford to let her go, we sit her and goes through all the business,
and I says to Jim, ‘Get one from the window,’ and then he takes the
first specimen that comes to hand. Then we fold it up in paper, and
don’t allow her to see it until she pays for it, and tell her not to
expose it to the air for three days, and that if then she doesn’t
approve of it and will call again we will take her another. Of course
they in general comes back. We have made some queer mistakes
doing this. One day a young lady came in, and wouldn’t wait, so Jim
takes a specimen from the window, and, as luck would have it, it
was the portrait of a widow in her cap. She insisted upon opening,
and then she said, ‘This isn’t me; it’s got a widow’s cap, and I was
never married in all my life!’ Jim answers, ‘Oh, miss! why it’s a
beautiful picture, and a correct likeness,’—and so it was, and no lies,
but it wasn’t of her.—Jim talked to her, and says he, ‘Why this ain’t a
cap, it’s the shadow of the hair,’—for she had ringlets,—and she
positively took it away believing that such was the case; and even
promised to send us customers, which she did.
“There was another lady that came in a hurry, and would stop if we
were not more than a minute; so Jim ups with a specimen, without
looking at it, and it was the picture of a woman and her child. We
went through the business of focussing the camera, and then gave
her the portrait and took the 6d. When she saw it she cries out,
‘Bless me! there’s a child: I haven’t ne’er a child!’ Jim looked at her,
and then at the picture, as if comparing, and says he, ‘It is certainly
a wonderful likeness, miss, and one of the best we ever took. It’s the
way you sat; and what has occasioned it was a child passing through
the yard.’ She said she supposed it must be so, and took the portrait
away highly delighted.
“Once a sailor came in, and as he was in haste, I shoved on to him
the picture of a carpenter, who was to call in the afternoon for his

portrait. The jacket was dark, but there was a white waistcoat; still I
persuaded him that it was his blue Guernsey which had come up
very light, and he was so pleased that he gave us 9d. instead of 6d.
The fact is, people don’t know their own faces. Half of ’em have
never looked in a glass half a dozen times in their life, and directly
they see a pair of eyes and a nose, they fancy they are their own.
“The only time we were done was with an old woman. We had only
one specimen left, and that was a sailor man, very dark—one of our
black pictures. But she put on her spectacles, and she looked at it up
and down, and says, ‘Eh?’ I said, ‘Did you speak, ma’am?’ and she
cries, ‘Why, this is a man! here’s the whiskers.’ I left, and Jim tried to
humbug her, for I was bursting with laughing. Jim said, ‘It’s you
ma’am; and a very excellent likeness, I assure you.’ But she kept on
saying, ‘Nonsense, I ain’t a man,’ and wouldn’t have it. Jim wanted
her to leave a deposit, and come next day, but she never called. It
was a little too strong.
“There was an old woman come in once and wanted to be taken
with a favourite hen in her lap. It was a very bad picture, and so
black there was nothing but the outline of her face and a white
speck for the beak of the bird. When she saw it, she asked where
the bird was? So Jim took a pin and scratched in an eye, and said,
‘There it is, ma’am—that’s her eye, it’s coming out,’ and then he
made a line for the comb on the head, and she kept saying,
‘Wonderful!’ and was quite delighted.
“The only bad money we have taken was from a Methodist
clergyman, who came in for a 1s. 6d. portrait. He gave us a bad
sixpence.
“For colouring we charge 3d. more. If the portraits are bad or dark
we tell them, that if they have them coloured the likeness will be
perfect. We flesh the face, scratch the eye in, and blue the coat and
colour the tablecloth. Sometimes the girl who does it puts in such a
lot of flesh paint, that you can scarcely distinguish a feature of the
person. If they grumble, we tell them it will be all right when the
picture’s dry. If it’s a good picture, the colour looks very nice, but in

