Replies: 2 comments 6 replies
-
|
mmm I missed this other discussion which might already give me a sort of answer to my problem... Just keeping this alive in case anyone wants to chime in 😺 |
Beta Was this translation helpful? Give feedback.
-
|
Hi there. Could you share what operating system this is on, and whether you are running this on cpu/mps/cuda devices? |
Beta Was this translation helpful? Give feedback.
-
|
@rasbt Just politely checking whether there is anything to add to this discussion, or if I can safely live with the different results on my side... If you need further info on my machine please do let me know and I will try my best to give a timely feedback. Otherwise all good :) |
Beta Was this translation helpful? Give feedback.
-
|
Hi there, and thanks for the additional info, I totally missed your follow-up. If you used "cpu" then it (highly likely) didn't use the internal GPU. Hm, I am actually not sure why the text differs. It's likely due to very small numerical differences (e.g., via a mix between processor, operating system, and PyTorch version). May I ask if you retyped the code from the book or whether you used the notebook from this repo? |
Beta Was this translation helpful? Give feedback.
-
Thanks for your reply appreciate that! I actually retyped the code myself, trying to make sure everything was aligned with the code in the book. I tracked my code and progress in this repository. I even used programs like Meld to check whether there were differences between my code and the one in the book, and I could not spot anything wrong... The fact that the output I get is the same as the one in #632 also makes me think it's a matter of processor/OS difference (if I am not mistaken you used a macOS in the book, while as said I am using a Windows 11 machine with an AMD Ryzen 7 processor...). The only test I am missing is running the code straight out of your repository and check whether that gives me the same output or not. I don't have much time to do it now, but will try to do it when I can. |
Beta Was this translation helpful? Give feedback.
-
|
Out of curiosity, I just ran the code from your repo and am getting:
So, yeah, it likely is due to
I wouldn't worry too much about that :) |
Beta Was this translation helpful? Give feedback.
-
|
That's great news, thanks @rasbt for checking it out! Closing this now 😄 |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
First of all, I think the book is great, so, chapeau to you, seriously! :)
I have just finished chapter 5 and realized that my output are slightly different than the ones in the book.
My
For instance, in the book, the very last generated sentence is:
In my code, the resulting sentence is the following instead:
I also notices in the repository notebook that a variable not present in the book (
inference_device) was introduced to "ensure readers get same results in the remaining sections of this book". I tried to add that to my code but the result did not match that of the book still... (the one I pasted above).Did anyone else experience the same? My code does not throw any error and the result is still somewhat meaningful, so I wonder if wither I did something wrong or if this difference is expected. I would want to make sure of this before proceeding to chapter 6. Here is the repository where I am keeping track of my progress (together with the version of the various packages in
requirements.txtwhich I am using, which is the same of the book except for torch (2.4.1 instead of 2.4.0) because I had some issues with that.https://github.com/umbe1987/build-LLM-from-scratch
Beta Was this translation helpful? Give feedback.
All reactions