Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

issue with Tinyllama-1.1B #95

Open
BrunoArsioli opened this issue May 28, 2024 · 0 comments
Open

issue with Tinyllama-1.1B #95

BrunoArsioli opened this issue May 28, 2024 · 0 comments

Comments

@BrunoArsioli
Copy link

When testing the latest version of llama2.mojo, the Stories models are working fine, but the Tinyllama-1.1B model is producing strange results. Here is an example:

mojo llama2.mojo tl-chat.bin \
    -z tok_tl-chat.bin \
    -n 256 -t 0 -s 100 -i "<|im_start|>user\nHello. What is your name?<|im_end|>\n<|im_start|>assistant\n"

num parallel workers: 8 SIMD width: 16
Total bytes read: 4400767004 Estimated checkpoint size: 4196 MB
n layers: 22 | vocab size: 32003
<|im_start|>user\nHello. What is your name?<|im_end|>\n<|im_start|>assistant\nqué<0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2><0xE2>
achieved tok/s: 29.109589041095887

or

mojo llama2.mojo tl-chat.bin \
    -z tok_tl-chat.bin \
    -n 256 -t 0 -s 100 -i "<|im_start|>user\nList all integer numbers from 0 to 150.<|im_end|>\n<|im_start|>assistant\n"

num parallel workers: 8 SIMD width: 16
Total bytes read: 4400767004 Estimated checkpoint size: 4196 MB
n layers: 22 | vocab size: 32003
<|im_start|>user\nList all integer numbers from 0 to 150.<|im_end|>\n<|im_start|>assistant\n<0xE2>сосо<0x0A>Que me puedes ayudar?<|im_end|><0x0A><|im_end|><0x0A><|im_start|>assistant<0x0A>¡Claro! ¿Cuál es tu tarea?<|im_end|><0x0A>
achieved tok/s: 28.91844997108155

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant