Instructions to use allenai/MolmoAct2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use allenai/MolmoAct2 with Transformers:
# Load model directly from transformers import AutoModelForImageTextToText model = AutoModelForImageTextToText.from_pretrained("allenai/MolmoAct2", trust_remote_code=True, dtype="auto") - Notebooks
- Google Colab
- Kaggle
| { | |
| "action_end_token_id": 151933, | |
| "action_expert_config": { | |
| "attn_dropout": 0.0, | |
| "causal_attn": false, | |
| "context_layer_norm": true, | |
| "dropout": 0.0, | |
| "ffn_multiple_of": 256, | |
| "hidden_size": 768, | |
| "mlp_ratio": 4.0, | |
| "model_type": "molmoact2_action_expert", | |
| "num_heads": 8, | |
| "num_layers": 36, | |
| "qk_norm": true, | |
| "qk_norm_eps": 1e-06, | |
| "rope": true, | |
| "timestep_embed_dim": 256 | |
| }, | |
| "action_expert_depth_gate": false, | |
| "action_expert_depth_gate_init_bias": -4.0, | |
| "action_expert_depth_gate_per_layer": false, | |
| "action_mode": "both", | |
| "max_action_horizon": 30, | |
| "action_output_token_id": 151931, | |
| "action_start_token_id": 151932, | |
| "action_token_start_id": 151934, | |
| "adapter_config": { | |
| "attention_dropout": 0.0, | |
| "attn_implementation": "sdpa", | |
| "float32_attention": true, | |
| "head_dim": 72, | |
| "hidden_act": "silu", | |
| "hidden_size": 1152, | |
| "image_feature_dropout": 0.0, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 9728, | |
| "model_type": "molmoact2", | |
| "num_attention_heads": 16, | |
| "num_key_value_heads": 16, | |
| "pooling_attention_mask": true, | |
| "residual_dropout": 0.0, | |
| "text_hidden_size": 2560, | |
| "vit_layers": [ | |
| -3, | |
| -9 | |
| ] | |
| }, | |
| "add_action_expert": true, | |
| "add_control_tokens": true, | |
| "add_setup_tokens": true, | |
| "architectures": [ | |
| "MolmoAct2ForConditionalGeneration" | |
| ], | |
| "auto_map": { | |
| "AutoConfig": "configuration_molmoact2.MolmoAct2Config", | |
| "AutoModelForImageTextToText": "modeling_molmoact2.MolmoAct2ForConditionalGeneration" | |
| }, | |
| "depth_end_token_id": null, | |
| "depth_mode": 2, | |
| "depth_output_token_id": null, | |
| "depth_start_token_id": null, | |
| "depth_token_start_id": null, | |
| "dtype": "float32", | |
| "enable_depth_reasoning": false, | |
| "flow_matching_beta_alpha": 1.0, | |
| "flow_matching_beta_beta": 1.5, | |
| "flow_matching_cutoff": 1.0, | |
| "flow_matching_num_steps": 10, | |
| "flow_matching_time_offset": 0.001, | |
| "flow_matching_time_scale": 0.999, | |
| "frame_end_token_id": 154632, | |
| "frame_start_token_id": 154631, | |
| "image_col_id": 154627, | |
| "image_end_token_id": 154625, | |
| "image_high_res_id": 154626, | |
| "image_low_res_id": 154630, | |
| "image_patch_id": 154626, | |
| "image_start_token_id": 154624, | |
| "initializer_range": 0.02, | |
| "low_res_image_start_token_id": 154628, | |
| "mask_action_dim_padding": true, | |
| "max_action_dim": 32, | |
| "model_type": "molmoact2", | |
| "n_obs_steps": 1, | |
| "norm_stats_filename": "norm_stats.json", | |
| "num_action_tokens": 2048, | |
| "num_depth_codes": 100, | |
| "num_depth_tokens": 0, | |
| "num_state_tokens": 256, | |
| "state_end_token_id": 151674, | |
| "state_format": "discrete", | |
| "state_start_token_id": 151673, | |
| "state_token_start_id": 151675, | |
| "text_config": { | |
| "additional_vocab_size": 128, | |
| "attention_dropout": 0.0, | |
| "attn_implementation": "sdpa", | |
| "embedding_dropout": 0.0, | |
| "head_dim": 128, | |
| "hidden_act": "silu", | |
| "hidden_size": 2560, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 9728, | |
| "layer_norm_eps": 1e-06, | |
| "max_position_embeddings": 16384, | |
| "model_type": "molmoact2_text", | |
| "norm_after": false, | |
| "num_attention_heads": 32, | |
| "num_hidden_layers": 36, | |
| "num_key_value_heads": 8, | |
| "qk_norm_type": "qwen3", | |
| "qkv_bias": false, | |
| "residual_dropout": 0.0, | |
| "rope_parameters": { | |
| "rope_theta": 5000000.0, | |
| "rope_type": "default" | |
| }, | |
| "rope_scaling_layers": null, | |
| "rope_theta": 5000000.0, | |
| "tie_word_embeddings": false, | |
| "use_cache": true, | |
| "use_qk_norm": true, | |
| "vocab_size": 154624 | |
| }, | |
| "tie_word_embeddings": false, | |
| "transformers_version": "5.3.0", | |
| "use_frame_special_tokens": true, | |
| "vit_config": { | |
| "attention_dropout": 0.0, | |
| "attn_implementation": "sdpa", | |
| "float32_attention": true, | |
| "head_dim": 72, | |
| "hidden_act": "gelu_pytorch_tanh", | |
| "hidden_size": 1152, | |
| "image_default_input_size": [ | |
| 378, | |
| 378 | |
| ], | |
| "image_num_pos": 729, | |
| "image_patch_size": 14, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 4304, | |
| "layer_norm_eps": 1e-06, | |
| "model_type": "molmoact2", | |
| "num_attention_heads": 16, | |
| "num_hidden_layers": 27, | |
| "num_key_value_heads": 16, | |
| "residual_dropout": 0.0 | |
| }, | |
| "bos_token_id": 151645, | |
| "eos_token_id": 151645, | |
| "pad_token_id": 151643 | |
| } | |