dagger it1 checkpoint-5500
Browse files- checkpoint-5500/config.json +88 -0
- checkpoint-5500/embodiment_id.json +57 -0
- checkpoint-5500/experiment_cfg/conf.yaml +230 -0
- checkpoint-5500/experiment_cfg/config.yaml +257 -0
- checkpoint-5500/experiment_cfg/dataset_statistics.json +257 -0
- checkpoint-5500/experiment_cfg/final_model_config.json +55 -0
- checkpoint-5500/experiment_cfg/final_processor_config.json +0 -0
- checkpoint-5500/model-00001-of-00002.safetensors +3 -0
- checkpoint-5500/model-00002-of-00002.safetensors +3 -0
- checkpoint-5500/model.safetensors.index.json +0 -0
- checkpoint-5500/processor_config.json +1124 -0
- checkpoint-5500/statistics.json +0 -0
- checkpoint-5500/trainer_state.json +3334 -0
- checkpoint-5500/training_args.bin +3 -0
- checkpoint-5500/wandb_config.json +1 -0
checkpoint-5500/config.json
ADDED
|
@@ -0,0 +1,88 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"action_horizon": 40,
|
| 3 |
+
"add_pos_embed": true,
|
| 4 |
+
"apply_sincos_state_encoding": false,
|
| 5 |
+
"architectures": [
|
| 6 |
+
"Gr00tN1d7"
|
| 7 |
+
],
|
| 8 |
+
"attn_dropout": 0.2,
|
| 9 |
+
"attn_implementation": null,
|
| 10 |
+
"backbone_embedding_dim": 2048,
|
| 11 |
+
"backbone_trainable_params_fp32": true,
|
| 12 |
+
"color_jitter_params": {
|
| 13 |
+
"brightness": 0.3,
|
| 14 |
+
"contrast": 0.4,
|
| 15 |
+
"hue": 0.08,
|
| 16 |
+
"saturation": 0.5
|
| 17 |
+
},
|
| 18 |
+
"crop_fraction": 1.0,
|
| 19 |
+
"diffusion_model_cfg": {
|
| 20 |
+
"attention_head_dim": 48,
|
| 21 |
+
"dropout": 0.2,
|
| 22 |
+
"final_dropout": true,
|
| 23 |
+
"interleave_self_attention": true,
|
| 24 |
+
"norm_type": "ada_norm",
|
| 25 |
+
"num_attention_heads": 32,
|
| 26 |
+
"num_layers": 32,
|
| 27 |
+
"output_dim": 1024,
|
| 28 |
+
"positional_embeddings": null
|
| 29 |
+
},
|
| 30 |
+
"dtype": "bfloat16",
|
| 31 |
+
"exclude_state": false,
|
| 32 |
+
"formalize_language": true,
|
| 33 |
+
"hidden_size": 1024,
|
| 34 |
+
"image_crop_size": [
|
| 35 |
+
230,
|
| 36 |
+
230
|
| 37 |
+
],
|
| 38 |
+
"image_target_size": [
|
| 39 |
+
256,
|
| 40 |
+
256
|
| 41 |
+
],
|
| 42 |
+
"letter_box_transform": false,
|
| 43 |
+
"load_bf16": false,
|
| 44 |
+
"max_action_dim": 132,
|
| 45 |
+
"max_num_embodiments": 32,
|
| 46 |
+
"max_seq_len": 1024,
|
| 47 |
+
"max_state_dim": 132,
|
| 48 |
+
"model_dtype": "bfloat16",
|
| 49 |
+
"model_name": "nvidia/Cosmos-Reason2-2B",
|
| 50 |
+
"model_type": "Gr00tN1d7",
|
| 51 |
+
"noise_beta_alpha": 1.5,
|
| 52 |
+
"noise_beta_beta": 1.0,
|
| 53 |
+
"noise_s": 0.999,
|
| 54 |
+
"num_inference_timesteps": 4,
|
| 55 |
+
"num_timestep_buckets": 1000,
|
| 56 |
+
"random_history_crop": true,
|
| 57 |
+
"random_rotation_angle": 0,
|
| 58 |
+
"reproject_vision": false,
|
| 59 |
+
"rtc_ramp_rate": 6.0,
|
| 60 |
+
"select_layer": 16,
|
| 61 |
+
"shortest_image_edge": 256,
|
| 62 |
+
"state_dropout_prob": 0.2,
|
| 63 |
+
"state_gaussian_noise_std": 0.0,
|
| 64 |
+
"transformers_version": "4.57.3",
|
| 65 |
+
"tune_diffusion_model": true,
|
| 66 |
+
"tune_linear": true,
|
| 67 |
+
"tune_llm": false,
|
| 68 |
+
"tune_projector": true,
|
| 69 |
+
"tune_top_llm_layers": 0,
|
| 70 |
+
"tune_visual": false,
|
| 71 |
+
"tune_vlln": true,
|
| 72 |
+
"use_albumentations": true,
|
| 73 |
+
"use_alternate_vl_dit": true,
|
| 74 |
+
"use_flash_attention": true,
|
| 75 |
+
"use_future_tokens": false,
|
| 76 |
+
"use_mean_std": false,
|
| 77 |
+
"use_percentiles": false,
|
| 78 |
+
"use_vl_self_attention": true,
|
| 79 |
+
"use_vlln": true,
|
| 80 |
+
"vl_self_attention_cfg": {
|
| 81 |
+
"attention_head_dim": 64,
|
| 82 |
+
"dropout": 0.2,
|
| 83 |
+
"final_dropout": true,
|
| 84 |
+
"num_attention_heads": 32,
|
| 85 |
+
"num_layers": 4,
|
| 86 |
+
"positional_embeddings": null
|
| 87 |
+
}
|
| 88 |
+
}
|
checkpoint-5500/embodiment_id.json
ADDED
|
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"robocasa_panda_omron": 13,
|
| 3 |
+
"oxe_droid": 17,
|
| 4 |
+
"oxe_fractal": 18,
|
| 5 |
+
"oxe_language_table": 19,
|
| 6 |
+
"oxe_bridge": 20,
|
| 7 |
+
"unknown": 22,
|
| 8 |
+
"gr1_unified": 20,
|
| 9 |
+
"agibot": 26,
|
| 10 |
+
"sim_behavior_r1_pro": 23,
|
| 11 |
+
"xdof": 24,
|
| 12 |
+
"xdof_oss_data": 25,
|
| 13 |
+
"unitree_g1_full_body_with_waist_height_nav_cmd": 25,
|
| 14 |
+
"real_r1_pro_sharpa": 27,
|
| 15 |
+
"real_r1_pro_sharpa_add_view": 27,
|
| 16 |
+
"real_r1_pro_sharpa_relative_arm_joint": 26,
|
| 17 |
+
"real_r1_pro_sharpa_delta_eef": 26,
|
| 18 |
+
"real_r1_pro_sharpa_absolute_eef": 26,
|
| 19 |
+
"real_r1_pro_sharpa_meanstd": 26,
|
| 20 |
+
"real_r1_pro_sharpa_relative_eef": 26,
|
| 21 |
+
"real_r1_pro_sharpa_relative_eef_add_view": 26,
|
| 22 |
+
"real_r1_pro_sharpa_relative_eef_relative_hand": 26,
|
| 23 |
+
"real_r1_pro_sharpa_relative_eef_human": 26,
|
| 24 |
+
"real_r1_pro_sharpa_relative_eef_human_add_view": 26,
|
| 25 |
+
"real_r1_pro_sharpa_relative_eef_human_relative_hand": 26,
|
| 26 |
+
"real_r1_pro_sharpa_relative_eef_egodex": 26,
|
| 27 |
+
"real_r1_pro_sharpa_relative_eef_egodex_relative_hand": 26,
|
| 28 |
+
"real_r1_pro_sharpa_relative_eef_egodex_wrist_only": 26,
|
| 29 |
+
"real_r1_pro_sharpa_relative_eef_maxinsights": 26,
|
| 30 |
+
"real_r1_pro_sharpa_relative_eef_maxinsights_relative_hand": 26,
|
| 31 |
+
"real_r1_pro_sharpa_relative_eef_mecka": 26,
|
| 32 |
+
"real_r1_pro_sharpa_relative_eef_mecka_relative_hand": 26,
|
| 33 |
+
"real_g1_relative_eef_absolute_joints": 25,
|
| 34 |
+
"real_g1_relative_eef_absolute_joints_wrist_cam": 25,
|
| 35 |
+
"real_g1_relative_eef_relative_joints": 25,
|
| 36 |
+
"real_r1_pro_sharpa_relative_eef_relative_hand_relative_joint": 26,
|
| 37 |
+
"real_r1_pro_sharpa_relative_joint": 29,
|
| 38 |
+
"oxe_droid_relative_eef_relative_joint": 24,
|
| 39 |
+
"oxe_droid_relative_eef_relative_joint_swapped": 24,
|
| 40 |
+
"oxe_droid_relative_eef_relative_joint_upweight_z": 24,
|
| 41 |
+
"oxe_droid_relative_eef_relative_joint_upweight_z_swapped": 24,
|
| 42 |
+
"oxe_droid_relative_eef_relative_joint_3view": 24,
|
| 43 |
+
"oxe_droid_relative_eef_relative_joint_3view_swapped": 24,
|
| 44 |
+
"oxe_droid_relative_eef": 24,
|
| 45 |
+
"oxe_droid_joint_position_relative": 24,
|
| 46 |
+
"xdof_relative_eef_relative_joint": 27,
|
| 47 |
+
"xdof_relative_eef_relative_joint_subtask": 27,
|
| 48 |
+
"xdof_relative_eef": 27,
|
| 49 |
+
"xdof_relative_joint": 28,
|
| 50 |
+
"simpler_env_google": 0,
|
| 51 |
+
"simpler_env_widowx": 1,
|
| 52 |
+
"libero_sim": 2,
|
| 53 |
+
"droid_sim": 3,
|
| 54 |
+
"unitree_g1_sonic": 11,
|
| 55 |
+
"new_embodiment": 10,
|
| 56 |
+
"robocasa_gr1_tabletop": 10
|
| 57 |
+
}
|
checkpoint-5500/experiment_cfg/conf.yaml
ADDED
|
@@ -0,0 +1,230 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
load_config_path: null
|
| 2 |
+
model:
|
| 3 |
+
model_type: Gr00tN1d7
|
| 4 |
+
model_dtype: bfloat16
|
| 5 |
+
model_name: nvidia/Cosmos-Reason2-2B
|
| 6 |
+
backbone_model_type: qwen
|
| 7 |
+
model_revision: null
|
| 8 |
+
tune_top_llm_layers: 0
|
| 9 |
+
backbone_embedding_dim: 2048
|
| 10 |
+
tune_llm: false
|
| 11 |
+
tune_visual: false
|
| 12 |
+
select_layer: 12
|
| 13 |
+
reproject_vision: false
|
| 14 |
+
use_flash_attention: true
|
| 15 |
+
load_bf16: false
|
| 16 |
+
backbone_trainable_params_fp32: true
|
| 17 |
+
image_crop_size:
|
| 18 |
+
- 230
|
| 19 |
+
- 230
|
| 20 |
+
image_target_size:
|
| 21 |
+
- 256
|
| 22 |
+
- 256
|
| 23 |
+
shortest_image_edge: null
|
| 24 |
+
crop_fraction: null
|
| 25 |
+
random_rotation_angle: null
|
| 26 |
+
color_jitter_params:
|
| 27 |
+
brightness: 0.2
|
| 28 |
+
contrast: 0.2
|
| 29 |
+
saturation: 0.2
|
| 30 |
+
hue: 0.03
|
| 31 |
+
use_albumentations_transforms: true
|
| 32 |
+
extra_augmentation_config: null
|
| 33 |
+
formalize_language: true
|
| 34 |
+
apply_sincos_state_encoding: false
|
| 35 |
+
use_percentiles: true
|
| 36 |
+
use_relative_action: true
|
| 37 |
+
max_state_dim: 132
|
| 38 |
+
max_action_dim: 132
|
| 39 |
+
action_horizon: 40
|
| 40 |
+
hidden_size: 1024
|
| 41 |
+
input_embedding_dim: 1536
|
| 42 |
+
state_history_length: 1
|
| 43 |
+
add_pos_embed: true
|
| 44 |
+
attn_dropout: 0.2
|
| 45 |
+
use_vlln: true
|
| 46 |
+
max_seq_len: 1024
|
| 47 |
+
use_alternate_vl_dit: true
|
| 48 |
+
attend_text_every_n_blocks: 2
|
| 49 |
+
diffusion_model_cfg:
|
| 50 |
+
positional_embeddings: null
|
| 51 |
+
num_layers: 16
|
| 52 |
+
num_attention_heads: 32
|
| 53 |
+
attention_head_dim: 48
|
| 54 |
+
norm_type: ada_norm
|
| 55 |
+
dropout: 0.2
|
| 56 |
+
final_dropout: true
|
| 57 |
+
output_dim: 1024
|
| 58 |
+
interleave_self_attention: true
|
| 59 |
+
num_inference_timesteps: 4
|
| 60 |
+
noise_beta_alpha: 1.5
|
| 61 |
+
noise_beta_beta: 1.0
|
| 62 |
+
noise_s: 0.999
|
| 63 |
+
num_timestep_buckets: 1000
|
| 64 |
+
tune_projector: true
|
| 65 |
+
tune_diffusion_model: true
|
| 66 |
+
tune_vlln: true
|
| 67 |
+
state_dropout_prob: 0.2
|
| 68 |
+
exclude_state: false
|
| 69 |
+
use_mean_std: false
|
| 70 |
+
max_num_embodiments: 32
|
| 71 |
+
data:
|
| 72 |
+
datasets:
|
| 73 |
+
- dataset_paths:
|
| 74 |
+
- /root/screwing_train/datasets/screwing35_follower_train
|
| 75 |
+
embodiment_tag: new_embodiment
|
| 76 |
+
mix_ratio: 0.6
|
| 77 |
+
dataset_type: physical_embodiment
|
| 78 |
+
val_dataset_path: null
|
| 79 |
+
- dataset_paths:
|
| 80 |
+
- /root/screwing_train/datasets/dagger_human
|
| 81 |
+
embodiment_tag: new_embodiment
|
| 82 |
+
mix_ratio: 0.3
|
| 83 |
+
dataset_type: physical_embodiment
|
| 84 |
+
val_dataset_path: null
|
| 85 |
+
- dataset_paths:
|
| 86 |
+
- /root/screwing_train/datasets/dagger_auto
|
| 87 |
+
embodiment_tag: new_embodiment
|
| 88 |
+
mix_ratio: 0.1
|
| 89 |
+
dataset_type: physical_embodiment
|
| 90 |
+
val_dataset_path: null
|
| 91 |
+
modality_configs:
|
| 92 |
+
new_embodiment:
|
| 93 |
+
video:
|
| 94 |
+
delta_indices:
|
| 95 |
+
- 0
|
| 96 |
+
modality_keys:
|
| 97 |
+
- cam_left_head
|
| 98 |
+
- cam_left_wrist
|
| 99 |
+
- cam_right_wrist
|
| 100 |
+
sin_cos_embedding_keys: null
|
| 101 |
+
mean_std_embedding_keys: null
|
| 102 |
+
action_configs: null
|
| 103 |
+
state:
|
| 104 |
+
delta_indices:
|
| 105 |
+
- 0
|
| 106 |
+
modality_keys:
|
| 107 |
+
- arm_left
|
| 108 |
+
- arm_right
|
| 109 |
+
sin_cos_embedding_keys: null
|
| 110 |
+
mean_std_embedding_keys: null
|
| 111 |
+
action_configs: null
|
| 112 |
+
action:
|
| 113 |
+
delta_indices:
|
| 114 |
+
- 0
|
| 115 |
+
- 1
|
| 116 |
+
- 2
|
| 117 |
+
- 3
|
| 118 |
+
- 4
|
| 119 |
+
- 5
|
| 120 |
+
- 6
|
| 121 |
+
- 7
|
| 122 |
+
- 8
|
| 123 |
+
- 9
|
| 124 |
+
- 10
|
| 125 |
+
- 11
|
| 126 |
+
- 12
|
| 127 |
+
- 13
|
| 128 |
+
- 14
|
| 129 |
+
- 15
|
| 130 |
+
modality_keys:
|
| 131 |
+
- arm_left
|
| 132 |
+
- arm_right
|
| 133 |
+
sin_cos_embedding_keys: null
|
| 134 |
+
mean_std_embedding_keys: null
|
| 135 |
+
action_configs:
|
| 136 |
+
- rep: ABSOLUTE
|
| 137 |
+
type: NON_EEF
|
| 138 |
+
format: DEFAULT
|
| 139 |
+
state_key: null
|
| 140 |
+
- rep: ABSOLUTE
|
| 141 |
+
type: NON_EEF
|
| 142 |
+
format: DEFAULT
|
| 143 |
+
state_key: null
|
| 144 |
+
language:
|
| 145 |
+
delta_indices:
|
| 146 |
+
- 0
|
| 147 |
+
modality_keys:
|
| 148 |
+
- annotation.human.primitive_instruction
|
| 149 |
+
sin_cos_embedding_keys: null
|
| 150 |
+
mean_std_embedding_keys: null
|
| 151 |
+
action_configs: null
|
| 152 |
+
download_cache: false
|
| 153 |
+
shard_size: 1024
|
| 154 |
+
episode_sampling_rate: 0.1
|
| 155 |
+
num_shards_per_epoch: 100000
|
| 156 |
+
override_pretraining_statistics: true
|
| 157 |
+
mode: single_turn
|
| 158 |
+
random_chop: 0.0
|
| 159 |
+
mock_dataset_mode: false
|
| 160 |
+
shuffle: true
|
| 161 |
+
seed: 42
|
| 162 |
+
multiprocessing_context: fork
|
| 163 |
+
allow_padding: false
|
| 164 |
+
subsample_ratio: 1.0
|
| 165 |
+
image_crop_size:
|
| 166 |
+
- 244
|
| 167 |
+
- 244
|
| 168 |
+
image_target_size:
|
| 169 |
+
- 224
|
| 170 |
+
- 224
|
| 171 |
+
video_backend: torchcodec
|
| 172 |
+
training:
|
| 173 |
+
output_dir: /root/screwing_train/runs/groot_dagger_it1
|
| 174 |
+
experiment_name: null
|
| 175 |
+
max_steps: 6000
|
| 176 |
+
global_batch_size: 128
|
| 177 |
+
batch_size: null
|
| 178 |
+
gradient_accumulation_steps: 1
|
| 179 |
+
learning_rate: 3.0e-05
|
| 180 |
+
lr_scheduler_type: cosine
|
| 181 |
+
weight_decay: 1.0e-05
|
| 182 |
+
warmup_ratio: 0.03
|
| 183 |
+
warmup_steps: 0
|
| 184 |
+
max_grad_norm: 1.0
|
| 185 |
+
optim: adamw_torch
|
| 186 |
+
start_from_checkpoint: /root/screwing_train/runs/16d_30k/checkpoint-30000
|
| 187 |
+
skip_weight_loading: false
|
| 188 |
+
tf32: true
|
| 189 |
+
fp16: false
|
| 190 |
+
bf16: true
|
| 191 |
+
eval_bf16: true
|
| 192 |
+
logging_steps: 10
|
| 193 |
+
save_steps: 250
|
| 194 |
+
save_total_limit: 24
|
| 195 |
+
save_vl_model: false
|
| 196 |
+
save_only_model: true
|
| 197 |
+
upload_checkpoints: false
|
| 198 |
+
upload_every: 1000
|
| 199 |
+
upload_last_n_checkpoints: 5
|
| 200 |
+
max_concurrent_uploads: 2
|
| 201 |
+
eval_strategy: 'no'
|
| 202 |
+
eval_steps: 500
|
| 203 |
+
eval_set_split_ratio: 0.1
|
| 204 |
+
eval_batch_size: 2
|
| 205 |
+
save_best_eval_metric_name: ''
|
| 206 |
+
save_best_eval_metric_greater_is_better: true
|
| 207 |
+
deepspeed_stage: 2
|
| 208 |
+
gradient_checkpointing: false
|
| 209 |
+
transformers_trust_remote_code: true
|
| 210 |
+
transformers_local_files_only: false
|
| 211 |
+
transformers_cache_dir: null
|
| 212 |
+
transformers_access_token: null
|
| 213 |
+
use_ddp: false
|
| 214 |
+
ddp_bucket_cap_mb: 100
|
| 215 |
+
num_gpus: 2
|
| 216 |
+
dataloader_num_workers: 32
|
| 217 |
+
remove_unused_columns: false
|
| 218 |
+
use_wandb: true
|
| 219 |
+
wandb_project: screwing_hil
|
| 220 |
+
enable_profiling: false
|
| 221 |
+
max_retries: 3
|
| 222 |
+
assert_loss_less_than: null
|
| 223 |
+
add_rl_callback: false
|
| 224 |
+
enable_open_loop_eval: false
|
| 225 |
+
open_loop_eval_traj_ids:
|
| 226 |
+
- 0
|
| 227 |
+
open_loop_eval_steps_per_traj: 100
|
| 228 |
+
open_loop_eval_plot_indices: null
|
| 229 |
+
max_steps: 6000
|
| 230 |
+
save_steps: 250
|
checkpoint-5500/experiment_cfg/config.yaml
ADDED
|
@@ -0,0 +1,257 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
!!python/object:gr00t.configs.base_config.Config
|
| 2 |
+
data: !!python/object:gr00t.configs.data.data_config.DataConfig
|
| 3 |
+
allow_padding: false
|
| 4 |
+
datasets:
|
| 5 |
+
- !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
|
| 6 |
+
dataset_paths:
|
| 7 |
+
- /root/screwing_train/datasets/screwing35_follower_train
|
| 8 |
+
dataset_type: physical_embodiment
|
| 9 |
+
embodiment_tag: new_embodiment
|
| 10 |
+
mix_ratio: 0.6
|
| 11 |
+
val_dataset_path: null
|
| 12 |
+
- !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
|
| 13 |
+
dataset_paths:
|
| 14 |
+
- /root/screwing_train/datasets/dagger_human
|
| 15 |
+
dataset_type: physical_embodiment
|
| 16 |
+
embodiment_tag: new_embodiment
|
| 17 |
+
mix_ratio: 0.3
|
| 18 |
+
