Describe the bug
Traceback (most recent call last):
File "evaclip_infer.py", line 86, in
main()
File "evaclip_infer.py", line 61, in main
image_features = model.encode_image(image)
File "/mnt/EVA/EVA-CLIP/rei/eva_clip/model.py", line 297, in encode_image
features = self.visual(image)
File "/opt/conda/envs/torch2/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "/mnt/EVA/EVA-CLIP/rei/eva_clip/eva_vit_model.py", line 523, in forward
x = self.forward_features(x)
File "/mnt/EVA/EVA-CLIP/rei/eva_clip/eva_vit_model.py", line 510, in forward_features
x = blk(x, rel_pos_bias=rel_pos_bias)
File "/opt/conda/envs/torch2/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "/mnt/EVA/EVA-CLIP/rei/eva_clip/eva_vit_model.py", line 287, in forward
x = x + self.drop_path(self.attn(self.norm1(x), rel_pos_bias=rel_pos_bias, attn_mask=attn_mask))
File "/opt/conda/envs/torch2/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "/mnt/EVA/EVA-CLIP/rei/eva_clip/eva_vit_model.py", line 202, in forward
x = xops.memory_efficient_attention(
File "/opt/conda/envs/torch2/lib/python3.8/site-packages/xformers/ops/fmha/init.py", line 192, in memory_efficient_attention
return _memory_efficient_attention(
File "/opt/conda/envs/torch2/lib/python3.8/site-packages/xformers/ops/fmha/init.py", line 295, in _memory_efficient_attention
return _fMHA.apply(
File "/opt/conda/envs/torch2/lib/python3.8/site-packages/torch/autograd/function.py", line 506, in apply
return super().apply(*args, **kwargs) # type: ignore[misc]
File "/opt/conda/envs/torch2/lib/python3.8/site-packages/xformers/ops/fmha/init.py", line 41, in forward
out, op_ctx = _memory_efficient_attention_forward_requires_grad(
File "/opt/conda/envs/torch2/lib/python3.8/site-packages/xformers/ops/fmha/init.py", line 320, in _memory_efficient_attention_forward_requires_grad
op = _dispatch_fw(inp)
File "/opt/conda/envs/torch2/lib/python3.8/site-packages/xformers/ops/fmha/dispatch.py", line 104, in _dispatch_fw
return _run_priority_list(
File "/opt/conda/envs/torch2/lib/python3.8/site-packages/xformers/ops/fmha/dispatch.py", line 79, in _run_priority_list
raise NotImplementedError(msg)
NotImplementedError: No operator found for memory_efficient_attention_forward with inputs:
query : shape=(1, 197, 12, 64) (torch.float32)
key : shape=(1, 197, 12, 64) (torch.float32)
value : shape=(1, 197, 12, 64) (torch.float32)
attn_bias : <class 'NoneType'>
p : 0.0
cutlassF is not supported because:
device=cpu (supported: {'cuda'})
flshattF is not supported because:
device=cpu (supported: {'cuda'})
dtype=torch.float32 (supported: {torch.bfloat16, torch.float16})
tritonflashattF is not supported because:
device=cpu (supported: {'cuda'})
dtype=torch.float32 (supported: {torch.bfloat16, torch.float16})
smallkF is not supported because:
max(query.shape[-1] != value.shape[-1]) > 32
has custom scale
To Reproduce
The command you executed.
import torch
from eva_clip import create_model_and_transforms, get_tokenizer
from PIL import Image
model_name = "EVA02-CLIP-B-16"
pretrained = "eva_clip" # or "/path/to/EVA02_CLIP_B_psz16_s8B.pt"
image_path = "CLIP.png"
caption = ["a diagram", "a dog", "a cat"]
device = "cuda" if torch.cuda.is_available() else "cpu"
model, _, preprocess = create_model_and_transforms(model_name, pretrained, force_custom_clip=True)
tokenizer = get_tokenizer(model_name)
model = model.to(device)
image = preprocess(Image.open(image_path)).unsqueeze(0).to(device)
text = tokenizer(["a diagram", "a dog", "a cat"]).to(device)
with torch.no_grad(), torch.cuda.amp.autocast():
image_features = model.encode_image(image)
text_features = model.encode_text(text)
image_features /= image_features.norm(dim=-1, keepdim=True)
text_features /= text_features.norm(dim=-1, keepdim=True)
text_probs = (100.0 * image_features @ text_features.T).softmax(dim=-1)
print("Label probs:", text_probs) # prints: [[0.8275, 0.1372, 0.0352]]
Post related information
PyTorch: 2.0.1+cu116
Additional context
[here]
Add any other context about the problem here.
[here]
baaivision