the black ones we are obliged to stick it on at a tremendous rate, to
make it show.
“Jim stands at the door, and he keeps on saying, ‘A correct portrait,
framed and glazed, for sixpence, beautifully enamelled.’ Then, when
they are listening, he shows the specimen in his hands, and adds, ‘If
not approved of, no charge made.’
“One morning, when we had been doing ‘quisby,’ that is, stopping
idle, we hit upon another dodge. Some friends dropped in to see
me, and as I left to accompany them to a tavern close by, I cried to
Jim, ‘Take that public-house opposite.’ He brought the camera and
stand to the door, and a mob soon collected. He kept saying, ‘Stand
back, gentlemen, stand back! I am about to take the public-house in
front by this wonderful process.’ Then he went over to the house,
and asked the landlord, and asked some gentlemen drinking there to
step into the road whilst he took the house with them facing it. Then
he went to a policeman and asked him to stop the carts from
passing, and he actually did. By this way he got up a tremendous
mob. He then put in the slide, pulled off the cap of the camera, and
focussed the house, and pretended to take the picture, though he
had no prepared glass, nor nothing. When he had done, he called
out, ‘Portraits taken in one minute. We are now taking portraits for
6d. only. Time of sitting, two seconds only. Step inside and have
your’n taken immediately.’ There was a regular rush, and I had to be
fetched, and we took 6s. worth right off.
“People seem to think the camera will do anything. We actually
persuade them that it will mesmerise them. After their portrait is
taken, we ask them if they would like to be mesmerised by the
camera, and the charge is only 2d. We then focus the camera, and
tell them to look firm at the tube; and they stop there for two or
three minutes staring, till their eyes begin to water, and then they
complain of a dizziness in the head, and give it up, saying they ‘can’t
stand it.’ I always tell them the operation was beginning, and they
were just going off, only they didn’t stay long enough. They always
remark, ‘Well, it certainly is a wonderful machine, and a most

curious invention.’ Once a coalheaver came in to be mesmerised, but
he got into a rage after five or six minutes, and said, ‘Strike me
dead, ain’t you keeping me a while!’ He wouldn’t stop still, so Jim
told him his sensitive nerves was too powerful, and sent him off
cursing and swearing because he couldn’t be mesmerised. We don’t
have many of these mesmerism customers, not more than four in
these five months; but it’s a curious circumstance, proving what
fools people is. Jim says he only introduces these games when
business is dull, to keep my spirits up—and they certainly are most
laughable.
“I also profess to remove warts, which I do by touching them with
nitric acid. My price is a penny a wart, or a shilling for the job; for
some of the hands is pretty well smothered with them. You see, we
never turn money away, for it’s hard work to make a living at
sixpenny portraits. My wart patients seldom come twice, for they
screams out ten thousand blue murders when the acid bites them.
“Another of my callings is to dye the hair. You see I have a good
many refuse baths, which is mostly nitrate of silver, the same as all
hair-dyes is composed of. I dyes the whiskers and moustache for 1s.
The worst of it is, that nitrate of silver also blacks the skin wherever
it touches. One fellow with carroty hair came in one day to have his
whiskers died, and I went clumsily to work and let the stuff trickle
down his chin and on his cheeks, as well as making the flesh at the
roots as black as a hat. He came the next day to have it taken off,
and I made him pay 3d. more, and then removed it with cyanide,
which certainly did clean him, but made him smart awfully.
“I have been told that there are near upon 250 houses in London
now getting a livelihood taking sixpenny portraits. There’s ninety of
’em I’m personally acquainted with, and one man I know has ten
different shops of his own. There’s eight in the Whitechapel-road
alone, from Butcher-row to the Mile-end turnpike. Bless you, yes!
they all make a good living at it. Why, I could go to-morrow, and
they would be glad to employ me at 2l. a-week—indeed they have
told me so.

Welcome to our website – the perfect destination for book lovers and
knowledge seekers. We believe that every book holds a new world,
offering opportunities for learning, discovery, and personal growth.
That’s why we are dedicated to bringing you a diverse collection of
books, ranging from classic literature and specialized publications to
self-development guides and children's books.
More than just a book-buying platform, we strive to be a bridge
connecting you with timeless cultural and intellectual values. With an
elegant, user-friendly interface and a smart search system, you can
quickly find the books that best suit your interests. Additionally,
our special promotions and home delivery services help you save time
and fully enjoy the joy of reading.
Join us on a journey of knowledge exploration, passion nurturing, and
personal growth every day!
ebookbell.com

Quick Start Guide To Large Language Models Second Edition Sinan Ozdemir

More Related Content

Similar to Quick Start Guide To Large Language Models Second Edition Sinan Ozdemir

Recently uploaded

Quick Start Guide To Large Language Models Second Edition Sinan Ozdemir