val_dataset_path: null
|
| 19 |
+
- !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
|
| 20 |
+
dataset_paths:
|
| 21 |
+
- /root/screwing_train/datasets/dagger_auto
|
| 22 |
+
dataset_type: physical_embodiment
|
| 23 |
+
embodiment_tag: new_embodiment
|
| 24 |
+
mix_ratio: 0.1
|
| 25 |
+
val_dataset_path: null
|
| 26 |
+
download_cache: false
|
| 27 |
+
episode_sampling_rate: 0.1
|
| 28 |
+
image_crop_size:
|
| 29 |
+
- 244
|
| 30 |
+
- 244
|
| 31 |
+
image_target_size:
|
| 32 |
+
- 224
|
| 33 |
+
- 224
|
| 34 |
+
mock_dataset_mode: false
|
| 35 |
+
modality_configs:
|
| 36 |
+
new_embodiment:
|
| 37 |
+
action: !!python/object:gr00t.data.types.ModalityConfig
|
| 38 |
+
action_configs:
|
| 39 |
+
- !!python/object:gr00t.data.types.ActionConfig
|
| 40 |
+
format: &id001 !!python/object/apply:gr00t.data.types.ActionFormat
|
| 41 |
+
- default
|
| 42 |
+
rep: &id002 !!python/object/apply:gr00t.data.types.ActionRepresentation
|
| 43 |
+
- absolute
|
| 44 |
+
state_key: null
|
| 45 |
+
type: &id003 !!python/object/apply:gr00t.data.types.ActionType
|
| 46 |
+
- non_eef
|
| 47 |
+
- !!python/object:gr00t.data.types.ActionConfig
|
| 48 |
+
format: *id001
|
| 49 |
+
rep: *id002
|
| 50 |
+
state_key: null
|
| 51 |
+
type: *id003
|
| 52 |
+
delta_indices:
|
| 53 |
+
- 0
|
| 54 |
+
- 1
|
| 55 |
+
- 2
|
| 56 |
+
- 3
|
| 57 |
+
- 4
|
| 58 |
+
- 5
|
| 59 |
+
- 6
|
| 60 |
+
- 7
|
| 61 |
+
- 8
|
| 62 |
+
- 9
|
| 63 |
+
- 10
|
| 64 |
+
- 11
|
| 65 |
+
- 12
|
| 66 |
+
- 13
|
| 67 |
+
- 14
|
| 68 |
+
- 15
|
| 69 |
+
mean_std_embedding_keys: null
|
| 70 |
+
modality_keys:
|
| 71 |
+
- arm_left
|
| 72 |
+
- arm_right
|
| 73 |
+
sin_cos_embedding_keys: null
|
| 74 |
+
language: !!python/object:gr00t.data.types.ModalityConfig
|
| 75 |
+
action_configs: null
|
| 76 |
+
delta_indices:
|
| 77 |
+
- 0
|
| 78 |
+
mean_std_embedding_keys: null
|
| 79 |
+
modality_keys:
|
| 80 |
+
- annotation.human.primitive_instruction
|
| 81 |
+
sin_cos_embedding_keys: null
|
| 82 |
+
state: !!python/object:gr00t.data.types.ModalityConfig
|
| 83 |
+
action_configs: null
|
| 84 |
+
delta_indices:
|
| 85 |
+
- 0
|
| 86 |
+
mean_std_embedding_keys: null
|
| 87 |
+
modality_keys:
|
| 88 |
+
- arm_left
|
| 89 |
+
- arm_right
|
| 90 |
+
sin_cos_embedding_keys: null
|
| 91 |
+
video: !!python/object:gr00t.data.types.ModalityConfig
|
| 92 |
+
action_configs: null
|
| 93 |
+
delta_indices:
|
| 94 |
+
- 0
|
| 95 |
+
mean_std_embedding_keys: null
|
| 96 |
+
modality_keys:
|
| 97 |
+
- cam_left_head
|
| 98 |
+
- cam_left_wrist
|
| 99 |
+
- cam_right_wrist
|
| 100 |
+
sin_cos_embedding_keys: null
|
| 101 |
+
mode: single_turn
|
| 102 |
+
multiprocessing_context: fork
|
| 103 |
+
num_shards_per_epoch: 100000
|
| 104 |
+
override_pretraining_statistics: true
|
| 105 |
+
random_chop: 0.0
|
| 106 |
+
seed: 42
|
| 107 |
+
shard_size: 1024
|
| 108 |
+
shuffle: true
|
| 109 |
+
subsample_ratio: 1.0
|
| 110 |
+
video_backend: torchcodec
|
| 111 |
+
load_config_path: null
|
| 112 |
+
model: !!python/object:gr00t.configs.model.gr00t_n1d7.Gr00tN1d7Config
|
| 113 |
+
_attn_implementation_internal: null
|
| 114 |
+
_commit_hash: null
|
| 115 |
+
_name_or_path: ''
|
| 116 |
+
_output_attentions: false
|
| 117 |
+
add_cross_attention: false
|
| 118 |
+
architectures: null
|
| 119 |
+
backbone_trainable_params_fp32: true
|
| 120 |
+
bad_words_ids: null
|
| 121 |
+
begin_suppress_tokens: null
|
| 122 |
+
bos_token_id: null
|
| 123 |
+
chunk_size_feed_forward: 0
|
| 124 |
+
color_jitter_params:
|
| 125 |
+
brightness: 0.2
|
| 126 |
+
contrast: 0.2
|
| 127 |
+
hue: 0.03
|
| 128 |
+
saturation: 0.2
|
| 129 |
+
cross_attention_hidden_size: null
|
| 130 |
+
decoder_start_token_id: null
|
| 131 |
+
diffusion_model_cfg:
|
| 132 |
+
attention_head_dim: 48
|
| 133 |
+
dropout: 0.2
|
| 134 |
+
final_dropout: true
|
| 135 |
+
interleave_self_attention: true
|
| 136 |
+
norm_type: ada_norm
|
| 137 |
+
num_attention_heads: 32
|
| 138 |
+
num_layers: 16
|
| 139 |
+
output_dim: 1024
|
| 140 |
+
positional_embeddings: null
|
| 141 |
+
diversity_penalty: 0.0
|
| 142 |
+
do_sample: false
|
| 143 |
+
dtype: null
|
| 144 |
+
early_stopping: false
|
| 145 |
+
encoder_no_repeat_ngram_size: 0
|
| 146 |
+
eos_token_id: null
|
| 147 |
+
exponential_decay_length_penalty: null
|
| 148 |
+
extra_augmentation_config: null
|
| 149 |
+
finetuning_task: null
|
| 150 |
+
forced_bos_token_id: null
|
| 151 |
+
forced_eos_token_id: null
|
| 152 |
+
id2label:
|
| 153 |
+
0: LABEL_0
|
| 154 |
+
1: LABEL_1
|
| 155 |
+
is_decoder: false
|
| 156 |
+
is_encoder_decoder: false
|
| 157 |
+
label2id:
|
| 158 |
+
LABEL_0: 0
|
| 159 |
+
LABEL_1: 1
|
| 160 |
+
length_penalty: 1.0
|
| 161 |
+
load_bf16: false
|
| 162 |
+
max_length: 20
|
| 163 |
+
min_length: 0
|
| 164 |
+
model_name: nvidia/Cosmos-Reason2-2B
|
| 165 |
+
no_repeat_ngram_size: 0
|
| 166 |
+
num_beam_groups: 1
|
| 167 |
+
num_beams: 1
|
| 168 |
+
num_return_sequences: 1
|
| 169 |
+
output_hidden_states: false
|
| 170 |
+
output_scores: false
|
| 171 |
+
pad_token_id: null
|
| 172 |
+
prefix: null
|
| 173 |
+
problem_type: null
|
| 174 |
+
pruned_heads: {}
|
| 175 |
+
random_rotation_angle: null
|
| 176 |
+
remove_invalid_values: false
|
| 177 |
+
repetition_penalty: 1.0
|
| 178 |
+
reproject_vision: false
|
| 179 |
+
return_dict: true
|
| 180 |
+
return_dict_in_generate: false
|
| 181 |
+
sep_token_id: null
|
| 182 |
+
state_dropout_prob: 0.2
|
| 183 |
+
suppress_tokens: null
|
| 184 |
+
task_specific_params: null
|
| 185 |
+
temperature: 1.0
|
| 186 |
+
tf_legacy_loss: false
|
| 187 |
+
tie_encoder_decoder: false
|
| 188 |
+
tie_word_embeddings: true
|
| 189 |
+
tokenizer_class: null
|
| 190 |
+
top_k: 50
|
| 191 |
+
top_p: 1.0
|
| 192 |
+
torchscript: false
|
| 193 |
+
transformers_version: null
|
| 194 |
+
tune_diffusion_model: true
|
| 195 |
+
tune_llm: false
|
| 196 |
+
tune_projector: true
|
| 197 |
+
tune_visual: false
|
| 198 |
+
typical_p: 1.0
|
| 199 |
+
use_bfloat16: false
|
| 200 |
+
use_relative_action: true
|
| 201 |
+
training: !!python/object:gr00t.configs.training.training_config.TrainingConfig
|
| 202 |
+
add_rl_callback: false
|
| 203 |
+
assert_loss_less_than: null
|
| 204 |
+
batch_size: null
|
| 205 |
+
bf16: true
|
| 206 |
+
dataloader_num_workers: 32
|
| 207 |
+
ddp_bucket_cap_mb: 100
|
| 208 |
+
deepspeed_stage: 2
|
| 209 |
+
enable_open_loop_eval: false
|
| 210 |
+
enable_profiling: false
|
| 211 |
+
eval_batch_size: 2
|
| 212 |
+
eval_bf16: true
|
| 213 |
+
eval_set_split_ratio: 0.1
|
| 214 |
+
eval_steps: 500
|
| 215 |
+
eval_strategy: 'no'
|
| 216 |
+
experiment_name: null
|
| 217 |
+
fp16: false
|
| 218 |
+
global_batch_size: 128
|
| 219 |
+
gradient_accumulation_steps: 1
|
| 220 |
+
gradient_checkpointing: false
|
| 221 |
+
learning_rate: 3.0e-05
|
| 222 |
+
logging_steps: 10
|
| 223 |
+
lr_scheduler_type: cosine
|
| 224 |
+
max_concurrent_uploads: 2
|
| 225 |
+
max_grad_norm: 1.0
|
| 226 |
+
max_retries: 3
|
| 227 |
+
max_steps: 6000
|
| 228 |
+
num_gpus: 2
|
| 229 |
+
open_loop_eval_plot_indices: null
|
| 230 |
+
open_loop_eval_steps_per_traj: 100
|
| 231 |
+
open_loop_eval_traj_ids:
|
| 232 |
+
- 0
|
| 233 |
+
optim: adamw_torch
|
| 234 |
+
output_dir: /root/screwing_train/runs/groot_dagger_it1
|
| 235 |
+
remove_unused_columns: false
|
| 236 |
+
save_best_eval_metric_greater_is_better: true
|
| 237 |
+
save_best_eval_metric_name: ''
|
| 238 |
+
save_only_model: true
|
| 239 |
+
save_steps: 250
|
| 240 |
+
save_total_limit: 24
|
| 241 |
+
save_vl_model: false
|
| 242 |
+
skip_weight_loading: false
|
| 243 |
+
start_from_checkpoint: /root/screwing_train/runs/16d_30k/checkpoint-30000
|
| 244 |
+
tf32: true
|
| 245 |
+
transformers_access_token: null
|
| 246 |
+
transformers_cache_dir: null
|
| 247 |
+
transformers_local_files_only: false
|
| 248 |
+
transformers_trust_remote_code: true
|
| 249 |
+
upload_checkpoints: false
|
| 250 |
+
upload_every: 1000
|
| 251 |
+
upload_last_n_checkpoints: 5
|
| 252 |
+
use_ddp: false
|
| 253 |
+
use_wandb: true
|
| 254 |
+
wandb_project: screwing_hil
|
| 255 |
+
warmup_ratio: 0.03
|
| 256 |
+
warmup_steps: 0
|
| 257 |
+
weight_decay: 1.0e-05
|
checkpoint-5500/experiment_cfg/dataset_statistics.json
ADDED
|
@@ -0,0 +1,257 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"new_embodiment": {
|
| 3 |
+
"state": {
|
| 4 |
+
"arm_left": {
|
| 5 |
+
"min": [
|
| 6 |
+
-0.3512935936450958,
|
| 7 |
+
-0.0003355582884978503,
|
| 8 |
+
-0.4982561469078064,
|
| 9 |
+
-2.0985336303710938,
|
| 10 |
+
-0.30987611413002014,
|
| 11 |
+
-0.390062540769577,
|
| 12 |
+
-0.756526529788971,
|
| 13 |
+
-0.0015339808305725455
|
| 14 |
+
],
|
| 15 |
+
"max": [
|
| 16 |
+
0.4587441384792328,
|
| 17 |
+
0.4674566686153412,
|
| 18 |
+
0.2403915673494339,
|
| 19 |
+
-0.8097261190414429,
|
| 20 |
+
0.3366488516330719,
|
| 21 |
+
0.7219297289848328,
|
| 22 |
+
0.09214304387569427,
|
| 23 |
+
1.0001554489135742
|
| 24 |
+
],
|
| 25 |
+
"mean": [
|
| 26 |
+
0.04355059830193595,
|
| 27 |
+
0.10864019471928726,
|
| 28 |
+
-0.10750242693181013,
|
| 29 |
+
-1.7219742652123884,
|
| 30 |
+
0.013241997072442023,
|
| 31 |
+
0.24038831582343306,
|
| 32 |
+
-0.3289317812120216,
|
| 33 |
+
0.21972585904618444
|
| 34 |
+
],
|
| 35 |
+
"std": [
|
| 36 |
+
0.201977025178683,
|
| 37 |
+
0.08930106217863336,
|
| 38 |
+
0.1703593429065693,
|
| 39 |
+
0.2290651088739729,
|
| 40 |
+
0.13351265238651933,
|
| 41 |
+
0.2640519575529048,
|
| 42 |
+
0.2311836246595438,
|
| 43 |
+
0.3695154517778703
|
| 44 |
+
],
|
| 45 |
+
"q01": [
|
| 46 |
+
-0.28752409607172014,
|
| 47 |
+
0.0,
|
| 48 |
+
-0.4731247413158417,
|
| 49 |
+
-2.013697385787964,
|
| 50 |
+
-0.2523278594017029,
|
| 51 |
+
-0.2879210114479065,
|
| 52 |
+
-0.7209579944610596,
|
| 53 |
+
0.0
|
| 54 |
+
],
|
| 55 |
+
"q99": [
|
| 56 |
+
0.4115143120288849,
|
| 57 |
+
0.4564431607723236,
|
| 58 |
+
0.18591128289699554,
|
| 59 |
+
-0.932169679403305,
|
| 60 |
+
0.2769750881195063,
|
| 61 |
+
0.7118869423866272,
|
| 62 |
+
0.06595534831285477,
|
| 63 |
+
0.9970875382423401
|
| 64 |
+
]
|
| 65 |
+
},
|
| 66 |
+
"arm_right": {
|
| 67 |
+
"min": [
|
| 68 |
+
-0.9125747680664062,
|
| 69 |
+
-0.46813976764678955,
|
| 70 |
+
-0.08915065228939056,
|
| 71 |
+
-2.0727076530456543,
|
| 72 |
+
-0.9404261112213135,
|
| 73 |
+
-0.15235544741153717,
|
| 74 |
+
-0.053947482258081436,
|
| 75 |
+
0.0
|
| 76 |
+
],
|
| 77 |
+
"max": [
|
| 78 |
+
0.4171828627586365,
|
| 79 |
+
-0.031806133687496185,
|
| 80 |
+
1.0656253099441528,
|
| 81 |
+
-1.032512903213501,
|
| 82 |
+
0.13881327211856842,
|
| 83 |
+
1.1778095960617065,
|
| 84 |
+
0.8715579509735107,
|
| 85 |
+
1.277806043624878
|
| 86 |
+
],
|
| 87 |
+
"mean": [
|
| 88 |
+
-0.29372113303835257,
|
| 89 |
+
-0.25370953344792585,
|
| 90 |
+
0.3190493214782265,
|
| 91 |
+
-1.6728425137445484,
|
| 92 |
+
-0.24982502985306895,
|
| 93 |
+
0.49206255521233844,
|
| 94 |
+
0.42576361554110886,
|
| 95 |
+
0.4498966951422322
|
| 96 |
+
],
|
| 97 |
+
"std": [
|
| 98 |
+
0.24819130455260907,
|
| 99 |
+
0.05331376467400995,
|
| 100 |
+
0.17461899684902152,
|
| 101 |
+
0.1924666931220957,
|
| 102 |
+
0.18633659471924663,
|
| 103 |
+
0.28551332367395904,
|
| 104 |
+
0.17700279003797392,
|
| 105 |
+
0.3465478824834864
|
| 106 |
+
],
|
| 107 |
+
"q01": [
|
| 108 |
+
-0.8353484272956848,
|
| 109 |
+
-0.43082091212272644,
|
| 110 |
+
-0.01853959634900093,
|
| 111 |
+
-2.0143797326087953,
|
| 112 |
+
-0.7463775277137756,
|
| 113 |
+
-0.06663229316473007,
|
| 114 |
+
0.07988695055246353,
|
| 115 |
+
0.0015339808305725455
|
| 116 |
+
],
|
| 117 |
+
"q99": [
|
| 118 |
+
0.28748956322669983,
|
| 119 |
+
-0.14874820411205292,
|
| 120 |
+
0.8437014222145081,
|
| 121 |
+
-1.2486603260040283,
|
| 122 |
+
0.12267052382230759,
|
| 123 |
+
1.0415489673614502,
|
| 124 |
+
0.8113635182380676,
|
| 125 |
+
0.8682331442832947
|
| 126 |
+
]
|
| 127 |
+
}
|
| 128 |
+
},
|
| 129 |
+
"action": {
|
| 130 |
+
"arm_left": {
|
| 131 |
+
"min": [
|
| 132 |
+
-0.3512935936450958,
|
| 133 |
+
-0.0003355582884978503,
|
| 134 |
+
-0.49900805950164795,
|
| 135 |
+
-2.0985336303710938,
|
| 136 |
+
-0.30987611413002014,
|
| 137 |
+
-0.390062540769577,
|
| 138 |
+
-0.756526529788971,
|
| 139 |
+
-0.0015339808305725455
|
| 140 |
+
],
|
| 141 |
+
"max": [
|
| 142 |
+
0.4587441384792328,
|
| 143 |
+
0.4674566686153412,
|
| 144 |
+
0.2403915673494339,
|
| 145 |
+
-0.8097261190414429,
|
| 146 |
+
0.3393723964691162,
|
| 147 |
+
0.7219297289848328,
|
| 148 |
+
0.09214304387569427,
|
| 149 |
+
1.2387962341308594
|
| 150 |
+
],
|
| 151 |
+
"mean": [
|
| 152 |
+
0.04354117550648276,
|
| 153 |
+
0.10888600258373515,
|
| 154 |
+
-0.10755284350603393,
|
| 155 |
+
-1.7220248298756244,
|
| 156 |
+
0.013494755258123084,
|
| 157 |
+
0.24065059866611607,
|
| 158 |
+
-0.32936498765478983,
|
| 159 |
+
0.22581150584984677
|
| 160 |
+
],
|
| 161 |
+
"std": [
|
| 162 |
+
0.20216011477897333,
|
| 163 |
+
0.08923828878575918,
|
| 164 |
+
0.170728239388772,
|
| 165 |
+
0.22946179527618987,
|
| 166 |
+
0.13395482743250087,
|
| 167 |
+
0.2638166849754727,
|
| 168 |
+
0.23123107266252962,
|
| 169 |
+
0.3869448499567072
|
| 170 |
+
],
|
| 171 |
+
"q01": [
|
| 172 |
+
-0.2880094051361084,
|
| 173 |
+
0.0,
|
| 174 |
+
-0.4840488076210022,
|
| 175 |
+
-2.013697385787964,
|
| 176 |
+
-0.2523278594017029,
|
| 177 |
+
-0.2877771854400635,
|
| 178 |
+
-0.7209579944610596,
|
| 179 |
+
0.0
|
| 180 |
+
],
|
| 181 |
+
"q99": [
|
| 182 |
+
0.41167011857032776,
|
| 183 |
+
0.4564431607723236,
|
| 184 |
+
0.1860191375017166,
|
| 185 |
+
-0.9313069581985474,
|
| 186 |
+
0.2917774844169608,
|
| 187 |
+
0.7118869423866272,
|
| 188 |
+
0.06595534831285477,
|
| 189 |
+
1.23726224899292
|
| 190 |
+
]
|
| 191 |
+
},
|
| 192 |
+
"arm_right": {
|
| 193 |
+
"min": [
|
| 194 |
+
-0.9125747680664062,
|
| 195 |
+
-0.46813976764678955,
|
| 196 |
+
-0.08915065228939056,
|
| 197 |
+
-2.0727076530456543,
|
| 198 |
+
-0.9404261112213135,
|
| 199 |
+
-0.15235544741153717,
|
| 200 |
+
-0.053947482258081436,
|
| 201 |
+
0.0
|
| 202 |
+
],
|
| 203 |
+
"max": [
|
| 204 |
+
0.4171828627586365,
|
| 205 |
+
-0.031806133687496185,
|
| 206 |
+
1.0656253099441528,
|
| 207 |
+
-1.032512903213501,
|
| 208 |
+
0.13881327211856842,
|
| 209 |
+
1.1778095960617065,
|
| 210 |
+
0.8715579509735107,
|
| 211 |
+
1.3185632228851318
|
| 212 |
+
],
|
| 213 |
+
"mean": [
|
| 214 |
+
-0.2942656780309203,
|
| 215 |
+
-0.25352573533534634,
|
| 216 |
+
0.31921004536062647,
|
| 217 |
+
-1.6735023107714255,
|
| 218 |
+
-0.2505161447075832,
|
| 219 |
+
0.4924302416916429,
|
| 220 |
+
0.42694812914012836,
|
| 221 |
+
0.5413754129270169
|
| 222 |
+
],
|
| 223 |
+
"std": [
|
| 224 |
+
0.2486902262630544,
|
| 225 |
+
0.05344843369680529,
|
| 226 |
+
0.17478912708816832,
|
| 227 |
+
0.19249047992758175,
|
| 228 |
+
0.18634200598780004,
|
| 229 |
+
0.28484720108338507,
|
| 230 |
+
0.17626246445564833,
|
| 231 |
+
0.47102636928109104
|
| 232 |
+
],
|
| 233 |
+
"q01": [
|
| 234 |
