Spaces:
kadirnar
/
Running on Zero

ZeST / ip_adapter /test_resampler.py
fffiloni's picture
Upload 47 files
a9289c0
import torch
from resampler import Resampler
from transformers import CLIPVisionModel
BATCH_SIZE = 2
OUTPUT_DIM = 1280
NUM_QUERIES = 8
NUM_LATENTS_MEAN_POOLED = 4 # 0 for no mean pooling (previous behavior)
APPLY_POS_EMB = True # False for no positional embeddings (previous behavior)
IMAGE_ENCODER_NAME_OR_PATH = "laion/CLIP-ViT-H-14-laion2B-s32B-b79K"
def main():
image_encoder = CLIPVisionModel.from_pretrained(IMAGE_ENCODER_NAME_OR_PATH)
embedding_dim = image_encoder.config.hidden_size
print(f"image_encoder hidden size: ", embedding_dim)
image_proj_model = Resampler(
dim=1024,
depth=2,
dim_head=64,
heads=16,
num_queries=NUM_QUERIES,
embedding_dim=embedding_dim,
output_dim=OUTPUT_DIM,
ff_mult=2,
max_seq_len=257,
apply_pos_emb=APPLY_POS_EMB,
num_latents_mean_pooled=NUM_LATENTS_MEAN_POOLED,
)
dummy_images = torch.randn(BATCH_SIZE, 3, 224, 224)
with torch.no_grad():
image_embeds = image_encoder(dummy_images, output_hidden_states=True).hidden_states[-2]
print("image_embds shape: ", image_embeds.shape)
with torch.no_grad():
ip_tokens = image_proj_model(image_embeds)
print("ip_tokens shape:", ip_tokens.shape)
assert ip_tokens.shape == (BATCH_SIZE, NUM_QUERIES + NUM_LATENTS_MEAN_POOLED, OUTPUT_DIM)
if __name__ == "__main__":
main()