You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I'm attempting to integrate LLaVA with vLLM for image processing, but I'm encountering a tensor size mismatch error when executing my script.
Setup:
I installed vLLM along with other required packages using the following command: !pip install vllm==0.4.1 kaleido python-multipart torch==2.2.1 torchvision==0.17.1 torchaudio==2.2.1
Code:
Here's the script I used to run LLaVA:
import torch
from vllm import LLM
from vllm.sequence import MultiModalData
def run_llava_pixel_values_debug():
llm = LLM(
model="llava-hf/llava-1.5-7b-hf",
enforce_eager=True,
tensor_parallel_size=1,
image_input_type="pixel_values",
image_token_id=32000,
image_input_shape=f"1,3,224,224",
image_feature_size=576,
)
prompt = "<image>" * 576 + (
"\nUSER: What is the content of this image?\nASSISTANT:")
# Load a smaller or test image file if available, or adjust the existing one to match the test size
with open("3d-background-with-hexagonal-shapes-texture_23-2150473185.jpg", "rb") as f:
image_file = f.read()
outputs = llm.generate(prompt,
multi_modal_data=MultiModalData(
type=MultiModalData.Type.IMAGE, data=encoded))
for o in outputs:
generated_text = o.outputs[0].text
print(generated_text)
run_llava_pixel_values_debug()
Error:
Upon running this script, I receive the following error: RuntimeError: The size of tensor a (257) must match the size of tensor b (577) at non-singleton dimension 1.
Could anyone assist in identifying the source of this issue and suggest how I might correct the tensor size mismatch? Any help or suggestions would be greatly appreciated.
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
-
I'm attempting to integrate LLaVA with vLLM for image processing, but I'm encountering a tensor size mismatch error when executing my script.
Setup:
I installed vLLM along with other required packages using the following command:
!pip install vllm==0.4.1 kaleido python-multipart torch==2.2.1 torchvision==0.17.1 torchaudio==2.2.1
Code:
Here's the script I used to run LLaVA:
Error:
Upon running this script, I receive the following error:
RuntimeError: The size of tensor a (257) must match the size of tensor b (577) at non-singleton dimension 1.
Could anyone assist in identifying the source of this issue and suggest how I might correct the tensor size mismatch? Any help or suggestions would be greatly appreciated.
Beta Was this translation helpful? Give feedback.
All reactions