+
-0.8353484272956848,
|
| 235 |
+
-0.43082091212272644,
|
| 236 |
+
-0.01853959634900093,
|
| 237 |
+
-2.014399766921997,
|
| 238 |
+
-0.7463775277137756,
|
| 239 |
+
-0.06663229316473007,
|
| 240 |
+
0.07988695055246353,
|
| 241 |
+
0.0015339808305725455
|
| 242 |
+
],
|
| 243 |
+
"q99": [
|
| 244 |
+
0.288448303937912,
|
| 245 |
+
-0.14868827164173126,
|
| 246 |
+
0.8437014222145081,
|
| 247 |
+
-1.2486603260040283,
|
| 248 |
+
0.12267052382230759,
|
| 249 |
+
1.0415489673614502,
|
| 250 |
+
0.8113635182380676,
|
| 251 |
+
1.3185632228851318
|
| 252 |
+
]
|
| 253 |
+
}
|
| 254 |
+
},
|
| 255 |
+
"relative_action": {}
|
| 256 |
+
}
|
| 257 |
+
}
|
checkpoint-5500/experiment_cfg/final_model_config.json
ADDED
|
@@ -0,0 +1,55 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_type": "Gr00tN1d7",
|
| 3 |
+
"model_dtype": "bfloat16",
|
| 4 |
+
"model_name": "nvidia/Cosmos-Reason2-2B",
|
| 5 |
+
"backbone_model_type": "qwen",
|
| 6 |
+
"model_revision": null,
|
| 7 |
+
"tune_top_llm_layers": 0,
|
| 8 |
+
"backbone_embedding_dim": 2048,
|
| 9 |
+
"tune_llm": false,
|
| 10 |
+
"tune_visual": false,
|
| 11 |
+
"select_layer": 16,
|
| 12 |
+
"reproject_vision": false,
|
| 13 |
+
"use_flash_attention": true,
|
| 14 |
+
"load_bf16": false,
|
| 15 |
+
"backbone_trainable_params_fp32": true,
|
| 16 |
+
"extra_augmentation_config": null,
|
| 17 |
+
"apply_sincos_state_encoding": false,
|
| 18 |
+
"use_percentiles": false,
|
| 19 |
+
"use_relative_action": false,
|
| 20 |
+
"max_state_dim": 132,
|
| 21 |
+
"max_action_dim": 132,
|
| 22 |
+
"action_horizon": 40,
|
| 23 |
+
"hidden_size": 1024,
|
| 24 |
+
"input_embedding_dim": 1536,
|
| 25 |
+
"state_history_length": 1,
|
| 26 |
+
"add_pos_embed": true,
|
| 27 |
+
"attn_dropout": 0.2,
|
| 28 |
+
"use_vlln": true,
|
| 29 |
+
"max_seq_len": 1024,
|
| 30 |
+
"use_alternate_vl_dit": true,
|
| 31 |
+
"attend_text_every_n_blocks": 2,
|
| 32 |
+
"diffusion_model_cfg": {
|
| 33 |
+
"attention_head_dim": 48,
|
| 34 |
+
"dropout": 0.2,
|
| 35 |
+
"final_dropout": true,
|
| 36 |
+
"interleave_self_attention": true,
|
| 37 |
+
"norm_type": "ada_norm",
|
| 38 |
+
"num_attention_heads": 32,
|
| 39 |
+
"num_layers": 32,
|
| 40 |
+
"output_dim": 1024,
|
| 41 |
+
"positional_embeddings": null
|
| 42 |
+
},
|
| 43 |
+
"num_inference_timesteps": 4,
|
| 44 |
+
"noise_beta_alpha": 1.5,
|
| 45 |
+
"noise_beta_beta": 1.0,
|
| 46 |
+
"noise_s": 0.999,
|
| 47 |
+
"num_timestep_buckets": 1000,
|
| 48 |
+
"tune_projector": true,
|
| 49 |
+
"tune_diffusion_model": true,
|
| 50 |
+
"tune_vlln": true,
|
| 51 |
+
"state_dropout_prob": 0.2,
|
| 52 |
+
"exclude_state": false,
|
| 53 |
+
"use_mean_std": false,
|
| 54 |
+
"max_num_embodiments": 32
|
| 55 |
+
}
|
checkpoint-5500/experiment_cfg/final_processor_config.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
checkpoint-5500/model-00001-of-00002.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7f7d446eecb2b7d9b3317a9ead36731958425b2a496ed63f201c1f2b8b3b0def
|
| 3 |
+
size 4990519232
|
checkpoint-5500/model-00002-of-00002.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:942563f16e6951b34e4dc82b1c7a1d98fd17562c50e94dc79cfaccb3119c02fb
|
| 3 |
+
size 1919980184
|
checkpoint-5500/model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
checkpoint-5500/processor_config.json
ADDED
|
@@ -0,0 +1,1124 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"processor_class": "Gr00tN1d7Processor",
|
| 3 |
+
"processor_kwargs": {
|
| 4 |
+
"modality_configs": {
|
| 5 |
+
"real_g1_relative_eef_relative_joints": {
|
| 6 |
+
"video": {
|
| 7 |
+
"delta_indices": [
|
| 8 |
+
-20,
|
| 9 |
+
0
|
| 10 |
+
],
|
| 11 |
+
"modality_keys": [
|
| 12 |
+
"ego_view"
|
| 13 |
+
],
|
| 14 |
+
"sin_cos_embedding_keys": null,
|
| 15 |
+
"mean_std_embedding_keys": null,
|
| 16 |
+
"action_configs": null
|
| 17 |
+
},
|
| 18 |
+
"state": {
|
| 19 |
+
"delta_indices": [
|
| 20 |
+
0
|
| 21 |
+
],
|
| 22 |
+
"modality_keys": [
|
| 23 |
+
"left_wrist_eef_9d",
|
| 24 |
+
"right_wrist_eef_9d",
|
| 25 |
+
"left_hand",
|
| 26 |
+
"right_hand",
|
| 27 |
+
"left_arm",
|
| 28 |
+
"right_arm",
|
| 29 |
+
"waist"
|
| 30 |
+
],
|
| 31 |
+
"sin_cos_embedding_keys": null,
|
| 32 |
+
"mean_std_embedding_keys": null,
|
| 33 |
+
"action_configs": null
|
| 34 |
+
},
|
| 35 |
+
"action": {
|
| 36 |
+
"delta_indices": [
|
| 37 |
+
0,
|
| 38 |
+
1,
|
| 39 |
+
2,
|
| 40 |
+
3,
|
| 41 |
+
4,
|
| 42 |
+
5,
|
| 43 |
+
6,
|
| 44 |
+
7,
|
| 45 |
+
8,
|
| 46 |
+
9,
|
| 47 |
+
10,
|
| 48 |
+
11,
|
| 49 |
+
12,
|
| 50 |
+
13,
|
| 51 |
+
14,
|
| 52 |
+
15,
|
| 53 |
+
16,
|
| 54 |
+
17,
|
| 55 |
+
18,
|
| 56 |
+
19,
|
| 57 |
+
20,
|
| 58 |
+
21,
|
| 59 |
+
22,
|
| 60 |
+
23,
|
| 61 |
+
24,
|
| 62 |
+
25,
|
| 63 |
+
26,
|
| 64 |
+
27,
|
| 65 |
+
28,
|
| 66 |
+
29,
|
| 67 |
+
30,
|
| 68 |
+
31,
|
| 69 |
+
32,
|
| 70 |
+
33,
|
| 71 |
+
34,
|
| 72 |
+
35,
|
| 73 |
+
36,
|
| 74 |
+
37,
|
| 75 |
+
38,
|
| 76 |
+
39
|
| 77 |
+
],
|
| 78 |
+
"modality_keys": [
|
| 79 |
+
"left_wrist_eef_9d",
|
| 80 |
+
"right_wrist_eef_9d",
|
| 81 |
+
"left_hand",
|
| 82 |
+
"right_hand",
|
| 83 |
+
"left_arm",
|
| 84 |
+
"right_arm",
|
| 85 |
+
"waist",
|
| 86 |
+
"base_height_command",
|
| 87 |
+
"navigate_command"
|
| 88 |
+
],
|
| 89 |
+
"sin_cos_embedding_keys": null,
|
| 90 |
+
"mean_std_embedding_keys": null,
|
| 91 |
+
"action_configs": [
|
| 92 |
+
{
|
| 93 |
+
"rep": "RELATIVE",
|
| 94 |
+
"type": "EEF",
|
| 95 |
+
"format": "XYZ_ROT6D",
|
| 96 |
+
"state_key": "left_wrist_eef_9d"
|
| 97 |
+
},
|
| 98 |
+
{
|
| 99 |
+
"rep": "RELATIVE",
|
| 100 |
+
"type": "EEF",
|
| 101 |
+
"format": "XYZ_ROT6D",
|
| 102 |
+
"state_key": "right_wrist_eef_9d"
|
| 103 |
+
},
|
| 104 |
+
{
|
| 105 |
+
"rep": "ABSOLUTE",
|
| 106 |
+
"type": "NON_EEF",
|
| 107 |
+
"format": "DEFAULT",
|
| 108 |
+
"state_key": "left_hand"
|
| 109 |
+
},
|
| 110 |
+
{
|
| 111 |
+
"rep": "ABSOLUTE",
|
| 112 |
+
"type": "NON_EEF",
|
| 113 |
+
"format": "DEFAULT",
|
| 114 |
+
"state_key": "right_hand"
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"rep": "RELATIVE",
|
| 118 |
+
"type": "NON_EEF",
|
| 119 |
+
"format": "DEFAULT",
|
| 120 |
+
"state_key": "left_arm"
|
| 121 |
+
},
|
| 122 |
+
{
|
| 123 |
+
"rep": "RELATIVE",
|
| 124 |
+
"type": "NON_EEF",
|
| 125 |
+
"format": "DEFAULT",
|
| 126 |
+
"state_key": "right_arm"
|
| 127 |
+
},
|
| 128 |
+
{
|
| 129 |
+
"rep": "ABSOLUTE",
|
| 130 |
+
"type": "NON_EEF",
|
| 131 |
+
"format": "DEFAULT",
|
| 132 |
+
"state_key": "waist"
|
| 133 |
+
},
|
| 134 |
+
{
|
| 135 |
+
"rep": "ABSOLUTE",
|
| 136 |
+
"type": "NON_EEF",
|
| 137 |
+
"format": "DEFAULT",
|
| 138 |
+
"state_key": "base_height_command"
|
| 139 |
+
},
|
| 140 |
+
{
|
| 141 |
+
"rep": "ABSOLUTE",
|
| 142 |
+
"type": "NON_EEF",
|
| 143 |
+
"format": "DEFAULT",
|
| 144 |
+
"state_key": "navigate_command"
|
| 145 |
+
}
|
| 146 |
+
]
|
| 147 |
+
},
|
| 148 |
+
"language": {
|
| 149 |
+
"delta_indices": [
|
| 150 |
+
0
|
| 151 |
+
],
|
| 152 |
+
"modality_keys": [
|
| 153 |
+
"annotation.human.task_description"
|
| 154 |
+
],
|
| 155 |
+
"sin_cos_embedding_keys": null,
|
| 156 |
+
"mean_std_embedding_keys": null,
|
| 157 |
+
"action_configs": null
|
| 158 |
+
}
|
| 159 |
+
},
|
| 160 |
+
"real_r1_pro_sharpa_relative_eef_mecka": {
|
| 161 |
+
"video": {
|
| 162 |
+
"delta_indices": [
|
| 163 |
+
-30,
|
| 164 |
+
0
|
| 165 |
+
],
|
| 166 |
+
"modality_keys": [
|
| 167 |
+
"ego_view_cropratio_res320x240_freq30"
|
| 168 |
+
],
|
| 169 |
+
"sin_cos_embedding_keys": null,
|
| 170 |
+
"mean_std_embedding_keys": null,
|
| 171 |
+
"action_configs": null
|
| 172 |
+
},
|
| 173 |
+
"state": {
|
| 174 |
+
"delta_indices": [
|
| 175 |
+
0
|
| 176 |
+
],
|
| 177 |
+
"modality_keys": [
|
| 178 |
+
"left_wrist_eef",
|
| 179 |
+
"right_wrist_eef",
|
| 180 |
+
"left_hand_joints",
|
| 181 |
+
"right_hand_joints"
|
| 182 |
+
],
|
| 183 |
+
"sin_cos_embedding_keys": null,
|
| 184 |
+
"mean_std_embedding_keys": null,
|
| 185 |
+
"action_configs": null
|
| 186 |
+
},
|
| 187 |
+
"action": {
|
| 188 |
+
"delta_indices": [
|
| 189 |
+
0,
|
| 190 |
+
1,
|
| 191 |
+
2,
|
| 192 |
+
3,
|
| 193 |
+
4,
|
| 194 |
+
5,
|
| 195 |
+
6,
|
| 196 |
+
7,
|
| 197 |
+
8,
|
| 198 |
+
9,
|
| 199 |
+
10,
|
| 200 |
+
11,
|
| 201 |
+
12,
|
| 202 |
+
13,
|
| 203 |
+
14,
|
| 204 |
+
15,
|
| 205 |
+
16,
|
| 206 |
+
17,
|
| 207 |
+
18,
|
| 208 |
+
19,
|
| 209 |
+
20,
|
| 210 |
+
21,
|
| 211 |
+
22,
|
| 212 |
+
23,
|
| 213 |
+
24,
|
| 214 |
+
25,
|
| 215 |
+
26,
|
| 216 |
+
27,
|
| 217 |
+
28,
|
| 218 |
+
29,
|
| 219 |
+
30,
|
| 220 |
+
31,
|
| 221 |
+
32,
|
| 222 |
+
33,
|
| 223 |
+
34,
|
| 224 |
+
35,
|
| 225 |
+
36,
|
| 226 |
+
37,
|
| 227 |
+
38,
|
| 228 |
+
39
|
| 229 |
+
],
|
| 230 |
+
"modality_keys": [
|
| 231 |
+
"left_wrist_eef",
|
| 232 |
+
"right_wrist_eef",
|
| 233 |
+
"left_hand_joints",
|
| 234 |
+
"right_hand_joints"
|
| 235 |
+
],
|
| 236 |
+
"sin_cos_embedding_keys": null,
|
| 237 |
+
"mean_std_embedding_keys": null,
|
| 238 |
+
"action_configs": [
|
| 239 |
+
{
|
| 240 |
+
"rep": "RELATIVE",
|
| 241 |
+
"type": "EEF",
|
| 242 |
+
"format": "XYZ_ROT6D",
|
| 243 |
+
"state_key": "left_wrist_eef"
|
| 244 |
+
},
|
| 245 |
+
{
|
| 246 |
+
"rep": "RELATIVE",
|
| 247 |
+
"type": "EEF",
|
| 248 |
+
"format": "XYZ_ROT6D",
|
| 249 |
+
"state_key": "right_wrist_eef"
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"rep": "ABSOLUTE",
|
| 253 |
+
"type": "NON_EEF",
|
| 254 |
+
"format": "DEFAULT",
|
| 255 |
+
"state_key": "left_hand_joints"
|
| 256 |
+
},
|
| 257 |
+
{
|
| 258 |
+
"rep": "ABSOLUTE",
|
| 259 |
+
"type": "NON_EEF",
|
| 260 |
+
"format": "DEFAULT",
|
| 261 |
+
"state_key": "right_hand_joints"
|
| 262 |
+
}
|
| 263 |
+
]
|
| 264 |
+
},
|
| 265 |
+
"language": {
|
| 266 |
+
"delta_indices": [
|
| 267 |
+
0
|
| 268 |
+
],
|
| 269 |
+
"modality_keys": [
|
| 270 |
+
"annotation.human.coarse_action"
|
| 271 |
+
],
|
| 272 |
+
"sin_cos_embedding_keys": null,
|
| 273 |
+
"mean_std_embedding_keys": null,
|
| 274 |
+
"action_configs": null
|
| 275 |
+
}
|
| 276 |
+
},
|
| 277 |
+
"real_r1_pro_sharpa_relative_eef_human": {
|
| 278 |
+
"video": {
|
| 279 |
+
"delta_indices": [
|
| 280 |
+
-20,
|
| 281 |
+
0
|
| 282 |
+
],
|
| 283 |
+
"modality_keys": [
|
| 284 |
+
"ego_view_res320x240_freq20",
|
| 285 |
+
"left_wrist_view_res320x240_freq20",
|
| 286 |
+
"right_wrist_view_res320x240_freq20"
|
| 287 |
+
],
|
| 288 |
+
"sin_cos_embedding_keys": null,
|
| 289 |
+
"mean_std_embedding_keys": null,
|
| 290 |
+
"action_configs": null
|
| 291 |
+
},
|
| 292 |
+
"state": {
|
| 293 |
+
"delta_indices": [
|
| 294 |
+
0
|
| 295 |
+
],
|
| 296 |
+
"modality_keys": [
|
| 297 |
+
"left_wrist_eef",
|
| 298 |
+
"right_wrist_eef",
|
| 299 |
+
"left_hand_joints",
|
| 300 |
+
"right_hand_joints"
|
| 301 |
+
],
|
| 302 |
+
"sin_cos_embedding_keys": null,
|
| 303 |
+
"mean_std_embedding_keys": null,
|
| 304 |
+
"action_configs": null
|
| 305 |
+
},
|
| 306 |
+
"action": {
|
| 307 |
+
"delta_indices": [
|
| 308 |
+
0,
|
| 309 |
+
1,
|
| 310 |
+
2,
|
| 311 |
+
3,
|
| 312 |
+
4,
|
| 313 |
+
5,
|
| 314 |
+
6,
|
| 315 |
+
7,
|
| 316 |
+
8,
|
| 317 |
+
9,
|
| 318 |
+
10,
|
| 319 |
+
11,
|
| 320 |
+
12,
|
| 321 |
+
13,
|
| 322 |
+
14,
|
| 323 |
+
15,
|
| 324 |
+
16,
|
| 325 |
+
17,
|
| 326 |
+
18,
|
| 327 |
+
19,
|
| 328 |
+
20,
|
| 329 |
+
21,
|
| 330 |
+
22,
|
| 331 |
+
23,
|
| 332 |
+
24,
|
| 333 |
+
25,
|
| 334 |
+
26,
|
| 335 |
+
27,
|
| 336 |
+
28,
|
| 337 |
+
29,
|
| 338 |
+
30,
|
| 339 |
+
31,
|
| 340 |
+
32,
|
| 341 |
+
33,
|
| 342 |
+
34,
|
| 343 |
+
35,
|
| 344 |
+
36,
|
| 345 |
+
37,
|
| 346 |
+
38,
|
| 347 |
+
39
|
| 348 |
+
],
|
| 349 |
+
"modality_keys": [
|
| 350 |
+
"left_wrist_eef",
|
| 351 |
+
"right_wrist_eef",
|
| 352 |
+
"left_hand_joints",
|
| 353 |
+
"right_hand_joints"
|
| 354 |
+
],
|
| 355 |
+
"sin_cos_embedding_keys": null,
|
| 356 |
+
"mean_std_embedding_keys": null,
|
| 357 |
+
"action_configs": [
|
| 358 |
+
{
|
| 359 |
+
"rep": "RELATIVE",
|
| 360 |
+
"type": "EEF",
|
| 361 |
+
"format": "XYZ_ROT6D",
|
| 362 |
+
"state_key": "left_wrist_eef"
|
| 363 |
+
},
|
| 364 |
+
{
|
| 365 |
+
"rep": "RELATIVE",
|
| 366 |
+
"type": "EEF",
|
| 367 |
+
"format": "XYZ_ROT6D",
|
| 368 |
+
"state_key": "right_wrist_eef"
|
| 369 |
+
},
|
| 370 |
+
{
|
| 371 |
+
"rep": "ABSOLUTE",
|
| 372 |
+
"type": "NON_EEF",
|
| 373 |
+
"format": "DEFAULT",
|
| 374 |
+
"state_key": "left_hand_joints"
|
| 375 |
+
},
|
| 376 |
+
{
|
| 377 |
+
"rep": "ABSOLUTE",
|
| 378 |
+
"type": "NON_EEF",
|
| 379 |
+
"format": "DEFAULT",
|
| 380 |
+
"state_key": "right_hand_joints"
|
| 381 |
+
}
|
| 382 |
+
]
|
| 383 |
+
},
|
| 384 |
+
"language": {
|
| 385 |
+
"delta_indices": [
|
| 386 |
+
0
|
| 387 |
+
],
|
| 388 |
+
"modality_keys": [
|
| 389 |
+
"annotation.human.coarse_action"
|
| 390 |
+
],
|
| 391 |
+
"sin_cos_embedding_keys": null,
|
| 392 |
+
"mean_std_embedding_keys": null,
|
| 393 |
+
"action_configs": null
|
| 394 |
+
}
|
| 395 |
+
},
|
| 396 |
+
"real_r1_pro_sharpa_relative_eef": {
|
| 397 |
+
"video": {
|
| 398 |
+
"delta_indices": [
|
| 399 |
+
-20,
|
| 400 |
+
0
|
| 401 |
+
],
|
| 402 |
+
"modality_keys": [
|
| 403 |
+
"ego_view_res320x240_freq20",
|
| 404 |
+
"left_wrist_view_res320x240_freq20",
|
| 405 |
+
"right_wrist_view_res320x240_freq20"
|
| 406 |
+
],
|
| 407 |
+
"sin_cos_embedding_keys": null,
|
| 408 |
+
"mean_std_embedding_keys": null,
|
| 409 |
+
"action_configs": null
|
| 410 |
+
},
|
| 411 |
+
"state": {
|
| 412 |
+
"delta_indices": [
|
| 413 |
+
0
|
| 414 |
+
],
|
| 415 |
+
"modality_keys": [
|
| 416 |
+
"left_wrist_eef",
|
| 417 |
+
"right_wrist_eef",
|
| 418 |
+
"left_hand_joints",
|
| 419 |
+
"right_hand_joints"
|
| 420 |
+
],
|
| 421 |
+
"sin_cos_embedding_keys": null,
|
| 422 |
+
"mean_std_embedding_keys": null,
|
| 423 |
+
"action_configs": null
|
| 424 |
+
},
|
| 425 |
+
"action": {
|
| 426 |
+
"delta_indices": [
|
| 427 |
+
0,
|
| 428 |
+
1,
|
| 429 |
+
2,
|
| 430 |
+
3,
|
| 431 |
+
4,
|
| 432 |
+
5,
|
| 433 |
+
6,
|
| 434 |
+
7,
|
| 435 |
+
8,
|
| 436 |
+
9,
|
| 437 |
+
10,
|
| 438 |
+
11,
|
| 439 |
+
12,
|
| 440 |
+
13,
|
| 441 |
+
14,
|
| 442 |
+
15,
|
| 443 |
+
16,
|
| 444 |
+
17,
|
| 445 |
+
18,
|
| 446 |
+
19,
|
| 447 |
+
20,
|
| 448 |
+
21,
|
| 449 |
+
22,
|
| 450 |
+
23,
|
| 451 |
+
24,
|
| 452 |
+
25,
|
| 453 |
+
26,
|
| 454 |
+
27,
|
| 455 |
+
28,
|
| 456 |
+
29,
|
| 457 |
+
30,
|
| 458 |
+
31,
|
| 459 |
+
32,
|
| 460 |
+
33,
|
| 461 |
+
34,
|
| 462 |
+
35,
|
| 463 |
+
36,
|
| 464 |
+
37,
|
| 465 |
+
38,
|
| 466 |
+
39
|
| 467 |
+
],
|
| 468 |
+
"modality_keys": [
|
| 469 |
+
"left_wrist_eef",
|
| 470 |
+
"right_wrist_eef",
|
| 471 |
+
"left_hand_joints",
|
| 472 |
+
"right_hand_joints"
|
| 473 |
+
],
|
| 474 |
+
"sin_cos_embedding_keys": null,
|
| 475 |
+
"mean_std_embedding_keys": null,
|
| 476 |
+
"action_configs": [
|
| 477 |
+
{
|
| 478 |
+
"rep": "RELATIVE",
|
| 479 |
+
"type": "EEF",
|
| 480 |
+
"format": "XYZ_ROT6D",
|
| 481 |
+
