返回 JoyAI-Echo
inference.fp4.yaml
根目录 / echo_longvideo / configs / inference.fp4.yaml
1 # Echo 1.5 packed ModelOpt FP4 DMD inference configuration.
2
3 paths:
4 checkpoint: checkpoints/echo15_fp4
5 gemma_path: checkpoints/gemma-3-12b
6 requests_dir: examples/the_last_visa/requests
7 requests_glob: "*.json"
8 output_root: inference_result
9 conditioning_cache_dir:
10
11 video:
12 num_frames: 241
13 height: 736
14 width: 1280
15 fps: 25
16 seed: 42
17
18 video_vae:
19 decode_mode: tiled
20 tile_size_frames: 64
21 tile_overlap_frames: 24
22 tile_size_pixels: 512
23 tile_overlap_pixels: 64
24
25 denoising:
26 steps: [1000, 994, 988, 981, 975, 909, 725, 422, 0]
27 sigmas: [1.0, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875, 0.0]
28
29 memory:
30 max_size: 7
31 downscale_factor: 1
32 enable_audio: true
33 position_mode: slot_center
34 position_offset: 500.0
35 position_slot_stride: 50.0
36 voice_filter:
37 enabled: true
38 backend: msst_speech
39 min_output_rms: 0.004
40 msst_dir: third_party/MSST-WebUI
41 msst_model_path: checkpoints/msst/model_bandit_plus_dnr_sdr_11.47.chpt
42 msst_config_path: third_party/MSST-WebUI/configs_backup/multi_stem_models/model_bandit_plus_dnr_sdr_11.47.chpt.yaml
43 msst_model_type: bandit
44 msst_sample_rate: 44100
45 msst_device: auto
46 msst_local_rank_env: LOCAL_RANK
47
48 inference:
49 device: cuda
50 dtype: bfloat16
51 prompt_max_chars: 1500
52 text_batch_size: 1
53 image_batch_size: 1
54 audio_batch_size: 1
55
55 lines YAML