Runware/Cosmos3-Nano

1{2  "allow_patterns_overrides": [3    "*/*.safetensors"4  ],5  "architectures": [6    "Cosmos3ForConditionalGeneration"7  ],8  "image_token_id": 151655,9  "model": {10    "_recursive_": false,11    "_target": "omni_mot_model",12    "config": {13      "_type": "omni_mot_model_config",14      "action_gen": true,15      "activation_checkpointing": {16        "_type": "activation_checkpointing_config",17        "determinism_check": "default",18        "mode": "selective",19        "preserve_rng_state": true,20        "save_ops_regex": [21          "fmha"22        ]23      },24      "causal_training_strategy": "none",25      "diffusion_expert_config": {26        "_type": "diffusion_expert_config",27        "base_fps": 24,28        "enable_fps_modulation": true,29        "load_weights_from_pretrained": false,30        "max_vae_latent_side_after_patchify": 20,31        "patch_spatial": 2,32        "position_embedding_type": "unified_3d_mrope",33        "rope_h_extrapolation_ratio": 1.0,34        "rope_t_extrapolation_ratio": 1.0,35        "rope_w_extrapolation_ratio": 1.0,36        "timestep_range": 1.0,37        "unified_3d_mrope_reset_spatial_ids": true,38        "unified_3d_mrope_temporal_modality_margin": 1500039      },40      "ema": {41        "_type": "ema_config",42        "enabled": false,43        "iteration_shift": 0,44        "rate": 0.145      },46      "fixed_step_sampler_config": null,47      "input_caption_key": "ai_caption",48      "input_image_key": "images",49      "input_video_key": "video",50      "joint_attn_implementation": "two_way",51      "latent_downsample_factor": 16,52      "lbl": {53        "_type": "lbl_config",54        "coeff_gen": null,55        "coeff_und": null,56        "method": "local"57      },58      "log_enc_time_every_n": 100,59      "lora_alpha": 32,60      "lora_enabled": false,61      "lora_rank": 16,62      "lora_target_modules": "q_proj_moe_gen,k_proj_moe_gen,v_proj_moe_gen,o_proj_moe_gen",63      "max_action_dim": 64,64      "max_num_tokens_after_packing": 74000,65      "natten_parameter_list": null,66      "net": null,67      "num_embodiment_domains": 32,68      "parallelism": {69        "_type": "parallelism_config",70        "cfg_parallel_shard_degree": 1,71        "compile_dynamic": true,72        "compiled_region": "language",73        "context_parallel_shard_degree": 1,74        "coordinate_descent_tuning": false,75        "data_parallel_replicate_degree": 1,76        "data_parallel_shard_degree": 8,77        "enable_inference_mode": false,78        "max_autotune_pointwise": false,79        "precision": "bfloat16",80        "use_cuda_graphs": false,81        "use_torch_compile": true82      },83      "rectified_flow_inference_config": {84        "_type": "rectified_flow_inference_config",85        "num_train_timesteps": 1000,86        "scheduler_type": "unipc",87        "shift": 1,88        "use_dynamic_shifting": false89      },90      "rectified_flow_training_config": {91        "_type": "rectified_flow_training_config",92        "action_loss_weight": 10.0,93        "high_sigma_ratio": 0.05,94        "high_sigma_timesteps_max": 1000,95        "high_sigma_timesteps_min": 995,96        "image_loss_scale": null,97        "independent_action_schedule": false,98        "independent_sound_schedule": false,99        "loss_scale": 10.0,100        "normalize_loss_by_active": false,101        "shift": {102          "256": 3,103          "480": 5,104          "720": 10105        },106        "shift_action": null,107        "shift_sound": null,108        "sound_loss_scale": 2.0,109        "train_time_action_distribution": "logitnormal",110        "train_time_image_distribution": "logitnormal",111        "train_time_sound_distribution": "logitnormal",112        "train_time_video_distribution": "waver",113        "train_time_weight": "uniform",114        "use_discrete_rf": false,115        "use_dynamic_shift": false,116        "use_high_sigma_strategy": false,117        "use_high_sigma_strategy_action": false,118        "use_high_sigma_strategy_sound": false119      },120      "resolution": "720",121      "sound_dim": 64,122      "sound_gen": true,123      "sound_latent_fps": 25,124      "sound_tokenizer": {125        "_target": "avae_interface",126        "audio_channels": 2,127        "avae_config_path": "",128        "avae_path": "pretrained/tokenizers/audio/avae/avae_48k_noncausal_25hz_64ch.ckpt",129        "bucket_name": "bucket",130        "hop_size": 1920,131        "io_channels": 64,132        "latent_mean": null,133        "latent_std": null,134        "normalization_type": "none",135        "normalize_latents": false,136        "object_store_credential_path_pretrained": "credentials/gcp_training.secret",137        "sample_rate": 48000,138        "tanh_clamp": 0.995,139        "tanh_input_scale": 1.5,140        "tanh_output_scale": 3.5141      },142      "state_ch": 48,143      "state_t": 300,144      "tokenizer": {145        "_target": "wan2pt2_vae_interface",146        "bucket_name": "bucket",147        "chunk_duration": 93,148        "encode_bucket_multiple": null,149        "encode_chunk_frames": {150          "256": 68,151          "480": 24,152          "720": 12153        },154        "encode_exact_durations": [155          17,156          61,157          73158        ],159        "keep_decoder_cache": false,160        "object_store_credential_path_pretrained": "credentials/gcp_training.secret",161        "spatial_compression_factor": 16,162        "temporal_compression_factor": 4,163        "temporal_window": null,164        "use_streaming_encode": false,165        "vae_path": "pretrained/tokenizers/video/wan2pt2/Wan2.2_VAE.pth"166      },167      "video_temporal_causal": false,168      "vision_gen": true,169      "vlm_config": {170        "_type": "vlm_config",171        "layer_module": null,172        "model_instance": {173          "_target": "qwen3_vl_text_for_causal_lm",174          "config": {175            "_target": "create_vlm_config",176            "base_config": {177              "_target": "qwen3_vl_mot_config_from_json_file",178              "json_file": "cosmos3://vfm/models/vlm/qwen3_vl/configs/Qwen3-VL-8B-Instruct.json"179            },180            "qk_norm_for_text": true181          }182        },183        "model_name": "nvidia/Cosmos3-Nano-Reasoner",184        "pretrained_weights": {185          "_type": "pretrained_weights_config",186          "backbone_path": "s3://bucket/cosmos3/pretrained/huggingface/Cosmos-Reason/Cosmos3-Nano-Reasoner-bb9c6f5/",187          "checkpoint_format": null,188          "credentials_path": "credentials/gcp_checkpoint.secret",189          "enable_gcs_patch_in_boto3": true,190          "enabled": false191        },192        "qk_norm": false,193        "tie_word_embeddings": false,194        "tokenizer": {195          "_target": "create_qwen2_tokenizer_with_download",196          "config_variant": "gcp",197          "pretrained_model_name": "Qwen/Qwen3-VL-8B-Instruct"198        },199        "use_system_prompt": false200      }201    }202  },203  "model_type": "cosmos3_omni",204  "text_config": {205    "attention_bias": false,206    "attention_dropout": 0.0,207    "bos_token_id": 151643,208    "dtype": "bfloat16",209    "eos_token_id": 151645,210    "head_dim": 128,211    "hidden_act": "silu",212    "hidden_size": 4096,213    "initializer_range": 0.02,214    "intermediate_size": 12288,215    "max_position_embeddings": 262144,216    "model_type": "qwen3_vl_text",217    "num_attention_heads": 32,218    "num_hidden_layers": 36,219    "num_key_value_heads": 8,220    "rms_norm_eps": 1e-06,221    "rope_scaling": {222      "mrope_interleaved": true,223      "mrope_section": [224        24,225        20,226        20227      ],228      "rope_type": "default"229    },230    "rope_theta": 5000000,231    "use_cache": true,232    "vocab_size": 151936233  },234  "tie_word_embeddings": false,235  "transformers_version": "4.57.0.dev0",236  "video_token_id": 151656,237  "vision_config": {238    "deepstack_visual_indexes": [239      8,240      16,241      24242    ],243    "depth": 27,244    "hidden_act": "gelu_pytorch_tanh",245    "hidden_size": 1152,246    "in_channels": 3,247    "initializer_range": 0.02,248    "intermediate_size": 4304,249    "model_type": "qwen3_vl",250    "num_heads": 16,251    "num_position_embeddings": 2304,252    "out_hidden_size": 4096,253    "patch_size": 16,254    "spatial_merge_size": 2,255    "temporal_patch_size": 2256  },257  "vision_end_token_id": 151653,258  "vision_start_token_id": 151652259}260