"state_key": "left_wrist_eef"
|
| 482 |
+
},
|
| 483 |
+
{
|
| 484 |
+
"rep": "RELATIVE",
|
| 485 |
+
"type": "EEF",
|
| 486 |
+
"format": "XYZ_ROT6D",
|
| 487 |
+
"state_key": "right_wrist_eef"
|
| 488 |
+
},
|
| 489 |
+
{
|
| 490 |
+
"rep": "ABSOLUTE",
|
| 491 |
+
"type": "NON_EEF",
|
| 492 |
+
"format": "DEFAULT",
|
| 493 |
+
"state_key": "left_hand_joints"
|
| 494 |
+
},
|
| 495 |
+
{
|
| 496 |
+
"rep": "ABSOLUTE",
|
| 497 |
+
"type": "NON_EEF",
|
| 498 |
+
"format": "DEFAULT",
|
| 499 |
+
"state_key": "right_hand_joints"
|
| 500 |
+
}
|
| 501 |
+
]
|
| 502 |
+
},
|
| 503 |
+
"language": {
|
| 504 |
+
"delta_indices": [
|
| 505 |
+
0
|
| 506 |
+
],
|
| 507 |
+
"modality_keys": [
|
| 508 |
+
"annotation.human.coarse_action"
|
| 509 |
+
],
|
| 510 |
+
"sin_cos_embedding_keys": null,
|
| 511 |
+
"mean_std_embedding_keys": null,
|
| 512 |
+
"action_configs": null
|
| 513 |
+
}
|
| 514 |
+
},
|
| 515 |
+
"xdof_relative_eef_relative_joint": {
|
| 516 |
+
"video": {
|
| 517 |
+
"delta_indices": [
|
| 518 |
+
-30,
|
| 519 |
+
0
|
| 520 |
+
],
|
| 521 |
+
"modality_keys": [
|
| 522 |
+
"top_camera-images-rgb_320_240",
|
| 523 |
+
"left_camera-images-rgb_320_240",
|
| 524 |
+
"right_camera-images-rgb_320_240"
|
| 525 |
+
],
|
| 526 |
+
"sin_cos_embedding_keys": null,
|
| 527 |
+
"mean_std_embedding_keys": null,
|
| 528 |
+
"action_configs": null
|
| 529 |
+
},
|
| 530 |
+
"state": {
|
| 531 |
+
"delta_indices": [
|
| 532 |
+
0
|
| 533 |
+
],
|
| 534 |
+
"modality_keys": [
|
| 535 |
+
"left_wrist_eef",
|
| 536 |
+
"right_wrist_eef",
|
| 537 |
+
"left_gripper_pos",
|
| 538 |
+
"right_gripper_pos",
|
| 539 |
+
"left_joint_pos",
|
| 540 |
+
"right_joint_pos"
|
| 541 |
+
],
|
| 542 |
+
"sin_cos_embedding_keys": null,
|
| 543 |
+
"mean_std_embedding_keys": null,
|
| 544 |
+
"action_configs": null
|
| 545 |
+
},
|
| 546 |
+
"action": {
|
| 547 |
+
"delta_indices": [
|
| 548 |
+
0,
|
| 549 |
+
1,
|
| 550 |
+
2,
|
| 551 |
+
3,
|
| 552 |
+
4,
|
| 553 |
+
5,
|
| 554 |
+
6,
|
| 555 |
+
7,
|
| 556 |
+
8,
|
| 557 |
+
9,
|
| 558 |
+
10,
|
| 559 |
+
11,
|
| 560 |
+
12,
|
| 561 |
+
13,
|
| 562 |
+
14,
|
| 563 |
+
15,
|
| 564 |
+
16,
|
| 565 |
+
17,
|
| 566 |
+
18,
|
| 567 |
+
19,
|
| 568 |
+
20,
|
| 569 |
+
21,
|
| 570 |
+
22,
|
| 571 |
+
23,
|
| 572 |
+
24,
|
| 573 |
+
25,
|
| 574 |
+
26,
|
| 575 |
+
27,
|
| 576 |
+
28,
|
| 577 |
+
29,
|
| 578 |
+
30,
|
| 579 |
+
31,
|
| 580 |
+
32,
|
| 581 |
+
33,
|
| 582 |
+
34,
|
| 583 |
+
35,
|
| 584 |
+
36,
|
| 585 |
+
37,
|
| 586 |
+
38,
|
| 587 |
+
39
|
| 588 |
+
],
|
| 589 |
+
"modality_keys": [
|
| 590 |
+
"left_wrist_eef",
|
| 591 |
+
"right_wrist_eef",
|
| 592 |
+
"left_gripper_pos",
|
| 593 |
+
"right_gripper_pos",
|
| 594 |
+
"left_joint_pos",
|
| 595 |
+
"right_joint_pos"
|
| 596 |
+
],
|
| 597 |
+
"sin_cos_embedding_keys": null,
|
| 598 |
+
"mean_std_embedding_keys": null,
|
| 599 |
+
"action_configs": [
|
| 600 |
+
{
|
| 601 |
+
"rep": "RELATIVE",
|
| 602 |
+
"type": "EEF",
|
| 603 |
+
"format": "XYZ_ROT6D",
|
| 604 |
+
"state_key": "left_wrist_eef"
|
| 605 |
+
},
|
| 606 |
+
{
|
| 607 |
+
"rep": "RELATIVE",
|
| 608 |
+
"type": "EEF",
|
| 609 |
+
"format": "XYZ_ROT6D",
|
| 610 |
+
"state_key": "right_wrist_eef"
|
| 611 |
+
},
|
| 612 |
+
{
|
| 613 |
+
"rep": "ABSOLUTE",
|
| 614 |
+
"type": "NON_EEF",
|
| 615 |
+
"format": "DEFAULT",
|
| 616 |
+
"state_key": "left_gripper_pos"
|
| 617 |
+
},
|
| 618 |
+
{
|
| 619 |
+
"rep": "ABSOLUTE",
|
| 620 |
+
"type": "NON_EEF",
|
| 621 |
+
"format": "DEFAULT",
|
| 622 |
+
"state_key": "right_gripper_pos"
|
| 623 |
+
},
|
| 624 |
+
{
|
| 625 |
+
"rep": "RELATIVE",
|
| 626 |
+
"type": "NON_EEF",
|
| 627 |
+
"format": "DEFAULT",
|
| 628 |
+
"state_key": "left_joint_pos"
|
| 629 |
+
},
|
| 630 |
+
{
|
| 631 |
+
"rep": "RELATIVE",
|
| 632 |
+
"type": "NON_EEF",
|
| 633 |
+
"format": "DEFAULT",
|
| 634 |
+
"state_key": "right_joint_pos"
|
| 635 |
+
}
|
| 636 |
+
]
|
| 637 |
+
},
|
| 638 |
+
"language": {
|
| 639 |
+
"delta_indices": [
|
| 640 |
+
0
|
| 641 |
+
],
|
| 642 |
+
"modality_keys": [
|
| 643 |
+
"annotation.task"
|
| 644 |
+
],
|
| 645 |
+
"sin_cos_embedding_keys": null,
|
| 646 |
+
"mean_std_embedding_keys": null,
|
| 647 |
+
"action_configs": null
|
| 648 |
+
}
|
| 649 |
+
},
|
| 650 |
+
"real_r1_pro_sharpa_relative_eef_maxinsights": {
|
| 651 |
+
"video": {
|
| 652 |
+
"delta_indices": [
|
| 653 |
+
-30,
|
| 654 |
+
0
|
| 655 |
+
],
|
| 656 |
+
"modality_keys": [
|
| 657 |
+
"ego_view_cropratio_res320x240_freq30"
|
| 658 |
+
],
|
| 659 |
+
"sin_cos_embedding_keys": null,
|
| 660 |
+
"mean_std_embedding_keys": null,
|
| 661 |
+
"action_configs": null
|
| 662 |
+
},
|
| 663 |
+
"state": {
|
| 664 |
+
"delta_indices": [
|
| 665 |
+
0
|
| 666 |
+
],
|
| 667 |
+
"modality_keys": [
|
| 668 |
+
"left_wrist_eef",
|
| 669 |
+
"right_wrist_eef",
|
| 670 |
+
"left_hand_joints",
|
| 671 |
+
"right_hand_joints"
|
| 672 |
+
],
|
| 673 |
+
"sin_cos_embedding_keys": null,
|
| 674 |
+
"mean_std_embedding_keys": null,
|
| 675 |
+
"action_configs": null
|
| 676 |
+
},
|
| 677 |
+
"action": {
|
| 678 |
+
"delta_indices": [
|
| 679 |
+
0,
|
| 680 |
+
1,
|
| 681 |
+
2,
|
| 682 |
+
3,
|
| 683 |
+
4,
|
| 684 |
+
5,
|
| 685 |
+
6,
|
| 686 |
+
7,
|
| 687 |
+
8,
|
| 688 |
+
9,
|
| 689 |
+
10,
|
| 690 |
+
11,
|
| 691 |
+
12,
|
| 692 |
+
13,
|
| 693 |
+
14,
|
| 694 |
+
15,
|
| 695 |
+
16,
|
| 696 |
+
17,
|
| 697 |
+
18,
|
| 698 |
+
19,
|
| 699 |
+
20,
|
| 700 |
+
21,
|
| 701 |
+
22,
|
| 702 |
+
23,
|
| 703 |
+
24,
|
| 704 |
+
25,
|
| 705 |
+
26,
|
| 706 |
+
27,
|
| 707 |
+
28,
|
| 708 |
+
29,
|
| 709 |
+
30,
|
| 710 |
+
31,
|
| 711 |
+
32,
|
| 712 |
+
33,
|
| 713 |
+
34,
|
| 714 |
+
35,
|
| 715 |
+
36,
|
| 716 |
+
37,
|
| 717 |
+
38,
|
| 718 |
+
39
|
| 719 |
+
],
|
| 720 |
+
"modality_keys": [
|
| 721 |
+
"left_wrist_eef",
|
| 722 |
+
"right_wrist_eef",
|
| 723 |
+
"left_hand_joints",
|
| 724 |
+
"right_hand_joints"
|
| 725 |
+
],
|
| 726 |
+
"sin_cos_embedding_keys": null,
|
| 727 |
+
"mean_std_embedding_keys": null,
|
| 728 |
+
"action_configs": [
|
| 729 |
+
{
|
| 730 |
+
"rep": "RELATIVE",
|
| 731 |
+
"type": "EEF",
|
| 732 |
+
"format": "XYZ_ROT6D",
|
| 733 |
+
"state_key": "left_wrist_eef"
|
| 734 |
+
},
|
| 735 |
+
{
|
| 736 |
+
"rep": "RELATIVE",
|
| 737 |
+
"type": "EEF",
|
| 738 |
+
"format": "XYZ_ROT6D",
|
| 739 |
+
"state_key": "right_wrist_eef"
|
| 740 |
+
},
|
| 741 |
+
{
|
| 742 |
+
"rep": "ABSOLUTE",
|
| 743 |
+
"type": "NON_EEF",
|
| 744 |
+
"format": "DEFAULT",
|
| 745 |
+
"state_key": "left_hand_joints"
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"rep": "ABSOLUTE",
|
| 749 |
+
"type": "NON_EEF",
|
| 750 |
+
"format": "DEFAULT",
|
| 751 |
+
"state_key": "right_hand_joints"
|
| 752 |
+
}
|
| 753 |
+
]
|
| 754 |
+
},
|
| 755 |
+
"language": {
|
| 756 |
+
"delta_indices": [
|
| 757 |
+
0
|
| 758 |
+
],
|
| 759 |
+
"modality_keys": [
|
| 760 |
+
"annotation.human.coarse_action"
|
| 761 |
+
],
|
| 762 |
+
"sin_cos_embedding_keys": null,
|
| 763 |
+
"mean_std_embedding_keys": null,
|
| 764 |
+
"action_configs": null
|
| 765 |
+
}
|
| 766 |
+
},
|
| 767 |
+
"xdof_relative_eef_relative_joint_subtask": {
|
| 768 |
+
"video": {
|
| 769 |
+
"delta_indices": [
|
| 770 |
+
-30,
|
| 771 |
+
0
|
| 772 |
+
],
|
| 773 |
+
"modality_keys": [
|
| 774 |
+
"top_camera-images-rgb_320_240",
|
| 775 |
+
"left_camera-images-rgb_320_240",
|
| 776 |
+
"right_camera-images-rgb_320_240"
|
| 777 |
+
],
|
| 778 |
+
"sin_cos_embedding_keys": null,
|
| 779 |
+
"mean_std_embedding_keys": null,
|
| 780 |
+
"action_configs": null
|
| 781 |
+
},
|
| 782 |
+
"state": {
|
| 783 |
+
"delta_indices": [
|
| 784 |
+
0
|
| 785 |
+
],
|
| 786 |
+
"modality_keys": [
|
| 787 |
+
"left_wrist_eef",
|
| 788 |
+
"right_wrist_eef",
|
| 789 |
+
"left_gripper_pos",
|
| 790 |
+
"right_gripper_pos",
|
| 791 |
+
"left_joint_pos",
|
| 792 |
+
"right_joint_pos"
|
| 793 |
+
],
|
| 794 |
+
"sin_cos_embedding_keys": null,
|
| 795 |
+
"mean_std_embedding_keys": null,
|
| 796 |
+
"action_configs": null
|
| 797 |
+
},
|
| 798 |
+
"action": {
|
| 799 |
+
"delta_indices": [
|
| 800 |
+
0,
|
| 801 |
+
1,
|
| 802 |
+
2,
|
| 803 |
+
3,
|
| 804 |
+
4,
|
| 805 |
+
5,
|
| 806 |
+
6,
|
| 807 |
+
7,
|
| 808 |
+
8,
|
| 809 |
+
9,
|
| 810 |
+
10,
|
| 811 |
+
11,
|
| 812 |
+
12,
|
| 813 |
+
13,
|
| 814 |
+
14,
|
| 815 |
+
15,
|
| 816 |
+
16,
|
| 817 |
+
17,
|
| 818 |
+
18,
|
| 819 |
+
19,
|
| 820 |
+
20,
|
| 821 |
+
21,
|
| 822 |
+
22,
|
| 823 |
+
23,
|
| 824 |
+
24,
|
| 825 |
+
25,
|
| 826 |
+
26,
|
| 827 |
+
27,
|
| 828 |
+
28,
|
| 829 |
+
29,
|
| 830 |
+
30,
|
| 831 |
+
31,
|
| 832 |
+
32,
|
| 833 |
+
33,
|
| 834 |
+
34,
|
| 835 |
+
35,
|
| 836 |
+
36,
|
| 837 |
+
37,
|
| 838 |
+
38,
|
| 839 |
+
39
|
| 840 |
+
],
|
| 841 |
+
"modality_keys": [
|
| 842 |
+
"left_wrist_eef",
|
| 843 |
+
"right_wrist_eef",
|
| 844 |
+
"left_gripper_pos",
|
| 845 |
+
"right_gripper_pos",
|
| 846 |
+
"left_joint_pos",
|
| 847 |
+
"right_joint_pos"
|
| 848 |
+
],
|
| 849 |
+
"sin_cos_embedding_keys": null,
|
| 850 |
+
"mean_std_embedding_keys": null,
|
| 851 |
+
"action_configs": [
|
| 852 |
+
{
|
| 853 |
+
"rep": "RELATIVE",
|
| 854 |
+
"type": "EEF",
|
| 855 |
+
"format": "XYZ_ROT6D",
|
| 856 |
+
"state_key": "left_wrist_eef"
|
| 857 |
+
},
|
| 858 |
+
{
|
| 859 |
+
"rep": "RELATIVE",
|
| 860 |
+
"type": "EEF",
|
| 861 |
+
"format": "XYZ_ROT6D",
|
| 862 |
+
"state_key": "right_wrist_eef"
|
| 863 |
+
},
|
| 864 |
+
{
|
| 865 |
+
"rep": "ABSOLUTE",
|
| 866 |
+
"type": "NON_EEF",
|
| 867 |
+
"format": "DEFAULT",
|
| 868 |
+
"state_key": "left_gripper_pos"
|
| 869 |
+
},
|
| 870 |
+
{
|
| 871 |
+
"rep": "ABSOLUTE",
|
| 872 |
+
"type": "NON_EEF",
|
| 873 |
+
"format": "DEFAULT",
|
| 874 |
+
"state_key": "right_gripper_pos"
|
| 875 |
+
},
|
| 876 |
+
{
|
| 877 |
+
"rep": "RELATIVE",
|
| 878 |
+
"type": "NON_EEF",
|
| 879 |
+
"format": "DEFAULT",
|
| 880 |
+
"state_key": "left_joint_pos"
|
| 881 |
+
},
|
| 882 |
+
{
|
| 883 |
+
"rep": "RELATIVE",
|
| 884 |
+
"type": "NON_EEF",
|
| 885 |
+
"format": "DEFAULT",
|
| 886 |
+
"state_key": "right_joint_pos"
|
| 887 |
+
}
|
| 888 |
+
]
|
| 889 |
+
},
|
| 890 |
+
"language": {
|
| 891 |
+
"delta_indices": [
|
| 892 |
+
0
|
| 893 |
+
],
|
| 894 |
+
"modality_keys": [
|
| 895 |
+
"annotation.sub_task"
|
| 896 |
+
],
|
| 897 |
+
"sin_cos_embedding_keys": null,
|
| 898 |
+
"mean_std_embedding_keys": null,
|
| 899 |
+
"action_configs": null
|
| 900 |
+
}
|
| 901 |
+
},
|
| 902 |
+
"oxe_droid_relative_eef_relative_joint": {
|
| 903 |
+
"video": {
|
| 904 |
+
"delta_indices": [
|
| 905 |
+
-15,
|
| 906 |
+
0
|
| 907 |
+
],
|
| 908 |
+
"modality_keys": [
|
| 909 |
+
"exterior_image_1_left",
|
| 910 |
+
"wrist_image_left"
|
| 911 |
+
],
|
| 912 |
+
"sin_cos_embedding_keys": null,
|
| 913 |
+
"mean_std_embedding_keys": null,
|
| 914 |
+
"action_configs": null
|
| 915 |
+
},
|
| 916 |
+
"state": {
|
| 917 |
+
"delta_indices": [
|
| 918 |
+
0
|
| 919 |
+
],
|
| 920 |
+
"modality_keys": [
|
| 921 |
+
"eef_9d",
|
| 922 |
+
"gripper_position",
|
| 923 |
+
"joint_position"
|
| 924 |
+
],
|
| 925 |
+
"sin_cos_embedding_keys": null,
|
| 926 |
+
"mean_std_embedding_keys": null,
|
| 927 |
+
"action_configs": null
|
| 928 |
+
},
|
| 929 |
+
"action": {
|
| 930 |
+
"delta_indices": [
|
| 931 |
+
0,
|
| 932 |
+
1,
|
| 933 |
+
2,
|
| 934 |
+
3,
|
| 935 |
+
4,
|
| 936 |
+
5,
|
| 937 |
+
6,
|
| 938 |
+
7,
|
| 939 |
+
8,
|
| 940 |
+
9,
|
| 941 |
+
10,
|
| 942 |
+
11,
|
| 943 |
+
12,
|
| 944 |
+
13,
|
| 945 |
+
14,
|
| 946 |
+
15,
|
| 947 |
+
16,
|
| 948 |
+
17,
|
| 949 |
+
18,
|
| 950 |
+
19,
|
| 951 |
+
20,
|
| 952 |
+
21,
|
| 953 |
+
22,
|
| 954 |
+
23,
|
| 955 |
+
24,
|
| 956 |
+
25,
|
| 957 |
+
26,
|
| 958 |
+
27,
|
| 959 |
+
28,
|
| 960 |
+
29,
|
| 961 |
+
30,
|
| 962 |
+
31,
|
| 963 |
+
32,
|
| 964 |
+
33,
|
| 965 |
+
34,
|
| 966 |
+
35,
|
| 967 |
+
36,
|
| 968 |
+
37,
|
| 969 |
+
38,
|
| 970 |
+
39
|
| 971 |
+
],
|
| 972 |
+
"modality_keys": [
|
| 973 |
+
"eef_9d",
|
| 974 |
+
"gripper_position",
|
| 975 |
+
"joint_position"
|
| 976 |
+
],
|
| 977 |
+
"sin_cos_embedding_keys": null,
|
| 978 |
+
"mean_std_embedding_keys": null,
|
| 979 |
+
"action_configs": [
|
| 980 |
+
{
|
| 981 |
+
"rep": "RELATIVE",
|
| 982 |
+
"type": "EEF",
|
| 983 |
+
"format": "XYZ_ROT6D",
|
| 984 |
+
"state_key": "eef_9d"
|
| 985 |
+
},
|
| 986 |
+
{
|
| 987 |
+
"rep": "ABSOLUTE",
|
| 988 |
+
"type": "NON_EEF",
|
| 989 |
+
"format": "DEFAULT",
|
| 990 |
+
"state_key": "gripper_position"
|
| 991 |
+
},
|
| 992 |
+
{
|
| 993 |
+
"rep": "RELATIVE",
|
| 994 |
+
"type": "NON_EEF",
|
| 995 |
+
"format": "DEFAULT",
|
| 996 |
+
"state_key": "joint_position"
|
| 997 |
+
}
|
| 998 |
+
]
|
| 999 |
+
},
|
| 1000 |
+
"language": {
|
| 1001 |
+
"delta_indices": [
|
| 1002 |
+
0
|
| 1003 |
+
],
|
| 1004 |
+
"modality_keys": [
|
| 1005 |
+
"annotation.language.language_instruction"
|
| 1006 |
+
],
|
| 1007 |
+
"sin_cos_embedding_keys": null,
|
| 1008 |
+
"mean_std_embedding_keys": null,
|
| 1009 |
+
"action_configs": null
|
| 1010 |
+
}
|
| 1011 |
+
},
|
| 1012 |
+
"new_embodiment": {
|
| 1013 |
+
"video": {
|
| 1014 |
+
"delta_indices": [
|
| 1015 |
+
0
|
| 1016 |
+
],
|
| 1017 |
+
"modality_keys": [
|
| 1018 |
+
"cam_left_head",
|
| 1019 |
+
"cam_left_wrist",
|
| 1020 |
+
"cam_right_wrist"
|
| 1021 |
+
],
|
| 1022 |
+
"sin_cos_embedding_keys": null,
|
| 1023 |
+
"mean_std_embedding_keys": null,
|
| 1024 |
+
"action_configs": null
|
| 1025 |
+
},
|
| 1026 |
+
"state": {
|
| 1027 |
+
"delta_indices": [
|
| 1028 |
+
0
|
| 1029 |
+
],
|
| 1030 |
+
"modality_keys": [
|
| 1031 |
+
"arm_left",
|
| 1032 |
+
"arm_right"
|
| 1033 |
+
],
|
| 1034 |
+
"sin_cos_embedding_keys": null,
|
| 1035 |
+
"mean_std_embedding_keys": null,
|
| 1036 |
+
"action_configs": null
|
| 1037 |
+
},
|
| 1038 |
+
"action": {
|
| 1039 |
+
"delta_indices": [
|
| 1040 |
+
0,
|
| 1041 |
+
1,
|
| 1042 |
+
2,
|
| 1043 |
+
3,
|
| 1044 |
+
4,
|
| 1045 |
+
5,
|
| 1046 |
+
6,
|
| 1047 |
+
7,
|
| 1048 |
+
8,
|
| 1049 |
+
9,
|
| 1050 |
+
10,
|
| 1051 |
+
11,
|
| 1052 |
+
12,
|
| 1053 |
+
13,
|
| 1054 |
+
14,
|
| 1055 |
+
15
|
| 1056 |
+
],
|
| 1057 |
+
"modality_keys": [
|
| 1058 |
+
"arm_left",
|
| 1059 |
+
"arm_right"
|
| 1060 |
+
],
|
| 1061 |
+
"sin_cos_embedding_keys": null,
|
| 1062 |
+
"mean_std_embedding_keys": null,
|
| 1063 |
+
"action_configs": [
|
| 1064 |
+
{
|
| 1065 |
+
"rep": "ABSOLUTE",
|
| 1066 |
+
"type": "NON_EEF",
|
| 1067 |
+
"format": "DEFAULT",
|
| 1068 |
+
"state_key": null
|
| 1069 |
+
},
|
| 1070 |
+
{
|
| 1071 |
+
"rep": "ABSOLUTE",
|
| 1072 |
+
"type": "NON_EEF",
|
| 1073 |
+
"format": "DEFAULT",
|
| 1074 |
+
"state_key": null
|
| 1075 |
+
}
|
| 1076 |
+
]
|
| 1077 |
+
},
|
| 1078 |
+
"language": {
|
| 1079 |
+
"delta_indices": [
|
| 1080 |
+
0
|
| 1081 |
+
],
|
| 1082 |
+
"modality_keys": [
|
| 1083 |
+
"annotation.human.primitive_instruction"
|
| 1084 |
+
],
|
| 1085 |
+
"sin_cos_embedding_keys": null,
|
| 1086 |
+
"mean_std_embedding_keys": null,
|
