| Flax classes are deprecated and will be removed in Diffusers v0.40.0. We recommend migrating to PyTorch classes or pinning your version of Diffusers. |
| Flax classes are deprecated and will be removed in Diffusers v0.40.0. We recommend migrating to PyTorch classes or pinning your version of Diffusers. |
|
|
| Fetching 17 files: 0%| | 0/17 [00:00<?, ?it/s] |
| Fetching 17 files: 100%|██████████| 17/17 [00:00<00:00, 8112.77it/s] |
|
|
| Loading pipeline components...: 0%| | 0/7 [00:00<?, ?it/s] |
| Loading pipeline components...: 29%|██▊ | 2/7 [00:00<00:00, 5.42it/s] |
| Loading pipeline components...: 43%|████▎ | 3/7 [00:02<00:03, 1.13it/s] |
| Loading pipeline components...: 71%|███████▏ | 5/7 [00:02<00:00, 2.22it/s]There are modules in UNet2DConditionModel that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
|
|
| Loading pipeline components...: 86%|████████▌ | 6/7 [00:02<00:00, 2.33it/s]There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
|
|
| Loading pipeline components...: 100%|██████████| 7/7 [00:04<00:00, 1.45it/s] |
| Loading pipeline components...: 100%|██████████| 7/7 [00:04<00:00, 1.70it/s] |
| /usr/local/lib/python3.12/dist-packages/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl.py:748: FutureWarning: `upcast_vae` is deprecated and will be removed in version 1.0.0. `upcast_vae` is deprecated. Please use `pipe.vae.to(torch.float32)`. For more details, please refer to: https://github.com/huggingface/diffusers/pull/12619 |
| deprecate( |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| dense case 0: 1.977s |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| dense case 1: 1.923s |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| dense case 2: 1.922s |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| dense case 3: 1.913s |
|
|
| Fetching 17 files: 0%| | 0/17 [00:00<?, ?it/s] |
| Fetching 17 files: 100%|██████████| 17/17 [00:00<00:00, 23339.83it/s] |
|
|
| Loading pipeline components...: 0%| | 0/7 [00:00<?, ?it/s] |
| Loading pipeline components...: 29%|██▊ | 2/7 [00:00<00:00, 19.92it/s] |
| Loading pipeline components...: 57%|█████▋ | 4/7 [00:00<00:00, 7.16it/s]There are modules in UNet2DConditionModel that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
|
|
| Loading pipeline components...: 86%|████████▌ | 6/7 [00:00<00:00, 5.55it/s]There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
|
|
| Loading pipeline components...: 100%|██████████| 7/7 [00:01<00:00, 6.63it/s] |
|
|
| ======================================================================================== |
| structured_20 PACKAGE 1/7 targets=36 |
| ======================================================================================== |
| down_blocks.1.attentions.0.transformer_blocks.0.ff: H=2560 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.0.transformer_blocks.0.attn1: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.0.transformer_blocks.0.attn2: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.0.transformer_blocks.1.ff: H=2560 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.0.transformer_blocks.1.attn1: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.0.transformer_blocks.1.attn2: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.1.transformer_blocks.0.ff: H=2560 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.1.transformer_blocks.0.attn1: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.1.transformer_blocks.0.attn2: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.1.transformer_blocks.1.ff: H=2560 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.1.transformer_blocks.1.attn1: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.1.transformer_blocks.1.attn2: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.0.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.0.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.0.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.1.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.1.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.1.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.2.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.2.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.2.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.3.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.3.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.3.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.4.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.4.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.4.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.5.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.5.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.5.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.6.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.6.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.6.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.7.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.7.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.7.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune down_blocks.1.attentions.0.transformer_blocks.0.ff: kind=ff ratio=20% |
| prune down_blocks.1.attentions.0.transformer_blocks.0.attn1: kind=attention ratio=20% |
| prune down_blocks.1.attentions.0.transformer_blocks.0.attn2: kind=attention ratio=20% |
| prune down_blocks.1.attentions.0.transformer_blocks.1.ff: kind=ff ratio=20% |
| prune down_blocks.1.attentions.0.transformer_blocks.1.attn1: kind=attention ratio=20% |
| prune down_blocks.1.attentions.0.transformer_blocks.1.attn2: kind=attention ratio=20% |
| prune down_blocks.1.attentions.1.transformer_blocks.0.ff: kind=ff ratio=20% |
| prune down_blocks.1.attentions.1.transformer_blocks.0.attn1: kind=attention ratio=20% |
| prune down_blocks.1.attentions.1.transformer_blocks.0.attn2: kind=attention ratio=20% |
| prune down_blocks.1.attentions.1.transformer_blocks.1.ff: kind=ff ratio=20% |
| prune down_blocks.1.attentions.1.transformer_blocks.1.attn1: kind=attention ratio=20% |
| prune down_blocks.1.attentions.1.transformer_blocks.1.attn2: kind=attention ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.0.ff: kind=ff ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.0.attn1: kind=attention ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.0.attn2: kind=attention ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.1.ff: kind=ff ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.1.attn1: kind=attention ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.1.attn2: kind=attention ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.2.ff: kind=ff ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.2.attn1: kind=attention ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.2.attn2: kind=attention ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.3.ff: kind=ff ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.3.attn1: kind=attention ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.3.attn2: kind=attention ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.4.ff: kind=ff ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.4.attn1: kind=attention ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.4.attn2: kind=attention ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.5.ff: kind=ff ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.5.attn1: kind=attention ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.5.attn2: kind=attention ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.6.ff: kind=ff ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.6.attn1: kind=attention ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.6.attn2: kind=attention ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.7.ff: kind=ff ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.7.attn1: kind=attention ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.7.attn2: kind=attention ratio=20% |
|
|
| ======================================================================================== |
| structured_20 PACKAGE 2/7 targets=27 |
| ======================================================================================== |
| down_blocks.2.attentions.0.transformer_blocks.8.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.8.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.8.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.9.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.9.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.9.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.0.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.0.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.0.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.1.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.1.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.1.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.2.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.2.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.2.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.3.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.3.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.3.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.4.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.4.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.4.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.5.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.5.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.5.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.6.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.6.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.6.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune down_blocks.2.attentions.0.transformer_blocks.8.ff: kind=ff ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.8.attn1: kind=attention ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.8.attn2: kind=attention ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.9.ff: kind=ff ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.9.attn1: kind=attention ratio=20% |
| prune down_blocks.2.attentions.0.transformer_blocks.9.attn2: kind=attention ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.0.ff: kind=ff ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.0.attn1: kind=attention ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.0.attn2: kind=attention ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.1.ff: kind=ff ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.1.attn1: kind=attention ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.1.attn2: kind=attention ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.2.ff: kind=ff ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.2.attn1: kind=attention ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.2.attn2: kind=attention ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.3.ff: kind=ff ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.3.attn1: kind=attention ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.3.attn2: kind=attention ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.4.ff: kind=ff ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.4.attn1: kind=attention ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.4.attn2: kind=attention ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.5.ff: kind=ff ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.5.attn1: kind=attention ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.5.attn2: kind=attention ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.6.ff: kind=ff ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.6.attn1: kind=attention ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.6.attn2: kind=attention ratio=20% |
|
|
| ======================================================================================== |
| structured_20 PACKAGE 3/7 targets=27 |
| ======================================================================================== |
| down_blocks.2.attentions.1.transformer_blocks.7.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.7.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.7.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.8.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.8.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.8.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.9.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.9.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.9.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.0.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.0.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.0.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.1.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.1.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.1.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.2.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.2.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.2.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.3.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.3.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.3.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.4.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.4.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.4.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.5.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.5.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.5.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune down_blocks.2.attentions.1.transformer_blocks.7.ff: kind=ff ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.7.attn1: kind=attention ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.7.attn2: kind=attention ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.8.ff: kind=ff ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.8.attn1: kind=attention ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.8.attn2: kind=attention ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.9.ff: kind=ff ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.9.attn1: kind=attention ratio=20% |
| prune down_blocks.2.attentions.1.transformer_blocks.9.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.0.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.0.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.0.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.1.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.1.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.1.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.2.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.2.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.2.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.3.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.3.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.3.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.4.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.4.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.4.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.5.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.5.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.5.attn2: kind=attention ratio=20% |
|
|
| ======================================================================================== |
| structured_20 PACKAGE 4/7 targets=27 |
| ======================================================================================== |
