Video-Text-to-Text
Transformers
Safetensors
aviot_qwen
text-generation
video-language-model
video-question-answering
video-understanding
token-compression
optimal-transport
qwen2
siglip
Instructions to use ernie-research/AVIOT with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use ernie-research/AVIOT with Transformers:
# Load model directly from transformers import AutoModelForCausalLM model = AutoModelForCausalLM.from_pretrained("ernie-research/AVIOT", device_map="auto") - Notebooks
- Google Colab
- Kaggle
| { | |
| "attention_dropout": 0.0, | |
| "bos_token_id": 151643, | |
| "eos_token_id": 151645, | |
| "hidden_act": "silu", | |
| "hidden_size": 3584, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 18944, | |
| "max_position_embeddings": 32768, | |
| "max_window_layers": 28, | |
| "num_attention_heads": 28, | |
| "num_hidden_layers": 28, | |
| "num_key_value_heads": 4, | |
| "rms_norm_eps": 1e-06, | |
| "rope_scaling": null, | |
| "rope_theta": 1000000.0, | |
| "sliding_window": 131072, | |
| "tie_word_embeddings": false, | |
| "use_cache": false, | |
| "use_sliding_window": false, | |
| "vocab_size": 152064, | |
| "model_type": "aviot_qwen", | |
| "architectures": [ | |
| "AVIOTQwenForCausalLM" | |
| ], | |
| "vision_hidden_size": 1152, | |
| "multimodal_projector_type": "mlp2x_gelu", | |
| "spatial_pool_mode": "bilinear", | |
| "spatial_pool_stride": 2, | |
| "vision_tower_batch_size": 0, | |
| "tokenizer_padding_side": "right", | |
| "tokenizer_model_max_length": 32768, | |
| "vision_tower": "vision_tower", | |
| "aviot_vision_weights_embedded": true, | |
| "aviot_default_supports": 8, | |
| "aviot_cost_dim": 256, | |
| "aviot_temporal_segments": 4, | |
| "aviot_entropy": 0.1, | |
| "aviot_source_relaxation": 0.5, | |
| "aviot_target_relaxation": 5.0, | |
| "aviot_sinkhorn_iterations": 20, | |
| "aviot_global_refinement_rounds": 5, | |
| "aviot_transport_loss_weight": 1.0, | |
| "aviot_mass_loss_weight": 0.01, | |
| "aviot_progressive_ratios": [ | |
| 0.75, | |
| 0.5, | |
| 0.25 | |
| ], | |
| "aviot_progressive_round_to": 8, | |
| "aviot_max_input_frames": 224, | |
| "aviot_training_ratios": [ | |
| 2.0, | |
| 2.5, | |
| 3.0, | |
| 3.5, | |
| 4.0, | |
| 4.5, | |
| 5.0, | |
| 5.5, | |
| 6.0, | |
| 6.5, | |
| 7.0, | |
| 7.5, | |
| 8.0, | |
| 8.5, | |
| 9.0, | |
| 9.5, | |
| 10.0 | |
| ], | |
| "aviot_ratio_policy": "random", | |
| "aviot_thw_num_bands": 16, | |
| "aviot_thw_temporal_scale": 1.0, | |
| "aviot_thw_spatial_scale": 0.5, | |
| "aviot_thw_warmup_steps": 500, | |
| "aviot_allocation_temperature_initial": 1.0, | |
| "aviot_allocation_temperature_final": 0.1, | |
| "aviot_allocation_warmup_steps": 500, | |
| "aviot_question_allocation_weight": 0.3, | |
| "aviot_multiscale_entropy": [ | |
| 0.1, | |
| 0.12, | |
| 0.15 | |
| ], | |
| "aviot_multiscale_refinement_rounds": [ | |
| 3, | |
| 2, | |
| 1 | |
| ], | |
| "aviot_multiscale_warmup_steps": 500, | |
| "aviot_medium_parent_weight": 0.5, | |
| "aviot_local_parent_weight": 0.25, | |
| "aviot_regional_transport_weight": 1.0, | |
| "aviot_continuity_weight": 0.01, | |
| "aviot_continuity_scale": 0.2, | |
| "aviot_gate_hidden_size": 32, | |
| "aviot_gate_global_floor": 0.2, | |
| "aviot_gate_temperature": 0.05, | |
| "aviot_gate_tv_weight": 0.001, | |
| "aviot_gate_balance_weight": 0.5, | |
| "aviot_gate_balance_target": [ | |
| 0.35, | |
| 0.45, | |
| 0.2 | |
| ], | |
| "aviot_gate_entropy_weight": 0.5, | |
| "aviot_gate_entropy_floor": 0.85 | |
| } | |