mirror of
https://github.com/harvard-edge/cs249r_book.git
synced 2026-04-29 00:59:07 -05:00
Scales cover images to 100% width
Ensures cover images in Vol. 2 chapters fill the available width, improving visual presentation across different screen sizes. Removes duplicate cover image from the introduction chapter. Corrects a typographical error in Appendix Machine regarding energy ratios.
This commit is contained in:
Vijay Janapa Reddi
@@ -49,7 +49,7 @@ Conventions used here follow the book-wide notation (for example, we reserve \(B
|
||||
# │
|
||||
# │ Goal: Compute all invariant ratios and current hardware specs for the
|
||||
# │ "Numbers Every ML Systems Engineer Should Know" reference section.
|
||||
# │ Show: Energy ratios (~580x DRAM vs FP16 compute), latency hierarchy,
|
||||
# │ Show: Energy ratios (~580$\times$ DRAM vs FP16 compute), latency hierarchy,
|
||||
# │ bandwidth specs, ridge points for A100/H100.
|
||||
# │ How: Scalar extraction via .m_as(unit); ratios remain stable across
|
||||
# │ hardware generations.
|
||||
|
||||
@@ -10,7 +10,7 @@ engine: jupyter
|
||||
:::
|
||||
|
||||
\noindent
|
||||
{fig-alt="Three communication topologies: parameter server (top) with nodes connecting through aggregators to central server, ring AllReduce (bottom left) with 8 GPUs in circular data flow, and all-to-all mesh (bottom right) with fully connected nodes."}
|
||||
{fig-alt="Three communication topologies: parameter server (top) with nodes connecting through aggregators to central server, ring AllReduce (bottom left) with 8 GPUs in circular data flow, and all-to-all mesh (bottom right) with fully connected nodes." width=100%}
|
||||
|
||||
:::
|
||||
|
||||
|
||||
@@ -49,7 +49,7 @@ start_chapter("vol2:compute_infrastructure")
|
||||
:::
|
||||
|
||||
\noindent
|
||||
{fig-alt="Isometric datacenter visualization with rows of server racks connected by glowing blue network paths, cooling infrastructure below, and monitoring dashboards displaying utilization graphs above."}
|
||||
{fig-alt="Isometric datacenter visualization with rows of server racks connected by glowing blue network paths, cooling infrastructure below, and monitoring dashboards displaying utilization graphs above." width=100%}
|
||||
|
||||
:::
|
||||
|
||||
|
||||
@@ -9,7 +9,7 @@
|
||||
:::
|
||||
|
||||
\noindent
|
||||
{fig-alt="Mountain path ascending through four zones: infrastructure at base, distributed systems mid-elevation, production challenges higher, responsible deployment at summit with future horizons view."}
|
||||
{fig-alt="Mountain path ascending through four zones: infrastructure at base, distributed systems mid-elevation, production challenges higher, responsible deployment at summit with future horizons view." width=100%}
|
||||
|
||||
:::
|
||||
|
||||
|
||||
@@ -43,7 +43,7 @@ start_chapter("vol2:storage")
|
||||
:::
|
||||
|
||||
\noindent
|
||||
{fig-alt="Distributed data storage and scalable storage systems for ML workloads."}
|
||||
{fig-alt="Distributed data storage and scalable storage systems for ML workloads." width=100%}
|
||||
|
||||
:::
|
||||
|
||||
|
||||
@@ -36,7 +36,7 @@ start_chapter("vol2:distributed_training")
|
||||
:::
|
||||
|
||||
\noindent
|
||||
{fig-alt="Distributed training across multiple nodes and accelerators."}
|
||||
{fig-alt="Distributed training across multiple nodes and accelerators." width=100%}
|
||||
|
||||
:::
|
||||
|
||||
|
||||
@@ -13,7 +13,7 @@ engine: jupyter
|
||||
:::
|
||||
|
||||
\noindent
|
||||
{fig-alt="Edge intelligence and on-device learning."}
|
||||
{fig-alt="Edge intelligence and on-device learning." width=100%}
|
||||
|
||||
:::
|
||||
|
||||
|
