diff --git a/01_introduction.ipynb b/01_introduction.ipynb
index b11fae2..d405868 100644
--- a/01_introduction.ipynb
+++ b/01_introduction.ipynb
@@ -31,6 +31,13 @@
     "# Hello Transformers"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"transformer-timeline\" caption=\"The transformers timeline\" src=\"images/chapter01_timeline.png\" id=\"transformer-timeline\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -38,6 +45,20 @@
     "## The Encoder-Decoder Framework"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"rnn\" caption=\"Unrolling an RNN in time.\" src=\"images/chapter01_rnn.png\" id=\"rnn\"/>"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"enc-dec\" caption=\"Encoder-decoder architecture with a pair of RNNs. In general, there are many more recurrent layers than those shown.\" src=\"images/chapter01_enc-dec.png\" id=\"enc-dec\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -45,6 +66,27 @@
     "## Attention Mechanisms"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"enc-dec-attn\" caption=\"Encoder-decoder architecture with an attention mechanism for a pair of RNNs.\" src=\"images/chapter01_enc-dec-attn.png\" id=\"enc-dec-attn\"/> "
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"attention-alignment\" width=\"500\" caption=\"RNN encoder-decoder alignment of words in English and the generated translation in French (courtesy of Dzmitry Bahdanau).\" src=\"images/chapter02_attention-alignment.png\" id=\"attention-alignment\"/> "
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"transformer-self-attn\" caption=\"Encoder-decoder architecture of the original Transformer.\" src=\"images/chapter01_self-attention.png\" id=\"transformer-self-attn\"/> "
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -52,6 +94,20 @@
     "## Transfer Learning in NLP"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"transfer-learning\" caption=\"Comparison of traditional supervised learning (left) and transfer learning (right).\" src=\"images/chapter01_transfer-learning.png\" id=\"transfer-learning\"/>  "
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"ulmfit\" width=\"500\" caption=\"The ULMFiT process (courtesy of Jeremy Howard).\" src=\"images/chapter01_ulmfit.png\" id=\"ulmfit\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -497,6 +553,13 @@
     "## The Hugging Face Ecosystem"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"ecosystem\" width=\"500\" caption=\"An overview of the Hugging Face ecosystem of libraries and the Hub.\" src=\"images/chapter01_hf-ecosystem.png\" id=\"ecosystem\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -504,6 +567,20 @@
     "### The Hugging Face Hub"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"hub-overview\" width=\"1000\" caption=\"The models page of the Hugging Face Hub, showing filters on the left and a list of models on the right.\" src=\"images/chapter01_hub-overview.png\" id=\"hub-overview\"/> "
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"hub-model-card\" width=\"1000\" caption=\"A example model card from the Hugging Face Hub. The inference widget is shown on the right, where you can interact with the model.\" src=\"images/chapter01_hub-model-card.png\" id=\"hub-model-card\"/> "
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
diff --git a/03_transformer-anatomy.ipynb b/03_transformer-anatomy.ipynb
index eb150e0..7c125cf 100644
--- a/03_transformer-anatomy.ipynb
+++ b/03_transformer-anatomy.ipynb
@@ -48,6 +48,13 @@
     "## The Transformer Architecture"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"transformer-encoder-decoder\" caption=\"Encoder-decoder architecture of the transformer, with the encoder shown in the upper half of the figure and the decoder in the lower half\" src=\"images/chapter03_transformer-encoder-decoder.png\" id=\"transformer-encoder-decoder\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -55,6 +62,13 @@
     "## The Encoder"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"encoder-zoom\" caption=\"Zooming into the encoder layer\" src=\"images/chapter03_encoder-zoom.png\" id=\"encoder-zoom\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -62,6 +76,13 @@
     "### Self-Attention"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Contextualized embeddings\" caption=\"Diagram showing how self-attention updates raw token embeddings (upper) into contextualized embeddings (lower) to create representations that incorporate information from the whole sequence\" src=\"images/chapter03_contextualized-embedding.png\" id=\"contextualized-embeddings\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -181,6 +202,13 @@
     "### End sidebar"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Operations in scaled dot-product attention\" height=\"125\" caption=\"Operations in scaled dot-product attention\" src=\"images/chapter03_attention-ops.png\" id=\"attention-ops\"/>"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -351,6 +379,13 @@
     "#### Multi-headed attention"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Multi-head attention\" height=\"125\" caption=\"Multi-head attention\" src=\"images/chapter03_multihead-attention.png\" id=\"multihead-attention\"/>"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -557,6 +592,13 @@
     "### Adding Layer Normalization"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Transformer layer normalization\" height=\"500\" caption=\"Different arrangements of layer normalization in a transformer encoder layer\" src=\"images/chapter03_layer-norm.png\" id=\"layer-norm\"/>"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -757,6 +799,13 @@