| 1087 |
+
"action_configs": null
|
| 1088 |
+
}
|
| 1089 |
+
}
|
| 1090 |
+
},
|
| 1091 |
+
"image_crop_size": [
|
| 1092 |
+
230,
|
| 1093 |
+
230
|
| 1094 |
+
],
|
| 1095 |
+
"image_target_size": [
|
| 1096 |
+
256,
|
| 1097 |
+
256
|
| 1098 |
+
],
|
| 1099 |
+
"use_albumentations": true,
|
| 1100 |
+
"random_rotation_angle": 0,
|
| 1101 |
+
"color_jitter_params": {
|
| 1102 |
+
"brightness": 0.2,
|
| 1103 |
+
"contrast": 0.2,
|
| 1104 |
+
"saturation": 0.2,
|
| 1105 |
+
"hue": 0.03
|
| 1106 |
+
},
|
| 1107 |
+
"shortest_image_edge": 256,
|
| 1108 |
+
"crop_fraction": 1.0,
|
| 1109 |
+
"letter_box_transform": false,
|
| 1110 |
+
"model_name": "nvidia/Cosmos-Reason2-2B",
|
| 1111 |
+
"model_type": "qwen",
|
| 1112 |
+
"formalize_language": true,
|
| 1113 |
+
"max_state_dim": 132,
|
| 1114 |
+
"max_action_dim": 132,
|
| 1115 |
+
"max_action_horizon": 40,
|
| 1116 |
+
"use_percentiles": false,
|
| 1117 |
+
"use_mean_std": false,
|
| 1118 |
+
"clip_outliers": true,
|
| 1119 |
+
"apply_sincos_state_encoding": false,
|
| 1120 |
+
"use_relative_action": true,
|
| 1121 |
+
"exclude_state": false,
|
| 1122 |
+
"state_dropout_prob": 0.2
|
| 1123 |
+
}
|
| 1124 |
+
}
|
checkpoint-5500/statistics.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
checkpoint-5500/trainer_state.json
ADDED
|
@@ -0,0 +1,3334 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": null,
|
| 3 |
+
"best_metric": null,
|
| 4 |
+
"best_model_checkpoint": null,
|
| 5 |
+
"epoch": 0.9166666666666666,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 5500,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"grad_norm": 1.1998956203460693,
|
| 14 |
+
"learning_rate": 1.5e-06,
|
| 15 |
+
"loss": 0.0948,
|
| 16 |
+
"step": 10
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"grad_norm": 1.149483561515808,
|
| 20 |
+
"learning_rate": 3.1666666666666667e-06,
|
| 21 |
+
"loss": 0.1133,
|
| 22 |
+
"step": 20
|
| 23 |
+
},
|
| 24 |
+
{
|
| 25 |
+
"grad_norm": 0.5785359144210815,
|
| 26 |
+
"learning_rate": 4.833333333333333e-06,
|
| 27 |
+
"loss": 0.0921,
|
| 28 |
+
"step": 30
|
| 29 |
+
},
|
| 30 |
+
{
|
| 31 |
+
"grad_norm": 0.5978452563285828,
|
| 32 |
+
"learning_rate": 6.5000000000000004e-06,
|
| 33 |
+
"loss": 0.0424,
|
| 34 |
+
"step": 40
|
| 35 |
+
},
|
| 36 |
+
{
|
| 37 |
+
"grad_norm": 0.5148229598999023,
|
| 38 |
+
"learning_rate": 8.166666666666666e-06,
|
| 39 |
+
"loss": 0.0314,
|
| 40 |
+
"step": 50
|
| 41 |
+
},
|
| 42 |
+
{
|
| 43 |
+
"grad_norm": 0.456398606300354,
|
| 44 |
+
"learning_rate": 9.833333333333333e-06,
|
| 45 |
+
"loss": 0.0311,
|
| 46 |
+
"step": 60
|
| 47 |
+
},
|
| 48 |
+
{
|
| 49 |
+
"grad_norm": 0.2702748775482178,
|
| 50 |
+
"learning_rate": 1.1500000000000002e-05,
|
| 51 |
+
"loss": 0.0249,
|
| 52 |
+
"step": 70
|
| 53 |
+
},
|
| 54 |
+
{
|
| 55 |
+
"grad_norm": 0.2341022789478302,
|
| 56 |
+
"learning_rate": 1.3166666666666667e-05,
|
| 57 |
+
"loss": 0.0208,
|
| 58 |
+
"step": 80
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"grad_norm": 0.24386702477931976,
|
| 62 |
+
"learning_rate": 1.4833333333333334e-05,
|
| 63 |
+
"loss": 0.0168,
|
| 64 |
+
"step": 90
|
| 65 |
+
},
|
| 66 |
+
{
|
| 67 |
+
"grad_norm": 0.15167136490345,
|
| 68 |
+
"learning_rate": 1.65e-05,
|
| 69 |
+
"loss": 0.0114,
|
| 70 |
+
"step": 100
|
| 71 |
+
},
|
| 72 |
+
{
|
| 73 |
+
"grad_norm": 0.12867538630962372,
|
| 74 |
+
"learning_rate": 1.8166666666666667e-05,
|
| 75 |
+
"loss": 0.015,
|
| 76 |
+
"step": 110
|
| 77 |
+
},
|
| 78 |
+
{
|
| 79 |
+
"grad_norm": 0.11734911054372787,
|
| 80 |
+
"learning_rate": 1.9833333333333332e-05,
|
| 81 |
+
"loss": 0.0108,
|
| 82 |
+
"step": 120
|
| 83 |
+
},
|
| 84 |
+
{
|
| 85 |
+
"grad_norm": 0.10739504545927048,
|
| 86 |
+
"learning_rate": 2.15e-05,
|
| 87 |
+
"loss": 0.0085,
|
| 88 |
+
"step": 130
|
| 89 |
+
},
|
| 90 |
+
{
|
| 91 |
+
"grad_norm": 0.11796178668737411,
|
| 92 |
+
"learning_rate": 2.316666666666667e-05,
|
| 93 |
+
"loss": 0.0108,
|
| 94 |
+
"step": 140
|
| 95 |
+
},
|
| 96 |
+
{
|
| 97 |
+
"grad_norm": 0.1394447386264801,
|
| 98 |
+
"learning_rate": 2.483333333333333e-05,
|
| 99 |
+
"loss": 0.0111,
|
| 100 |
+
"step": 150
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"grad_norm": 0.1239742562174797,
|
| 104 |
+
"learning_rate": 2.65e-05,
|
| 105 |
+
"loss": 0.01,
|
| 106 |
+
"step": 160
|
| 107 |
+
},
|
| 108 |
+
{
|
| 109 |
+
"grad_norm": 0.15108101069927216,
|
| 110 |
+
"learning_rate": 2.8166666666666666e-05,
|
| 111 |
+
"loss": 0.0088,
|
| 112 |
+
"step": 170
|
| 113 |
+
},
|
| 114 |
+
{
|
| 115 |
+
"grad_norm": 0.10625599324703217,
|
| 116 |
+
"learning_rate": 2.9833333333333335e-05,
|
| 117 |
+
"loss": 0.0073,
|
| 118 |
+
"step": 180
|
| 119 |
+
},
|
| 120 |
+
{
|
| 121 |
+
"grad_norm": 0.18963077664375305,
|
| 122 |
+
"learning_rate": 2.9999822989428527e-05,
|
| 123 |
+
"loss": 0.0086,
|
| 124 |
+
"step": 190
|
| 125 |
+
},
|
| 126 |
+
{
|
| 127 |
+
"grad_norm": 0.13748835027217865,
|
| 128 |
+
"learning_rate": 2.9999211106396838e-05,
|
| 129 |
+
"loss": 0.008,
|
| 130 |
+
"step": 200
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"grad_norm": 0.11299511790275574,
|
| 134 |
+
"learning_rate": 2.9998162183413746e-05,
|
| 135 |
+
"loss": 0.0084,
|
| 136 |
+
"step": 210
|
| 137 |
+
},
|
| 138 |
+
{
|
| 139 |
+
"grad_norm": 0.09556759148836136,
|
| 140 |
+
"learning_rate": 2.9996676251042288e-05,
|
| 141 |
+
"loss": 0.0075,
|
| 142 |
+
"step": 220
|
| 143 |
+
},
|
| 144 |
+
{
|
| 145 |
+
"grad_norm": 0.13729563355445862,
|
| 146 |
+
"learning_rate": 2.9994753352578854e-05,
|
| 147 |
+
"loss": 0.0078,
|
| 148 |
+
"step": 230
|
| 149 |
+
},
|
| 150 |
+
{
|
| 151 |
+
"grad_norm": 0.13592101633548737,
|
| 152 |
+
"learning_rate": 2.999239354405195e-05,
|
| 153 |
+
"loss": 0.0087,
|
| 154 |
+
"step": 240
|
| 155 |
+
},
|
| 156 |
+
{
|
| 157 |
+
"grad_norm": 0.13999280333518982,
|
| 158 |
+
"learning_rate": 2.9989596894220574e-05,
|
| 159 |
+
"loss": 0.0072,
|
| 160 |
+
"step": 250
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"grad_norm": 0.20841121673583984,
|
| 164 |
+
"learning_rate": 2.998636348457219e-05,
|
| 165 |
+
"loss": 0.0078,
|
| 166 |
+
"step": 260
|
| 167 |
+
},
|
| 168 |
+
{
|
| 169 |
+
"grad_norm": 0.1364365965127945,
|
| 170 |
+
"learning_rate": 2.998269340932036e-05,
|
| 171 |
+
"loss": 0.0086,
|
| 172 |
+
"step": 270
|
| 173 |
+
},
|
| 174 |
+
{
|
| 175 |
+
"grad_norm": 0.1565951108932495,
|
| 176 |
+
"learning_rate": 2.9978586775402017e-05,
|
| 177 |
+
"loss": 0.0066,
|
| 178 |
+
"step": 280
|
| 179 |
+
},
|
| 180 |
+
{
|
| 181 |
+
"grad_norm": 0.1358197182416916,
|
| 182 |
+
"learning_rate": 2.9974043702474322e-05,
|
| 183 |
+
"loss": 0.0057,
|
| 184 |
+
"step": 290
|
| 185 |
+
},
|
| 186 |
+
{
|
| 187 |
+
"grad_norm": 0.11862654238939285,
|
| 188 |
+
"learning_rate": 2.9969064322911198e-05,
|
| 189 |
+
"loss": 0.0074,
|
| 190 |
+
"step": 300
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"grad_norm": 0.13869383931159973,
|
| 194 |
+
"learning_rate": 2.9963648781799457e-05,
|
| 195 |
+
"loss": 0.0073,
|
| 196 |
+
"step": 310
|
| 197 |
+
},
|
| 198 |
+
{
|
| 199 |
+
"grad_norm": 0.10812313109636307,
|
| 200 |
+
"learning_rate": 2.9957797236934595e-05,
|
| 201 |
+
"loss": 0.0055,
|
| 202 |
+
"step": 320
|
| 203 |
+
},
|
| 204 |
+
{
|
| 205 |
+
"grad_norm": 0.10657292604446411,
|
| 206 |
+
"learning_rate": 2.995150985881616e-05,
|
| 207 |
+
"loss": 0.0066,
|
| 208 |
+
"step": 330
|
| 209 |
+
},
|
| 210 |
+
{
|
| 211 |
+
"grad_norm": 0.1428467333316803,
|
| 212 |
+
"learning_rate": 2.9944786830642822e-05,
|
| 213 |
+
"loss": 0.0075,
|
| 214 |
+
"step": 340
|
| 215 |
+
},
|
| 216 |
+
{
|
| 217 |
+
"grad_norm": 0.2037973701953888,
|
| 218 |
+
"learning_rate": 2.9937628348307015e-05,
|
| 219 |
+
"loss": 0.0068,
|
| 220 |
+
"step": 350
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"grad_norm": 0.17431488633155823,
|
| 224 |
+
"learning_rate": 2.9930034620389217e-05,
|
| 225 |
+
"loss": 0.0085,
|
| 226 |
+
"step": 360
|
| 227 |
+
},
|
| 228 |
+
{
|
| 229 |
+
"grad_norm": 0.11183494329452515,
|
| 230 |
+
"learning_rate": 2.9922005868151907e-05,
|
| 231 |
+
"loss": 0.0075,
|
| 232 |
+
"step": 370
|
| 233 |
+
},
|
| 234 |
+
{
|
| 235 |
+
"grad_norm": 0.15082184970378876,
|
| 236 |
+
"learning_rate": 2.991354232553308e-05,
|
| 237 |
+
"loss": 0.0062,
|
| 238 |
+
"step": 380
|
| 239 |
+
},
|
| 240 |
+
{
|
| 241 |
+
"grad_norm": 0.18662261962890625,
|
| 242 |
+
"learning_rate": 2.9904644239139453e-05,
|
| 243 |
+
"loss": 0.0065,
|
| 244 |
+
"step": 390
|
| 245 |
+
},
|
| 246 |
+
{
|
| 247 |
+
"grad_norm": 0.08828644454479218,
|
| 248 |
+
"learning_rate": 2.9895311868239293e-05,
|
| 249 |
+
"loss": 0.0063,
|
| 250 |
+
"step": 400
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"grad_norm": 0.09585582464933395,
|
| 254 |
+
"learning_rate": 2.988554548475481e-05,
|
| 255 |
+
"loss": 0.0085,
|
| 256 |
+
"step": 410
|
| 257 |
+
},
|
| 258 |
+
{
|
| 259 |
+
"grad_norm": 0.1154177337884903,
|
| 260 |
+
"learning_rate": 2.9875345373254302e-05,
|
| 261 |
+
"loss": 0.0062,
|
| 262 |
+
"step": 420
|
| 263 |
+
},
|
| 264 |
+
{
|
| 265 |
+
"grad_norm": 0.13848039507865906,
|
| 266 |
+
"learning_rate": 2.986471183094382e-05,
|
| 267 |
+
"loss": 0.0072,
|
| 268 |
+
"step": 430
|
| 269 |
+
},
|
| 270 |
+
{
|
| 271 |
+
"grad_norm": 0.11135748773813248,
|
| 272 |
+
"learning_rate": 2.9853645167658494e-05,
|
| 273 |
+
"loss": 0.0052,
|
| 274 |
+
"step": 440
|
| 275 |
+
},
|
| 276 |
+
{
|
| 277 |
+
"grad_norm": 0.15800616145133972,
|
| 278 |
+
"learning_rate": 2.9842145705853565e-05,
|
| 279 |
+
"loss": 0.0057,
|
| 280 |
+
"step": 450
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"grad_norm": 0.11519774794578552,
|
| 284 |
+
"learning_rate": 2.9830213780594925e-05,
|
| 285 |
+
"loss": 0.0041,
|
| 286 |
+
"step": 460
|
| 287 |
+
},
|
| 288 |
+
{
|
| 289 |
+
"grad_norm": 0.06121053546667099,
|
| 290 |
+
"learning_rate": 2.9817849739549408e-05,
|
| 291 |
+
"loss": 0.0045,
|
| 292 |
+
"step": 470
|
| 293 |
+
},
|
| 294 |
+
{
|
| 295 |
+
"grad_norm": 0.11012984067201614,
|
| 296 |
+
"learning_rate": 2.980505394297461e-05,
|
| 297 |
+
"loss": 0.0047,
|
| 298 |
+
"step": 480
|
| 299 |
+
},
|
| 300 |
+
{
|
| 301 |
+
"grad_norm": 0.12504155933856964,
|
| 302 |
+
"learning_rate": 2.9791826763708432e-05,
|
| 303 |
+
"loss": 0.0043,
|
| 304 |
+
"step": 490
|
| 305 |
+
},
|
| 306 |
+
{
|
| 307 |
+
"grad_norm": 0.10626716166734695,
|
| 308 |
+
"learning_rate": 2.9778168587158196e-05,
|
| 309 |
+
"loss": 0.005,
|
| 310 |
+
"step": 500
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"grad_norm": 0.100404754281044,
|
| 314 |
+
"learning_rate": 2.976407981128942e-05,
|
| 315 |
+
"loss": 0.0037,
|
| 316 |
+
"step": 510
|
| 317 |
+
},
|
| 318 |
+
{
|
| 319 |
+
"grad_norm": 0.11580510437488556,
|
| 320 |
+
"learning_rate": 2.9749560846614216e-05,
|
| 321 |
+
"loss": 0.0042,
|
| 322 |
+
"step": 520
|
| 323 |
+
},
|
| 324 |
+
{
|
| 325 |
+
"grad_norm": 0.1404998004436493,
|
| 326 |
+
"learning_rate": 2.973461211617935e-05,
|
| 327 |
+
"loss": 0.0055,
|
| 328 |
+
"step": 530
|
| 329 |
+
},
|
| 330 |
+
{
|
| 331 |
+
"grad_norm": 0.1066383644938469,
|
| 332 |
+
"learning_rate": 2.9719234055553884e-05,
|
| 333 |
+
"loss": 0.0055,
|
| 334 |
+
"step": 540
|
| 335 |
+
},
|
| 336 |
+
{
|
| 337 |
+
"grad_norm": 0.16468606889247894,
|
| 338 |
+
"learning_rate": 2.970342711281651e-05,
|
| 339 |
+
"loss": 0.0038,
|
| 340 |
+
"step": 550
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"grad_norm": 0.1270781308412552,
|
| 344 |
+
"learning_rate": 2.9687191748542485e-05,
|
| 345 |
+
"loss": 0.0056,
|
| 346 |
+
"step": 560
|
| 347 |
+
},
|
| 348 |
+
{
|
| 349 |
+
"grad_norm": 0.13871441781520844,
|
| 350 |
+
"learning_rate": 2.967052843579021e-05,
|
| 351 |
+
"loss": 0.0079,
|
| 352 |
+
"step": 570
|
| 353 |
+
},
|
| 354 |
+
{
|
| 355 |
+
"grad_norm": 0.13760028779506683,
|
| 356 |
+
"learning_rate": 2.965343766008745e-05,
|
| 357 |
+
"loss": 0.0047,
|
| 358 |
+
"step": 580
|
| 359 |
+
},
|
| 360 |
+
{
|
| 361 |
+
"grad_norm": 0.23738646507263184,
|
| 362 |
+
"learning_rate": 2.9635919919417175e-05,
|
| 363 |
+
"loss": 0.0058,
|
| 364 |
+
"step": 590
|
| 365 |
+
},
|
| 366 |
+
{
|
| 367 |
+
"grad_norm": 0.12334813177585602,
|
| 368 |
+
"learning_rate": 2.9617975724203065e-05,
|
| 369 |
+
"loss": 0.006,
|
| 370 |
+
"step": 600
|
| 371 |
+
},
|
| 372 |
+
{
|
| 373 |
+
"grad_norm": 0.11787013709545135,
|
| 374 |
+
"learning_rate": 2.9599605597294635e-05,
|
| 375 |
+
"loss": 0.0042,
|
| 376 |
+
"step": 610
|
| 377 |
+
},
|
| 378 |
+
{
|
| 379 |
+
"grad_norm": 0.10075175762176514,
|
| 380 |
+
"learning_rate": 2.958081007395199e-05,
|
| 381 |
+
"loss": 0.0051,
|
| 382 |
+
"step": 620
|
| 383 |
+
},
|
| 384 |
+
{
|
| 385 |
+
"grad_norm": 0.12161751091480255,
|
| 386 |
+
"learning_rate": 2.956158970183024e-05,
|
| 387 |
+
"loss": 0.006,
|
| 388 |
+
"step": 630
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"grad_norm": 0.1339486688375473,
|
| 392 |
+
"learning_rate": 2.954194504096355e-05,
|
| 393 |
+
"loss": 0.0053,
|
| 394 |
+
"step": 640
|
| 395 |
+
},
|
| 396 |
+
{
|
| 397 |
+
"grad_norm": 0.09474559128284454,
|
| 398 |
+
"learning_rate": 2.9521876663748787e-05,
|
| 399 |
+
"loss": 0.0069,
|
| 400 |
+
"step": 650
|
| 401 |
+
},
|
| 402 |
+
{
|
| 403 |
+
"grad_norm": 0.10630160570144653,
|
| 404 |
+
"learning_rate": 2.950138515492888e-05,
|
| 405 |
+
"loss": 0.0053,
|
| 406 |
+
"step": 660
|
| 407 |
+
},
|
| 408 |
+
{
|
| 409 |
+
"grad_norm": 0.1059340387582779,
|
| 410 |
+
"learning_rate": 2.9480471111575768e-05,
|
| 411 |
+
"loss": 0.0047,
|
| 412 |
+
"step": 670
|
| 413 |
+
},
|
| 414 |
+
{
|
| 415 |
+
"grad_norm": 0.13061285018920898,
|
| 416 |
+
"learning_rate": 2.9459135143073e-05,
|
| 417 |
+
"loss": 0.0044,
|
| 418 |
+
"step": 680
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"grad_norm": 0.11301890015602112,
|
| 422 |
+
"learning_rate": 2.9437377871097974e-05,
|
| 423 |
+
"loss": 0.004,
|
| 424 |
+
"step": 690
|
| 425 |
+
},
|
| 426 |
+
{
|
| 427 |
+
"grad_norm": 0.11071354895830154,
|
| 428 |
+
"learning_rate": 2.9415199929603843e-05,
|
| 429 |
+
"loss": 0.0048,
|
| 430 |
+
"step": 700
|
| 431 |
+
},
|
| 432 |
+
{
|
| 433 |
+
"grad_norm": 0.10549204051494598,
|
| 434 |
+
"learning_rate": 2.939260196480103e-05,
|
| 435 |
+
"loss": 0.0053,
|
| 436 |
+
"step": 710
|
| 437 |
+
},
|
| 438 |
+
{
|
| 439 |
+
"grad_norm": 0.13164091110229492,
|
| 440 |
+
"learning_rate": 2.9369584635138386e-05,
|
| 441 |
+
"loss": 0.0051,
|
| 442 |
+
"step": 720
|
| 443 |
+
},
|
| 444 |
+
{
|
| 445 |
+
"grad_norm": 0.1299622356891632,
|
| 446 |
+
"learning_rate": 2.9346148611284025e-05,
|
| 447 |
+
"loss": 0.0057,
|
| 448 |
+
"step": 730
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"grad_norm": 0.12443673610687256,
|
| 452 |
+
"learning_rate": 2.9322294576105785e-05,
|
| 453 |
+
"loss": 0.0054,
|
| 454 |
+
"step": 740
|
| 455 |
+
},
|
| 456 |
