The errata list is a list of errors and their corrections that were found after the product was released. If the error was corrected in a later version or reprint the date of the correction will be displayed in the column titled "Date Corrected".
The following errata were submitted by our customers and approved as valid errors by the author or editor.
Color key: Serious technical mistake Minor technical mistake Language or formatting error Typo Question Note Update
Version |
Location |
Description |
Submitted By |
Date submitted |
Date corrected |
|
Page Chapter 4. Memory and Compute Optimizations
section : Memory Challenges, last paragraph |
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."
|
Jagadish Kavuturu |
Dec 06, 2023 |
|
|
Page Chapter 6: Target Modules and Layers
Is a picture afther the third paragraph |
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.
|
Gonzalo albornoz |
Dec 07, 2023 |
|
|
Page 68
2nd paragraph |
..., 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!!
|
Wayne Scarano |
Jan 23, 2024 |
|
|
Page 85
python code example |
The `convert_row_to_instruction` python function creates a formats a prompt like
prompt = prompt_template.format(
...
and never closes the format function with a right paren
Note from the Author or Editor: We will fix for the next printing. Thanks for the report!
|
Ari Roffe |
Dec 21, 2023 |
|
|
Page 167
code snippet |
Hi,
In the langchain code on 167, when adding `year` to the document metadata there's a ')' that should be a '}'
`document_fragment.metadata = {"year": year, "source": filename)` should be
`document_fragment.metadata = {"year": year, "source": filename}`
Note from the Author or Editor: Thanks for catching this! We’ll update for the next printing.
Thanks!
|
Ari Roffe |
Dec 29, 2023 |
|