     "## The Decoder"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Transformer decoder zoom\" caption=\"Zooming into the transformer decoder layer\" src=\"images/chapter03_decoder-zoom.png\" id=\"decoder-zoom\"/> "
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -851,6 +900,13 @@
     "### The Transformer Tree of Life"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Transformer family tree\" caption=\"An overview of some of the most prominent transformer architectures\" src=\"images/chapter03_transformers-compact.png\" id=\"family-tree\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
diff --git a/04_multilingual-ner.ipynb b/04_multilingual-ner.ipynb
index 3f23705..a725151 100644
--- a/04_multilingual-ner.ipynb
+++ b/04_multilingual-ner.ipynb
@@ -960,6 +960,13 @@
     "### The Tokenizer Pipeline"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Tokenizer pipeline\" caption=\"The steps in the tokenization pipeline\" src=\"images/chapter04_tokenizer-pipeline.png\" id=\"toknizer-pipeline\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -1001,6 +1008,20 @@
     "## Transformers for Named Entity Recognition"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Architecture of a transformer encoder for classification.\" caption=\"Fine-tuning an encoder-based transformer for sequence classification\" src=\"images/chapter04_clf-architecture.png\" id=\"clf-arch\"/>"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Architecture of a transformer encoder for named entity recognition. The wide linear layer shows that the same linear layer is applied to all hidden states.\" caption=\"Fine-tuning an encoder-based transformer for named entity recognition\" src=\"images/chapter04_ner-architecture.png\" id=\"ner-arch\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -1015,6 +1036,13 @@
     "### Bodies and Heads"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"bert-body-head\" caption=\"The `BertModel` class only contains the body of the model, while the `BertFor&lt;Task&gt;` classes combine the body with a dedicated head for a given task\" src=\"images/chapter04_bert-body-head.png\" id=\"bert-body-head\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -4096,6 +4124,13 @@
     "## Interacting with Model Widgets"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"A Hub widget\" caption=\"Example of a widget on the Hugging Face Hub\" src=\"images/chapter04_ner-widget.png\" id=\"ner-widget\"/>  "
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
diff --git a/05_text-generation.ipynb b/05_text-generation.ipynb
index 6076bac..e44f8b1 100644
--- a/05_text-generation.ipynb
+++ b/05_text-generation.ipynb
@@ -41,6 +41,20 @@
     "# Text Generation"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"LM Meta Learning\" width=\"800\" caption=\"During pretraining, language models are exposed to sequences of tasks that can be adapted during inference (courtesy of Tom B. Brown)\" src=\"images/chapter05_lm-meta-learning.png\" id=\"lm-meta-learning\"/>"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Meena\" width=\"300\" caption=\"Meena on the left telling a corny joke to a human on the right (courtesy of Daniel Adiwardana and Thang Luong)\" src=\"images/chapter05_meena.png\" id=\"meena\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -48,6 +62,13 @@
     "## The Challenge with Generating Coherent Text"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Text generation\" width=\"700\" caption=\"Generating text from an input sequence by adding a new word to the input at each step\" src=\"images/chapter05_text-generation.png\" id=\"text-generation\"/> "
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -320,6 +341,13 @@
     "## Beam Search Decoding"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Beam search\" width=\"700\" caption=\"Beam search with two beams—the most probable sequences at each timestep are highlighted in blue\" src=\"images/chapter05_beam-search.png\" id=\"beam-search\"/> "
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
diff --git a/06_summarization.ipynb b/06_summarization.ipynb
index a5f80ff..e526a6a 100644
--- a/06_summarization.ipynb
+++ b/06_summarization.ipynb
@@ -269,6 +269,13 @@
     "### T5"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"T5\" width=\"700\" caption=\"Diagram of T5's text-to-text framework (courtesy of Colin Raffel); besides translation and summarization, the CoLA (linguistic acceptability) and STSB (semantic similarity) tasks are shown\" src=\"images/chapter08_t5.png\" id=\"T5\"/>"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -307,6 +314,13 @@
     "### PEGASUS"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"pegasus\" width=\"700\" caption=\"Diagram of PEGASUS architecture (courtesy of Jingqing Zhang et al.)\" src=\"images/chapter08_pegasus.png\" id=\"pegasus\"/>"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
diff --git a/07_question-answering.ipynb b/07_question-answering.ipynb
index 7276ee9..42dfd4a 100644
--- a/07_question-answering.ipynb
+++ b/07_question-answering.ipynb
@@ -68,6 +68,13 @@
     "# Question Answering"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Marie Curie\" width=\"500\" caption=\"A Google search query and corresponding answer snippet\" src=\"images/chapter07_marie-curie.png\" id=\"marie-curie\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -82,6 +89,13 @@
