EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

[Fix] Fix clean_up_tokenization_spaces in tokenizer (#1510)

Browse Source
This commit is contained in:
Lianmin Zheng
2024-09-24 21:37:33 -07:00
committed by GitHub
parent 067d8e16fc
commit fb2d0680e0
4 changed files with 12 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
python/sglang/srt/hf_transformers_utils.py
View File

@@ -129,6 +129,7 @@ def get_tokenizer(
*args, *args,
trust_remote_code=trust_remote_code, trust_remote_code=trust_remote_code,
tokenizer_revision=tokenizer_revision, tokenizer_revision=tokenizer_revision,
clean_up_tokenization_spaces=False,
**kwargs, **kwargs,
) )
except TypeError as e: except TypeError as e:

9
python/sglang/test/runners.py
View File

@@ -21,8 +21,9 @@ from typing import List, Union
import torch import torch
import torch.nn.functional as F import torch.nn.functional as F
from transformers import AutoModelForCausalLM, AutoTokenizer from transformers import AutoModelForCausalLM
from sglang.srt.hf_transformers_utils import get_tokenizer
from sglang.srt.server import Runtime from sglang.srt.server import Runtime
from sglang.test.test_utils import DEFAULT_PORT_FOR_SRT_TEST_RUNNER from sglang.test.test_utils import DEFAULT_PORT_FOR_SRT_TEST_RUNNER
@@ -92,11 +93,7 @@ class HFRunner:
self.model_proc.start() self.model_proc.start()
def start_model_process(self, in_queue, out_queue, model_path, torch_dtype): def start_model_process(self, in_queue, out_queue, model_path, torch_dtype):
self.tokenizer = AutoTokenizer.from_pretrained( self.tokenizer = get_tokenizer(model_path)
model_path,
torch_dtype=torch_dtype,
)
if self.is_generation: if self.is_generation:
self.base_model = AutoModelForCausalLM.from_pretrained( self.base_model = AutoModelForCausalLM.from_pretrained(
model_path, model_path,

6
scripts/playground/reference_hf.py
View File

@@ -26,12 +26,14 @@ I'm going to the
import argparse import argparse
import torch import torch
from transformers import AutoModelForCausalLM, AutoTokenizer from transformers import AutoModelForCausalLM
from sglang.srt.hf_transformers_utils import get_tokenizer
@torch.inference_mode() @torch.inference_mode()
def normal_text(args): def normal_text(args):
t = AutoTokenizer.from_pretrained(args.model_path, trust_remote_code=True) t = get_tokenizer(args.model_path, trust_remote_code=True)
m = AutoModelForCausalLM.from_pretrained( m = AutoModelForCausalLM.from_pretrained(
args.model_path, args.model_path,
torch_dtype=torch.float16, torch_dtype=torch.float16,

5
test/srt/models/test_generation_models.py
View File

@@ -30,7 +30,7 @@ from typing import List
import torch import torch
from sglang.test.runners import DEFAULT_PROMPTS, HFRunner, SRTRunner from sglang.test.runners import DEFAULT_PROMPTS, HFRunner, SRTRunner
from sglang.test.test_utils import calculate_rouge_l from sglang.test.test_utils import calculate_rouge_l, is_in_ci
@dataclasses.dataclass @dataclasses.dataclass
@@ -132,6 +132,9 @@ class TestGenerationModels(unittest.TestCase):
) )
def test_others(self): def test_others(self):
if is_in_ci():
return
for model_case in ALL_OTHER_MODELS: for model_case in ALL_OTHER_MODELS:
if ( if (
"ONLY_RUN" in os.environ "ONLY_RUN" in os.environ