upgrade cache-dit api

[DefTruth]

upgrade cache-dit api to 1.0.x

python run_benchmark.py \
    --ckpt ${CKPT} \
    --trace-file optimized_cache_dit.json.gz \
    --compile_export_mode compile \
    --disable_fa3 \
    --num_inference_steps 28 \
    --cache_dit_config cache_config.yaml \
    --output-file optimized_cache_dit.png --disable_quant

Loading checkpoint shards: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:00<00:00, 40.76it/s]
Loading pipeline components...:  71%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████▏                                                | 5/7 [00:00<00:00, 14.82it/s]You set `add_prefix_space`. The tokenizer needs to be converted from the slow tokenizers
Loading checkpoint shards: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 78.88it/s]
Loading pipeline components...: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 10.54it/s]
INFO 10-11 06:19:58 [cache_adapter.py:46] FluxPipeline is officially supported by cache-dit. Use it's pre-defined BlockAdapter directly!
INFO 10-11 06:19:58 [block_adapters.py:201] Auto fill blocks_name: ['transformer_blocks', 'single_transformer_blocks'].
INFO 10-11 06:19:58 [block_adapters.py:482] Match Block Forward Pattern: FluxTransformerBlock, ForwardPattern.Pattern_1
INFO 10-11 06:19:58 [block_adapters.py:482] IN:('hidden_states', 'encoder_hidden_states'), OUT:('encoder_hidden_states', 'hidden_states'))
INFO 10-11 06:19:58 [block_adapters.py:482] Match Block Forward Pattern: FluxSingleTransformerBlock, ForwardPattern.Pattern_1
INFO 10-11 06:19:58 [block_adapters.py:482] IN:('hidden_states', 'encoder_hidden_states'), OUT:('encoder_hidden_states', 'hidden_states'))
INFO 10-11 06:19:58 [cache_adapter.py:141] Use default 'enable_separate_cfg' from block adapter register: False, Pipeline: FluxPipeline.
INFO 10-11 06:19:58 [cache_adapter.py:275] Collected Cache Config: DBCACHE_F1B0_W0M0MC2_R0.3, Calibrator Config: TaylorSeer_O(2)
INFO 10-11 06:19:58 [cache_adapter.py:275] Collected Cache Config: DBCACHE_F1B0_W0M0MC2_R0.3, Calibrator Config: TaylorSeer_O(2)
INFO 10-11 06:19:58 [pattern_base.py:51] Match Cached Blocks: CachedBlocks_Pattern_0_1_2, for transformer_blocks, cache_context: transformer_blocks_140567643384928, cache_manager: FluxPipeline_140567649235936.
INFO 10-11 06:19:58 [pattern_base.py:51] Match Cached Blocks: CachedBlocks_Pattern_0_1_2, for single_transformer_blocks, cache_context: single_transformer_blocks_140567528265904, cache_manager: FluxPipeline_140567649235936.
time mean/var: tensor([13.3521, 13.3886, 13.4104, 13.4335, 13.4646, 13.4767, 13.4747, 13.4828,
        13.4852, 13.4868]) 13.445541381835938 0.0022438988089561462

🤗Cache Options: FluxSingleTransformerBlock

{'cache_config': BasicCacheConfig(Fn_compute_blocks=1, Bn_compute_blocks=0, residual_diff_threshold=0.3, max_warmup_steps=0, max_cached_steps=-1, max_continuous_cached_steps=2, enable_separate_cfg=False, cfg_compute_first=False, cfg_diff_compute_separate=True), 'calibrator_config': TaylorSeerCalibratorConfig(enable_calibrator=True, enable_encoder_calibrator=True, calibrator_type='taylorseer', calibrator_cache_type='residual', calibrator_kwargs={}, taylorseer_order=2), 'name': 'single_transformer_blocks_140567528265904'}

⚡️Cache Steps and Residual Diffs Statistics: FluxSingleTransformerBlock

| Cache Steps | Diffs P00 | Diffs P25 | Diffs P50 | Diffs P75 | Diffs P95 | Diffs Min | Diffs Max |
|-------------|-----------|-----------|-----------|-----------|-----------|-----------|-----------|
| 7           | 0.117     | 0.304     | 0.523     | 0.6       | 0.769     | 0.117     | 0.793     |


🤗Cache Options: FluxTransformerBlock

{'cache_config': BasicCacheConfig(Fn_compute_blocks=1, Bn_compute_blocks=0, residual_diff_threshold=0.3, max_warmup_steps=0, max_cached_steps=-1, max_continuous_cached_steps=2, enable_separate_cfg=False, cfg_compute_first=False, cfg_diff_compute_separate=True), 'calibrator_config': TaylorSeerCalibratorConfig(enable_calibrator=True, enable_encoder_calibrator=True, calibrator_type='taylorseer', calibrator_cache_type='residual', calibrator_kwargs={}, taylorseer_order=2), 'name': 'transformer_blocks_140567643384928'}

⚡️Cache Steps and Residual Diffs Statistics: FluxTransformerBlock

| Cache Steps | Diffs P00 | Diffs P25 | Diffs P50 | Diffs P75 | Diffs P95 | Diffs Min | Diffs Max |
|-------------|-----------|-----------|-----------|-----------|-----------|-----------|-----------|
| 17          | 0.034     | 0.064     | 0.111     | 0.163     | 0.312     | 0.034     | 0.32      |

• baseline: NVIDIA L20, 25s

• w/ cache-dit: NVIDIA L20, 13s

[Met4physics]

Here comes an error:

Traceback (most recent call last):
  File "/storage/flux-fast/run_benchmark.py", line 92, in <module>
    main(args)
    ~~~~^^^^^^
  File "/storage/flux-fast/run_benchmark.py", line 27, in main
    pipeline = load_pipeline(args)
  File "/storage/utils/pipeline_utils.py", line 476, in load_pipeline
    pipeline = optimize(pipeline, args)
  File "/storage/flux-fast/utils/pipeline_utils.py", line 413, in optimize
    pipeline, **cache_dit.load_options(args.cache_dit_config),
                ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^
  File "/storage/miniconda3/envs/fastflux/lib/python3.13/site-packages/cache_dit/cache_factory/utils.py", line 57, in load_options
    return load_cache_options_from_yaml(path)
  File "/storage/miniconda3/envs/fastflux/lib/python3.13/site-packages/cache_dit/cache_factory/utils.py", line 20, in load_cache_options_from_yaml
    raise ValueError(
        f"Configuration file missing required item: {key}"
    )
ValueError: Configuration file missing required item: cache_type

[DefTruth]

@Met4physics please install cache-dit==1.0.3 and re-try