     "### The Dataset"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Phone with Query\" width=\"400\" caption=\"A question about a product and the corresponding review (the answer span is underlined)\" src=\"images/chapter07_phone.png\" id=\"phone\"/>"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -327,6 +341,13 @@
     "### Sidebar: The Stanford Question Answering Dataset"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"SQuAD SotA\" width=\"600\" caption=\"Progress on the SQuAD 2.0 benchmark (image from Papers with Code)\" src=\"images/chapter07_squad-sota.png\" id=\"squad-sota\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -348,6 +369,20 @@
     "#### Span classification"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"QA Head\" caption=\"The span classification head for QA tasks\" src=\"images/chapter07_qa-head.png\" id=\"qa-head\"/>"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"SQuAD models\" width=\"600\" caption=\"A selection of extractive QA models on the Hugging Face Hub\" src=\"images/chapter07_squad-models.png\" id=\"squad-models\"/> "
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -706,6 +741,13 @@
     "plt.show()"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Sliding window\" caption=\"How the sliding window creates multiple question-context pairs for long documents—the first bar corresponds to the question, while the second bar is the context captured in each window\" src=\"images/chapter07_sliding-window.png\" id=\"sliding-window\"/>"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -774,6 +816,13 @@
     "### Using Haystack to Build a QA Pipeline"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"QA Architecture\" caption=\"The retriever-reader architecture for modern QA systems\" src=\"images/chapter07_retriever-reader.png\" id=\"retriever-reader\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -1362,6 +1411,13 @@
     "#### Dense Passage Retrieval"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"DPR Architecture\" caption=\"DPR's bi-encoder architecture for computing the relevance of a document and query\" src=\"images/chapter07_dpr.png\" id=\"dpr\"/>"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -1562,6 +1618,13 @@
     "### Domain Adaptation"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"SQuAD Schema\" caption=\"Visualization of the SQuAD JSON format\" src=\"images/chapter07_squad-schema.png\" id=\"squad-schema\"/>"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -1902,6 +1965,13 @@
     "## Going Beyond Extractive QA"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"RAG Architecture\" width=\"600\" caption=\"The RAG architecture for fine-tuning a retriever and generator end-to-end (courtesy of Ethan Perez)\" src=\"images/chapter07_rag-architecture.png\" id=\"rag\"/>"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -2000,6 +2070,18 @@
    "source": [
     "## Conclusion"
    ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"QA Pyramid\" caption=\"The QA hierarchy of needs\" src=\"images/chapter07_qa-pyramid.png\" id=\"qa-pyramid\"/> "
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": []
   }
  ],
  "metadata": {
diff --git a/08_model-compression.ipynb b/08_model-compression.ipynb
index af986ec..be50ba7 100644
--- a/08_model-compression.ipynb
+++ b/08_model-compression.ipynb
@@ -49,6 +49,13 @@
     "# Making Transformers Efficient in Production"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Scaling BERT at Roblox\" caption=\"How Roblox scaled BERT with knowledge distillation, dynamic padding, and weight quantization (photo courtesy of Roblox employees Quoc N. Le and Kip Kaehler)\" src=\"images/chapter08_roblox.png\" id=\"roblox\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -56,6 +63,13 @@
     "## Intent Detection as a Case Study"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Out of Scope Query\" width=\"400\" caption=\"Three exchanges between a human (right) and a text-based assistant (left) for personal finance (courtesy of Stefan Larson et al.)\" src=\"images/chapter08_oos.png\" id=\"oos\"/> "
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -561,6 +575,20 @@
     "### Knowledge Distillation for Fine-Tuning"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Soft Probabilities\" caption=\"Comparison of a hard label that is one-hot encoded (left), softmax probabilities (middle), and softened class probabilities (right)\" src=\"images/chapter08_soft-probs.png\" id=\"soft-probs\"/> "
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Knowledge distillation\" caption=\"The knowledge distillation process\" src=\"images/chapter08_kd.png\" id=\"kd\"/> "
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -1322,6 +1350,13 @@
     "### End sidebar"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Mapping floating-point numbers to 8-bit integers\" width=\"800\" caption=\"Quantizing floating-point numbers as unsigned 8-bit integers (courtesy of Manas Sahni)\" src=\"images/chapter08_fp32-to-int8.png\" id=\"fp32toint8\"/>"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -1605,6 +1640,20 @@
     "## Optimizing Inference with ONNX and the ONNX Runtime"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Example ONNX graph\" width=\"500\" caption=\"A section of the ONNX graph for BERT-base, visualized in Netron\" src=\"images/chapter08_bert-onnx.png\" id=\"bert-onnx\"/>"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Architecture of the ONNX and ONNX Runtime ecosystem\" width=\"500\" caption=\"Architecture of the ONNX and ONNX Runtime ecosystem (courtesy of the ONNX Runtime team)\" src=\"images/chapter08_onnx-ort.png\" id=\"onnx-ort\"/>"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -1940,6 +1989,13 @@