+
{
|
| 457 |
+
"grad_norm": 0.2119351178407669,
|
| 458 |
+
"learning_rate": 2.92980232246513e-05,
|
| 459 |
+
"loss": 0.006,
|
| 460 |
+
"step": 750
|
| 461 |
+
},
|
| 462 |
+
{
|
| 463 |
+
"grad_norm": 0.17323265969753265,
|
| 464 |
+
"learning_rate": 2.927333526412779e-05,
|
| 465 |
+
"loss": 0.0051,
|
| 466 |
+
"step": 760
|
| 467 |
+
},
|
| 468 |
+
{
|
| 469 |
+
"grad_norm": 0.14540384709835052,
|
| 470 |
+
"learning_rate": 2.9248231413881423e-05,
|
| 471 |
+
"loss": 0.0051,
|
| 472 |
+
"step": 770
|
| 473 |
+
},
|
| 474 |
+
{
|
| 475 |
+
"grad_norm": 0.1042582094669342,
|
| 476 |
+
"learning_rate": 2.9222712405376372e-05,
|
| 477 |
+
"loss": 0.0058,
|
| 478 |
+
"step": 780
|
| 479 |
+
},
|
| 480 |
+
{
|
| 481 |
+
"grad_norm": 0.10313008725643158,
|
| 482 |
+
"learning_rate": 2.919677898217348e-05,
|
| 483 |
+
"loss": 0.0047,
|
| 484 |
+
"step": 790
|
| 485 |
+
},
|
| 486 |
+
{
|
| 487 |
+
"grad_norm": 0.09979265928268433,
|
| 488 |
+
"learning_rate": 2.917043189990864e-05,
|
| 489 |
+
"loss": 0.0046,
|
| 490 |
+
"step": 800
|
| 491 |
+
},
|
| 492 |
+
{
|
| 493 |
+
"grad_norm": 0.08516388386487961,
|
| 494 |
+
"learning_rate": 2.914367192627072e-05,
|
| 495 |
+
"loss": 0.0041,
|
| 496 |
+
"step": 810
|
| 497 |
+
},
|
| 498 |
+
{
|
| 499 |
+
"grad_norm": 0.12445282191038132,
|
| 500 |
+
"learning_rate": 2.9116499840979232e-05,
|
| 501 |
+
"loss": 0.0042,
|
| 502 |
+
"step": 820
|
| 503 |
+
},
|
| 504 |
+
{
|
| 505 |
+
"grad_norm": 0.07022149115800858,
|
| 506 |
+
"learning_rate": 2.9088916435761604e-05,
|
| 507 |
+
"loss": 0.0043,
|
| 508 |
+
"step": 830
|
| 509 |
+
},
|
| 510 |
+
{
|
| 511 |
+
"grad_norm": 0.12865330278873444,
|
| 512 |
+
"learning_rate": 2.9060922514330117e-05,
|
| 513 |
+
"loss": 0.0049,
|
| 514 |
+
"step": 840
|
| 515 |
+
},
|
| 516 |
+
{
|
| 517 |
+
"grad_norm": 0.16881509125232697,
|
| 518 |
+
"learning_rate": 2.903251889235847e-05,
|
| 519 |
+
"loss": 0.0054,
|
| 520 |
+
"step": 850
|
| 521 |
+
},
|
| 522 |
+
{
|
| 523 |
+
"grad_norm": 0.12935705482959747,
|
| 524 |
+
"learning_rate": 2.9003706397458025e-05,
|
| 525 |
+
"loss": 0.0053,
|
| 526 |
+
"step": 860
|
| 527 |
+
},
|
| 528 |
+
{
|
| 529 |
+
"grad_norm": 0.11892224848270416,
|
| 530 |
+
"learning_rate": 2.8974485869153698e-05,
|
| 531 |
+
"loss": 0.0039,
|
| 532 |
+
"step": 870
|
| 533 |
+
},
|
| 534 |
+
{
|
| 535 |
+
"grad_norm": 0.12481220811605453,
|
| 536 |
+
"learning_rate": 2.8944858158859484e-05,
|
| 537 |
+
"loss": 0.0059,
|
| 538 |
+
"step": 880
|
| 539 |
+
},
|
| 540 |
+
{
|
| 541 |
+
"grad_norm": 0.09241459518671036,
|
| 542 |
+
"learning_rate": 2.8914824129853653e-05,
|
| 543 |
+
"loss": 0.0051,
|
| 544 |
+
"step": 890
|
| 545 |
+
},
|
| 546 |
+
{
|
| 547 |
+
"grad_norm": 0.08305121958255768,
|
| 548 |
+
"learning_rate": 2.8884384657253603e-05,
|
| 549 |
+
"loss": 0.0057,
|
| 550 |
+
"step": 900
|
| 551 |
+
},
|
| 552 |
+
{
|
| 553 |
+
"grad_norm": 0.12699495255947113,
|
| 554 |
+
"learning_rate": 2.8853540627990356e-05,
|
| 555 |
+
"loss": 0.005,
|
| 556 |
+
"step": 910
|
| 557 |
+
},
|
| 558 |
+
{
|
| 559 |
+
"grad_norm": 0.11409413814544678,
|
| 560 |
+
"learning_rate": 2.8822292940782718e-05,
|
| 561 |
+
"loss": 0.0042,
|
| 562 |
+
"step": 920
|
| 563 |
+
},
|
| 564 |
+
{
|
| 565 |
+
"grad_norm": 0.1391240954399109,
|
| 566 |
+
"learning_rate": 2.8790642506111083e-05,
|
| 567 |
+
"loss": 0.0042,
|
| 568 |
+
"step": 930
|
| 569 |
+
},
|
| 570 |
+
{
|
| 571 |
+
"grad_norm": 0.12300487607717514,
|
| 572 |
+
"learning_rate": 2.8758590246190914e-05,
|
| 573 |
+
"loss": 0.005,
|
| 574 |
+
"step": 940
|
| 575 |
+
},
|
| 576 |
+
{
|
| 577 |
+
"grad_norm": 0.1318342536687851,
|
| 578 |
+
"learning_rate": 2.8726137094945875e-05,
|
| 579 |
+
"loss": 0.0058,
|
| 580 |
+
"step": 950
|
| 581 |
+
},
|
| 582 |
+
{
|
| 583 |
+
"grad_norm": 0.1381806582212448,
|
| 584 |
+
"learning_rate": 2.8693283997980604e-05,
|
| 585 |
+
"loss": 0.0049,
|
| 586 |
+
"step": 960
|
| 587 |
+
},
|
| 588 |
+
{
|
| 589 |
+
"grad_norm": 0.11659003049135208,
|
| 590 |
+
"learning_rate": 2.8660031912553177e-05,
|
| 591 |
+
"loss": 0.0059,
|
| 592 |
+
"step": 970
|
| 593 |
+
},
|
| 594 |
+
{
|
| 595 |
+
"grad_norm": 0.09772439301013947,
|
| 596 |
+
"learning_rate": 2.8626381807547197e-05,
|
| 597 |
+
"loss": 0.004,
|
| 598 |
+
"step": 980
|
| 599 |
+
},
|
| 600 |
+
{
|
| 601 |
+
"grad_norm": 0.11248010396957397,
|
| 602 |
+
"learning_rate": 2.859233466344358e-05,
|
| 603 |
+
"loss": 0.0062,
|
| 604 |
+
"step": 990
|
| 605 |
+
},
|
| 606 |
+
{
|
| 607 |
+
"grad_norm": 0.08497095108032227,
|
| 608 |
+
"learning_rate": 2.8557891472291994e-05,
|
| 609 |
+
"loss": 0.0051,
|
| 610 |
+
"step": 1000
|
| 611 |
+
},
|
| 612 |
+
{
|
| 613 |
+
"grad_norm": 0.10097184032201767,
|
| 614 |
+
"learning_rate": 2.852305323768191e-05,
|
| 615 |
+
"loss": 0.0048,
|
| 616 |
+
"step": 1010
|
| 617 |
+
},
|
| 618 |
+
{
|
| 619 |
+
"grad_norm": 0.12525945901870728,
|
| 620 |
+
"learning_rate": 2.8487820974713416e-05,
|
| 621 |
+
"loss": 0.0067,
|
| 622 |
+
"step": 1020
|
| 623 |
+
},
|
| 624 |
+
{
|
| 625 |
+
"grad_norm": 0.12628182768821716,
|
| 626 |
+
"learning_rate": 2.8452195709967595e-05,
|
| 627 |
+
"loss": 0.0051,
|
| 628 |
+
"step": 1030
|
| 629 |
+
},
|
| 630 |
+
{
|
| 631 |
+
"grad_norm": 0.09084980189800262,
|
| 632 |
+
"learning_rate": 2.8416178481476652e-05,
|
| 633 |
+
"loss": 0.0044,
|
| 634 |
+
"step": 1040
|
| 635 |
+
},
|
| 636 |
+
{
|
| 637 |
+
"grad_norm": 0.09692411869764328,
|
| 638 |
+
"learning_rate": 2.837977033869362e-05,
|
| 639 |
+
"loss": 0.0043,
|
| 640 |
+
"step": 1050
|
| 641 |
+
},
|
| 642 |
+
{
|
| 643 |
+
"grad_norm": 0.11363832652568817,
|
| 644 |
+
"learning_rate": 2.8342972342461827e-05,
|
| 645 |
+
"loss": 0.0046,
|
| 646 |
+
"step": 1060
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"grad_norm": 0.11651736497879028,
|
| 650 |
+
"learning_rate": 2.8305785564983964e-05,
|
| 651 |
+
"loss": 0.0049,
|
| 652 |
+
"step": 1070
|
| 653 |
+
},
|
| 654 |
+
{
|
| 655 |
+
"grad_norm": 0.11450709402561188,
|
| 656 |
+
"learning_rate": 2.8268211089790837e-05,
|
| 657 |
+
"loss": 0.006,
|
| 658 |
+
"step": 1080
|
| 659 |
+
},
|
| 660 |
+
{
|
| 661 |
+
"grad_norm": 0.13134150207042694,
|
| 662 |
+
"learning_rate": 2.8230250011709824e-05,
|
| 663 |
+
"loss": 0.0051,
|
| 664 |
+
"step": 1090
|
| 665 |
+
},
|
| 666 |
+
{
|
| 667 |
+
"grad_norm": 0.06729510426521301,
|
| 668 |
+
"learning_rate": 2.819190343683294e-05,
|
| 669 |
+
"loss": 0.0046,
|
| 670 |
+
"step": 1100
|
| 671 |
+
},
|
| 672 |
+
{
|
| 673 |
+
"grad_norm": 0.07857424765825272,
|
| 674 |
+
"learning_rate": 2.8153172482484634e-05,
|
| 675 |
+
"loss": 0.0041,
|
| 676 |
+
"step": 1110
|
| 677 |
+
},
|
| 678 |
+
{
|
| 679 |
+
"grad_norm": 0.12144170701503754,
|
| 680 |
+
"learning_rate": 2.811405827718921e-05,
|
| 681 |
+
"loss": 0.0042,
|
| 682 |
+
"step": 1120
|
| 683 |
+
},
|
| 684 |
+
{
|
| 685 |
+
"grad_norm": 0.1513696163892746,
|
| 686 |
+
"learning_rate": 2.8074561960637976e-05,
|
| 687 |
+
"loss": 0.0058,
|
| 688 |
+
"step": 1130
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"grad_norm": 0.10109832137823105,
|
| 692 |
+
"learning_rate": 2.8034684683656003e-05,
|
| 693 |
+
"loss": 0.0045,
|
| 694 |
+
"step": 1140
|
| 695 |
+
},
|
| 696 |
+
{
|
| 697 |
+
"grad_norm": 0.07752282917499542,
|
| 698 |
+
"learning_rate": 2.7994427608168617e-05,
|
| 699 |
+
"loss": 0.0044,
|
| 700 |
+
"step": 1150
|
| 701 |
+
},
|
| 702 |
+
{
|
| 703 |
+
"grad_norm": 0.0888523980975151,
|
| 704 |
+
"learning_rate": 2.7953791907167525e-05,
|
| 705 |
+
"loss": 0.0055,
|
| 706 |
+
"step": 1160
|
| 707 |
+
},
|
| 708 |
+
{
|
| 709 |
+
"grad_norm": 0.10832788050174713,
|
| 710 |
+
"learning_rate": 2.791277876467665e-05,
|
| 711 |
+
"loss": 0.0061,
|
| 712 |
+
"step": 1170
|
| 713 |
+
},
|
| 714 |
+
{
|
| 715 |
+
"grad_norm": 0.08513602614402771,
|
| 716 |
+
"learning_rate": 2.7871389375717637e-05,
|
| 717 |
+
"loss": 0.0058,
|
| 718 |
+
"step": 1180
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"grad_norm": 0.09278848767280579,
|
| 722 |
+
"learning_rate": 2.782962494627501e-05,
|
| 723 |
+
"loss": 0.0043,
|
| 724 |
+
"step": 1190
|
| 725 |
+
},
|
| 726 |
+
{
|
| 727 |
+
"grad_norm": 0.09228791296482086,
|
| 728 |
+
"learning_rate": 2.7787486693261047e-05,
|
| 729 |
+
"loss": 0.0053,
|
| 730 |
+
"step": 1200
|
| 731 |
+
},
|
| 732 |
+
{
|
| 733 |
+
"grad_norm": 0.11885028332471848,
|
| 734 |
+
"learning_rate": 2.7744975844480326e-05,
|
| 735 |
+
"loss": 0.0043,
|
| 736 |
+
"step": 1210
|
| 737 |
+
},
|
| 738 |
+
{
|
| 739 |
+
"grad_norm": 0.08647848665714264,
|
| 740 |
+
"learning_rate": 2.7702093638593952e-05,
|
| 741 |
+
"loss": 0.0043,
|
| 742 |
+
"step": 1220
|
| 743 |
+
},
|
| 744 |
+
{
|
| 745 |
+
"grad_norm": 0.1418542116880417,
|
| 746 |
+
"learning_rate": 2.7658841325083445e-05,
|
| 747 |
+
"loss": 0.0048,
|
| 748 |
+
"step": 1230
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"grad_norm": 0.1654549092054367,
|
| 752 |
+
"learning_rate": 2.761522016421435e-05,
|
| 753 |
+
"loss": 0.0047,
|
| 754 |
+
"step": 1240
|
| 755 |
+
},
|
| 756 |
+
{
|
| 757 |
+
"grad_norm": 0.13846398890018463,
|
| 758 |
+
"learning_rate": 2.757123142699952e-05,
|
| 759 |
+
"loss": 0.0044,
|
| 760 |
+
"step": 1250
|
| 761 |
+
},
|
| 762 |
+
{
|
| 763 |
+
"grad_norm": 0.11066365987062454,
|
| 764 |
+
"learning_rate": 2.7526876395162063e-05,
|
| 765 |
+
"loss": 0.0047,
|
| 766 |
+
"step": 1260
|
| 767 |
+
},
|
| 768 |
+
{
|
| 769 |
+
"grad_norm": 0.1423611342906952,
|
| 770 |
+
"learning_rate": 2.748215636109802e-05,
|
| 771 |
+
"loss": 0.0049,
|
| 772 |
+
"step": 1270
|
| 773 |
+
},
|
| 774 |
+
{
|
| 775 |
+
"grad_norm": 0.1386536806821823,
|
| 776 |
+
"learning_rate": 2.7437072627838685e-05,
|
| 777 |
+
"loss": 0.0056,
|
| 778 |
+
"step": 1280
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"grad_norm": 0.1230684295296669,
|
| 782 |
+
"learning_rate": 2.7391626509012657e-05,
|
| 783 |
+
"loss": 0.0051,
|
| 784 |
+
"step": 1290
|
| 785 |
+
},
|
| 786 |
+
{
|
| 787 |
+
"grad_norm": 0.1601068675518036,
|
| 788 |
+
"learning_rate": 2.7345819328807545e-05,
|
| 789 |
+
"loss": 0.0046,
|
| 790 |
+
"step": 1300
|
| 791 |
+
},
|
| 792 |
+
{
|
| 793 |
+
"grad_norm": 0.11355318874120712,
|
| 794 |
+
"learning_rate": 2.7299652421931398e-05,
|
| 795 |
+
"loss": 0.0045,
|
| 796 |
+
"step": 1310
|
| 797 |
+
},
|
| 798 |
+
{
|
| 799 |
+
"grad_norm": 0.09870899468660355,
|
| 800 |
+
"learning_rate": 2.7253127133573818e-05,
|
| 801 |
+
"loss": 0.004,
|
| 802 |
+
"step": 1320
|
| 803 |
+
},
|
| 804 |
+
{
|
| 805 |
+
"grad_norm": 0.13512560725212097,
|
| 806 |
+
"learning_rate": 2.7206244819366748e-05,
|
| 807 |
+
"loss": 0.0072,
|
| 808 |
+
"step": 1330
|
| 809 |
+
},
|
| 810 |
+
{
|
| 811 |
+
"grad_norm": 0.12584540247917175,
|
| 812 |
+
"learning_rate": 2.7159006845344985e-05,
|
| 813 |
+
"loss": 0.0059,
|
| 814 |
+
"step": 1340
|
| 815 |
+
},
|
| 816 |
+
{
|
| 817 |
+
"grad_norm": 0.1445896178483963,
|
| 818 |
+
"learning_rate": 2.7111414587906383e-05,
|
| 819 |
+
"loss": 0.0045,
|
| 820 |
+
"step": 1350
|
| 821 |
+
},
|
| 822 |
+
{
|
| 823 |
+
"grad_norm": 0.08940842747688293,
|
| 824 |
+
"learning_rate": 2.7063469433771726e-05,
|
| 825 |
+
"loss": 0.0049,
|
| 826 |
+
"step": 1360
|
| 827 |
+
},
|
| 828 |
+
{
|
| 829 |
+
"grad_norm": 0.07216592133045197,
|
| 830 |
+
"learning_rate": 2.701517277994434e-05,
|
| 831 |
+
"loss": 0.0041,
|
| 832 |
+
"step": 1370
|
| 833 |
+
},
|
| 834 |
+
{
|
| 835 |
+
"grad_norm": 0.0641094371676445,
|
| 836 |
+
"learning_rate": 2.6966526033669398e-05,
|
| 837 |
+
"loss": 0.0037,
|
| 838 |
+
"step": 1380
|
| 839 |
+
},
|
| 840 |
+
{
|
| 841 |
+
"grad_norm": 0.1202094703912735,
|
| 842 |
+
"learning_rate": 2.6917530612392874e-05,
|
| 843 |
+
"loss": 0.004,
|
| 844 |
+
"step": 1390
|
| 845 |
+
},
|
| 846 |
+
{
|
| 847 |
+
"grad_norm": 0.10160712152719498,
|
| 848 |
+
"learning_rate": 2.6868187943720294e-05,
|
| 849 |
+
"loss": 0.0039,
|
| 850 |
+
"step": 1400
|
| 851 |
+
},
|
| 852 |
+
{
|
| 853 |
+
"grad_norm": 0.0838804617524147,
|
| 854 |
+
"learning_rate": 2.681849946537511e-05,
|
| 855 |
+
"loss": 0.0044,
|
| 856 |
+
"step": 1410
|
| 857 |
+
},
|
| 858 |
+
{
|
| 859 |
+
"grad_norm": 0.07244281470775604,
|
| 860 |
+
"learning_rate": 2.6768466625156797e-05,
|
| 861 |
+
"loss": 0.0042,
|
| 862 |
+
"step": 1420
|
| 863 |
+
},
|
| 864 |
+
{
|
| 865 |
+
"grad_norm": 0.09038420021533966,
|
| 866 |
+
"learning_rate": 2.671809088089871e-05,
|
| 867 |
+
"loss": 0.0043,
|
| 868 |
+
"step": 1430
|
| 869 |
+
},
|
| 870 |
+
{
|
| 871 |
+
"grad_norm": 0.1908736675977707,
|
| 872 |
+
"learning_rate": 2.666737370042556e-05,
|
| 873 |
+
"loss": 0.0037,
|
| 874 |
+
"step": 1440
|
| 875 |
+
},
|
| 876 |
+
{
|
| 877 |
+
"grad_norm": 0.07205282896757126,
|
| 878 |
+
"learning_rate": 2.661631656151068e-05,
|
| 879 |
+
"loss": 0.0039,
|
| 880 |
+
"step": 1450
|
| 881 |
+
},
|
| 882 |
+
{
|
| 883 |
+
"grad_norm": 0.14247483015060425,
|
| 884 |
+
"learning_rate": 2.6564920951832937e-05,
|
| 885 |
+
"loss": 0.0053,
|
| 886 |
+
"step": 1460
|
| 887 |
+
},
|
| 888 |
+
{
|
| 889 |
+
"grad_norm": 0.09355881810188293,
|
| 890 |
+
"learning_rate": 2.6513188368933416e-05,
|
| 891 |
+
"loss": 0.0035,
|
| 892 |
+
"step": 1470
|
| 893 |
+
},
|
| 894 |
+
{
|
| 895 |
+
"grad_norm": 0.1190556213259697,
|
| 896 |
+
"learning_rate": 2.6461120320171757e-05,
|
| 897 |
+
"loss": 0.0049,
|
| 898 |
+
"step": 1480
|
| 899 |
+
},
|
| 900 |
+
{
|
| 901 |
+
"grad_norm": 0.1366739571094513,
|
| 902 |
+
"learning_rate": 2.6408718322682246e-05,
|
| 903 |
+
"loss": 0.0047,
|
| 904 |
+
"step": 1490
|
| 905 |
+
},
|
| 906 |
+
{
|
| 907 |
+
"grad_norm": 0.12044065445661545,
|
| 908 |
+
"learning_rate": 2.6355983903329627e-05,
|
| 909 |
+
"loss": 0.0036,
|
| 910 |
+
"step": 1500
|
| 911 |
+
},
|
| 912 |
+
{
|
| 913 |
+
"grad_norm": 0.11228612065315247,
|
| 914 |
+
"learning_rate": 2.630291859866457e-05,
|
| 915 |
+
"loss": 0.005,
|
| 916 |
+
"step": 1510
|
| 917 |
+
},
|
| 918 |
+
{
|
| 919 |
+
"grad_norm": 0.09218034148216248,
|
| 920 |
+
"learning_rate": 2.624952395487895e-05,
|
| 921 |
+
"loss": 0.0033,
|
| 922 |
+
"step": 1520
|
| 923 |
+
},
|
| 924 |
+
{
|
| 925 |
+
"grad_norm": 0.07770651578903198,
|
| 926 |
+
"learning_rate": 2.619580152776075e-05,
|
| 927 |
+
"loss": 0.0047,
|
| 928 |
+
"step": 1530
|
| 929 |
+
},
|
| 930 |
+
{
|
| 931 |
+
"grad_norm": 0.15504707396030426,
|
| 932 |
+
"learning_rate": 2.6141752882648758e-05,
|
| 933 |
+
"loss": 0.0041,
|
| 934 |
+
"step": 1540
|
| 935 |
+
},
|
| 936 |
+
{
|
| 937 |
+
"grad_norm": 0.16510525345802307,
|
| 938 |
+
"learning_rate": 2.6087379594386956e-05,
|
| 939 |
+
"loss": 0.0041,
|
| 940 |
+
"step": 1550
|
| 941 |
+
},
|
| 942 |
+
{
|
| 943 |
+
"grad_norm": 0.09923221915960312,
|
| 944 |
+
"learning_rate": 2.6032683247278613e-05,
|
| 945 |
+
"loss": 0.0046,
|
| 946 |
+
"step": 1560
|
| 947 |
+
},
|
| 948 |
+
{
|
| 949 |
+
"grad_norm": 0.10440414398908615,
