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

fix cuda UT torch_dtype #423

Merged
merged 2 commits into from
Jan 24, 2025
Merged

fix cuda UT torch_dtype #423

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
4337f29
fix cuda UT torch_dtype
WeiweiZhang1 Jan 24, 2025
ead838d
fix
n1ck-guo Jan 24, 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
11 changes: 6 additions & 5 deletions test_cuda/test_support_vlms.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,7 +35,7 @@ def test_qwen2(self):
from transformers import Qwen2VLForConditionalGeneration, AutoProcessor
model = Qwen2VLForConditionalGeneration.from_pretrained(
quantized_model_path,
torch_dtype="auto",
torch_dtype="float16",
device_map=f"cuda:{self.device}",
)
processor = AutoProcessor.from_pretrained(quantized_model_path)
Expand Down Expand Up @@ -92,7 +92,7 @@ def test_phi3(self):
quantized_model_path,
device_map=f"cuda:{self.device}",
trust_remote_code=True,
torch_dtype="auto"
torch_dtype="float16",
)
processor = AutoProcessor.from_pretrained(quantized_model_path,
trust_remote_code=True,
Expand Down Expand Up @@ -244,7 +244,7 @@ def test_llama(self):
quantized_model_path = os.path.join(self.save_dir, "Llama-3.2-11B-Vision-Instruct-w4g128-auto_round")
model = MllamaForConditionalGeneration.from_pretrained(
quantized_model_path,
torch_dtype="auto",
torch_dtype="float16",
device_map=f"cuda:{self.device}",
)
processor = AutoProcessor.from_pretrained(quantized_model_path)
Expand Down Expand Up @@ -288,7 +288,7 @@ def test_cogvlm(self):
)
model = AutoModelForCausalLM.from_pretrained(
quantized_model_path,
torch_dtype="auto",
torch_dtype="float16",
trust_remote_code=True,
device_map=DEVICE,
).to(DEVICE).eval()
Expand Down Expand Up @@ -352,7 +352,7 @@ def test_deepseek_vl2(self):
quantized_model_path,
trust_remote_code=True,
device_map=f"cuda:{self.device}",
torch_dtype="auto",
torch_dtype="float16",
)
vl_gpt = vl_gpt.eval()

Expand Down Expand Up @@ -399,3 +399,4 @@ def test_deepseek_vl2(self):

if __name__ == "__main__":
unittest.main()

7 changes: 4 additions & 3 deletions test_cuda/test_vlms.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
import copy
iimport copy
import shutil
import sys
import unittest
Expand Down Expand Up @@ -46,7 +46,7 @@ def qwen_inference(self, quantized_model_dir):
processor = AutoProcessor.from_pretrained(quantized_model_dir, trust_remote_code=True)
model = Qwen2VLForConditionalGeneration.from_pretrained(
quantized_model_dir,
torch_dtype="auto",
torch_dtype="float16",
device_map="auto",
##revision="df7f44c" ##AutoGPTQ format
)
Expand Down Expand Up @@ -120,7 +120,7 @@ def phi3_infernece(self, quantized_model_dir):
quantized_model_path,
device_map="auto",
trust_remote_code=True,
torch_dtype="auto"
torch_dtype="float16",
)
processor = AutoProcessor.from_pretrained(quantized_model_path,
trust_remote_code=True,
Expand Down Expand Up @@ -198,3 +198,4 @@ def test_quant_not_text_fp_layers(self):

if __name__ == "__main__":
unittest.main()