Instructions to use Intel/Cosmos3-Super-int4-AutoRound with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use Intel/Cosmos3-Super-int4-AutoRound with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("Intel/Cosmos3-Super-int4-AutoRound", torch_dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
| { | |
| "_class_name": "Cosmos3VFMTransformer", | |
| "_diffusers_version": "0.37.1", | |
| "action_dim": 64, | |
| "action_gen": true, | |
| "architectures": [ | |
| "Cosmos3VFMTransformer" | |
| ], | |
| "attention_bias": false, | |
| "attention_dropout": 0.0, | |
| "base_fps": 24, | |
| "dtype": "bfloat16", | |
| "enable_fps_modulation": true, | |
| "freeze_und": false, | |
| "head_dim": 128, | |
| "hidden_act": "silu", | |
| "hidden_size": 5120, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 25600, | |
| "joint_attn_implementation": "two_way", | |
| "latent_channel": 48, | |
| "latent_patch_size": 2, | |
| "max_action_dim": 64, | |
| "max_position_embeddings": 262144, | |
| "model_type": "cosmos3_vllm_omni", | |
| "num_attention_heads": 64, | |
| "num_embodiment_domains": 32, | |
| "num_hidden_layers": 64, | |
| "num_key_value_heads": 8, | |
| "patch_latent_dim": 192, | |
| "position_embedding_type": "unified_3d_mrope", | |
| "qk_norm": false, | |
| "qk_norm_for_diffusion": true, | |
| "qk_norm_for_text": true, | |
| "quantization_config": { | |
| "autoround_version": "0.14.0", | |
| "batch_size": 1, | |
| "bits": 4, | |
| "block_name_to_quantize": [ | |
| "language_model.layers", | |
| "gen_layers", | |
| "blocks", | |
| "deepstack_merger_list" | |
| ], | |
| "data_type": "int", | |
| "extra_config": { | |
| "action_proj_in.bias": { | |
| "bits": 16, | |
| "data_type": "fp" | |
| }, | |
| "action_proj_in.fc": { | |
| "bits": 16, | |
| "data_type": "fp" | |
| }, | |
| "action_proj_out.bias": { | |
| "bits": 16, | |
| "data_type": "fp" | |
| }, | |
| "action_proj_out.fc": { | |
| "bits": 16, | |
| "data_type": "fp" | |
| }, | |
| "audio_proj_in": { | |
| "bits": 16, | |
| "data_type": "fp" | |
| }, | |
| "audio_proj_out": { | |
| "bits": 16, | |
| "data_type": "fp" | |
| }, | |
| "embed_tokens": { | |
| "bits": 16, | |
| "data_type": "fp" | |
| }, | |
| "merger.linear_fc1": { | |
| "bits": 16, | |
| "data_type": "fp" | |
| }, | |
| "merger.linear_fc2": { | |
| "bits": 16, | |
| "data_type": "fp" | |
| }, | |
| "pos_embed": { | |
| "bits": 16, | |
| "data_type": "fp" | |
| }, | |
| "proj_in": { | |
| "bits": 16, | |
| "data_type": "fp" | |
| }, | |
| "proj_out": { | |
| "bits": 16, | |
| "data_type": "fp" | |
| }, | |
| "time_embedder.linear_1": { | |
| "bits": 16, | |
| "data_type": "fp" | |
| }, | |
| "time_embedder.linear_2": { | |
| "bits": 16, | |
| "data_type": "fp" | |
| } | |
| }, | |
| "group_size": 128, | |
| "nsamples": 16, | |
| "packing_format": "auto_round:auto_gptq", | |
| "quant_method": "auto-round", | |
| "sym": true | |
| }, | |
| "rms_norm_eps": 1e-06, | |
| "rope_parameters": { | |
| "mrope_interleaved": true, | |
| "mrope_section": [ | |
| 24, | |
| 20, | |
| 20 | |
| ], | |
| "rope_theta": 5000000, | |
| "rope_type": "default" | |
| }, | |
| "sound_dim": 64, | |
| "sound_gen": true, | |
| "sound_latent_fps": 25, | |
| "temporal_compression_factor_sound": 1, | |
| "timestep_scale": 0.001, | |
| "transformers_version": "5.13.0", | |
| "unified_3d_mrope_reset_spatial_ids": true, | |
| "unified_3d_mrope_temporal_modality_margin": 15000, | |
| "use_moe": true, | |
| "video_temporal_causal": false, | |
| "vocab_size": 151936 | |
| } |