     "### Sparsity in Deep Neural Networks"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Network Pruning\" width=\"500\" caption=\"Weights and neurons before and after pruning (courtesy of Song Han)\" src=\"images/chapter08_network-pruning.png\" id=\"network-pruning\"/> "
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -2004,6 +2060,20 @@
     "#### Movement pruning"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Magnitude vs Movement Pruning\" width=\"700\" caption=\"Comparison of weights removed (in gray) during magnitude pruning (left) and movement pruning (right)\" src=\"images/chapter08_magnitude-vs-movement.png\" id=\"magnitude-vs-movement\"/> "
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Pruning Distributions\" width=\"500\" caption=\"Distribution of remaining weights for magnitude pruning (MaP) and movement pruning (MvP)\" src=\"images/chapter08_pruning-dists.png\" id=\"pruning-dists\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
diff --git a/09_few-to-no-labels.ipynb b/09_few-to-no-labels.ipynb
index 3540848..2ba9eaa 100644
--- a/09_few-to-no-labels.ipynb
+++ b/09_few-to-no-labels.ipynb
@@ -49,6 +49,13 @@
     "# Dealing with Few to No Labels"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"decision-tree\" caption=\"Several techniques that can be used to improve model performance in the absence of large amounts of labeled data\" src=\"images/chapter09_decision-tree.png\" id=\"decision-tree\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -56,6 +63,13 @@
     "## Building a GitHub Issues Tagger"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"issue-example\" caption=\"A typical GitHub issue on the Transformers repository\" src=\"images/chapter09_issue-example-v2.png\" id=\"issue-example\"/> "
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -1762,6 +1776,13 @@
     "### Using Embeddings as a Lookup Table"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"nearest-neighbours\" caption=\"An illustration of nearest neighbor embedding lookup\" src=\"images/chapter09_nearest-neighbours.png\" id=\"nearest-neighbours\"/>"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -2197,6 +2218,13 @@
     "### Sidebar: Efficient Similarity Search with FAISS"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"faiss-index\" caption=\"The structure of a FAISS index: the gray points represent data points added to the index, the bold black points are the cluster centers found via k-means clustering, and the colored areas represent the regions belonging to a cluster center\" src=\"images/chapter09_faiss-index.png\" id=\"faiss-index\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -2831,6 +2859,13 @@
     "#### Unsupervised data augmentation"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"uda\" width=\"600\" caption=\"Training a model M with UDA (courtesy of Qizhe Xie)\" src=\"images/chapter09_uda.png\" id=\"uda\"/>    "
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -2838,6 +2873,13 @@
     "#### Uncertainty-aware self-training"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"ust\" width=\"600\" caption=\"The UST method consists of a teacher that generates pseudo-labels and a student that is subsequently trained on those labels; after the student is trained it becomes the teacher and the step is repeated (courtesy of Subhabrata Mukherjee)footnote:[S. Mukherjee and A.H. Awadallah, [\"Uncertainty-Aware Self-Training for Few-Shot Text Classification\"](https://arxiv.org/abs/2006.15315), (2020).]\" src=\"images/chapter09_ust.png\" id=\"ust\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
diff --git a/10_transformers-from-scratch.ipynb b/10_transformers-from-scratch.ipynb
index 33c4d81..cbb576d 100644
--- a/10_transformers-from-scratch.ipynb
+++ b/10_transformers-from-scratch.ipynb
@@ -1053,6 +1053,13 @@
     "### A Tale of Pretraining Objectives"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Code snippet\" caption=\"An example of a Python function that could be found in our dataset\" src=\"images/chapter10_code-snippet.png\" id=\"code-snippet\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -1060,6 +1067,13 @@
     "#### Causal language modeling"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"CLM pretraining\" caption=\"In causal language modeling, the future tokens are masked and the model has to predict them; typically a decoder model such as GPT is used for such a task\" src=\"images/chapter10_pretraining-clm.png\" id=\"pretraining-clm\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -1067,6 +1081,13 @@
     "#### Masked language modeling"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"MLM pretraining\" caption=\"In masked language modeling some of the input tokens are either masked or replaced, and the model's task is to predict the original tokens; this is the architecture underlying the encoder branch of transformer models\" src=\"images/chapter10_pretraining-mlm.png\" id=\"pretraining-mlm\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -1074,6 +1095,13 @@
     "#### Sequence-to-sequence training"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Seq2seq pretraining\" caption=\"Using an encoder-decoder architecture for a sequence-to-sequence task where the inputs are split into comment/code pairs using heuristics: the model gets one element as input and needs to generate the other one\" src=\"images/chapter10_pretraining-seq2seq.png\" id=\"pretraining-seq2seq\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -1199,6 +1227,13 @@