| up_blocks.0.attentions.0.transformer_blocks.6.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.6.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.6.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.7.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.7.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.7.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.8.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.8.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.8.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.9.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.9.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.9.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.0.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.0.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.0.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.1.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.1.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.1.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.2.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.2.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.2.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.3.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.3.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.3.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.4.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.4.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.4.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune up_blocks.0.attentions.0.transformer_blocks.6.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.6.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.6.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.7.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.7.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.7.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.8.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.8.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.8.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.9.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.9.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.0.transformer_blocks.9.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.0.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.0.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.0.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.1.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.1.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.1.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.2.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.2.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.2.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.3.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.3.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.3.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.4.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.4.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.4.attn2: kind=attention ratio=20% |
|
|
| ======================================================================================== |
| structured_20 PACKAGE 5/7 targets=27 |
| ======================================================================================== |
| up_blocks.0.attentions.1.transformer_blocks.5.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.5.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.5.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.6.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.6.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.6.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.7.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.7.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.7.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.8.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.8.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.8.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.9.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.9.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.9.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.0.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.0.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.0.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.1.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.1.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.1.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.2.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.2.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.2.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.3.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.3.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.3.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune up_blocks.0.attentions.1.transformer_blocks.5.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.5.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.5.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.6.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.6.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.6.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.7.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.7.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.7.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.8.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.8.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.8.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.9.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.9.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.1.transformer_blocks.9.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.0.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.0.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.0.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.1.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.1.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.1.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.2.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.2.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.2.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.3.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.3.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.3.attn2: kind=attention ratio=20% |
|
|
| ======================================================================================== |
| structured_20 PACKAGE 6/7 targets=39 |
| ======================================================================================== |
| up_blocks.0.attentions.2.transformer_blocks.4.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.4.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.4.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.5.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.5.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.5.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.6.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.6.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.6.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.7.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.7.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.7.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.8.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.8.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.8.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.9.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.9.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.9.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.0.transformer_blocks.0.ff: H=2560 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.0.transformer_blocks.0.attn1: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.0.transformer_blocks.0.attn2: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.0.transformer_blocks.1.ff: H=2560 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.0.transformer_blocks.1.attn1: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.0.transformer_blocks.1.attn2: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.1.transformer_blocks.0.ff: H=2560 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.1.transformer_blocks.0.attn1: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.1.transformer_blocks.0.attn2: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.1.transformer_blocks.1.ff: H=2560 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.1.transformer_blocks.1.attn1: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.1.transformer_blocks.1.attn2: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.2.transformer_blocks.0.ff: H=2560 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.2.transformer_blocks.0.attn1: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.2.transformer_blocks.0.attn2: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.2.transformer_blocks.1.ff: H=2560 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.2.transformer_blocks.1.attn1: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.2.transformer_blocks.1.attn2: H=640 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.0.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.0.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.0.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune up_blocks.0.attentions.2.transformer_blocks.4.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.4.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.4.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.5.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.5.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.5.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.6.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.6.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.6.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.7.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.7.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.7.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.8.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.8.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.8.attn2: kind=attention ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.9.ff: kind=ff ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.9.attn1: kind=attention ratio=20% |
| prune up_blocks.0.attentions.2.transformer_blocks.9.attn2: kind=attention ratio=20% |
| prune up_blocks.1.attentions.0.transformer_blocks.0.ff: kind=ff ratio=20% |
| prune up_blocks.1.attentions.0.transformer_blocks.0.attn1: kind=attention ratio=20% |
| prune up_blocks.1.attentions.0.transformer_blocks.0.attn2: kind=attention ratio=20% |
| prune up_blocks.1.attentions.0.transformer_blocks.1.ff: kind=ff ratio=20% |
| prune up_blocks.1.attentions.0.transformer_blocks.1.attn1: kind=attention ratio=20% |
| prune up_blocks.1.attentions.0.transformer_blocks.1.attn2: kind=attention ratio=20% |
| prune up_blocks.1.attentions.1.transformer_blocks.0.ff: kind=ff ratio=20% |
| prune up_blocks.1.attentions.1.transformer_blocks.0.attn1: kind=attention ratio=20% |
| prune up_blocks.1.attentions.1.transformer_blocks.0.attn2: kind=attention ratio=20% |
| prune up_blocks.1.attentions.1.transformer_blocks.1.ff: kind=ff ratio=20% |
| prune up_blocks.1.attentions.1.transformer_blocks.1.attn1: kind=attention ratio=20% |
| prune up_blocks.1.attentions.1.transformer_blocks.1.attn2: kind=attention ratio=20% |
| prune up_blocks.1.attentions.2.transformer_blocks.0.ff: kind=ff ratio=20% |
| prune up_blocks.1.attentions.2.transformer_blocks.0.attn1: kind=attention ratio=20% |
| prune up_blocks.1.attentions.2.transformer_blocks.0.attn2: kind=attention ratio=20% |
| prune up_blocks.1.attentions.2.transformer_blocks.1.ff: kind=ff ratio=20% |
| prune up_blocks.1.attentions.2.transformer_blocks.1.attn1: kind=attention ratio=20% |
| prune up_blocks.1.attentions.2.transformer_blocks.1.attn2: kind=attention ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.0.ff: kind=ff ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.0.attn1: kind=attention ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.0.attn2: kind=attention ratio=20% |
|
|
| ======================================================================================== |
| structured_20 PACKAGE 7/7 targets=27 |
| ======================================================================================== |
| mid_block.attentions.0.transformer_blocks.1.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.1.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.1.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.2.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.2.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.2.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.3.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.3.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.3.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.4.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.4.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.4.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.5.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.5.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.5.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.6.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.6.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.6.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.7.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.7.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.7.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.8.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.8.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.8.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.9.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.9.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.9.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune mid_block.attentions.0.transformer_blocks.1.ff: kind=ff ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.1.attn1: kind=attention ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.1.attn2: kind=attention ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.2.ff: kind=ff ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.2.attn1: kind=attention ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.2.attn2: kind=attention ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.3.ff: kind=ff ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.3.attn1: kind=attention ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.3.attn2: kind=attention ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.4.ff: kind=ff ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.4.attn1: kind=attention ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.4.attn2: kind=attention ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.5.ff: kind=ff ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.5.attn1: kind=attention ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.5.attn2: kind=attention ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.6.ff: kind=ff ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.6.attn1: kind=attention ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.6.attn2: kind=attention ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.7.ff: kind=ff ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.7.attn1: kind=attention ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.7.attn2: kind=attention ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.8.ff: kind=ff ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.8.attn1: kind=attention ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.8.attn2: kind=attention ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.9.ff: kind=ff ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.9.attn1: kind=attention ratio=20% |
| prune mid_block.attentions.0.transformer_blocks.9.attn2: kind=attention ratio=20% |
| structured_20: 2,130,533,124 params (2.1305B), physical reduction 17.0180% |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| structured_20 case 0: 1.902s |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| structured_20 case 1: 1.872s |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| structured_20 case 2: 1.871s |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| structured_20 case 3: 1.856s |
| There are modules in UNet2DConditionModel that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| standalone torch.load verification: PASS |
|
|
| Fetching 17 files: 0%| | 0/17 [00:00<?, ?it/s] |
| Fetching 17 files: 100%|██████████| 17/17 [00:00<00:00, 22773.29it/s] |
|
|
| Loading pipeline components...: 0%| | 0/7 [00:00<?, ?it/s] |
| Loading pipeline components...: 29%|██▊ | 2/7 [00:00<00:00, 19.89it/s] |
| Loading pipeline components...: 57%|█████▋ | 4/7 [00:00<00:00, 7.75it/s]There are modules in UNet2DConditionModel that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
|
|
| Loading pipeline components...: 86%|████████▌ | 6/7 [00:00<00:00, 5.79it/s]There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
|
|
| Loading pipeline components...: 100%|██████████| 7/7 [00:01<00:00, 7.00it/s] |
|
|
| ======================================================================================== |
| structured_30 PACKAGE 1/7 targets=36 |
| ======================================================================================== |
| down_blocks.1.attentions.0.transformer_blocks.0.ff: H=2560 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.0.transformer_blocks.0.attn1: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.0.transformer_blocks.0.attn2: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.0.transformer_blocks.1.ff: H=2560 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.0.transformer_blocks.1.attn1: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.0.transformer_blocks.1.attn2: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.1.transformer_blocks.0.ff: H=2560 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.1.transformer_blocks.0.attn1: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.1.transformer_blocks.0.attn2: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.1.transformer_blocks.1.ff: H=2560 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.1.transformer_blocks.1.attn1: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.1.transformer_blocks.1.attn2: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.0.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.0.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.0.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.1.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.1.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.1.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.2.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.2.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.2.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.3.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.3.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.3.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.4.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.4.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.4.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.5.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.5.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.5.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.6.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.6.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.6.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.7.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.7.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.7.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune down_blocks.1.attentions.0.transformer_blocks.0.ff: kind=ff ratio=30% |
| prune down_blocks.1.attentions.0.transformer_blocks.0.attn1: kind=attention ratio=30% |
| prune down_blocks.1.attentions.0.transformer_blocks.0.attn2: kind=attention ratio=30% |
| prune down_blocks.1.attentions.0.transformer_blocks.1.ff: kind=ff ratio=30% |
| prune down_blocks.1.attentions.0.transformer_blocks.1.attn1: kind=attention ratio=30% |
| prune down_blocks.1.attentions.0.transformer_blocks.1.attn2: kind=attention ratio=30% |
| prune down_blocks.1.attentions.1.transformer_blocks.0.ff: kind=ff ratio=30% |
| prune down_blocks.1.attentions.1.transformer_blocks.0.attn1: kind=attention ratio=30% |
| prune down_blocks.1.attentions.1.transformer_blocks.0.attn2: kind=attention ratio=30% |
| prune down_blocks.1.attentions.1.transformer_blocks.1.ff: kind=ff ratio=30% |
| prune down_blocks.1.attentions.1.transformer_blocks.1.attn1: kind=attention ratio=30% |
| prune down_blocks.1.attentions.1.transformer_blocks.1.attn2: kind=attention ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.0.ff: kind=ff ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.0.attn1: kind=attention ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.0.attn2: kind=attention ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.1.ff: kind=ff ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.1.attn1: kind=attention ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.1.attn2: kind=attention ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.2.ff: kind=ff ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.2.attn1: kind=attention ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.2.attn2: kind=attention ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.3.ff: kind=ff ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.3.attn1: kind=attention ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.3.attn2: kind=attention ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.4.ff: kind=ff ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.4.attn1: kind=attention ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.4.attn2: kind=attention ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.5.ff: kind=ff ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.5.attn1: kind=attention ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.5.attn2: kind=attention ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.6.ff: kind=ff ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.6.attn1: kind=attention ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.6.attn2: kind=attention ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.7.ff: kind=ff ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.7.attn1: kind=attention ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.7.attn2: kind=attention ratio=30% |
|
|
| ======================================================================================== |
| structured_30 PACKAGE 2/7 targets=27 |
| ======================================================================================== |
| down_blocks.2.attentions.0.transformer_blocks.8.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.8.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.8.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.9.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.9.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.9.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.0.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.0.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.0.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.1.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.1.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.1.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.2.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.2.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.2.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.3.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.3.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.3.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.4.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.4.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.4.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.5.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.5.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.5.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.6.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.6.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.6.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune down_blocks.2.attentions.0.transformer_blocks.8.ff: kind=ff ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.8.attn1: kind=attention ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.8.attn2: kind=attention ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.9.ff: kind=ff ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.9.attn1: kind=attention ratio=30% |
| prune down_blocks.2.attentions.0.transformer_blocks.9.attn2: kind=attention ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.0.ff: kind=ff ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.0.attn1: kind=attention ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.0.attn2: kind=attention ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.1.ff: kind=ff ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.1.attn1: kind=attention ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.1.attn2: kind=attention ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.2.ff: kind=ff ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.2.attn1: kind=attention ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.2.attn2: kind=attention ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.3.ff: kind=ff ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.3.attn1: kind=attention ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.3.attn2: kind=attention ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.4.ff: kind=ff ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.4.attn1: kind=attention ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.4.attn2: kind=attention ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.5.ff: kind=ff ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.5.attn1: kind=attention ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.5.attn2: kind=attention ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.6.ff: kind=ff ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.6.attn1: kind=attention ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.6.attn2: kind=attention ratio=30% |
|
|
| ======================================================================================== |
| structured_30 PACKAGE 3/7 targets=27 |
| ======================================================================================== |
| down_blocks.2.attentions.1.transformer_blocks.7.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.7.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.7.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.8.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.8.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.8.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.9.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.9.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.9.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.0.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.0.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.0.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.1.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.1.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.1.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.2.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.2.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.2.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.3.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.3.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.3.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.4.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.4.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.4.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.5.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.5.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.5.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune down_blocks.2.attentions.1.transformer_blocks.7.ff: kind=ff ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.7.attn1: kind=attention ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.7.attn2: kind=attention ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.8.ff: kind=ff ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.8.attn1: kind=attention ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.8.attn2: kind=attention ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.9.ff: kind=ff ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.9.attn1: kind=attention ratio=30% |
| prune down_blocks.2.attentions.1.transformer_blocks.9.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.0.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.0.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.0.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.1.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.1.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.1.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.2.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.2.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.2.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.3.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.3.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.3.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.4.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.4.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.4.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.5.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.5.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.5.attn2: kind=attention ratio=30% |
|
|
| ======================================================================================== |
| structured_30 PACKAGE 4/7 targets=27 |
| ======================================================================================== |
