Errata

Hi Chris,

From your book : Generative AI on AWS,
Chapter 4. Memory and Compute Optimizations,

Figure 4-2. says : to train a 1-billion-parameter model, you will need approximately 24 GB of GPU RAM at 32-bit full precision.

where as, the last paragraph of the same section says :
By quantizing your model weights from 32-bit full precision down to 16-bit half precision, you can quickly reduce your 1-billion-parameter-model memory requirement down 50% to only 2 GB for loading and 40 GB for training.

For 1 Billion parameter model, shouldn't the memory at 16-bit precision be approximately half of 24 GB (at 32-bit precision) = 12 GB ?

Please confirm. Thanks !

Note from the Author or Editor:
Yes, you are correct. The last sentence of that paragraph should say "and 12 GB for training." instead of "and 40 GB for training."

Describes a matrix 64*512 and with Lora and rank 4 you would have a matrix A= 44*64 and a matrix B=512*4. In the next image is the error it refers to B as a matrix 64*4 (in the representation of the two low rank matrices).

Note from the Author or Editor:
Figure 6-5 should be changed to show the horizontal dimension of B = 512 instead of 64.

described more generically, the number below the blue B bar in the upper left (right below "Example: rank = 4") should be 512 instead of 64.

Yup! Hey I, uh, forgot where you live."
has a redundant double quotes(") at its end.

Note from the Author or Editor:
Pls fix

XXX: Yup! Hey I, uh, forgot where you live."
has a redundant double quotes (") at its end.

Note from the Author or Editor:
Pls fix

The link for "Beyond the Imitation Game (BIG-bench)" goes to GLUE page while it should go to BIG-bench page.

Note from the Author or Editor:
Please fix

Increased storage requirements model
should be
Increased model storage requirements

Note from the Author or Editor:
Please fix

# Save the quantize model to disk
should be
# Save the quantized model to disk

Note from the Author or Editor:
Please update

s3_model_path = "s3://<your-private-s3-location/"
should be
s3_model_path = "s3://<your-private-s3-location>/"

Note from the Author or Editor:
Please fix

Output has a redundant double quote (“) at the end

Note from the Author or Editor:
Please fix

The section title “Forward Diffusion” seems not correct as it is not directly relevant to this section. (The correct 'Forward Diffusion' section appears 2 pages later)

Note from the Author or Editor:
Change to “Image-to-Text Generative Tasks”

the process image information
should be
the processed image information

Note from the Author or Editor:
Yes please update

drag-and-drop UI interface
should be
drag-and-drop user interface

Note from the Author or Editor:
Yes that’s fine

The 2nd paragraph says

"The code uses the OpenCV library to extract the edges using the Canny edge map
ControlNet control:"

but the code does not use Canny edge map ControlNet control so it should be

"The code uses the OpenCV library to extract the edges:"

This line "from diffusers import StableDiffusionControlNetPipeline" in the code is unused and unnecessary.

Also

# Load the image
image = load_image(""
)

should be

# Load the image
image = load_image("")

Note from the Author or Editor:
Yes that’s fine

The explanation of the code says "Here is the code to implement the chain-of-thought version of this prompt:" making it clear that it is CoT, but the question part of the prompts has only "Who produced the movie that features this character?" lacking "Think step-by-step" unlike the code sample in Page 215.

Note from the Author or Editor:
Pls add “think step by step”

..., MQA is particularly useful.

The section refers to GQA not MQA. Assume it's a typo. Please confirm.

Note from the Author or Editor:
Yes you are correct. We will fix this on the next printing.

Thanks for the note!!

The order of the matrix dimensions is swapped, indicating "column x row" instead of "row x column" (I am inferring the actual dimensions from the figures).

I am not sure if this was intended, but the second option (row x column) is more common in mathematical usage.

Note from the Author or Editor:
We could swap these on the next release of the book. That’s for pointing this out!

Completion prompt:
should be
Completion:

Note from the Author or Editor:
Yes that’s fine