Upload 001-lora_sft_baseline-merged_ckpt-1084/video_preprocessor_config.json with huggingface_hub
Browse files
001-lora_sft_baseline-merged_ckpt-1084/video_preprocessor_config.json
ADDED
|
@@ -0,0 +1,98 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_valid_kwargs_names": [
|
| 3 |
+
"do_convert_rgb",
|
| 4 |
+
"do_resize",
|
| 5 |
+
"size",
|
| 6 |
+
"size_divisor",
|
| 7 |
+
"default_to_square",
|
| 8 |
+
"resample",
|
| 9 |
+
"do_rescale",
|
| 10 |
+
"rescale_factor",
|
| 11 |
+
"do_normalize",
|
| 12 |
+
"image_mean",
|
| 13 |
+
"image_std",
|
| 14 |
+
"do_pad",
|
| 15 |
+
"do_center_crop",
|
| 16 |
+
"crop_size",
|
| 17 |
+
"data_format",
|
| 18 |
+
"input_data_format",
|
| 19 |
+
"device",
|
| 20 |
+
"min_pixels",
|
| 21 |
+
"max_pixels",
|
| 22 |
+
"patch_size",
|
| 23 |
+
"temporal_patch_size",
|
| 24 |
+
"merge_size"
|
| 25 |
+
],
|
| 26 |
+
"chunk_length": 300,
|
| 27 |
+
"crop_size": null,
|
| 28 |
+
"data_format": "channels_first",
|
| 29 |
+
"default_to_square": true,
|
| 30 |
+
"device": null,
|
| 31 |
+
"dither": 0.0,
|
| 32 |
+
"do_center_crop": null,
|
| 33 |
+
"do_convert_rgb": true,
|
| 34 |
+
"do_normalize": true,
|
| 35 |
+
"do_pad": null,
|
| 36 |
+
"do_rescale": true,
|
| 37 |
+
"do_resize": true,
|
| 38 |
+
"feature_extractor_type": "WhisperFeatureExtractor",
|
| 39 |
+
"feature_size": 128,
|
| 40 |
+
"hop_length": 160,
|
| 41 |
+
"image_mean": [
|
| 42 |
+
0.48145466,
|
| 43 |
+
0.4578275,
|
| 44 |
+
0.40821073
|
| 45 |
+
],
|
| 46 |
+
"image_processor_type": "Qwen2VLImageProcessor",
|
| 47 |
+
"image_std": [
|
| 48 |
+
0.26862954,
|
| 49 |
+
0.26130258,
|
| 50 |
+
0.27577711
|
| 51 |
+
],
|
| 52 |
+
"input_data_format": null,
|
| 53 |
+
"max_pixels": 12845056,
|
| 54 |
+
"merge_size": 2,
|
| 55 |
+
"min_pixels": 3136,
|
| 56 |
+
"model_valid_processing_keys": [
|
| 57 |
+
"do_convert_rgb",
|
| 58 |
+
"do_resize",
|
| 59 |
+
"size",
|
| 60 |
+
"size_divisor",
|
| 61 |
+
"default_to_square",
|
| 62 |
+
"resample",
|
| 63 |
+
"do_rescale",
|
| 64 |
+
"rescale_factor",
|
| 65 |
+
"do_normalize",
|
| 66 |
+
"image_mean",
|
| 67 |
+
"image_std",
|
| 68 |
+
"do_pad",
|
| 69 |
+
"do_center_crop",
|
| 70 |
+
"crop_size",
|
| 71 |
+
"data_format",
|
| 72 |
+
"input_data_format",
|
| 73 |
+
"device",
|
| 74 |
+
"min_pixels",
|
| 75 |
+
"max_pixels",
|
| 76 |
+
"patch_size",
|
| 77 |
+
"temporal_patch_size",
|
| 78 |
+
"merge_size"
|
| 79 |
+
],
|
| 80 |
+
"n_fft": 400,
|
| 81 |
+
"n_samples": 4800000,
|
| 82 |
+
"nb_max_frames": 30000,
|
| 83 |
+
"padding_side": "right",
|
| 84 |
+
"padding_value": 0.0,
|
| 85 |
+
"patch_size": 14,
|
| 86 |
+
"processor_class": "Qwen2_5OmniProcessor",
|
| 87 |
+
"resample": 3,
|
| 88 |
+
"rescale_factor": 0.00392156862745098,
|
| 89 |
+
"return_attention_mask": true,
|
| 90 |
+
"sampling_rate": 16000,
|
| 91 |
+
"size": {
|
| 92 |
+
"longest_edge": 12845056,
|
| 93 |
+
"shortest_edge": 3136
|
| 94 |
+
},
|
| 95 |
+
"size_divisor": null,
|
| 96 |
+
"temporal_patch_size": 2,
|
| 97 |
+
"video_processor_type": "Qwen2VLVideoProcessor"
|
| 98 |
+
}
|