| up_blocks.0.attentions.0.transformer_blocks.6.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.6.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.6.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.7.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.7.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.7.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.8.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.8.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.8.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.9.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.9.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.9.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.0.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.0.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.0.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.1.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.1.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.1.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.2.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.2.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.2.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.3.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.3.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.3.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.4.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.4.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.4.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune up_blocks.0.attentions.0.transformer_blocks.6.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.6.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.6.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.7.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.7.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.7.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.8.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.8.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.8.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.9.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.9.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.0.transformer_blocks.9.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.0.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.0.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.0.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.1.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.1.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.1.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.2.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.2.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.2.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.3.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.3.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.3.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.4.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.4.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.4.attn2: kind=attention ratio=30% |
|
|
| ======================================================================================== |
| structured_30 PACKAGE 5/7 targets=27 |
| ======================================================================================== |
| up_blocks.0.attentions.1.transformer_blocks.5.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.5.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.5.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.6.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.6.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.6.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.7.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.7.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.7.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.8.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.8.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.8.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.9.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.9.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.9.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.0.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.0.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.0.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.1.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.1.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.1.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.2.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.2.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.2.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.3.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.3.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.3.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune up_blocks.0.attentions.1.transformer_blocks.5.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.5.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.5.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.6.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.6.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.6.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.7.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.7.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.7.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.8.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.8.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.8.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.9.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.9.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.1.transformer_blocks.9.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.0.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.0.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.0.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.1.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.1.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.1.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.2.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.2.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.2.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.3.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.3.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.3.attn2: kind=attention ratio=30% |
|
|
| ======================================================================================== |
| structured_30 PACKAGE 6/7 targets=39 |
| ======================================================================================== |
| up_blocks.0.attentions.2.transformer_blocks.4.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.4.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.4.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.5.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.5.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.5.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.6.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.6.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.6.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.7.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.7.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.7.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.8.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.8.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.8.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.9.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.9.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.9.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.0.transformer_blocks.0.ff: H=2560 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.0.transformer_blocks.0.attn1: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.0.transformer_blocks.0.attn2: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.0.transformer_blocks.1.ff: H=2560 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.0.transformer_blocks.1.attn1: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.0.transformer_blocks.1.attn2: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.1.transformer_blocks.0.ff: H=2560 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.1.transformer_blocks.0.attn1: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.1.transformer_blocks.0.attn2: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.1.transformer_blocks.1.ff: H=2560 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.1.transformer_blocks.1.attn1: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.1.transformer_blocks.1.attn2: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.2.transformer_blocks.0.ff: H=2560 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.2.transformer_blocks.0.attn1: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.2.transformer_blocks.0.attn2: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.2.transformer_blocks.1.ff: H=2560 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.2.transformer_blocks.1.attn1: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.2.transformer_blocks.1.attn2: H=640 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.0.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.0.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.0.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune up_blocks.0.attentions.2.transformer_blocks.4.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.4.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.4.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.5.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.5.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.5.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.6.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.6.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.6.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.7.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.7.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.7.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.8.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.8.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.8.attn2: kind=attention ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.9.ff: kind=ff ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.9.attn1: kind=attention ratio=30% |
| prune up_blocks.0.attentions.2.transformer_blocks.9.attn2: kind=attention ratio=30% |
| prune up_blocks.1.attentions.0.transformer_blocks.0.ff: kind=ff ratio=30% |
| prune up_blocks.1.attentions.0.transformer_blocks.0.attn1: kind=attention ratio=30% |
| prune up_blocks.1.attentions.0.transformer_blocks.0.attn2: kind=attention ratio=30% |
| prune up_blocks.1.attentions.0.transformer_blocks.1.ff: kind=ff ratio=30% |
| prune up_blocks.1.attentions.0.transformer_blocks.1.attn1: kind=attention ratio=30% |
| prune up_blocks.1.attentions.0.transformer_blocks.1.attn2: kind=attention ratio=30% |
| prune up_blocks.1.attentions.1.transformer_blocks.0.ff: kind=ff ratio=30% |
| prune up_blocks.1.attentions.1.transformer_blocks.0.attn1: kind=attention ratio=30% |
| prune up_blocks.1.attentions.1.transformer_blocks.0.attn2: kind=attention ratio=30% |
| prune up_blocks.1.attentions.1.transformer_blocks.1.ff: kind=ff ratio=30% |
| prune up_blocks.1.attentions.1.transformer_blocks.1.attn1: kind=attention ratio=30% |
| prune up_blocks.1.attentions.1.transformer_blocks.1.attn2: kind=attention ratio=30% |
| prune up_blocks.1.attentions.2.transformer_blocks.0.ff: kind=ff ratio=30% |
| prune up_blocks.1.attentions.2.transformer_blocks.0.attn1: kind=attention ratio=30% |
| prune up_blocks.1.attentions.2.transformer_blocks.0.attn2: kind=attention ratio=30% |
| prune up_blocks.1.attentions.2.transformer_blocks.1.ff: kind=ff ratio=30% |
| prune up_blocks.1.attentions.2.transformer_blocks.1.attn1: kind=attention ratio=30% |
| prune up_blocks.1.attentions.2.transformer_blocks.1.attn2: kind=attention ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.0.ff: kind=ff ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.0.attn1: kind=attention ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.0.attn2: kind=attention ratio=30% |
|
|
| ======================================================================================== |
| structured_30 PACKAGE 7/7 targets=27 |
| ======================================================================================== |
| mid_block.attentions.0.transformer_blocks.1.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.1.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.1.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.2.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.2.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.2.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.3.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.3.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.3.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.4.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.4.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.4.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.5.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.5.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.5.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.6.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.6.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.6.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.7.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.7.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.7.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.8.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.8.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.8.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.9.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.9.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.9.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune mid_block.attentions.0.transformer_blocks.1.ff: kind=ff ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.1.attn1: kind=attention ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.1.attn2: kind=attention ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.2.ff: kind=ff ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.2.attn1: kind=attention ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.2.attn2: kind=attention ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.3.ff: kind=ff ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.3.attn1: kind=attention ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.3.attn2: kind=attention ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.4.ff: kind=ff ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.4.attn1: kind=attention ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.4.attn2: kind=attention ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.5.ff: kind=ff ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.5.attn1: kind=attention ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.5.attn2: kind=attention ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.6.ff: kind=ff ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.6.attn1: kind=attention ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.6.attn2: kind=attention ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.7.ff: kind=ff ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.7.attn1: kind=attention ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.7.attn2: kind=attention ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.8.ff: kind=ff ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.8.attn1: kind=attention ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.8.attn2: kind=attention ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.9.ff: kind=ff ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.9.attn1: kind=attention ratio=30% |
| prune mid_block.attentions.0.transformer_blocks.9.attn2: kind=attention ratio=30% |
| structured_30: 1,912,067,844 params (1.9121B), physical reduction 25.5270% |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| structured_30 case 0: 1.808s |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| structured_30 case 1: 1.773s |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| structured_30 case 2: 1.779s |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| structured_30 case 3: 1.778s |
