In order to run SVD-Series in a GPU-cuda-memory-constrained environment, I choose CPU by change device to cpu, as shown before：
def sample(
input_path: str = "assets/test_image.png", # Can either be image file or folder with image files
num_frames: Optional[int] = None,
num_steps: Optional[int] = None,
version: str = "svd",
fps_id: int = 6,
motion_bucket_id: int = 127,
cond_aug: float = 0.02,
seed: int = 23,
decoding_t: int = 14, # Number of frames decoded at a time! This eats most VRAM. Reduce if necessary.
device: str = "cpu",#cuda
output_folder: Optional[str] = None,
):
And got Error:
Restored from checkpoints/svd-fp16.safetensors with 0 missing and 0 unexpected keys
Traceback (most recent call last):
File "/home/dongsongb/PYTHON-PROJ/openai/generative-models/scripts/sampling/simple_video_sample.py", line 284, in
Fire(sample)
File "/home/dongsongb/anaconda3/envs/tinygptv/lib/python3.9/site-packages/fire/core.py", line 141, in Fire
component_trace = _Fire(component, args, parsed_flag_args, context, name)
File "/home/dongsongb/anaconda3/envs/tinygptv/lib/python3.9/site-packages/fire/core.py", line 475, in _Fire
component, remaining_args = _CallAndUpdateTrace(
File "/home/dongsongb/anaconda3/envs/tinygptv/lib/python3.9/site-packages/fire/core.py", line 691, in _CallAndUpdateTrace
component = fn(*varargs, **kwargs)
File "/home/dongsongb/PYTHON-PROJ/openai/generative-models/scripts/sampling/simple_video_sample.py", line 155, in sample
c, uc = model.conditioner.get_unconditional_conditioning(
File "/home/dongsongb/anaconda3/envs/tinygptv/lib/python3.9/site-packages/sgm/modules/encoders/modules.py", line 179, in get_unconditional_conditioning
c = self(batch_c, force_cond_zero_embeddings)
File "/home/dongsongb/anaconda3/envs/tinygptv/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "/home/dongsongb/anaconda3/envs/tinygptv/lib/python3.9/site-packages/sgm/modules/encoders/modules.py", line 132, in forward
emb_out = embedder(batch[embedder.input_key])
File "/home/dongsongb/anaconda3/envs/tinygptv/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl
return forward_call(*args, **kwargs)
File "/home/dongsongb/anaconda3/envs/tinygptv/lib/python3.9/site-packages/sgm/modules/encoders/modules.py", line 1012, in forward
out = self.encoder.encode(vid[n * n_samples : (n + 1) * n_samples])
File "/home/dongsongb/anaconda3/envs/tinygptv/lib/python3.9/site-packages/sgm/models/autoencoder.py", line 472, in encode
z = self.encoder(x)
File "/home/dongsongb/anaconda3/envs/tinygptv/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1501, in call_impl
return forward_call(*args, **kwargs)
File "/home/dongsongb/anaconda3/envs/tinygptv/lib/python3.9/site-packages/sgm/modules/diffusionmodules/model.py", line 594, in forward
h = self.mid.attn_1(h)
File "/home/dongsongb/anaconda3/envs/tinygptv/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1501, in call_impl
return forward_call(*args, **kwargs)
File "/home/dongsongb/anaconda3/envs/tinygptv/lib/python3.9/site-packages/sgm/modules/diffusionmodules/model.py", line 263, in forward
h = self.attention(h)
File "/home/dongsongb/anaconda3/envs/tinygptv/lib/python3.9/site-packages/sgm/modules/diffusionmodules/model.py", line 249, in attention
out = xformers.ops.memory_efficient_attention(
File "/home/dongsongb/anaconda3/envs/tinygptv/lib/python3.9/site-packages/xformers/ops/fmha/init.py", line 223, in memory_efficient_attention
return _memory_efficient_attention(
File "/home/dongsongb/anaconda3/envs/tinygptv/lib/python3.9/site-packages/xformers/ops/fmha/init.py", line 321, in _memory_efficient_attention
return _memory_efficient_attention_forward(
File "/home/dongsongb/anaconda3/envs/tinygptv/lib/python3.9/site-packages/xformers/ops/fmha/init.py", line 337, in _memory_efficient_attention_forward
op = _dispatch_fw(inp, False)
File "/home/dongsongb/anaconda3/envs/tinygptv/lib/python3.9/site-packages/xformers/ops/fmha/dispatch.py", line 120, in _dispatch_fw
return _run_priority_list(
File "/home/dongsongb/anaconda3/envs/tinygptv/lib/python3.9/site-packages/xformers/ops/fmha/dispatch.py", line 63, in _run_priority_list
raise NotImplementedError(msg)
NotImplementedError: No operator found for memory_efficient_attention_forward with inputs:
query : shape=(1, 9216, 1, 512) (torch.bfloat16)
key : shape=(1, 9216, 1, 512) (torch.bfloat16)
value : shape=(1, 9216, 1, 512) (torch.bfloat16)
attn_bias : <class 'NoneType'>
p : 0.0
decoderF is not supported because:
max(query.shape[-1] != value.shape[-1]) > 128
device=cpu (supported: {'cuda'})
attn_bias type is <class 'NoneType'>
bf16 is only supported on A100+ GPUs
[email protected] is not supported because:
max(query.shape[-1] != value.shape[-1]) > 256
device=cpu (supported: {'cuda'})
bf16 is only supported on A100+ GPUs
tritonflashattF is not supported because:
max(query.shape[-1] != value.shape[-1]) > 128
device=cpu (supported: {'cuda'})
bf16 is only supported on A100+ GPUs
operator wasn't built - see python -m xformers.info for more info
triton is not available
cutlassF is not supported because:
device=cpu (supported: {'cuda'})
bf16 is only supported on A100+ GPUs
smallkF is not supported because:
max(query.shape[-1] != value.shape[-1]) > 32
device=cpu (supported: {'cuda'})
dtype=torch.bfloat16 (supported: {torch.float32})
bf16 is only supported on A100+ GPUs
unsupported embed per head: 512
Same problem here, @DaBaiTuu did you happen to find the workaround solution already?
Stability-AI
