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.

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

Add an option to use the relevant tokenizer #81

Open
wants to merge 3 commits into
base: main
Choose a base branch
from
+14 −4
Open

Add an option to use the relevant tokenizer #81

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
ab84ceb
adding a tokenizer flag to use the relevant tokenizer
pinak-p Jan 6, 2025
ec97f3c
use AutoTokenizer class
pinak-p Jan 14, 2025
a24f20c
revert to using a tokenizer passed through the tokenizer argument
pinak-p Jan 21, 2025
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
18 changes: 14 additions & 4 deletions token_benchmark_ray.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,7 +24,7 @@
)
from tqdm import tqdm

from transformers import LlamaTokenizerFast
from transformers import AutoTokenizer

def get_token_throughput_latencies(
model: str,
Expand All @@ -34,6 +34,7 @@ def get_token_throughput_latencies(
stddev_output_tokens: int,
additional_sampling_params: Optional[Dict[str, Any]] = None,
num_concurrent_requests: int = 1,
tokenizer: str = "hf-internal-testing/llama-tokenizer",
max_num_completed_requests: int = 500,
test_timeout_s=90,
llm_api="openai",
Expand All @@ -60,9 +61,7 @@ def get_token_throughput_latencies(
"""
random.seed(11111)

tokenizer = LlamaTokenizerFast.from_pretrained(
"hf-internal-testing/llama-tokenizer"
)
tokenizer = AutoTokenizer.from_pretrained(tokenizer)
get_token_length = lambda text: len(tokenizer.encode(text))

if not additional_sampling_params:
Expand Down Expand Up @@ -292,6 +291,7 @@ def run_token_benchmark(
additional_sampling_params: str,
results_dir: str,
user_metadata: Dict[str, Any],
tokenizer: str
):
"""
Args:
Expand Down Expand Up @@ -327,6 +327,7 @@ def run_token_benchmark(
stddev_output_tokens=stddev_output_tokens,
num_concurrent_requests=num_concurrent_requests,
additional_sampling_params=json.loads(additional_sampling_params),
tokenizer=tokenizer
)

if results_dir:
Expand Down Expand Up @@ -462,6 +463,14 @@ def run_token_benchmark(
"name=foo,bar=1. These will be added to the metadata field of the results. "
),
)
args.add_argument(
"--tokenizer",
type=str,
default="hf-internal-testing/llama-tokenizer",
help=(
"Tokenizer to use for counting tokens"
),
)

if __name__ == "__main__":
env_vars = dict(os.environ)
Expand All @@ -488,4 +497,5 @@ def run_token_benchmark(
additional_sampling_params=args.additional_sampling_params,
results_dir=args.results_dir,
user_metadata=user_metadata,
tokenizer=args.tokenizer
)