| There are modules in UNet2DConditionModel that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| standalone torch.load verification: PASS |
|
|
| Fetching 17 files: 0%| | 0/17 [00:00<?, ?it/s] |
| Fetching 17 files: 100%|██████████| 17/17 [00:00<00:00, 19911.52it/s] |
|
|
| Loading pipeline components...: 0%| | 0/7 [00:00<?, ?it/s] |
| Loading pipeline components...: 29%|██▊ | 2/7 [00:00<00:00, 19.92it/s] |
| Loading pipeline components...: 57%|█████▋ | 4/7 [00:00<00:00, 7.30it/s]There are modules in UNet2DConditionModel that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
|
|
| Loading pipeline components...: 86%|████████▌ | 6/7 [00:00<00:00, 5.54it/s]There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
|
|
| Loading pipeline components...: 100%|██████████| 7/7 [00:01<00:00, 6.75it/s] |
|
|
| ======================================================================================== |
| structured_40 PACKAGE 1/7 targets=36 |
| ======================================================================================== |
| down_blocks.1.attentions.0.transformer_blocks.0.ff: H=2560 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.0.transformer_blocks.0.attn1: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.0.transformer_blocks.0.attn2: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.0.transformer_blocks.1.ff: H=2560 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.0.transformer_blocks.1.attn1: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.0.transformer_blocks.1.attn2: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.1.transformer_blocks.0.ff: H=2560 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.1.transformer_blocks.0.attn1: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.1.transformer_blocks.0.attn2: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.1.transformer_blocks.1.ff: H=2560 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.1.transformer_blocks.1.attn1: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.1.transformer_blocks.1.attn2: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.0.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.0.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.0.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.1.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.1.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.1.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.2.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.2.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.2.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.3.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.3.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.3.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.4.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.4.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.4.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.5.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.5.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.5.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.6.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.6.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.6.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.7.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.7.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.7.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune down_blocks.1.attentions.0.transformer_blocks.0.ff: kind=ff ratio=40% |
| prune down_blocks.1.attentions.0.transformer_blocks.0.attn1: kind=attention ratio=40% |
| prune down_blocks.1.attentions.0.transformer_blocks.0.attn2: kind=attention ratio=40% |
| prune down_blocks.1.attentions.0.transformer_blocks.1.ff: kind=ff ratio=40% |
| prune down_blocks.1.attentions.0.transformer_blocks.1.attn1: kind=attention ratio=40% |
| prune down_blocks.1.attentions.0.transformer_blocks.1.attn2: kind=attention ratio=40% |
| prune down_blocks.1.attentions.1.transformer_blocks.0.ff: kind=ff ratio=40% |
| prune down_blocks.1.attentions.1.transformer_blocks.0.attn1: kind=attention ratio=40% |
| prune down_blocks.1.attentions.1.transformer_blocks.0.attn2: kind=attention ratio=40% |
| prune down_blocks.1.attentions.1.transformer_blocks.1.ff: kind=ff ratio=40% |
| prune down_blocks.1.attentions.1.transformer_blocks.1.attn1: kind=attention ratio=40% |
| prune down_blocks.1.attentions.1.transformer_blocks.1.attn2: kind=attention ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.0.ff: kind=ff ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.0.attn1: kind=attention ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.0.attn2: kind=attention ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.1.ff: kind=ff ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.1.attn1: kind=attention ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.1.attn2: kind=attention ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.2.ff: kind=ff ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.2.attn1: kind=attention ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.2.attn2: kind=attention ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.3.ff: kind=ff ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.3.attn1: kind=attention ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.3.attn2: kind=attention ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.4.ff: kind=ff ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.4.attn1: kind=attention ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.4.attn2: kind=attention ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.5.ff: kind=ff ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.5.attn1: kind=attention ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.5.attn2: kind=attention ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.6.ff: kind=ff ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.6.attn1: kind=attention ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.6.attn2: kind=attention ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.7.ff: kind=ff ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.7.attn1: kind=attention ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.7.attn2: kind=attention ratio=40% |
|
|
| ======================================================================================== |
| structured_40 PACKAGE 2/7 targets=27 |
| ======================================================================================== |
| down_blocks.2.attentions.0.transformer_blocks.8.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.8.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.8.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.9.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.9.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.9.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.0.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.0.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.0.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.1.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.1.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.1.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.2.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.2.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.2.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.3.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.3.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.3.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.4.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.4.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.4.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.5.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.5.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.5.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.6.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.6.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.6.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune down_blocks.2.attentions.0.transformer_blocks.8.ff: kind=ff ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.8.attn1: kind=attention ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.8.attn2: kind=attention ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.9.ff: kind=ff ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.9.attn1: kind=attention ratio=40% |
| prune down_blocks.2.attentions.0.transformer_blocks.9.attn2: kind=attention ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.0.ff: kind=ff ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.0.attn1: kind=attention ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.0.attn2: kind=attention ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.1.ff: kind=ff ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.1.attn1: kind=attention ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.1.attn2: kind=attention ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.2.ff: kind=ff ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.2.attn1: kind=attention ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.2.attn2: kind=attention ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.3.ff: kind=ff ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.3.attn1: kind=attention ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.3.attn2: kind=attention ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.4.ff: kind=ff ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.4.attn1: kind=attention ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.4.attn2: kind=attention ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.5.ff: kind=ff ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.5.attn1: kind=attention ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.5.attn2: kind=attention ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.6.ff: kind=ff ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.6.attn1: kind=attention ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.6.attn2: kind=attention ratio=40% |
|
|
| ======================================================================================== |
| structured_40 PACKAGE 3/7 targets=27 |
| ======================================================================================== |
| down_blocks.2.attentions.1.transformer_blocks.7.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.7.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.7.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.8.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.8.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.8.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.9.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.9.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.9.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.0.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.0.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.0.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.1.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.1.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.1.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.2.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.2.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.2.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.3.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.3.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.3.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.4.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.4.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.4.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.5.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.5.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.5.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune down_blocks.2.attentions.1.transformer_blocks.7.ff: kind=ff ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.7.attn1: kind=attention ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.7.attn2: kind=attention ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.8.ff: kind=ff ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.8.attn1: kind=attention ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.8.attn2: kind=attention ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.9.ff: kind=ff ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.9.attn1: kind=attention ratio=40% |
| prune down_blocks.2.attentions.1.transformer_blocks.9.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.0.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.0.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.0.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.1.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.1.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.1.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.2.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.2.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.2.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.3.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.3.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.3.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.4.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.4.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.4.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.5.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.5.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.5.attn2: kind=attention ratio=40% |
|
|
| ======================================================================================== |
| structured_40 PACKAGE 4/7 targets=27 |
| ======================================================================================== |
| up_blocks.0.attentions.0.transformer_blocks.6.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.6.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.6.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.7.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.7.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.7.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.8.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.8.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.8.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.9.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.9.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.9.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.0.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.0.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.0.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.1.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.1.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.1.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.2.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.2.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.2.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.3.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.3.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.3.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.4.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.4.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.4.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune up_blocks.0.attentions.0.transformer_blocks.6.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.6.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.6.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.7.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.7.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.7.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.8.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.8.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.8.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.9.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.9.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.0.transformer_blocks.9.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.0.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.0.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.0.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.1.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.1.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.1.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.2.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.2.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.2.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.3.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.3.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.3.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.4.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.4.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.4.attn2: kind=attention ratio=40% |
