[Issue]: Benchmark throwing errors when switching model to different model type #3161
Closed
2 tasks done
Comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Issue Description
when I run the benchmark with SD1.5, and then again with SDXL, SD.Next throws a lot of errors for SXDL and doesn't do anything. If I first run SDXL and then SD1.5, the same. I can restart SD.next and run SD1.5 benchmark without error, restart it again and run SDXL benchmark without error.
Possibly something like the prompt caching is interfering here?
Version Platform Description
WSL-2, Win11
Relevant log output
14:55:01-931077 DEBUG Diffuser pipeline: StableDiffusionPipeline task=DiffusersTaskType.TEXT_2_IMAGE batch=1/1x1 set={'prompt_embeds': torch.Size([1, 77, 2048]), 'negative_prompt_embeds': torch.Size([1, 77, 2048]), 'guidance_scale': 15.0, 'num_inference_steps': 20, 'eta': 1.0, 'guidance_rescale': 0.7, 'output_type': 'latent', 'width': 512, 'height': 512, 'parser': 'Full parser'} 14:55:01-938147 DEBUG Sampler: sampler="Euler a" config={'num_train_timesteps': 1000, 'beta_start': 0.00085, 'beta_end': 0.012, 'beta_schedule': 'scaled_linear', 'prediction_type': 'epsilon', 'rescale_betas_zero_snr': False, 'timestep_spacing': 'leading'} Progress ?it/s 0% 0/20 00:00 ? Base 14:55:02-001562 ERROR Processing: args={'prompt_embeds': tensor([[[-3.8203, -2.3984, 4.2031, ..., 0.1333, 0.4436, -0.2432], [-0.3086, 0.0493, -0.1787, ..., 0.1885, 0.2786, 0.2222], [-1.1934, 0.5762, -0.2339, ..., 1.2129, -0.2021, -0.3577], ..., [-0.0967, 0.7129, -0.1975, ..., -0.7070, 1.1299, 1.1602], [-0.0812, 0.7148, -0.2534, ..., -0.7793, 0.9316, 1.1914], [-0.1299, 0.6274, -0.2267, ..., -0.6494, 1.2090, 1.0352]]], device='cuda:0', dtype=torch.float16), 'negative_prompt_embeds': tensor([[[-3.8203, -2.3984, 4.2031, ..., 0.1333, 0.4436, -0.2432], [ 0.1702, 0.5649, 2.3184, ..., -1.7305, 1.9375, -1.0293], [ 0.8711, 1.4121, 1.7031, ..., -0.6777, 1.1064, 0.4819], ..., [ 0.2856, 0.1072, -0.0791, ..., 0.0754, 0.2136, 0.9551], [ 0.2856, 0.1211, -0.0730, ..., 0.0415, 0.1313, 0.9951], [ 0.3125, 0.1519, -0.0571, ..., 0.1489, 0.2529, 1.0762]]], device='cuda:0', dtype=torch.float16), 'guidance_scale': 15.0, 'generator': [<torch._C.Generator object at 0x7f3a1b582010>], 'callback_on_step_end': <function diffusers_callback at 0x7f3a474f09d0>, 'callback_on_step_end_tensor_inputs': ['latents', 'prompt_embeds', 'negative_prompt_embeds'], 'num_inference_steps': 20, 'eta': 1.0, 'guidance_rescale': 0.7, 'output_type': 'latent', 'width': 512, 'height': 512} mat1 and mat2 shapes cannot be multiplied (154x2048 and 768x320) 14:55:02-005809 ERROR Processing: RuntimeError ╭─────────────────────────────────────────────────── Traceback (most recent call last) ────────────────────────────────────────────────────╮ │ /home/tillerz/dev/modules/processing_diffusers.py:118 in process_diffusers │ │ │ │ 117 │ │ # base_args['image'] = set_latents(p) │ │ ❱ 118 │ │ output = shared.sd_model(**base_args) # pylint: disable=not-callable │ │ 119 │ │ if isinstance(output, dict): │ │ │ │ /home/tillerz/dev/venv/lib/python3.10/site-packages/torch/utils/_contextlib.py:115 in decorate_context │ │ │ │ 114 │ │ with ctx_factory(): │ │ ❱ 115 │ │ │ return func(*args, **kwargs) │ │ 116 │ │ │ │ /home/tillerz/dev/venv/lib/python3.10/site-packages/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py:971 in __call__ │ │ │ │ 970 │ │ │ │ # predict the noise residual │ │ ❱ 971 │ │ │ │ noise_pred = self.unet( │ │ 972 │ │ │ │ │ latent_model_input, │ │ │ │ /home/tillerz/dev/venv/lib/python3.10/site-packages/torch/nn/modules/module.py:1532 in _wrapped_call_impl │ │ │ │ 1531 │ │ else: │ │ ❱ 1532 │ │ │ return self._call_impl(*args, **kwargs) │ │ 1533 │ │ │ │ /home/tillerz/dev/venv/lib/python3.10/site-packages/torch/nn/modules/module.py:1541 in _call_impl │ │ │ │ 1540 │ │ │ │ or _global_forward_hooks or _global_forward_pre_hooks): │ │ ❱ 1541 │ │ │ return forward_call(*args, **kwargs) │ │ 1542 │ │ │ │ ... 12 frames hidden ... │ │ │ │ /home/tillerz/dev/venv/lib/python3.10/site-packages/diffusers/models/attention_processor.py:522 in forward │ │ │ │ 521 │ │ │ │ ❱ 522 │ │ return self.processor( │ │ 523 │ │ │ self, │ │ │ │ /home/tillerz/dev/venv/lib/python3.10/site-packages/diffusers/models/attention_processor.py:1266 in __call__ │ │ │ │ 1265 │ │ │ │ ❱ 1266 │ │ key = attn.to_k(encoder_hidden_states) │ │ 1267 │ │ value = attn.to_v(encoder_hidden_states) │ │ │ │ /home/tillerz/dev/venv/lib/python3.10/site-packages/torch/nn/modules/module.py:1532 in _wrapped_call_impl │ │ │ │ 1531 │ │ else: │ │ ❱ 1532 │ │ │ return self._call_impl(*args, **kwargs) │ │ 1533 │ │ │ │ /home/tillerz/dev/venv/lib/python3.10/site-packages/torch/nn/modules/module.py:1541 in _call_impl │ │ │ │ 1540 │ │ │ │ or _global_forward_hooks or _global_forward_pre_hooks): │ │ ❱ 1541 │ │ │ return forward_call(*args, **kwargs) │ │ 1542 │ │ │ │ /home/tillerz/dev/venv/lib/python3.10/site-packages/torch/nn/modules/linear.py:116 in forward │ │ │ │ 115 │ def forward(self, input: Tensor) -> Tensor: │ │ ❱ 116 │ │ return F.linear(input, self.weight, self.bias) │ │ 117 │ ╰──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯ RuntimeError: mat1 and mat2 shapes cannot be multiplied (154x2048 and 768x320) 14:55:02-272125 DEBUG VAE load: type=taesd model=/home/tillerz/models/TAESD/taesd_decoder.pth 14:55:02-511637 INFO Processed: images=0 time=5.36 its=0.00 memory={'ram': {'used': 8.94, 'total': 31.35}, 'gpu': {'used': 3.99, 'total': 23.99}, 'retries': 0, 'oom': 0} 14:55:02-519414 ERROR SD-System-Info benchmark error: 1 results mismatchBackend
Diffusers
Branch
Dev
Model
SD 1.5
Acknowledgements
The text was updated successfully, but these errors were encountered: