Robotics
Safetensors
PyTorch
English
alpamayo2_edge
alpamayo
autonomous-driving
vision-language-action
Alpamayo2-Edge / config.json
yurongyou's picture polluxxx's picture
Update to the Cosmos3-Edge checkpoint (#1)
aaaa32c
Raw History Blame
8.18 kB
{
"architectures": [
"Alpamayo2Edge"
],
"cotrain_detach_expert_kv": true,
"cotrain_expert_vlm": true,
"dtype": "bfloat16",
"enable_expert": true,
"enable_thinking": false,
"encode_as_video": false,
"expert_config": {
"action_in_proj_cfg": {
"_target_": "alpamayo2_edge.models.action_in_proj.PerWaypointActionInProjV2",
"hidden_size": 512,
"max_freq": 100.0,
"num_enc_layers": 2,
"num_fourier_feats": 20
},
"action_out_proj_cfg": {
"_target_": "torch.nn.Linear"
},
"action_space_cfg": {
"_target_": "alpamayo2_edge.action_space.UnicycleAccelCurvatureCourseActionSpace",
"a_lambda": 0.0001,
"a_ridge": 0.0001,
"accel_bounds": [
-9.8,
9.8
],
"accel_mean": 0.02902694707164455,
"accel_std": 0.6810426736454882,
"curvature_bounds": [
-0.33,
0.33
],
"curvature_mean": 0.0002692167976330542,
"curvature_std": 0.026148280660833106,
"dt": 0.1,
"kappa_lambda": 0.0001,
"kappa_ridge": 0.0001,
"n_waypoints": 64,
"slip_gradient": 0.0025,
"theta_lambda": 3e-06,
"theta_ridge": 1e-08,
"v_lambda": 1e-06,
"v_ridge": 0.0001
},
"architectures": null,
"chunk_size_feed_forward": 0,
"diffusion_cfg": {
"_target_": "alpamayo2_edge.diffusion.flow_matching.FlowMatching",
"inference_guidance_weight": 3.0,
"int_method": "euler",
"train_ignore_guidance_rate": 0.1,
"train_timestep_sampler": "beta",
"use_classifier_free_guidance": false,
"x_dims": null
},
"dtype": "bfloat16",
"expert_non_causal_attention": true,
"expert_update_cfg": {
"head_dim": 128,
"hidden_size": 512,
"intermediate_size": 3840,
"num_attention_heads": 16,
"num_hidden_layers": 28,
"num_key_value_heads": 8
},
"id2label": {
"0": "LABEL_0",
"1": "LABEL_1"
},
"is_encoder_decoder": false,
"label2id": {
"LABEL_0": 0,
"LABEL_1": 1
},
"llm_config": {
"architectures": null,
"attention_bias": false,
"attention_dropout": 0.0,
"bos_token_id": 1,
"chunk_size_feed_forward": 0,
"dtype": "bfloat16",
"eos_token_id": 11,
"head_dim": 128,
"hidden_act": "relu2",
"hidden_size": 512,
"id2label": {
"0": "LABEL_0",
"1": "LABEL_1"
},
"initializer_range": 0.02,
"intermediate_size": 3840,
"is_encoder_decoder": false,
"label2id": {
"LABEL_0": 0,
"LABEL_1": 1
},
"max_position_embeddings": 131072,
"mlp_bias": false,
"model_type": "cosmos3_edge_text",
"num_attention_heads": 16,
"num_hidden_layers": 28,
"num_key_value_heads": 8,
"output_attentions": false,
"output_hidden_states": false,
"pad_token_id": 0,
"pretraining_tp": 1,
"problem_type": null,
"return_dict": true,
"rms_norm_eps": 1e-05,
"rope_parameters": {
"mrope_section": [
24,
20,
20
],
"rope_theta": 100000000,
"rope_type": "default"
},
"tie_word_embeddings": false,
"use_cache": true,
"vocab_size": 135168
},
"model_type": "alpamayo2_edge_expert",
"output_attentions": false,
"output_hidden_states": false,
"problem_type": null,
"return_dict": true
},
"frame_label": "relative_time",
"future_traj_tokenizer_cfg": {
"_recursive_": false,
"_target_": "alpamayo2_edge.action_space.discrete_action_space.DiscreteTrajectoryTokenizer",
"action_space_cfg": {
"_target_": "alpamayo2_edge.action_space.UnicycleAccelCurvatureCourseActionSpace",
"a_lambda": 0.0001,
"a_ridge": 0.0001,
"accel_bounds": [
-9.8,
9.8
],
"accel_mean": 0.02902694707164455,
"accel_std": 0.6810426736454882,
"curvature_bounds": [
-0.33,
0.33
],
"curvature_mean": 0.0002692167976330542,
"curvature_std": 0.026148280660833106,
"dt": 0.1,
"kappa_lambda": 0.0001,
"kappa_ridge": 0.0001,
"n_waypoints": 64,
"slip_gradient": 0.0025,
"theta_lambda": 3e-06,
"theta_ridge": 1e-08,
"v_lambda": 1e-06,
"v_ridge": 0.0001
},
"dims_max": [
10,
10
],
"dims_min": [
-10,
-10
],
"num_bins": 3000
},
"future_vocab_size": 3000,
"hist_traj_tokenizer_cfg": {
"_target_": "alpamayo2_edge.models.delta_tokenizer.DeltaTrajectoryTokenizer",
"pad_origin_at_beginning": false
},
"history_vocab_size": 1000,
"include_camera_ids": true,
"loss_weights": {
"future_traj": 1.0,
"others": 1.0
},
"max_pixels": 196608,
"min_pixels": 163840,
"model_type": "alpamayo2_edge",
"padding_side": "left",
"token_layout": "camera_ts",
"tokens_per_future_traj": 128,
"tokens_per_history_traj": 57,
"traj_ids": {
"future_end": 135086,
"future_id0": 132072,
"future_pad": 135088,
"future_start": 135084,
"history_end": 135079,
"history_id0": 131072,
"history_pad": 135087,
"history_start": 135077
},
"traj_vocab_size": 4000,
"transformers_version": "5.17.0",
"video_max_pixels": null,
"video_min_pixels": null,
"video_pixels_per_frame": false,
"vlm_class": "Cosmos3EdgeForConditionalGeneration",
"vlm_config": {
"architectures": [
"Cosmos3EdgeForConditionalGeneration"
],
"chunk_size_feed_forward": 0,
"dtype": "bfloat16",
"id2label": {
"0": "LABEL_0",
"1": "LABEL_1"
},
"image_token_id": 19,
"is_encoder_decoder": false,
"label2id": {
"LABEL_0": 0,
"LABEL_1": 1
},
"model_type": "cosmos3_edge",
"output_attentions": false,
"output_hidden_states": false,
"problem_type": null,
"projector_hidden_size": 11520,
"return_dict": true,
"text_config": {
"architectures": null,
"attention_bias": false,
"attention_dropout": 0.0,
"bos_token_id": 1,
"chunk_size_feed_forward": 0,
"dtype": "bfloat16",
"eos_token_id": 11,
"head_dim": 128,
"hidden_act": "relu2",
"hidden_size": 2048,
"id2label": {
"0": "LABEL_0",
"1": "LABEL_1"
},
"initializer_range": 0.02,
"intermediate_size": 9216,
"is_encoder_decoder": false,
"label2id": {
"LABEL_0": 0,
"LABEL_1": 1
},
"max_position_embeddings": 131072,
"mlp_bias": false,
"model_type": "cosmos3_edge_text",
"num_attention_heads": 16,
"num_hidden_layers": 28,
"num_key_value_heads": 8,
"output_attentions": false,
"output_hidden_states": false,
"pad_token_id": 0,
"pretraining_tp": 1,
"problem_type": null,
"return_dict": true,
"rms_norm_eps": 1e-05,
"rope_parameters": {
"mrope_section": [
24,
20,
20
],
"rope_theta": 100000000,
"rope_type": "default"
},
"tie_word_embeddings": false,
"use_cache": true,
"vocab_size": 135168
},
"tie_word_embeddings": false,
"video_token_id": 18,
"vision_config": {
"architectures": null,
"attention_dropout": 0.0,
"chunk_size_feed_forward": 0,
"dtype": "bfloat16",
"hidden_act": "gelu_pytorch_tanh",
"hidden_size": 1152,
"id2label": {
"0": "LABEL_0",
"1": "LABEL_1"
},
"intermediate_size": 4304,
"is_encoder_decoder": false,
"label2id": {
"LABEL_0": 0,
"LABEL_1": 1
},
"layer_norm_eps": 1e-06,
"model_type": "cosmos3_edge_vision",
"num_attention_heads": 16,
"num_channels": 3,
"num_hidden_layers": 27,
"num_patches": 256,
"output_attentions": false,
"output_hidden_states": false,
"patch_size": 16,
"problem_type": null,
"return_dict": true,
"spatial_merge_size": 2
},
"vision_end_token_id": 21,
"vision_start_token_id": 20
},
"vlm_name_or_path": null
}