|
| 950 |
+
"learning_rate": 2.5977665435040135e-05,
|
| 951 |
+
"loss": 0.0045,
|
| 952 |
+
"step": 1570
|
| 953 |
+
},
|
| 954 |
+
{
|
| 955 |
+
"grad_norm": 0.11448866128921509,
|
| 956 |
+
"learning_rate": 2.592232776075464e-05,
|
| 957 |
+
"loss": 0.0038,
|
| 958 |
+
"step": 1580
|
| 959 |
+
},
|
| 960 |
+
{
|
| 961 |
+
"grad_norm": 0.10592176020145416,
|
| 962 |
+
"learning_rate": 2.5866671836825226e-05,
|
| 963 |
+
"loss": 0.0041,
|
| 964 |
+
"step": 1590
|
| 965 |
+
},
|
| 966 |
+
{
|
| 967 |
+
"grad_norm": 0.10503924638032913,
|
| 968 |
+
"learning_rate": 2.5810699284928e-05,
|
| 969 |
+
"loss": 0.0048,
|
| 970 |
+
"step": 1600
|
| 971 |
+
},
|
| 972 |
+
{
|
| 973 |
+
"grad_norm": 0.08060912042856216,
|
| 974 |
+
"learning_rate": 2.575441173596482e-05,
|
| 975 |
+
"loss": 0.0042,
|
| 976 |
+
"step": 1610
|
| 977 |
+
},
|
| 978 |
+
{
|
| 979 |
+
"grad_norm": 0.08904369175434113,
|
| 980 |
+
"learning_rate": 2.569781083001579e-05,
|
| 981 |
+
"loss": 0.0035,
|
| 982 |
+
"step": 1620
|
| 983 |
+
},
|
| 984 |
+
{
|
| 985 |
+
"grad_norm": 0.10122060775756836,
|
| 986 |
+
"learning_rate": 2.5640898216291467e-05,
|
| 987 |
+
"loss": 0.005,
|
| 988 |
+
"step": 1630
|
| 989 |
+
},
|
| 990 |
+
{
|
| 991 |
+
"grad_norm": 0.07322278618812561,
|
| 992 |
+
"learning_rate": 2.558367555308479e-05,
|
| 993 |
+
"loss": 0.0036,
|
| 994 |
+
"step": 1640
|
| 995 |
+
},
|
| 996 |
+
{
|
| 997 |
+
"grad_norm": 0.10172346234321594,
|
| 998 |
+
"learning_rate": 2.5526144507722777e-05,
|
| 999 |
+
"loss": 0.0044,
|
| 1000 |
+
"step": 1650
|
| 1001 |
+
},
|
| 1002 |
+
{
|
| 1003 |
+
"grad_norm": 0.1091361865401268,
|
| 1004 |
+
"learning_rate": 2.546830675651794e-05,
|
| 1005 |
+
"loss": 0.0035,
|
| 1006 |
+
"step": 1660
|
| 1007 |
+
},
|
| 1008 |
+
{
|
| 1009 |
+
"grad_norm": 0.09636865556240082,
|
| 1010 |
+
"learning_rate": 2.5410163984719442e-05,
|
| 1011 |
+
"loss": 0.0037,
|
| 1012 |
+
"step": 1670
|
| 1013 |
+
},
|
| 1014 |
+
{
|
| 1015 |
+
"grad_norm": 0.11563992500305176,
|
| 1016 |
+
"learning_rate": 2.5351717886464e-05,
|
| 1017 |
+
"loss": 0.0048,
|
| 1018 |
+
"step": 1680
|
| 1019 |
+
},
|
| 1020 |
+
{
|
| 1021 |
+
"grad_norm": 0.0999450832605362,
|
| 1022 |
+
"learning_rate": 2.5292970164726503e-05,
|
| 1023 |
+
"loss": 0.0058,
|
| 1024 |
+
"step": 1690
|
| 1025 |
+
},
|
| 1026 |
+
{
|
| 1027 |
+
"grad_norm": 0.10548657923936844,
|
| 1028 |
+
"learning_rate": 2.5233922531270407e-05,
|
| 1029 |
+
"loss": 0.0057,
|
| 1030 |
+
"step": 1700
|
| 1031 |
+
},
|
| 1032 |
+
{
|
| 1033 |
+
"grad_norm": 0.11799237877130508,
|
| 1034 |
+
"learning_rate": 2.517457670659786e-05,
|
| 1035 |
+
"loss": 0.005,
|
| 1036 |
+
"step": 1710
|
| 1037 |
+
},
|
| 1038 |
+
{
|
| 1039 |
+
"grad_norm": 0.16543568670749664,
|
| 1040 |
+
"learning_rate": 2.5114934419899557e-05,
|
| 1041 |
+
"loss": 0.0047,
|
| 1042 |
+
"step": 1720
|
| 1043 |
+
},
|
| 1044 |
+
{
|
| 1045 |
+
"grad_norm": 0.13398373126983643,
|
| 1046 |
+
"learning_rate": 2.5054997409004368e-05,
|
| 1047 |
+
"loss": 0.0048,
|
| 1048 |
+
"step": 1730
|
| 1049 |
+
},
|
| 1050 |
+
{
|
| 1051 |
+
"grad_norm": 0.12377601861953735,
|
| 1052 |
+
"learning_rate": 2.4994767420328692e-05,
|
| 1053 |
+
"loss": 0.0042,
|
| 1054 |
+
"step": 1740
|
| 1055 |
+
},
|
| 1056 |
+
{
|
| 1057 |
+
"grad_norm": 0.1447695940732956,
|
| 1058 |
+
"learning_rate": 2.4934246208825594e-05,
|
| 1059 |
+
"loss": 0.0053,
|
| 1060 |
+
"step": 1750
|
| 1061 |
+
},
|
| 1062 |
+
{
|
| 1063 |
+
"grad_norm": 0.09021581709384918,
|
| 1064 |
+
"learning_rate": 2.4873435537933627e-05,
|
| 1065 |
+
"loss": 0.0041,
|
| 1066 |
+
"step": 1760
|
| 1067 |
+
},
|
| 1068 |
+
{
|
| 1069 |
+
"grad_norm": 0.2032671719789505,
|
| 1070 |
+
"learning_rate": 2.48123371795255e-05,
|
| 1071 |
+
"loss": 0.0059,
|
| 1072 |
+
"step": 1770
|
| 1073 |
+
},
|
| 1074 |
+
{
|
| 1075 |
+
"grad_norm": 0.12990808486938477,
|
| 1076 |
+
"learning_rate": 2.475095291385641e-05,
|
| 1077 |
+
"loss": 0.0045,
|
| 1078 |
+
"step": 1780
|
| 1079 |
+
},
|
| 1080 |
+
{
|
| 1081 |
+
"grad_norm": 0.1114439144730568,
|
| 1082 |
+
"learning_rate": 2.468928452951219e-05,
|
| 1083 |
+
"loss": 0.0061,
|
| 1084 |
+
"step": 1790
|
| 1085 |
+
},
|
| 1086 |
+
{
|
| 1087 |
+
"grad_norm": 0.13164660334587097,
|
| 1088 |
+
"learning_rate": 2.4627333823357185e-05,
|
| 1089 |
+
"loss": 0.0046,
|
| 1090 |
+
"step": 1800
|
| 1091 |
+
},
|
| 1092 |
+
{
|
| 1093 |
+
"grad_norm": 0.099468894302845,
|
| 1094 |
+
"learning_rate": 2.456510260048191e-05,
|
| 1095 |
+
"loss": 0.0038,
|
| 1096 |
+
"step": 1810
|
| 1097 |
+
},
|
| 1098 |
+
{
|
| 1099 |
+
"grad_norm": 0.13928678631782532,
|
| 1100 |
+
"learning_rate": 2.4502592674150436e-05,
|
| 1101 |
+
"loss": 0.005,
|
| 1102 |
+
"step": 1820
|
| 1103 |
+
},
|
| 1104 |
+
{
|
| 1105 |
+
"grad_norm": 0.08072284609079361,
|
| 1106 |
+
"learning_rate": 2.4439805865747564e-05,
|
| 1107 |
+
"loss": 0.0043,
|
| 1108 |
+
"step": 1830
|
| 1109 |
+
},
|
| 1110 |
+
{
|
| 1111 |
+
"grad_norm": 0.11410820484161377,
|
| 1112 |
+
"learning_rate": 2.4376744004725756e-05,
|
| 1113 |
+
"loss": 0.0043,
|
| 1114 |
+
"step": 1840
|
| 1115 |
+
},
|
| 1116 |
+
{
|
| 1117 |
+
"grad_norm": 0.11230126023292542,
|
| 1118 |
+
"learning_rate": 2.4313408928551834e-05,
|
| 1119 |
+
"loss": 0.0041,
|
| 1120 |
+
"step": 1850
|
| 1121 |
+
},
|
| 1122 |
+
{
|
| 1123 |
+
"grad_norm": 0.10740366578102112,
|
| 1124 |
+
"learning_rate": 2.424980248265342e-05,
|
| 1125 |
+
"loss": 0.0058,
|
| 1126 |
+
"step": 1860
|
| 1127 |
+
},
|
| 1128 |
+
{
|
| 1129 |
+
"grad_norm": 0.05973527953028679,
|
| 1130 |
+
"learning_rate": 2.418592652036519e-05,
|
| 1131 |
+
"loss": 0.0051,
|
| 1132 |
+
"step": 1870
|
| 1133 |
+
},
|
| 1134 |
+
{
|
| 1135 |
+
"grad_norm": 0.07472732663154602,
|
| 1136 |
+
"learning_rate": 2.4121782902874866e-05,
|
| 1137 |
+
"loss": 0.0043,
|
| 1138 |
+
"step": 1880
|
| 1139 |
+
},
|
| 1140 |
+
{
|
| 1141 |
+
"grad_norm": 0.09412290155887604,
|
| 1142 |
+
"learning_rate": 2.4057373499168968e-05,
|
| 1143 |
+
"loss": 0.0052,
|
| 1144 |
+
"step": 1890
|
| 1145 |
+
},
|
| 1146 |
+
{
|
| 1147 |
+
"grad_norm": 0.10615745931863785,
|
| 1148 |
+
"learning_rate": 2.399270018597838e-05,
|
| 1149 |
+
"loss": 0.0047,
|
| 1150 |
+
"step": 1900
|
| 1151 |
+
},
|
| 1152 |
+
{
|
| 1153 |
+
"grad_norm": 0.10771957784891129,
|
| 1154 |
+
"learning_rate": 2.392776484772365e-05,
|
| 1155 |
+
"loss": 0.0047,
|
| 1156 |
+
"step": 1910
|
| 1157 |
+
},
|
| 1158 |
+
{
|
| 1159 |
+
"grad_norm": 0.17450952529907227,
|
| 1160 |
+
"learning_rate": 2.3862569376460096e-05,
|
| 1161 |
+
"loss": 0.0046,
|
| 1162 |
+
"step": 1920
|
| 1163 |
+
},
|
| 1164 |
+
{
|
| 1165 |
+
"grad_norm": 0.14669440686702728,
|
| 1166 |
+
"learning_rate": 2.3797115671822665e-05,
|
| 1167 |
+
"loss": 0.0052,
|
| 1168 |
+
"step": 1930
|
| 1169 |
+
},
|
| 1170 |
+
{
|
| 1171 |
+
"grad_norm": 0.080330990254879,
|
| 1172 |
+
"learning_rate": 2.3731405640970578e-05,
|
| 1173 |
+
"loss": 0.0042,
|
| 1174 |
+
"step": 1940
|
| 1175 |
+
},
|
| 1176 |
+
{
|
| 1177 |
+
"grad_norm": 0.13788354396820068,
|
| 1178 |
+
"learning_rate": 2.366544119853178e-05,
|
| 1179 |
+
"loss": 0.0043,
|
| 1180 |
+
"step": 1950
|
| 1181 |
+
},
|
| 1182 |
+
{
|
| 1183 |
+
"grad_norm": 0.12239138782024384,
|
| 1184 |
+
"learning_rate": 2.3599224266547145e-05,
|
| 1185 |
+
"loss": 0.0039,
|
| 1186 |
+
"step": 1960
|
| 1187 |
+
},
|
| 1188 |
+
{
|
| 1189 |
+
"grad_norm": 0.12057246267795563,
|
| 1190 |
+
"learning_rate": 2.3532756774414454e-05,
|
| 1191 |
+
"loss": 0.0047,
|
| 1192 |
+
"step": 1970
|
| 1193 |
+
},
|
| 1194 |
+
{
|
| 1195 |
+
"grad_norm": 0.08931958675384521,
|
| 1196 |
+
"learning_rate": 2.3466040658832202e-05,
|
| 1197 |
+
"loss": 0.0043,
|
| 1198 |
+
"step": 1980
|
| 1199 |
+
},
|
| 1200 |
+
{
|
| 1201 |
+
"grad_norm": 0.11142227053642273,
|
| 1202 |
+
"learning_rate": 2.3399077863743154e-05,
|
| 1203 |
+
"loss": 0.0043,
|
| 1204 |
+
"step": 1990
|
| 1205 |
+
},
|
| 1206 |
+
{
|
| 1207 |
+
"grad_norm": 0.11959561705589294,
|
| 1208 |
+
"learning_rate": 2.3331870340277703e-05,
|
| 1209 |
+
"loss": 0.0043,
|
| 1210 |
+
"step": 2000
|
| 1211 |
+
},
|
| 1212 |
+
{
|
| 1213 |
+
"grad_norm": 0.07246878743171692,
|
| 1214 |
+
"learning_rate": 2.3264420046697033e-05,
|
| 1215 |
+
"loss": 0.0038,
|
| 1216 |
+
"step": 2010
|
| 1217 |
+
},
|
| 1218 |
+
{
|
| 1219 |
+
"grad_norm": 0.10344090312719345,
|
| 1220 |
+
"learning_rate": 2.3196728948336035e-05,
|
| 1221 |
+
"loss": 0.0051,
|
| 1222 |
+
"step": 2020
|
| 1223 |
+
},
|
| 1224 |
+
{
|
| 1225 |
+
"grad_norm": 0.11588168889284134,
|
| 1226 |
+
"learning_rate": 2.3128799017546066e-05,
|
| 1227 |
+
"loss": 0.0054,
|
| 1228 |
+
"step": 2030
|
| 1229 |
+
},
|
| 1230 |
+
{
|
| 1231 |
+
"grad_norm": 0.08791792392730713,
|
| 1232 |
+
"learning_rate": 2.306063223363747e-05,
|
| 1233 |
+
"loss": 0.0039,
|
| 1234 |
+
"step": 2040
|
| 1235 |
+
},
|
| 1236 |
+
{
|
| 1237 |
+
"grad_norm": 0.11206529289484024,
|
| 1238 |
+
"learning_rate": 2.299223058282189e-05,
|
| 1239 |
+
"loss": 0.0044,
|
| 1240 |
+
"step": 2050
|
| 1241 |
+
},
|
| 1242 |
+
{
|
| 1243 |
+
"grad_norm": 0.10127924382686615,
|
| 1244 |
+
"learning_rate": 2.2923596058154437e-05,
|
| 1245 |
+
"loss": 0.0049,
|
| 1246 |
+
"step": 2060
|
| 1247 |
+
},
|
| 1248 |
+
{
|
| 1249 |
+
"grad_norm": 0.09655720740556717,
|
| 1250 |
+
"learning_rate": 2.2854730659475575e-05,
|
| 1251 |
+
"loss": 0.004,
|
| 1252 |
+
"step": 2070
|
| 1253 |
+
},
|
| 1254 |
+
{
|
| 1255 |
+
"grad_norm": 0.0919470265507698,
|
| 1256 |
+
"learning_rate": 2.278563639335286e-05,
|
| 1257 |
+
"loss": 0.0037,
|
| 1258 |
+
"step": 2080
|
| 1259 |
+
},
|
| 1260 |
+
{
|
| 1261 |
+
"grad_norm": 0.1400183141231537,
|
| 1262 |
+
"learning_rate": 2.27163152730225e-05,
|
| 1263 |
+
"loss": 0.0061,
|
| 1264 |
+
"step": 2090
|
| 1265 |
+
},
|
| 1266 |
+
{
|
| 1267 |
+
"grad_norm": 0.09806305915117264,
|
| 1268 |
+
"learning_rate": 2.264676931833065e-05,
|
| 1269 |
+
"loss": 0.0044,
|
| 1270 |
+
"step": 2100
|
| 1271 |
+
},
|
| 1272 |
+
{
|
| 1273 |
+
"grad_norm": 0.07071761786937714,
|
| 1274 |
+
"learning_rate": 2.25770005556746e-05,
|
| 1275 |
+
"loss": 0.0027,
|
| 1276 |
+
"step": 2110
|
| 1277 |
+
},
|
| 1278 |
+
{
|
| 1279 |
+
"grad_norm": 0.08000872284173965,
|
| 1280 |
+
"learning_rate": 2.250701101794371e-05,
|
| 1281 |
+
"loss": 0.0039,
|
| 1282 |
+
"step": 2120
|
| 1283 |
+
},
|
| 1284 |
+
{
|
| 1285 |
+
"grad_norm": 0.09656523168087006,
|
| 1286 |
+
"learning_rate": 2.243680274446017e-05,
|
| 1287 |
+
"loss": 0.0057,
|
| 1288 |
+
"step": 2130
|
| 1289 |
+
},
|
| 1290 |
+
{
|
| 1291 |
+
"grad_norm": 0.09651938825845718,
|
| 1292 |
+
"learning_rate": 2.236637778091961e-05,
|
| 1293 |
+
"loss": 0.0034,
|
| 1294 |
+
"step": 2140
|
| 1295 |
+
},
|
| 1296 |
+
{
|
| 1297 |
+
"grad_norm": 0.09979426115751266,
|
| 1298 |
+
"learning_rate": 2.2295738179331452e-05,
|
| 1299 |
+
"loss": 0.004,
|
| 1300 |
+
"step": 2150
|
| 1301 |
+
},
|
| 1302 |
+
{
|
| 1303 |
+
"grad_norm": 0.05974041670560837,
|
| 1304 |
+
"learning_rate": 2.2224885997959155e-05,
|
| 1305 |
+
"loss": 0.0035,
|
| 1306 |
+
"step": 2160
|
| 1307 |
+
},
|
| 1308 |
+
{
|
| 1309 |
+
"grad_norm": 0.15712539851665497,
|
| 1310 |
+
"learning_rate": 2.215382330126021e-05,
|
| 1311 |
+
"loss": 0.0031,
|
| 1312 |
+
"step": 2170
|
| 1313 |
+
},
|
| 1314 |
+
{
|
| 1315 |
+
"grad_norm": 0.09726399183273315,
|
| 1316 |
+
"learning_rate": 2.2082552159826037e-05,
|
| 1317 |
+
"loss": 0.0031,
|
| 1318 |
+
"step": 2180
|
| 1319 |
+
},
|
| 1320 |
+
{
|
| 1321 |
+
"grad_norm": 0.06862549483776093,
|
| 1322 |
+
"learning_rate": 2.201107465032158e-05,
|
| 1323 |
+
"loss": 0.0037,
|
| 1324 |
+
"step": 2190
|
| 1325 |
+
},
|
| 1326 |
+
{
|
| 1327 |
+
"grad_norm": 0.18924099206924438,
|
| 1328 |
+
"learning_rate": 2.1939392855424865e-05,
|
| 1329 |
+
"loss": 0.0051,
|
| 1330 |
+
"step": 2200
|
| 1331 |
+
},
|
| 1332 |
+
{
|
| 1333 |
+
"grad_norm": 0.10739069432020187,
|
| 1334 |
+
"learning_rate": 2.186750886376629e-05,
|
| 1335 |
+
"loss": 0.0043,
|
| 1336 |
+
"step": 2210
|
| 1337 |
+
},
|
| 1338 |
+
{
|
| 1339 |
+
"grad_norm": 0.12378190457820892,
|
| 1340 |
+
"learning_rate": 2.1795424769867748e-05,
|
| 1341 |
+
"loss": 0.0065,
|
| 1342 |
+
"step": 2220
|
| 1343 |
+
},
|
| 1344 |
+
{
|
| 1345 |
+
"grad_norm": 0.09460093080997467,
|
| 1346 |
+
"learning_rate": 2.172314267408163e-05,
|
| 1347 |
+
"loss": 0.0033,
|
| 1348 |
+
"step": 2230
|
| 1349 |
+
},
|
| 1350 |
+
{
|
| 1351 |
+
"grad_norm": 0.0978519544005394,
|
| 1352 |
+
"learning_rate": 2.1650664682529626e-05,
|
| 1353 |
+
"loss": 0.003,
|
| 1354 |
+
"step": 2240
|
| 1355 |
+
},
|
| 1356 |
+
{
|
| 1357 |
+
"grad_norm": 0.09626927971839905,
|
| 1358 |
+
"learning_rate": 2.157799290704132e-05,
|
| 1359 |
+
"loss": 0.0046,
|
| 1360 |
+
"step": 2250
|
| 1361 |
+
},
|
| 1362 |
+
{
|
| 1363 |
+
"grad_norm": 0.1985124796628952,
|
| 1364 |
+
"learning_rate": 2.150512946509269e-05,
|
| 1365 |
+
"loss": 0.0044,
|
| 1366 |
+
"step": 2260
|
| 1367 |
+
},
|
| 1368 |
+
{
|
| 1369 |
+
"grad_norm": 0.17612029612064362,
|
| 1370 |
+
"learning_rate": 2.1432076479744403e-05,
|
| 1371 |
+
"loss": 0.0036,
|
| 1372 |
+
"step": 2270
|
| 1373 |
+
},
|
| 1374 |
+
{
|
| 1375 |
+
"grad_norm": 0.13746587932109833,
|
| 1376 |
+
"learning_rate": 2.1358836079579953e-05,
|
| 1377 |
+
"loss": 0.0036,
|
| 1378 |
+
"step": 2280
|
| 1379 |
+
},
|
| 1380 |
+
{
|
| 1381 |
+
"grad_norm": 0.1004326120018959,
|
| 1382 |
+
"learning_rate": 2.128541039864364e-05,
|
| 1383 |
+
"loss": 0.0044,
|
| 1384 |
+
"step": 2290
|
| 1385 |
+
},
|
| 1386 |
+
{
|
| 1387 |
+
"grad_norm": 0.126906156539917,
|
| 1388 |
+
"learning_rate": 2.1211801576378384e-05,
|
| 1389 |
+
"loss": 0.003,
|
| 1390 |
+
"step": 2300
|
| 1391 |
+
},
|
| 1392 |
+
{
|
| 1393 |
+
"grad_norm": 0.1311343014240265,
|
| 1394 |
+
"learning_rate": 2.1138011757563388e-05,
|
| 1395 |
+
"loss": 0.0048,
|
| 1396 |
+
"step": 2310
|
| 1397 |
+
},
|
| 1398 |
+
{
|
| 1399 |
+
"grad_norm": 0.13708554208278656,
|
| 1400 |
+
"learning_rate": 2.1064043092251654e-05,
|
| 1401 |
+
"loss": 0.0046,
|
| 1402 |
+
"step": 2320
|
| 1403 |
+
},
|
| 1404 |
+
{
|
| 1405 |
+
"grad_norm": 0.14371265470981598,
|
| 1406 |
+
"learning_rate": 2.098989773570733e-05,
|
| 1407 |
+
"loss": 0.0065,
|
| 1408 |
+
"step": 2330
|
| 1409 |
+
},
|
| 1410 |
+
{
|
| 1411 |
+
"grad_norm": 0.09575030952692032,
|
| 1412 |
+
"learning_rate": 2.0915577848342905e-05,
|
| 1413 |
+
"loss": 0.0032,
|
| 1414 |
+
"step": 2340
|
| 1415 |
+
},
|
| 1416 |
+
{
|
| 1417 |
+
"grad_norm": 0.09642798453569412,
|
| 1418 |
+
"learning_rate": 2.084108559565626e-05,
|
| 1419 |
+
"loss": 0.0058,
|
| 1420 |
+
"step": 2350
|
| 1421 |
+
},
|
| 1422 |
+
{
|
| 1423 |
+
"grad_norm": 0.09199727326631546,
|
| 1424 |
+
"learning_rate": 2.0766423148167592e-05,
|
| 1425 |
+
"loss": 0.0036,
|
| 1426 |
+
"step": 2360
|
| 1427 |
+
},
|
| 1428 |
+
{
|
| 1429 |
+
"grad_norm": 0.10715307295322418,
|