|
|
| ======================================================================================== |
| structured_40 PACKAGE 5/7 targets=27 |
| ======================================================================================== |
| up_blocks.0.attentions.1.transformer_blocks.5.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.5.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.5.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.6.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.6.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.6.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.7.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.7.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.7.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.8.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.8.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.8.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.9.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.9.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.9.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.0.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.0.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.0.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.1.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.1.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.1.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.2.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.2.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.2.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.3.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.3.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.3.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune up_blocks.0.attentions.1.transformer_blocks.5.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.5.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.5.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.6.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.6.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.6.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.7.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.7.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.7.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.8.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.8.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.8.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.9.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.9.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.1.transformer_blocks.9.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.0.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.0.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.0.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.1.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.1.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.1.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.2.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.2.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.2.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.3.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.3.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.3.attn2: kind=attention ratio=40% |
|
|
| ======================================================================================== |
| structured_40 PACKAGE 6/7 targets=39 |
| ======================================================================================== |
| up_blocks.0.attentions.2.transformer_blocks.4.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.4.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.4.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.5.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.5.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.5.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.6.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.6.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.6.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.7.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.7.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.7.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.8.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.8.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.8.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.9.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.9.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.9.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.0.transformer_blocks.0.ff: H=2560 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.0.transformer_blocks.0.attn1: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.0.transformer_blocks.0.attn2: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.0.transformer_blocks.1.ff: H=2560 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.0.transformer_blocks.1.attn1: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.0.transformer_blocks.1.attn2: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.1.transformer_blocks.0.ff: H=2560 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.1.transformer_blocks.0.attn1: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.1.transformer_blocks.0.attn2: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.1.transformer_blocks.1.ff: H=2560 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.1.transformer_blocks.1.attn1: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.1.transformer_blocks.1.attn2: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.2.transformer_blocks.0.ff: H=2560 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.2.transformer_blocks.0.attn1: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.2.transformer_blocks.0.attn2: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.2.transformer_blocks.1.ff: H=2560 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.2.transformer_blocks.1.attn1: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.2.transformer_blocks.1.attn2: H=640 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.0.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.0.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.0.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune up_blocks.0.attentions.2.transformer_blocks.4.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.4.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.4.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.5.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.5.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.5.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.6.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.6.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.6.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.7.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.7.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.7.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.8.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.8.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.8.attn2: kind=attention ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.9.ff: kind=ff ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.9.attn1: kind=attention ratio=40% |
| prune up_blocks.0.attentions.2.transformer_blocks.9.attn2: kind=attention ratio=40% |
| prune up_blocks.1.attentions.0.transformer_blocks.0.ff: kind=ff ratio=40% |
| prune up_blocks.1.attentions.0.transformer_blocks.0.attn1: kind=attention ratio=40% |
| prune up_blocks.1.attentions.0.transformer_blocks.0.attn2: kind=attention ratio=40% |
| prune up_blocks.1.attentions.0.transformer_blocks.1.ff: kind=ff ratio=40% |
| prune up_blocks.1.attentions.0.transformer_blocks.1.attn1: kind=attention ratio=40% |
| prune up_blocks.1.attentions.0.transformer_blocks.1.attn2: kind=attention ratio=40% |
| prune up_blocks.1.attentions.1.transformer_blocks.0.ff: kind=ff ratio=40% |
| prune up_blocks.1.attentions.1.transformer_blocks.0.attn1: kind=attention ratio=40% |
| prune up_blocks.1.attentions.1.transformer_blocks.0.attn2: kind=attention ratio=40% |
| prune up_blocks.1.attentions.1.transformer_blocks.1.ff: kind=ff ratio=40% |
| prune up_blocks.1.attentions.1.transformer_blocks.1.attn1: kind=attention ratio=40% |
| prune up_blocks.1.attentions.1.transformer_blocks.1.attn2: kind=attention ratio=40% |
| prune up_blocks.1.attentions.2.transformer_blocks.0.ff: kind=ff ratio=40% |
| prune up_blocks.1.attentions.2.transformer_blocks.0.attn1: kind=attention ratio=40% |
| prune up_blocks.1.attentions.2.transformer_blocks.0.attn2: kind=attention ratio=40% |
| prune up_blocks.1.attentions.2.transformer_blocks.1.ff: kind=ff ratio=40% |
| prune up_blocks.1.attentions.2.transformer_blocks.1.attn1: kind=attention ratio=40% |
| prune up_blocks.1.attentions.2.transformer_blocks.1.attn2: kind=attention ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.0.ff: kind=ff ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.0.attn1: kind=attention ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.0.attn2: kind=attention ratio=40% |
|
|
| ======================================================================================== |
| structured_40 PACKAGE 7/7 targets=27 |
| ======================================================================================== |
| mid_block.attentions.0.transformer_blocks.1.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.1.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.1.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.2.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.2.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.2.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.3.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.3.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.3.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.4.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.4.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.4.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.5.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.5.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.5.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.6.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.6.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.6.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.7.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.7.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.7.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.8.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.8.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.8.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.9.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.9.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.9.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune mid_block.attentions.0.transformer_blocks.1.ff: kind=ff ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.1.attn1: kind=attention ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.1.attn2: kind=attention ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.2.ff: kind=ff ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.2.attn1: kind=attention ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.2.attn2: kind=attention ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.3.ff: kind=ff ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.3.attn1: kind=attention ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.3.attn2: kind=attention ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.4.ff: kind=ff ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.4.attn1: kind=attention ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.4.attn2: kind=attention ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.5.ff: kind=ff ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.5.attn1: kind=attention ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.5.attn2: kind=attention ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.6.ff: kind=ff ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.6.attn1: kind=attention ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.6.attn2: kind=attention ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.7.ff: kind=ff ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.7.attn1: kind=attention ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.7.attn2: kind=attention ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.8.ff: kind=ff ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.8.attn1: kind=attention ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.8.attn2: kind=attention ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.9.ff: kind=ff ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.9.attn1: kind=attention ratio=40% |
| prune mid_block.attentions.0.transformer_blocks.9.attn2: kind=attention ratio=40% |
| structured_40: 1,693,602,564 params (1.6936B), physical reduction 34.0360% |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| structured_40 case 0: 1.742s |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| structured_40 case 1: 1.731s |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| structured_40 case 2: 1.732s |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| structured_40 case 3: 1.731s |
| There are modules in UNet2DConditionModel that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| standalone torch.load verification: PASS |
|
|
| Fetching 17 files: 0%| | 0/17 [00:00<?, ?it/s] |
| Fetching 17 files: 100%|██████████| 17/17 [00:00<00:00, 23736.07it/s] |
|
|
| Loading pipeline components...: 0%| | 0/7 [00:00<?, ?it/s] |
| Loading pipeline components...: 43%|████▎ | 3/7 [00:00<00:00, 7.69it/s] |
| Loading pipeline components...: 71%|███████▏ | 5/7 [00:00<00:00, 8.89it/s]There are modules in UNet2DConditionModel that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
|
|