     "### Implementing the Dataloader"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Preprocessing for CLM\" caption=\"Preparing sequences of varying length for causal language modeling by concatenating several tokenized examples with an EOS token  before chunking them\" src=\"images/chapter10_preprocessing-clm.png\" id=\"preprocessing-clm\"/>"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -1568,6 +1603,13 @@
     "    hf_repo.push_to_hub(commit_message=f'final model')"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<img alt=\"DDP\" caption=\"Illustration of the processing steps in DDP with four GPUs\" src=\"images/chapter10_ddp.png\" id=\"ddp\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
diff --git a/11_future-directions.ipynb b/11_future-directions.ipynb
index 04d5a7f..572d3c9 100644
--- a/11_future-directions.ipynb
+++ b/11_future-directions.ipynb
@@ -146,6 +146,22 @@
     "### Scaling Laws"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "ee076bd6",
+   "metadata": {},
+   "source": [
+    "<img alt=\"scaling-laws\" caption=\"Power-law scaling of test loss versus compute budget (left), dataset size (middle), and model size (right) (courtesy of Jared Kaplan)\" src=\"images/chapter11_scaling.png\" id=\"scaling-laws\"/>"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "59a31bd8",
+   "metadata": {},
+   "source": [
+    "<img alt=\"scaling-laws-modal\" caption=\"Power law scaling of test loss versus compute budget across a wide range of modalities (courtesy of Tom Henighan)\" src=\"images/chapter11_scaling-modal.png\" id=\"scaling-laws-modal\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "id": "foster-reading",
@@ -170,6 +186,22 @@
     "### Sparse Attention"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "2cfc7204",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Atomic sparsity patterns\" caption=\"Common atomic sparse attention patterns for self-attention: a colored square means the  attention score is calculated, while a blank square means the score is discarded (courtesy of Tianyang Lin)\" src=\"images/chapter11_atomic-sparse-attention.png\" id=\"sparsity-atomic\"/> "
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "133cabd9",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Attention with compound sparsity\" caption=\"Sparse attention patterns for recent transformer models (courtesy of Tianyang Lin)\" src=\"images/chapter11_compound-sparse-attention.png\" id=\"sparsity-compound\"/> "
+   ]
+  },
   {
    "cell_type": "markdown",
    "id": "simplified-viewer",
@@ -178,6 +210,14 @@
     "### Linearized Attention"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "d5126697",
+   "metadata": {},
+   "source": [
+    "<img alt=\"Linear attention\" caption=\"Complexity difference between standard self-attention and linearized self-attention (courtesy of Tianyang Lin)\" src=\"images/chapter11_linear-attention.png\" id=\"linear-attention\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "id": "vital-option",
@@ -202,6 +242,14 @@
     "#### iGPT"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "c3c7a08a",
+   "metadata": {},
+   "source": [
+    "<img alt=\"iGPT\" width=\"800\" caption=\"Examples of image completions with iGPT (courtesy of Mark Chen)\" src=\"images/chapter11_iGPT.png\" id=\"iGPT\"/> "
+   ]
+  },
   {
    "cell_type": "markdown",
    "id": "sporting-tuner",
@@ -210,6 +258,14 @@
     "#### ViT"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "b7c55d13",
+   "metadata": {},
+   "source": [
+    "<img alt=\"vit-architecture\" width=\"800\" caption=\"The ViT architecture (courtesy of Alexey Dosovitskiy et al.)\" src=\"images/chapter11_vit-architecture.png\" id=\"vit-architecture\"/> "
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -333,6 +389,22 @@
     "### Tables"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "7675b4fe",
+   "metadata": {},
+   "source": [
+    "<img alt=\"table-qa\" width=\"800\" caption=\"Question answering over a table (courtesy of Jonathan Herzig)\" src=\"images/chapter11_table-qa.png\" id=\"tapas-table\"/>"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "f43e79cd",
+   "metadata": {},
+   "source": [
+    "<img alt=\"tapas-architecture\" width=\"800\" caption=\"Architecture of TAPAS (courtesy of Jonathan Herzig)\" src=\"images/chapter11_tapas-architecture.png\" id=\"tapas-architecture\"/>"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -542,6 +614,14 @@
     "### Speech-to-Text"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "3a70c3fc",
+   "metadata": {},
+   "source": [
+    "<img alt=\"wav2vec2\" width=\"800\" caption=\"Architecture of wav2vec 2.0 (courtesy of Alexei Baevski)\" src=\"images/chapter11_wav2vec2.png\" id=\"wav2vec2\"/>"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -666,6 +746,14 @@
     "print(pred)"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "68f82db8",
+   "metadata": {},
+   "source": [
+    "<img alt=\"wav2vec-u\" width=\"800\" caption=\"Training scheme for wav2vec-U (courtesy of Alexsei Baevski)\" src=\"images/chapter11_wav2vec-u.png\" id=\"wav2vec-u\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "id": "fantastic-multiple",
@@ -682,6 +770,14 @@
     "#### VQA"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "7c0c1aeb",
+   "metadata": {},
+   "source": [
