{ "audio_end_token": "<|audio_end|>", "audio_start_token": "<|audio_start|>", "audio_token": "<|AUDIO_PAD|>", "auto_map": { "AutoFeatureExtractor": "audio_processing_hyperclovax_omni.HyperCLOVAXOmniAudioProcessor", "AutoImageProcessor": "image_processing_hyperclovax_omni.HyperCLOVAXOmniImageProcessor", "AutoProcessor": "processing_hyperclovax_omni.HyperCLOVAXOmniProcessor", "AutoVideoProcessor": "video_processing_hyperclovax_omni.HyperCLOVAXOmniVideoProcessor" }, "chunk_length": 30, "chunk_unit": 80, "discrete_audio_end_token": "<|discrete_audio_end|>", "discrete_audio_start_token": "<|discrete_audio_start|>", "discrete_audio_token": "<|DISCRETE_AUDIO_PAD|>", "discrete_image_end_token": "<|discrete_image_end|>", "discrete_image_ratios": [ [1, 1], [1, 2], [2, 1], [3, 4], [4, 3], [9, 16], [16, 9], [1, 3], [3, 1] ], "discrete_image_start_token": "<|discrete_image_start|>", "discrete_image_token": "<|DISCRETE_IMAGE_PAD|>", "discrete_token_size": 27, "dither": 0.0, "do_convert_rgb": true, "do_normalize": true, "do_rescale": true, "do_resize": true, "feature_extractor_type": "HyperCLOVAXOmniAudioProcessor", "feature_size": 128, "hop_length": 160, "image_end_token": "<|image_end|>", "image_mean": [ 0.48145466, 0.4578275, 0.40821073 ], "image_processor_type": "HyperCLOVAXOmniImageProcessor", "image_start_token": "<|image_start|>", "image_std": [ 0.26862954, 0.26130258, 0.27577711 ], "image_token": "<|IMAGE_PAD|>", "max_pixels": 2073600, "merge_size": 2, "min_chunk_size": 1600, "min_pixels": 3136, "n_fft": 400, "n_samples": 480000, "nb_max_frames": 3000, "padding_side": "right", "padding_value": 0.0, "patch_size": 14, "processor_class": "HyperCLOVAXOmniProcessor", "resample": 3, "rescale_factor": 0.00392156862745098, "return_attention_mask": true, "sampling_rate": 16000, "size": { "longest_edge": 2073600, "shortest_edge": 3136 }, "temporal_patch_size": 2, "use_discrete_audio_token": true, "use_discrete_image_token": true, "video_do_convert_rgb": true, "video_do_normalize": true, "video_do_rescale": true, "video_do_resize": true, "video_end_token": "<|video_end|>", "video_max_frames": 120, "video_max_pixels": 12845056, "video_mean": [ 0.48145466, 0.4578275, 0.40821073 ], "video_merge_size": 2, "video_min_pixels": 3136, "video_patch_size": 14, "video_processor_type": "HyperCLOVAXOmniVideoProcessor", "video_resample": 3, "video_rescale_factor": 0.00392156862745098, "video_start_token": "<|video_start|>", "video_std": [ 0.26862954, 0.26130258, 0.27577711 ], "video_temporal_patch_size": 2, "video_token": "<|VIDEO_PAD|>", "use_discrete_video_token": false, "vision_eof_token": "<|vision_eof|>", "vision_eol_token": "<|vision_eol|>" }