Skip to content

[fix] Grok-1 use tokenizer from the same pretrained path #5532

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

Merged
yuanheng-zhao merged 2 commits into from
Mar 28, 2024
Merged

[fix] Grok-1 use tokenizer from the same pretrained path #5532

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
cbdf0f1
[fix] use tokenizer from the same pretrained path
yuanheng-zhao Mar 28, 2024
61922c2
trust remote code
yuanheng-zhao Mar 28, 2024
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
9 changes: 4 additions & 5 deletions examples/language/grok-1/inference.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,14 +1,17 @@
import time

import torch
from transformers import AutoModelForCausalLM, LlamaTokenizerFast
from transformers import AutoModelForCausalLM, AutoTokenizer
from utils import get_defualt_parser, inference, print_output

if __name__ == "__main__":
parser = get_defualt_parser()
args = parser.parse_args()
start = time.time()
torch.set_default_dtype(torch.bfloat16)

tokenizer = AutoTokenizer.from_pretrained(args.pretrained, trust_remote_code=True)

model = AutoModelForCausalLM.from_pretrained(
args.pretrained,
trust_remote_code=True,
Expand All @@ -18,10 +21,6 @@
model.eval()
init_time = time.time() - start

# A transformers-compatible version of the grok-1 tokenizer by Xenova
# https://huggingface.co/Xenova/grok-1-tokenizer
tokenizer = LlamaTokenizerFast.from_pretrained("Xenova/grok-1-tokenizer")

for text in args.text:
output = inference(
model,
Expand Down
9 changes: 4 additions & 5 deletions examples/language/grok-1/inference_tp.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@

import torch
from grok1_policy import Grok1ForCausalLMPolicy
from transformers import AutoModelForCausalLM, LlamaTokenizerFast
from transformers import AutoModelForCausalLM, AutoTokenizer
from utils import get_defualt_parser, inference, print_output

import colossalai
Expand All @@ -27,6 +27,9 @@
)
booster = Booster(plugin=plugin)
torch.set_default_dtype(torch.bfloat16)

tokenizer = AutoTokenizer.from_pretrained(args.pretrained, trust_remote_code=True)

with LazyInitContext(default_device=get_current_device()):
model = AutoModelForCausalLM.from_pretrained(
args.pretrained, trust_remote_code=True, torch_dtype=torch.bfloat16
Expand All @@ -35,10 +38,6 @@
model.eval()
init_time = time.time() - start

# A transformers-compatible version of the grok-1 tokenizer by Xenova
# https://huggingface.co/Xenova/grok-1-tokenizer
tokenizer = LlamaTokenizerFast.from_pretrained("Xenova/grok-1-tokenizer")

for text in args.text:
output = inference(
model.unwrap(),
Expand Down