+    "<img alt=\"vqa\" width=\"800\" caption=\"Example of a visual question answering task from the VQA dataset (courtesy of Yash Goyal)\" src=\"images/chapter11_vqa.png\" id=\"vqa\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "id": "42701ffc-1148-4612-951d-54eee357647c",
@@ -690,6 +786,14 @@
     "#### LayoutLM"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "2fdba92a",
+   "metadata": {},
+   "source": [
+    "<img alt=\"layoutlm\" width=\"500\" caption=\"The model architecture and pretraining strategies for LayoutLMv2 (courtesy of Yang Xu)\" src=\"images/chapter11_layoutlm.png\" id=\"layoutlm\"/> "
+   ]
+  },
   {
    "cell_type": "markdown",
    "id": "colored-candle",
@@ -698,6 +802,14 @@
     "#### DALL·E"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "09edc043",
+   "metadata": {},
+   "source": [
+    "<img alt=\"dall-e\" width=\"800\" caption=\"Generation examples with DALL·E (courtesy of Aditya Ramesh)\" src=\"images/chapter11_dall-e.png\" id=\"dall-e\"/>"
+   ]
+  },
   {
    "cell_type": "markdown",
    "id": "measured-marathon",
@@ -706,6 +818,14 @@
     "#### CLIP"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "0c1ec7db",
+   "metadata": {},
+   "source": [
+    "<img alt=\"clip-arch\" width=\"800\" caption=\"Architecture of CLIP (courtesy of Alec Radford)\" src=\"images/chapter11_clip-arch.png\" id=\"clip-arch\"/>"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
diff --git a/images/chapter01_enc-dec-attn.png b/images/chapter01_enc-dec-attn.png
new file mode 100755
index 0000000..32f54c6
Binary files /dev/null and b/images/chapter01_enc-dec-attn.png differ
diff --git a/images/chapter01_enc-dec.png b/images/chapter01_enc-dec.png
new file mode 100755
index 0000000..39bfc6c
Binary files /dev/null and b/images/chapter01_enc-dec.png differ
diff --git a/images/chapter01_hf-ecosystem.png b/images/chapter01_hf-ecosystem.png
new file mode 100644
index 0000000..7263920
Binary files /dev/null and b/images/chapter01_hf-ecosystem.png differ
diff --git a/images/chapter01_hub-model-card.png b/images/chapter01_hub-model-card.png
new file mode 100644
index 0000000..c35592c
Binary files /dev/null and b/images/chapter01_hub-model-card.png differ
diff --git a/images/chapter01_hub-overview.png b/images/chapter01_hub-overview.png
new file mode 100644
index 0000000..db3c80b
Binary files /dev/null and b/images/chapter01_hub-overview.png differ
diff --git a/images/chapter01_rnn.png b/images/chapter01_rnn.png
new file mode 100755
index 0000000..4cc7a19
Binary files /dev/null and b/images/chapter01_rnn.png differ
diff --git a/images/chapter01_self-attention.png b/images/chapter01_self-attention.png
new file mode 100755
index 0000000..ea5c3e1
Binary files /dev/null and b/images/chapter01_self-attention.png differ
diff --git a/images/chapter01_timeline.png b/images/chapter01_timeline.png
new file mode 100644
index 0000000..0395269
Binary files /dev/null and b/images/chapter01_timeline.png differ
diff --git a/images/chapter01_transfer-learning.png b/images/chapter01_transfer-learning.png
new file mode 100755
index 0000000..99e3bda
Binary files /dev/null and b/images/chapter01_transfer-learning.png differ
diff --git a/images/chapter01_ulmfit.png b/images/chapter01_ulmfit.png
new file mode 100644
index 0000000..4bd4a96
Binary files /dev/null and b/images/chapter01_ulmfit.png differ
diff --git a/images/chapter03_attention-ops.png b/images/chapter03_attention-ops.png
new file mode 100755
index 0000000..2dbffef
Binary files /dev/null and b/images/chapter03_attention-ops.png differ
diff --git a/images/chapter03_contextualized-embedding.png b/images/chapter03_contextualized-embedding.png
new file mode 100644
index 0000000..3d97842
Binary files /dev/null and b/images/chapter03_contextualized-embedding.png differ
diff --git a/images/chapter03_decoder-zoom.png b/images/chapter03_decoder-zoom.png
new file mode 100755
index 0000000..b84a1ba
Binary files /dev/null and b/images/chapter03_decoder-zoom.png differ
diff --git a/images/chapter03_encoder-zoom.png b/images/chapter03_encoder-zoom.png
new file mode 100755
index 0000000..2e7c3d1
Binary files /dev/null and b/images/chapter03_encoder-zoom.png differ
diff --git a/images/chapter03_layer-norm.png b/images/chapter03_layer-norm.png
new file mode 100755
index 0000000..11517ea
Binary files /dev/null and b/images/chapter03_layer-norm.png differ
diff --git a/images/chapter03_multihead-attention.png b/images/chapter03_multihead-attention.png
new file mode 100755
index 0000000..ac85d53
Binary files /dev/null and b/images/chapter03_multihead-attention.png differ
diff --git a/images/chapter03_transformer-encoder-decoder.png b/images/chapter03_transformer-encoder-decoder.png
new file mode 100755
index 0000000..82420c8
Binary files /dev/null and b/images/chapter03_transformer-encoder-decoder.png differ
diff --git a/images/chapter03_transformers-compact.png b/images/chapter03_transformers-compact.png
new file mode 100755
index 0000000..d1cfc01
Binary files /dev/null and b/images/chapter03_transformers-compact.png differ
diff --git a/images/chapter04_bert-body-head.png b/images/chapter04_bert-body-head.png
new file mode 100755
index 0000000..4d7a5f6
Binary files /dev/null and b/images/chapter04_bert-body-head.png differ
diff --git a/images/chapter04_clf-architecture.png b/images/chapter04_clf-architecture.png
new file mode 100755
index 0000000..0604e4a
Binary files /dev/null and b/images/chapter04_clf-architecture.png differ
diff --git a/images/chapter04_ner-architecture.png b/images/chapter04_ner-architecture.png
new file mode 100755
index 0000000..fb2bbc7
Binary files /dev/null and b/images/chapter04_ner-architecture.png differ
diff --git a/images/chapter04_ner-widget.png b/images/chapter04_ner-widget.png
new file mode 100644
index 0000000..636a3d2
Binary files /dev/null and b/images/chapter04_ner-widget.png differ
diff --git a/images/chapter04_tokenizer-pipeline.png b/images/chapter04_tokenizer-pipeline.png
new file mode 100644
index 0000000..6c898d0
Binary files /dev/null and b/images/chapter04_tokenizer-pipeline.png differ
diff --git a/images/chapter05_beam-search.png b/images/chapter05_beam-search.png
new file mode 100644
index 0000000..a5fd0cf
Binary files /dev/null and b/images/chapter05_beam-search.png differ
diff --git a/images/chapter05_lm-meta-learning.png b/images/chapter05_lm-meta-learning.png
new file mode 100644
index 0000000..e24b4a3
Binary files /dev/null and b/images/chapter05_lm-meta-learning.png differ
diff --git a/images/chapter05_meena.png b/images/chapter05_meena.png
new file mode 100644
index 0000000..37a701a
Binary files /dev/null and b/images/chapter05_meena.png differ
diff --git a/images/chapter05_text-generation.png b/images/chapter05_text-generation.png
new file mode 100644
index 0000000..1c83f08
Binary files /dev/null and b/images/chapter05_text-generation.png differ
diff --git a/images/chapter07_dpr.png b/images/chapter07_dpr.png
new file mode 100644
index 0000000..e334e3d
Binary files /dev/null and b/images/chapter07_dpr.png differ
diff --git a/images/chapter07_marie-curie.png b/images/chapter07_marie-curie.png
new file mode 100644
index 0000000..8f2f8c9
Binary files /dev/null and b/images/chapter07_marie-curie.png differ
diff --git a/images/chapter07_phone.png b/images/chapter07_phone.png
new file mode 100644
index 0000000..d6ba222
Binary files /dev/null and b/images/chapter07_phone.png differ
diff --git a/images/chapter07_qa-head.png b/images/chapter07_qa-head.png
new file mode 100644
index 0000000..c809116
Binary files /dev/null and b/images/chapter07_qa-head.png differ
diff --git a/images/chapter07_qa-pyramid.png b/images/chapter07_qa-pyramid.png
new file mode 100755
index 0000000..33bb7cf
Binary files /dev/null and b/images/chapter07_qa-pyramid.png differ
diff --git a/images/chapter07_rag-architecture.png b/images/chapter07_rag-architecture.png
new file mode 100644
index 0000000..aef949a
Binary files /dev/null and b/images/chapter07_rag-architecture.png differ
diff --git a/images/chapter07_retriever-reader.png b/images/chapter07_retriever-reader.png
new file mode 100755
index 0000000..11d6077
Binary files /dev/null and b/images/chapter07_retriever-reader.png differ
diff --git a/images/chapter07_sliding-window.png b/images/chapter07_sliding-window.png
new file mode 100644
index 0000000..de1122c
Binary files /dev/null and b/images/chapter07_sliding-window.png differ
diff --git a/images/chapter07_squad-models.png b/images/chapter07_squad-models.png
new file mode 100644
index 0000000..3e0b04f
Binary files /dev/null and b/images/chapter07_squad-models.png differ
diff --git a/images/chapter07_squad-schema.png b/images/chapter07_squad-schema.png
new file mode 100644
index 0000000..d981fc8
Binary files /dev/null and b/images/chapter07_squad-schema.png differ
diff --git a/images/chapter07_squad-sota.png b/images/chapter07_squad-sota.png
new file mode 100644
index 0000000..12d85f6
Binary files /dev/null and b/images/chapter07_squad-sota.png differ
diff --git a/images/chapter08_bert-onnx.png b/images/chapter08_bert-onnx.png
new file mode 100644
index 0000000..d788fae
Binary files /dev/null and b/images/chapter08_bert-onnx.png differ
diff --git a/images/chapter08_fp32-to-int8.png b/images/chapter08_fp32-to-int8.png
new file mode 100644
index 0000000..00c7bcf
Binary files /dev/null and b/images/chapter08_fp32-to-int8.png differ
diff --git a/images/chapter08_kd.png b/images/chapter08_kd.png
new file mode 100644
index 0000000..9d5915d
Binary files /dev/null and b/images/chapter08_kd.png differ
diff --git a/images/chapter08_magnitude-vs-movement.png b/images/chapter08_magnitude-vs-movement.png
new file mode 100644
index 0000000..242fe85
Binary files /dev/null and b/images/chapter08_magnitude-vs-movement.png differ
diff --git a/images/chapter08_network-pruning.png b/images/chapter08_network-pruning.png
new file mode 100644
index 0000000..f0de487
Binary files /dev/null and b/images/chapter08_network-pruning.png differ
diff --git a/images/chapter08_onnx-ort.png b/images/chapter08_onnx-ort.png
new file mode 100644
index 0000000..d67f5f4
Binary files /dev/null and b/images/chapter08_onnx-ort.png differ
diff --git a/images/chapter08_oos.png b/images/chapter08_oos.png
new file mode 100644
index 0000000..3d621f1
Binary files /dev/null and b/images/chapter08_oos.png differ
diff --git a/images/chapter08_pegasus.png b/images/chapter08_pegasus.png
new file mode 100644
index 0000000..1d7713f
Binary files /dev/null and b/images/chapter08_pegasus.png differ
diff --git a/images/chapter08_pruning-dists.png b/images/chapter08_pruning-dists.png
new file mode 100644
index 0000000..b2c20c0
Binary files /dev/null and b/images/chapter08_pruning-dists.png differ
diff --git a/images/chapter08_roblox.png b/images/chapter08_roblox.png
new file mode 100644
index 0000000..d562075
Binary files /dev/null and b/images/chapter08_roblox.png differ
diff --git a/images/chapter08_soft-probs.png b/images/chapter08_soft-probs.png
new file mode 100644
index 0000000..c51f531
Binary files /dev/null and b/images/chapter08_soft-probs.png differ
diff --git a/images/chapter08_t5.png b/images/chapter08_t5.png
new file mode 100644
index 0000000..24aa2bf
Binary files /dev/null and b/images/chapter08_t5.png differ
diff --git a/images/chapter09_decision-tree.png b/images/chapter09_decision-tree.png
new file mode 100644
index 0000000..1f779e6
Binary files /dev/null and b/images/chapter09_decision-tree.png differ
diff --git a/images/chapter09_faiss-index.png b/images/chapter09_faiss-index.png
new file mode 100644
index 0000000..bffdc5e
Binary files /dev/null and b/images/chapter09_faiss-index.png differ
diff --git a/images/chapter09_issue-example-v2.png b/images/chapter09_issue-example-v2.png
new file mode 100644
index 0000000..b745592
Binary files /dev/null and b/images/chapter09_issue-example-v2.png differ
diff --git a/images/chapter09_nearest-neighbours.png b/images/chapter09_nearest-neighbours.png
new file mode 100755
index 0000000..6be10ba
Binary files /dev/null and b/images/chapter09_nearest-neighbours.png differ
diff --git a/images/chapter09_uda.png b/images/chapter09_uda.png
new file mode 100644
index 0000000..a5d6594
Binary files /dev/null and b/images/chapter09_uda.png differ
diff --git a/images/chapter09_ust.png b/images/chapter09_ust.png
new file mode 100644
index 0000000..6e8ee04
Binary files /dev/null and b/images/chapter09_ust.png differ
diff --git a/images/chapter10_code-snippet.png b/images/chapter10_code-snippet.png
new file mode 100755
index 0000000..8029342
Binary files /dev/null and b/images/chapter10_code-snippet.png differ
diff --git a/images/chapter10_ddp.png b/images/chapter10_ddp.png
new file mode 100755
index 0000000..b318ba6
Binary files /dev/null and b/images/chapter10_ddp.png differ
diff --git a/images/chapter10_preprocessing-clm.png b/images/chapter10_preprocessing-clm.png
new file mode 100755
index 0000000..4316232
Binary files /dev/null and b/images/chapter10_preprocessing-clm.png differ
diff --git a/images/chapter10_pretraining-clm.png b/images/chapter10_pretraining-clm.png
new file mode 100755
index 0000000..a11eb13
Binary files /dev/null and b/images/chapter10_pretraining-clm.png differ
diff --git a/images/chapter10_pretraining-mlm.png b/images/chapter10_pretraining-mlm.png
new file mode 100755
index 0000000..f5930bb
Binary files /dev/null and b/images/chapter10_pretraining-mlm.png differ
diff --git a/images/chapter10_pretraining-seq2seq.png b/images/chapter10_pretraining-seq2seq.png
new file mode 100755
index 0000000..f8d2e69
Binary files /dev/null and b/images/chapter10_pretraining-seq2seq.png differ
diff --git a/images/chapter11_atomic-sparse-attention.png b/images/chapter11_atomic-sparse-attention.png
new file mode 100644
index 0000000..0f35595
Binary files /dev/null and b/images/chapter11_atomic-sparse-attention.png differ
diff --git a/images/chapter11_clip-arch.png b/images/chapter11_clip-arch.png
new file mode 100644
index 0000000..48887a3
Binary files /dev/null and b/images/chapter11_clip-arch.png differ
diff --git a/images/chapter11_compound-sparse-attention.png b/images/chapter11_compound-sparse-attention.png
new file mode 100644
index 0000000..8c696dc
Binary files /dev/null and b/images/chapter11_compound-sparse-attention.png differ
diff --git a/images/chapter11_dall-e.png b/images/chapter11_dall-e.png
new file mode 100644
index 0000000..a154756
Binary files /dev/null and b/images/chapter11_dall-e.png differ
diff --git a/images/chapter11_efficient-attention.png b/images/chapter11_efficient-attention.png
new file mode 100644
index 0000000..2ba7a8d
Binary files /dev/null and b/images/chapter11_efficient-attention.png differ
diff --git a/images/chapter11_iGPT.png b/images/chapter11_iGPT.png
new file mode 100644
index 0000000..ec3122b
Binary files /dev/null and b/images/chapter11_iGPT.png differ
diff --git a/images/chapter11_layoutlm.png b/images/chapter11_layoutlm.png
new file mode 100644
index 0000000..089f714
Binary files /dev/null and b/images/chapter11_layoutlm.png differ
diff --git a/images/chapter11_linear-attention.png b/images/chapter11_linear-attention.png
new file mode 100644
index 0000000..e99a3bd
Binary files /dev/null and b/images/chapter11_linear-attention.png differ
diff --git a/images/chapter11_scaling-modal.png b/images/chapter11_scaling-modal.png
new file mode 100644
index 0000000..a98cbc6
Binary files /dev/null and b/images/chapter11_scaling-modal.png differ
diff --git a/images/chapter11_scaling.png b/images/chapter11_scaling.png
new file mode 100644
index 0000000..4796187
Binary files /dev/null and b/images/chapter11_scaling.png differ
diff --git a/images/chapter11_table-qa.png b/images/chapter11_table-qa.png
new file mode 100644
index 0000000..3280a28
Binary files /dev/null and b/images/chapter11_table-qa.png differ
diff --git a/images/chapter11_tapas-architecture.png b/images/chapter11_tapas-architecture.png
new file mode 100644
index 0000000..9154bff
Binary files /dev/null and b/images/chapter11_tapas-architecture.png differ
diff --git a/images/chapter11_vit-architecture.png b/images/chapter11_vit-architecture.png
new file mode 100644
index 0000000..4ceb8ac
Binary files /dev/null and b/images/chapter11_vit-architecture.png differ
diff --git a/images/chapter11_vqa.png b/images/chapter11_vqa.png
new file mode 100644
index 0000000..083d865
Binary files /dev/null and b/images/chapter11_vqa.png differ
diff --git a/images/chapter11_wav2vec-u.png b/images/chapter11_wav2vec-u.png
new file mode 100644
index 0000000..0a2b1fc
Binary files /dev/null and b/images/chapter11_wav2vec-u.png differ
diff --git a/images/chapter11_wav2vec2.png b/images/chapter11_wav2vec2.png
new file mode 100644
index 0000000..6e26442
Binary files /dev/null and b/images/chapter11_wav2vec2.png differ