||||
@@ -10,7 +10,7 @@ engine: jupyter
|
||||
:::
|
||||
|
||||
\noindent
|
||||
{fig-alt="Fault tolerance, checkpointing, and recovery in distributed training."}
|
||||
{fig-alt="Fault tolerance, checkpointing, and recovery in distributed training." width=100%}
|
||||
|
||||
:::
|
||||
|
||||
|
||||
@@ -39,7 +39,7 @@ start_chapter("vol2:fleet_orchestration")
|
||||
:::
|
||||
|
||||
\noindent
|
||||
{fig-alt="Stylized datacenter visualization with blue glowing server towers, hexagonal node clusters connected by bright network pathways, and circuit board patterns below."}
|
||||
{fig-alt="Stylized datacenter visualization with blue glowing server towers, hexagonal node clusters connected by bright network pathways, and circuit board patterns below." width=100%}
|
||||
|
||||
:::
|
||||
|
||||
|
||||
@@ -10,7 +10,7 @@ engine: jupyter
|
||||
:::
|
||||
|
||||
\noindent
|
||||
{fig-alt="Distributed inference and model serving at scale."}
|
||||
{fig-alt="Distributed inference and model serving at scale." width=100%}
|
||||
|
||||
:::
|
||||
|
||||
|
||||
@@ -10,11 +10,10 @@ engine: jupyter
|
||||
:::
|
||||
|
||||
\noindent
|
||||
{fig-alt="Abstract geometric composition with interconnected polygons, flowing lines, and node clusters in blue and gold gradients against a dark background." width=100%}
|
||||
|
||||
:::
|
||||
|
||||
{fig-alt="Abstract geometric composition with interconnected polygons, flowing lines, and node clusters in blue and gold gradients against a dark background."}
|
||||
|
||||
## Purpose {.unnumbered}
|
||||
|
||||
\begin{marginfigure}
|
||||
|
||||
@@ -10,7 +10,7 @@ engine: jupyter
|
||||
:::
|
||||
|
||||
\noindent
|
||||
{fig-alt="Network interconnects and fabric topology linking compute nodes in a datacenter."}
|
||||
{fig-alt="Network interconnects and fabric topology linking compute nodes in a datacenter." width=100%}
|
||||
|
||||
:::
|
||||
|
||||
|
||||
@@ -33,7 +33,7 @@ start_chapter("vol2:ops_scale")
|
||||
:::
|
||||
|
||||
\noindent
|
||||
{fig-alt="Operations and production management at fleet scale."}
|
||||
{fig-alt="Operations and production management at fleet scale." width=100%}
|
||||
|
||||
:::
|
||||
|
||||
|
||||
@@ -32,7 +32,7 @@ start_chapter("vol2:performance_engineering")
|
||||
:::
|
||||
|
||||
\noindent
|
||||
{fig-alt="Performance engineering and optimization at scale."}
|
||||
{fig-alt="Performance engineering and optimization at scale." width=100%}
|
||||
|
||||
:::
|
||||
|
||||
|
||||
@@ -37,7 +37,7 @@ start_chapter("vol2:responsible_engineering")
|
||||
:::
|
||||
|
||||
\noindent
|
||||
{fig-alt="Responsible AI governance, fairness, and accountability at fleet scale."}
|
||||
{fig-alt="Responsible AI governance, fairness, and accountability at fleet scale." width=100%}
|
||||
|
||||
:::
|
||||
|
||||
|
||||
@@ -13,7 +13,7 @@ engine: jupyter
|
||||
:::
|
||||
|
||||
\noindent
|
||||
{fig-alt="Adversarial robustness and reliable AI under distribution shift."}
|
||||
{fig-alt="Adversarial robustness and reliable AI under distribution shift." width=100%}
|
||||
|
||||
:::
|
||||
|
||||
|
||||
@@ -37,7 +37,7 @@ start_chapter("vol2:security_privacy")
|
||||
:::
|
||||
|
||||
\noindent
|
||||
{fig-alt="Security and privacy in ML systems at scale."}
|
||||
{fig-alt="Security and privacy in ML systems at scale." width=100%}
|
||||
|
||||
:::
|
||||
|
||||
|
||||
@@ -12,7 +12,7 @@ engine: jupyter
|
||||
:::
|
||||
|
||||
\noindent
|
||||
{fig-alt="Sustainable AI and energy-efficient computing."}
|
||||
{fig-alt="Sustainable AI and energy-efficient computing." width=100%}
|
||||
|
||||
:::
|
||||
|
||||
|
||||
Reference in New Issue
Block a user