| Loading pipeline components...: 86%|████████▌ | 6/7 [00:00<00:00, 5.56it/s]There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
|
|
| Loading pipeline components...: 100%|██████████| 7/7 [00:01<00:00, 6.83it/s] |
|
|
| ======================================================================================== |
| structured_50 PACKAGE 1/7 targets=36 |
| ======================================================================================== |
| down_blocks.1.attentions.0.transformer_blocks.0.ff: H=2560 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.0.transformer_blocks.0.attn1: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.0.transformer_blocks.0.attn2: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.0.transformer_blocks.1.ff: H=2560 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.0.transformer_blocks.1.attn1: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.0.transformer_blocks.1.attn2: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.1.transformer_blocks.0.ff: H=2560 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.1.transformer_blocks.0.attn1: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.1.transformer_blocks.0.attn2: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.1.transformer_blocks.1.ff: H=2560 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.1.transformer_blocks.1.attn1: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.1.attentions.1.transformer_blocks.1.attn2: H=640 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.0.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.0.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.0.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.1.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.1.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.1.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.2.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.2.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.2.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.3.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.3.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.3.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.4.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.4.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.4.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.5.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.5.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.5.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.6.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.6.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.6.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.7.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.7.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.7.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune down_blocks.1.attentions.0.transformer_blocks.0.ff: kind=ff ratio=50% |
| prune down_blocks.1.attentions.0.transformer_blocks.0.attn1: kind=attention ratio=50% |
| prune down_blocks.1.attentions.0.transformer_blocks.0.attn2: kind=attention ratio=50% |
| prune down_blocks.1.attentions.0.transformer_blocks.1.ff: kind=ff ratio=50% |
| prune down_blocks.1.attentions.0.transformer_blocks.1.attn1: kind=attention ratio=50% |
| prune down_blocks.1.attentions.0.transformer_blocks.1.attn2: kind=attention ratio=50% |
| prune down_blocks.1.attentions.1.transformer_blocks.0.ff: kind=ff ratio=50% |
| prune down_blocks.1.attentions.1.transformer_blocks.0.attn1: kind=attention ratio=50% |
| prune down_blocks.1.attentions.1.transformer_blocks.0.attn2: kind=attention ratio=50% |
| prune down_blocks.1.attentions.1.transformer_blocks.1.ff: kind=ff ratio=50% |
| prune down_blocks.1.attentions.1.transformer_blocks.1.attn1: kind=attention ratio=50% |
| prune down_blocks.1.attentions.1.transformer_blocks.1.attn2: kind=attention ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.0.ff: kind=ff ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.0.attn1: kind=attention ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.0.attn2: kind=attention ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.1.ff: kind=ff ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.1.attn1: kind=attention ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.1.attn2: kind=attention ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.2.ff: kind=ff ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.2.attn1: kind=attention ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.2.attn2: kind=attention ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.3.ff: kind=ff ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.3.attn1: kind=attention ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.3.attn2: kind=attention ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.4.ff: kind=ff ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.4.attn1: kind=attention ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.4.attn2: kind=attention ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.5.ff: kind=ff ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.5.attn1: kind=attention ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.5.attn2: kind=attention ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.6.ff: kind=ff ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.6.attn1: kind=attention ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.6.attn2: kind=attention ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.7.ff: kind=ff ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.7.attn1: kind=attention ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.7.attn2: kind=attention ratio=50% |
|
|
| ======================================================================================== |
| structured_50 PACKAGE 2/7 targets=27 |
| ======================================================================================== |
| down_blocks.2.attentions.0.transformer_blocks.8.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.8.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.8.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.9.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.9.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.0.transformer_blocks.9.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.0.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.0.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.0.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.1.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.1.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.1.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.2.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.2.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.2.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.3.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.3.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.3.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.4.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.4.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.4.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.5.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.5.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.5.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.6.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.6.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.6.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune down_blocks.2.attentions.0.transformer_blocks.8.ff: kind=ff ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.8.attn1: kind=attention ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.8.attn2: kind=attention ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.9.ff: kind=ff ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.9.attn1: kind=attention ratio=50% |
| prune down_blocks.2.attentions.0.transformer_blocks.9.attn2: kind=attention ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.0.ff: kind=ff ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.0.attn1: kind=attention ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.0.attn2: kind=attention ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.1.ff: kind=ff ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.1.attn1: kind=attention ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.1.attn2: kind=attention ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.2.ff: kind=ff ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.2.attn1: kind=attention ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.2.attn2: kind=attention ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.3.ff: kind=ff ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.3.attn1: kind=attention ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.3.attn2: kind=attention ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.4.ff: kind=ff ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.4.attn1: kind=attention ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.4.attn2: kind=attention ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.5.ff: kind=ff ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.5.attn1: kind=attention ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.5.attn2: kind=attention ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.6.ff: kind=ff ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.6.attn1: kind=attention ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.6.attn2: kind=attention ratio=50% |
|
|
| ======================================================================================== |
| structured_50 PACKAGE 3/7 targets=27 |
| ======================================================================================== |
| down_blocks.2.attentions.1.transformer_blocks.7.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.7.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.7.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.8.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.8.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.8.attn2: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.9.ff: H=5120 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.9.attn1: H=1280 calls=64 sampled_tokens=8192 |
| down_blocks.2.attentions.1.transformer_blocks.9.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.0.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.0.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.0.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.1.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.1.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.1.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.2.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.2.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.2.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.3.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.3.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.3.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.4.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.4.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.4.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.5.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.5.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.5.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune down_blocks.2.attentions.1.transformer_blocks.7.ff: kind=ff ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.7.attn1: kind=attention ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.7.attn2: kind=attention ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.8.ff: kind=ff ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.8.attn1: kind=attention ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.8.attn2: kind=attention ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.9.ff: kind=ff ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.9.attn1: kind=attention ratio=50% |
| prune down_blocks.2.attentions.1.transformer_blocks.9.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.0.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.0.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.0.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.1.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.1.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.1.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.2.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.2.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.2.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.3.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.3.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.3.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.4.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.4.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.4.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.5.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.5.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.5.attn2: kind=attention ratio=50% |
|
|
| ======================================================================================== |
| structured_50 PACKAGE 4/7 targets=27 |
| ======================================================================================== |
| up_blocks.0.attentions.0.transformer_blocks.6.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.6.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.6.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.7.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.7.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.7.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.8.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.8.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.8.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.9.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.9.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.0.transformer_blocks.9.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.0.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.0.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.0.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.1.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.1.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.1.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.2.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.2.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.2.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.3.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.3.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.3.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.4.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.4.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.4.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune up_blocks.0.attentions.0.transformer_blocks.6.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.6.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.6.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.7.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.7.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.7.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.8.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.8.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.8.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.9.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.9.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.0.transformer_blocks.9.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.0.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.0.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.0.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.1.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.1.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.1.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.2.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.2.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.2.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.3.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.3.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.3.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.4.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.4.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.4.attn2: kind=attention ratio=50% |
|
|
| ======================================================================================== |
| structured_50 PACKAGE 5/7 targets=27 |
| ======================================================================================== |
| up_blocks.0.attentions.1.transformer_blocks.5.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.5.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.5.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.6.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.6.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.6.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.7.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.7.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.7.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.8.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.8.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.8.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.9.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.9.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.1.transformer_blocks.9.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.0.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.0.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.0.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.1.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.1.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.1.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.2.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.2.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.2.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.3.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.3.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.3.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune up_blocks.0.attentions.1.transformer_blocks.5.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.5.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.5.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.6.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.6.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.6.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.7.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.7.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.7.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.8.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.8.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.8.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.9.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.9.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.1.transformer_blocks.9.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.0.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.0.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.0.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.1.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.1.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.1.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.2.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.2.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.2.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.3.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.3.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.3.attn2: kind=attention ratio=50% |
|
|
| ======================================================================================== |
| structured_50 PACKAGE 6/7 targets=39 |
| ======================================================================================== |
| up_blocks.0.attentions.2.transformer_blocks.4.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.4.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.4.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.5.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.5.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.5.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.6.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.6.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.6.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.7.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.7.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.7.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.8.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.8.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.8.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.9.ff: H=5120 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.9.attn1: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.0.attentions.2.transformer_blocks.9.attn2: H=1280 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.0.transformer_blocks.0.ff: H=2560 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.0.transformer_blocks.0.attn1: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.0.transformer_blocks.0.attn2: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.0.transformer_blocks.1.ff: H=2560 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.0.transformer_blocks.1.attn1: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.0.transformer_blocks.1.attn2: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.1.transformer_blocks.0.ff: H=2560 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.1.transformer_blocks.0.attn1: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.1.transformer_blocks.0.attn2: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.1.transformer_blocks.1.ff: H=2560 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.1.transformer_blocks.1.attn1: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.1.transformer_blocks.1.attn2: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.2.transformer_blocks.0.ff: H=2560 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.2.transformer_blocks.0.attn1: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.2.transformer_blocks.0.attn2: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.2.transformer_blocks.1.ff: H=2560 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.2.transformer_blocks.1.attn1: H=640 calls=64 sampled_tokens=8192 |
| up_blocks.1.attentions.2.transformer_blocks.1.attn2: H=640 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.0.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.0.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.0.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune up_blocks.0.attentions.2.transformer_blocks.4.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.4.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.4.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.5.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.5.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.5.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.6.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.6.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.6.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.7.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.7.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.7.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.8.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.8.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.8.attn2: kind=attention ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.9.ff: kind=ff ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.9.attn1: kind=attention ratio=50% |
| prune up_blocks.0.attentions.2.transformer_blocks.9.attn2: kind=attention ratio=50% |
| prune up_blocks.1.attentions.0.transformer_blocks.0.ff: kind=ff ratio=50% |
| prune up_blocks.1.attentions.0.transformer_blocks.0.attn1: kind=attention ratio=50% |
| prune up_blocks.1.attentions.0.transformer_blocks.0.attn2: kind=attention ratio=50% |
| prune up_blocks.1.attentions.0.transformer_blocks.1.ff: kind=ff ratio=50% |
| prune up_blocks.1.attentions.0.transformer_blocks.1.attn1: kind=attention ratio=50% |
| prune up_blocks.1.attentions.0.transformer_blocks.1.attn2: kind=attention ratio=50% |
| prune up_blocks.1.attentions.1.transformer_blocks.0.ff: kind=ff ratio=50% |
| prune up_blocks.1.attentions.1.transformer_blocks.0.attn1: kind=attention ratio=50% |
| prune up_blocks.1.attentions.1.transformer_blocks.0.attn2: kind=attention ratio=50% |
| prune up_blocks.1.attentions.1.transformer_blocks.1.ff: kind=ff ratio=50% |
| prune up_blocks.1.attentions.1.transformer_blocks.1.attn1: kind=attention ratio=50% |
| prune up_blocks.1.attentions.1.transformer_blocks.1.attn2: kind=attention ratio=50% |
| prune up_blocks.1.attentions.2.transformer_blocks.0.ff: kind=ff ratio=50% |
| prune up_blocks.1.attentions.2.transformer_blocks.0.attn1: kind=attention ratio=50% |
| prune up_blocks.1.attentions.2.transformer_blocks.0.attn2: kind=attention ratio=50% |
| prune up_blocks.1.attentions.2.transformer_blocks.1.ff: kind=ff ratio=50% |
| prune up_blocks.1.attentions.2.transformer_blocks.1.attn1: kind=attention ratio=50% |
| prune up_blocks.1.attentions.2.transformer_blocks.1.attn2: kind=attention ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.0.ff: kind=ff ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.0.attn1: kind=attention ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.0.attn2: kind=attention ratio=50% |
|
|
| ======================================================================================== |
| structured_50 PACKAGE 7/7 targets=27 |
| ======================================================================================== |
| mid_block.attentions.0.transformer_blocks.1.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.1.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.1.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.2.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.2.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.2.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.3.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.3.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.3.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.4.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.4.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.4.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.5.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.5.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.5.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.6.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.6.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.6.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.7.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.7.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.7.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.8.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.8.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.8.attn2: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.9.ff: H=5120 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.9.attn1: H=1280 calls=64 sampled_tokens=8192 |
| mid_block.attentions.0.transformer_blocks.9.attn2: H=1280 calls=64 sampled_tokens=8192 |
| prune mid_block.attentions.0.transformer_blocks.1.ff: kind=ff ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.1.attn1: kind=attention ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.1.attn2: kind=attention ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.2.ff: kind=ff ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.2.attn1: kind=attention ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.2.attn2: kind=attention ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.3.ff: kind=ff ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.3.attn1: kind=attention ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.3.attn2: kind=attention ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.4.ff: kind=ff ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.4.attn1: kind=attention ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.4.attn2: kind=attention ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.5.ff: kind=ff ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.5.attn1: kind=attention ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.5.attn2: kind=attention ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.6.ff: kind=ff ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.6.attn1: kind=attention ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.6.attn2: kind=attention ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.7.ff: kind=ff ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.7.attn1: kind=attention ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.7.attn2: kind=attention ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.8.ff: kind=ff ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.8.attn1: kind=attention ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.8.attn2: kind=attention ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.9.ff: kind=ff ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.9.attn1: kind=attention ratio=50% |
| prune mid_block.attentions.0.transformer_blocks.9.attn2: kind=attention ratio=50% |
| structured_50: 1,475,137,284 params (1.4751B), physical reduction 42.5450% |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| structured_50 case 0: 1.696s |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| structured_50 case 1: 1.688s |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| structured_50 case 2: 1.697s |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| There are modules in AutoencoderKL that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| structured_50 case 3: 1.691s |
| There are modules in UNet2DConditionModel that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in float32. |
| standalone torch.load verification: PASS |
|
|
| ======================================================================================== |
| STRUCTURED OBS-DIFF SDXL COMPLETE |
| ======================================================================================== |
| dense params=2,567,463,684 reduction=0.000% pth=0.000 GiB mean=1.9338s |
| structured_20 params=2,130,533,124 reduction=17.018% pth=3.969 GiB mean=1.8754s |
| structured_30 params=1,912,067,844 reduction=25.527% pth=3.562 GiB mean=1.7842s |
| structured_40 params=1,693,602,564 reduction=34.036% pth=3.155 GiB mean=1.7340s |
| structured_50 params=1,475,137,284 reduction=42.545% pth=2.749 GiB mean=1.6932s |
| JSON: /content/obs_diff_sdxl_structured_results/obs_sdxl_structured_compare.json |
| HTML: /content/obs_diff_sdxl_structured_results/obs_sdxl_structured_compare.html |
| UNets: /content/obs_diff_sdxl_structured_results/unets |
|
|
| RETURN CODE: 0 |
|
|
