Skip to content

Allow overriding the gated Llama-3 tokenizer used for the LLaMa3 prompt style #151

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.

Sign up for GitHub

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

Open
DrStranded wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Allow overriding the gated Llama-3 tokenizer used for the LLaMa3 prompt style #151

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 7 additions & 1 deletion lcb_runner/prompts/code_generation.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,5 @@
import json
import os

try:
from anthropic import HUMAN_PROMPT, AI_PROMPT
Expand Down Expand Up @@ -264,8 +265,13 @@ def format_prompt_generation(
]
from transformers import AutoTokenizer

# The tokenizer is only used to render the Llama 3 chat template. The default repo is gated, so any
# Llama 3 family checkpoint the user can access (same chat template) may be substituted via
# LCB_LLAMA3_TOKENIZER.
tokenizer = AutoTokenizer.from_pretrained(
"meta-llama/Meta-Llama-3-8B-Instruct", padding_side="left", use_fast=False
os.environ.get("LCB_LLAMA3_TOKENIZER", "meta-llama/Meta-Llama-3-8B-Instruct"),
padding_side="left",
use_fast=False,
)
return tokenizer.apply_chat_template(
chat_messages,
Expand Down