| 1430 |
+
"learning_rate": 2.0691592681356134e-05,
|
| 1431 |
+
"loss": 0.0038,
|
| 1432 |
+
"step": 2370
|
| 1433 |
+
},
|
| 1434 |
+
{
|
| 1435 |
+
"grad_norm": 0.13233180344104767,
|
| 1436 |
+
"learning_rate": 2.0616596375596805e-05,
|
| 1437 |
+
"loss": 0.0045,
|
| 1438 |
+
"step": 2380
|
| 1439 |
+
},
|
| 1440 |
+
{
|
| 1441 |
+
"grad_norm": 0.07768990844488144,
|
| 1442 |
+
"learning_rate": 2.054143641609665e-05,
|
| 1443 |
+
"loss": 0.0042,
|
| 1444 |
+
"step": 2390
|
| 1445 |
+
},
|
| 1446 |
+
{
|
| 1447 |
+
"grad_norm": 0.13608784973621368,
|
| 1448 |
+
"learning_rate": 2.0466114992831175e-05,
|
| 1449 |
+
"loss": 0.005,
|
| 1450 |
+
"step": 2400
|
| 1451 |
+
},
|
| 1452 |
+
{
|
| 1453 |
+
"grad_norm": 0.10314002633094788,
|
| 1454 |
+
"learning_rate": 2.039063430048056e-05,
|
| 1455 |
+
"loss": 0.0036,
|
| 1456 |
+
"step": 2410
|
| 1457 |
+
},
|
| 1458 |
+
{
|
| 1459 |
+
"grad_norm": 0.15518108010292053,
|
| 1460 |
+
"learning_rate": 2.031499653836568e-05,
|
| 1461 |
+
"loss": 0.0077,
|
| 1462 |
+
"step": 2420
|
| 1463 |
+
},
|
| 1464 |
+
{
|
| 1465 |
+
"grad_norm": 0.15236151218414307,
|
| 1466 |
+
"learning_rate": 2.0239203910384035e-05,
|
| 1467 |
+
"loss": 0.0055,
|
| 1468 |
+
"step": 2430
|
| 1469 |
+
},
|
| 1470 |
+
{
|
| 1471 |
+
"grad_norm": 0.11873604357242584,
|
| 1472 |
+
"learning_rate": 2.0163258624945532e-05,
|
| 1473 |
+
"loss": 0.0047,
|
| 1474 |
+
"step": 2440
|
| 1475 |
+
},
|
| 1476 |
+
{
|
| 1477 |
+
"grad_norm": 0.10288285464048386,
|
| 1478 |
+
"learning_rate": 2.008716289490814e-05,
|
| 1479 |
+
"loss": 0.0038,
|
| 1480 |
+
"step": 2450
|
| 1481 |
+
},
|
| 1482 |
+
{
|
| 1483 |
+
"grad_norm": 0.09240736067295074,
|
| 1484 |
+
"learning_rate": 2.0010918937513418e-05,
|
| 1485 |
+
"loss": 0.0032,
|
| 1486 |
+
"step": 2460
|
| 1487 |
+
},
|
| 1488 |
+
{
|
| 1489 |
+
"grad_norm": 0.10177815705537796,
|
| 1490 |
+
"learning_rate": 1.9934528974321904e-05,
|
| 1491 |
+
"loss": 0.0037,
|
| 1492 |
+
"step": 2470
|
| 1493 |
+
},
|
| 1494 |
+
{
|
| 1495 |
+
"grad_norm": 0.10068890452384949,
|
| 1496 |
+
"learning_rate": 1.9857995231148374e-05,
|
| 1497 |
+
"loss": 0.0034,
|
| 1498 |
+
"step": 2480
|
| 1499 |
+
},
|
| 1500 |
+
{
|
| 1501 |
+
"grad_norm": 0.09852269291877747,
|
| 1502 |
+
"learning_rate": 1.9781319937997003e-05,
|
| 1503 |
+
"loss": 0.0057,
|
| 1504 |
+
"step": 2490
|
| 1505 |
+
},
|
| 1506 |
+
{
|
| 1507 |
+
"grad_norm": 0.10190477967262268,
|
| 1508 |
+
"learning_rate": 1.9704505328996387e-05,
|
| 1509 |
+
"loss": 0.0034,
|
| 1510 |
+
"step": 2500
|
| 1511 |
+
},
|
| 1512 |
+
{
|
| 1513 |
+
"grad_norm": 0.11787329614162445,
|
| 1514 |
+
"learning_rate": 1.9627553642334435e-05,
|
| 1515 |
+
"loss": 0.0059,
|
| 1516 |
+
"step": 2510
|
| 1517 |
+
},
|
| 1518 |
+
{
|
| 1519 |
+
"grad_norm": 0.12991037964820862,
|
| 1520 |
+
"learning_rate": 1.9550467120193167e-05,
|
| 1521 |
+
"loss": 0.004,
|
| 1522 |
+
"step": 2520
|
| 1523 |
+
},
|
| 1524 |
+
{
|
| 1525 |
+
"grad_norm": 0.12583398818969727,
|
| 1526 |
+
"learning_rate": 1.947324800868337e-05,
|
| 1527 |
+
"loss": 0.0042,
|
| 1528 |
+
"step": 2530
|
| 1529 |
+
},
|
| 1530 |
+
{
|
| 1531 |
+
"grad_norm": 0.1513422727584839,
|
| 1532 |
+
"learning_rate": 1.9395898557779154e-05,
|
| 1533 |
+
"loss": 0.0039,
|
| 1534 |
+
"step": 2540
|
| 1535 |
+
},
|
| 1536 |
+
{
|
| 1537 |
+
"grad_norm": 0.10720251500606537,
|
| 1538 |
+
"learning_rate": 1.9318421021252408e-05,
|
| 1539 |
+
"loss": 0.0059,
|
| 1540 |
+
"step": 2550
|
| 1541 |
+
},
|
| 1542 |
+
{
|
| 1543 |
+
"grad_norm": 0.108665332198143,
|
| 1544 |
+
"learning_rate": 1.9240817656607114e-05,
|
| 1545 |
+
"loss": 0.0056,
|
| 1546 |
+
"step": 2560
|
| 1547 |
+
},
|
| 1548 |
+
{
|
| 1549 |
+
"grad_norm": 0.18657821416854858,
|
| 1550 |
+
"learning_rate": 1.9163090725013578e-05,
|
| 1551 |
+
"loss": 0.0044,
|
| 1552 |
+
"step": 2570
|
| 1553 |
+
},
|
| 1554 |
+
{
|
| 1555 |
+
"grad_norm": 0.12283720821142197,
|
| 1556 |
+
"learning_rate": 1.9085242491242524e-05,
|
| 1557 |
+
"loss": 0.0046,
|
| 1558 |
+
"step": 2580
|
| 1559 |
+
},
|
| 1560 |
+
{
|
| 1561 |
+
"grad_norm": 0.10580199956893921,
|
| 1562 |
+
"learning_rate": 1.900727522359915e-05,
|
| 1563 |
+
"loss": 0.0035,
|
| 1564 |
+
"step": 2590
|
| 1565 |
+
},
|
| 1566 |
+
{
|
| 1567 |
+
"grad_norm": 0.1246466338634491,
|
| 1568 |
+
"learning_rate": 1.8929191193856985e-05,
|
| 1569 |
+
"loss": 0.0059,
|
| 1570 |
+
"step": 2600
|
| 1571 |
+
},
|
| 1572 |
+
{
|
| 1573 |
+
"grad_norm": 0.11310068517923355,
|
| 1574 |
+
"learning_rate": 1.8850992677191726e-05,
|
| 1575 |
+
"loss": 0.0042,
|
| 1576 |
+
"step": 2610
|
| 1577 |
+
},
|
| 1578 |
+
{
|
| 1579 |
+
"grad_norm": 0.10423216968774796,
|
| 1580 |
+
"learning_rate": 1.877268195211494e-05,
|
| 1581 |
+
"loss": 0.0044,
|
| 1582 |
+
"step": 2620
|
| 1583 |
+
},
|
| 1584 |
+
{
|
| 1585 |
+
"grad_norm": 0.09140152484178543,
|
| 1586 |
+
"learning_rate": 1.8694261300407653e-05,
|
| 1587 |
+
"loss": 0.004,
|
| 1588 |
+
"step": 2630
|
| 1589 |
+
},
|
| 1590 |
+
{
|
| 1591 |
+
"grad_norm": 0.15965282917022705,
|
| 1592 |
+
"learning_rate": 1.861573300705391e-05,
|
| 1593 |
+
"loss": 0.0044,
|
| 1594 |
+
"step": 2640
|
| 1595 |
+
},
|
| 1596 |
+
{
|
| 1597 |
+
"grad_norm": 0.11887385696172714,
|
| 1598 |
+
"learning_rate": 1.8537099360174137e-05,
|
| 1599 |
+
"loss": 0.0042,
|
| 1600 |
+
"step": 2650
|
| 1601 |
+
},
|
| 1602 |
+
{
|
| 1603 |
+
"grad_norm": 0.09578254073858261,
|
| 1604 |
+
"learning_rate": 1.845836265095852e-05,
|
| 1605 |
+
"loss": 0.0038,
|
| 1606 |
+
"step": 2660
|
| 1607 |
+
},
|
| 1608 |
+
{
|
| 1609 |
+
"grad_norm": 0.09519214928150177,
|
| 1610 |
+
"learning_rate": 1.837952517360022e-05,
|
| 1611 |
+
"loss": 0.0041,
|
| 1612 |
+
"step": 2670
|
| 1613 |
+
},
|
| 1614 |
+
{
|
| 1615 |
+
"grad_norm": 0.12270296365022659,
|
| 1616 |
+
"learning_rate": 1.8300589225228533e-05,
|
| 1617 |
+
"loss": 0.0044,
|
| 1618 |
+
"step": 2680
|
| 1619 |
+
},
|
| 1620 |
+
{
|
| 1621 |
+
"grad_norm": 0.11333918571472168,
|
| 1622 |
+
"learning_rate": 1.8221557105841958e-05,
|
| 1623 |
+
"loss": 0.0047,
|
| 1624 |
+
"step": 2690
|
| 1625 |
+
},
|
| 1626 |
+
{
|
| 1627 |
+
"grad_norm": 0.08164770901203156,
|
| 1628 |
+
"learning_rate": 1.814243111824119e-05,
|
| 1629 |
+
"loss": 0.0032,
|
| 1630 |
+
"step": 2700
|
| 1631 |
+
},
|
| 1632 |
+
{
|
| 1633 |
+
"grad_norm": 0.11921174079179764,
|
| 1634 |
+
"learning_rate": 1.8063213567961994e-05,
|
| 1635 |
+
"loss": 0.0037,
|
| 1636 |
+
"step": 2710
|
| 1637 |
+
},
|
| 1638 |
+
{
|
| 1639 |
+
"grad_norm": 0.09655310958623886,
|
| 1640 |
+
"learning_rate": 1.7983906763208055e-05,
|
| 1641 |
+
"loss": 0.0041,
|
| 1642 |
+
"step": 2720
|
| 1643 |
+
},
|
| 1644 |
+
{
|
| 1645 |
+
"grad_norm": 0.0791718065738678,
|
| 1646 |
+
"learning_rate": 1.7904513014783706e-05,
|
| 1647 |
+
"loss": 0.0042,
|
| 1648 |
+
"step": 2730
|
| 1649 |
+
},
|
| 1650 |
+
{
|
| 1651 |
+
"grad_norm": 0.11868743598461151,
|
| 1652 |
+
"learning_rate": 1.782503463602662e-05,
|
| 1653 |
+
"loss": 0.0038,
|
| 1654 |
+
"step": 2740
|
| 1655 |
+
},
|
| 1656 |
+
{
|
| 1657 |
+
"grad_norm": 0.08265439420938492,
|
| 1658 |
+
"learning_rate": 1.7745473942740373e-05,
|
| 1659 |
+
"loss": 0.0035,
|
| 1660 |
+
"step": 2750
|
| 1661 |
+
},
|
| 1662 |
+
{
|
| 1663 |
+
"grad_norm": 0.10162845999002457,
|
| 1664 |
+
"learning_rate": 1.7665833253126987e-05,
|
| 1665 |
+
"loss": 0.0033,
|
| 1666 |
+
"step": 2760
|
| 1667 |
+
},
|
| 1668 |
+
{
|
| 1669 |
+
"grad_norm": 0.10096478462219238,
|
| 1670 |
+
"learning_rate": 1.7586114887719378e-05,
|
| 1671 |
+
"loss": 0.0062,
|
| 1672 |
+
"step": 2770
|
| 1673 |
+
},
|
| 1674 |
+
{
|
| 1675 |
+
"grad_norm": 0.11772424727678299,
|
| 1676 |
+
"learning_rate": 1.750632116931374e-05,
|
| 1677 |
+
"loss": 0.0032,
|
| 1678 |
+
"step": 2780
|
| 1679 |
+
},
|
| 1680 |
+
{
|
| 1681 |
+
"grad_norm": 0.10858657211065292,
|
| 1682 |
+
"learning_rate": 1.7426454422901888e-05,
|
| 1683 |
+
"loss": 0.0035,
|
| 1684 |
+
"step": 2790
|
| 1685 |
+
},
|
| 1686 |
+
{
|
| 1687 |
+
"grad_norm": 0.11779221147298813,
|
| 1688 |
+
"learning_rate": 1.7346516975603466e-05,
|
| 1689 |
+
"loss": 0.0042,
|
| 1690 |
+
"step": 2800
|
| 1691 |
+
},
|
| 1692 |
+
{
|
| 1693 |
+
"grad_norm": 0.09140808880329132,
|
| 1694 |
+
"learning_rate": 1.7266511156598176e-05,
|
| 1695 |
+
"loss": 0.0043,
|
| 1696 |
+
"step": 2810
|
| 1697 |
+
},
|
| 1698 |
+
{
|
| 1699 |
+
"grad_norm": 0.08126810193061829,
|
| 1700 |
+
"learning_rate": 1.7186439297057906e-05,
|
| 1701 |
+
"loss": 0.0042,
|
| 1702 |
+
"step": 2820
|
| 1703 |
+
},
|
| 1704 |
+
{
|
| 1705 |
+
"grad_norm": 0.09498797357082367,
|
| 1706 |
+
"learning_rate": 1.7106303730078806e-05,
|
| 1707 |
+
"loss": 0.0037,
|
| 1708 |
+
"step": 2830
|
| 1709 |
+
},
|
| 1710 |
+
{
|
| 1711 |
+
"grad_norm": 0.08488073945045471,
|
| 1712 |
+
"learning_rate": 1.7026106790613295e-05,
|
| 1713 |
+
"loss": 0.0034,
|
| 1714 |
+
"step": 2840
|
| 1715 |
+
},
|
| 1716 |
+
{
|
| 1717 |
+
"grad_norm": 0.07819471508264542,
|
| 1718 |
+
"learning_rate": 1.6945850815402033e-05,
|
| 1719 |
+
"loss": 0.0043,
|
| 1720 |
+
"step": 2850
|
| 1721 |
+
},
|
| 1722 |
+
{
|
| 1723 |
+
"grad_norm": 0.13870303332805634,
|
| 1724 |
+
"learning_rate": 1.6865538142905846e-05,
|
| 1725 |
+
"loss": 0.0047,
|
| 1726 |
+
"step": 2860
|
| 1727 |
+
},
|
| 1728 |
+
{
|
| 1729 |
+
"grad_norm": 0.0990184098482132,
|
| 1730 |
+
"learning_rate": 1.6785171113237562e-05,
|
| 1731 |
+
"loss": 0.0048,
|
| 1732 |
+
"step": 2870
|
| 1733 |
+
},
|
| 1734 |
+
{
|
| 1735 |
+
"grad_norm": 0.05612372234463692,
|
| 1736 |
+
"learning_rate": 1.670475206809387e-05,
|
| 1737 |
+
"loss": 0.0045,
|
| 1738 |
+
"step": 2880
|
| 1739 |
+
},
|
| 1740 |
+
{
|
| 1741 |
+
"grad_norm": 0.11454035341739655,
|
| 1742 |
+
"learning_rate": 1.662428335068704e-05,
|
| 1743 |
+
"loss": 0.0041,
|
| 1744 |
+
"step": 2890
|
| 1745 |
+
},
|
| 1746 |
+
{
|
| 1747 |
+
"grad_norm": 0.09158965945243835,
|
| 1748 |
+
"learning_rate": 1.6543767305676663e-05,
|
| 1749 |
+
"loss": 0.0039,
|
| 1750 |
+
"step": 2900
|
| 1751 |
+
},
|
| 1752 |
+
{
|
| 1753 |
+
"grad_norm": 0.09644990414381027,
|
| 1754 |
+
"learning_rate": 1.6463206279101363e-05,
|
| 1755 |
+
"loss": 0.004,
|
| 1756 |
+
"step": 2910
|
| 1757 |
+
},
|
| 1758 |
+
{
|
| 1759 |
+
"grad_norm": 0.09871806204319,
|
| 1760 |
+
"learning_rate": 1.6382602618310396e-05,
|
| 1761 |
+
"loss": 0.0046,
|
| 1762 |
+
"step": 2920
|
| 1763 |
+
},
|
| 1764 |
+
{
|
| 1765 |
+
"grad_norm": 0.1133292093873024,
|
| 1766 |
+
"learning_rate": 1.6301958671895288e-05,
|
| 1767 |
+
"loss": 0.0057,
|
| 1768 |
+
"step": 2930
|
| 1769 |
+
},
|
| 1770 |
+
{
|
| 1771 |
+
"grad_norm": 0.15669667720794678,
|
| 1772 |
+
"learning_rate": 1.6221276789621377e-05,
|
| 1773 |
+
"loss": 0.0056,
|
| 1774 |
+
"step": 2940
|
| 1775 |
+
},
|
| 1776 |
+
{
|
| 1777 |
+
"grad_norm": 0.07493694126605988,
|
| 1778 |
+
"learning_rate": 1.6140559322359358e-05,
|
| 1779 |
+
"loss": 0.0047,
|
| 1780 |
+
"step": 2950
|
| 1781 |
+
},
|
| 1782 |
+
{
|
| 1783 |
+
"grad_norm": 0.14167147874832153,
|
| 1784 |
+
"learning_rate": 1.605980862201679e-05,
|
| 1785 |
+
"loss": 0.004,
|
| 1786 |
+
"step": 2960
|
| 1787 |
+
},
|
| 1788 |
+
{
|
| 1789 |
+
"grad_norm": 0.12322021275758743,
|
| 1790 |
+
"learning_rate": 1.5979027041469576e-05,
|
| 1791 |
+
"loss": 0.0038,
|
| 1792 |
+
"step": 2970
|
| 1793 |
+
},
|
| 1794 |
+
{
|
| 1795 |
+
"grad_norm": 0.1101364865899086,
|
| 1796 |
+
"learning_rate": 1.5898216934493365e-05,
|
| 1797 |
+
"loss": 0.0035,
|
| 1798 |
+
"step": 2980
|
| 1799 |
+
},
|
| 1800 |
+
{
|
| 1801 |
+
"grad_norm": 0.1453101634979248,
|
| 1802 |
+
"learning_rate": 1.5817380655695005e-05,
|
| 1803 |
+
"loss": 0.0034,
|
| 1804 |
+
"step": 2990
|
| 1805 |
+
},
|
| 1806 |
+
{
|
| 1807 |
+
"grad_norm": 0.09624999016523361,
|
| 1808 |
+
"learning_rate": 1.5736520560443933e-05,
|
| 1809 |
+
"loss": 0.0039,
|
| 1810 |
+
"step": 3000
|
| 1811 |
+
},
|
| 1812 |
+
{
|
| 1813 |
+
"grad_norm": 0.10005682706832886,
|
| 1814 |
+
"learning_rate": 1.5655639004803537e-05,
|
| 1815 |
+
"loss": 0.0036,
|
| 1816 |
+
"step": 3010
|
| 1817 |
+
},
|
| 1818 |
+
{
|
| 1819 |
+
"grad_norm": 0.11586767435073853,
|
| 1820 |
+
"learning_rate": 1.55747383454625e-05,
|
| 1821 |
+
"loss": 0.0038,
|
| 1822 |
+
"step": 3020
|
| 1823 |
+
},
|
| 1824 |
+
{
|
| 1825 |
+
"grad_norm": 0.06066455319523811,
|
| 1826 |
+
"learning_rate": 1.5493820939666146e-05,
|
| 1827 |
+
"loss": 0.0037,
|
| 1828 |
+
"step": 3030
|
| 1829 |
+
},
|
| 1830 |
+
{
|
| 1831 |
+
"grad_norm": 0.06761951744556427,
|
| 1832 |
+
"learning_rate": 1.541288914514775e-05,
|
| 1833 |
+
"loss": 0.0045,
|
| 1834 |
+
"step": 3040
|
| 1835 |
+
},
|
| 1836 |
+
{
|
| 1837 |
+
"grad_norm": 0.08163075149059296,
|
| 1838 |
+
"learning_rate": 1.533194532005983e-05,
|
| 1839 |
+
"loss": 0.0037,
|
| 1840 |
+
"step": 3050
|
| 1841 |
+
},
|
| 1842 |
+
{
|
| 1843 |
+
"grad_norm": 0.1107611283659935,
|
| 1844 |
+
"learning_rate": 1.5250991822905468e-05,
|
| 1845 |
+
"loss": 0.0036,
|
| 1846 |
+
"step": 3060
|
| 1847 |
+
},
|
| 1848 |
+
{
|
| 1849 |
+
"grad_norm": 0.12838762998580933,
|
| 1850 |
+
"learning_rate": 1.5170031012469538e-05,
|
| 1851 |
+
"loss": 0.004,
|
| 1852 |
+
"step": 3070
|
| 1853 |
+
},
|
| 1854 |
+
{
|
| 1855 |
+
"grad_norm": 0.10671990364789963,
|
| 1856 |
+
"learning_rate": 1.508906524775002e-05,
|
| 1857 |
+
"loss": 0.0044,
|
| 1858 |
+
"step": 3080
|
| 1859 |
+
},
|
| 1860 |
+
{
|
| 1861 |
+
"grad_norm": 0.1331837922334671,
|
| 1862 |
+
"learning_rate": 1.5008096887889245e-05,
|
| 1863 |
+
"loss": 0.0048,
|
| 1864 |
+
"step": 3090
|
| 1865 |
+
},
|
| 1866 |
+
{
|
| 1867 |
+
"grad_norm": 0.1242435947060585,
|
| 1868 |
+
"learning_rate": 1.492712829210516e-05,
|
| 1869 |
+
"loss": 0.0042,
|
| 1870 |
+
"step": 3100
|
| 1871 |
+
},
|
| 1872 |
+
{
|
| 1873 |
+
"grad_norm": 0.08521351218223572,
|
| 1874 |
+
"learning_rate": 1.4846161819622594e-05,
|
| 1875 |
+
"loss": 0.0038,
|
| 1876 |
+
"step": 3110
|
| 1877 |
+
},
|
| 1878 |
+
{
|
| 1879 |
+
"grad_norm": 0.07415805011987686,
|
| 1880 |
+
"learning_rate": 1.4765199829604498e-05,
|
| 1881 |
+
"loss": 0.0032,
|
| 1882 |
+
"step": 3120
|
| 1883 |
+
},
|
| 1884 |
+
{
|
| 1885 |
+
"grad_norm": 0.06467261910438538,
|
| 1886 |
+
"learning_rate": 1.468424468108322e-05,
|
| 1887 |
+
"loss": 0.0061,
|
| 1888 |
+
"step": 3130
|
| 1889 |
+
},
|
| 1890 |
+
{
|
| 1891 |
+
"grad_norm": 0.09693094342947006,
|
| 1892 |
+
"learning_rate": 1.4603298732891762e-05,
|
| 1893 |
+
"loss": 0.0038,
|
| 1894 |
+
"step": 3140
|
| 1895 |
+
},
|
| 1896 |
+
{
|
| 1897 |
+
"grad_norm": 0.05952545627951622,
|
| 1898 |
+
"learning_rate": 1.4522364343595054e-05,
|
| 1899 |
+
"loss": 0.0044,
|
| 1900 |
+
"step": 3150
|
| 1901 |
+
},
|
| 1902 |
+
{
|
| 1903 |
+
"grad_norm": 0.0716821476817131,
|
| 1904 |
+
"learning_rate": 1.4441443871421228e-05,
|
| 1905 |
+
"loss": 0.0034,
|
| 1906 |
+
"step": 3160
|
| 1907 |
+
},
|
| 1908 |
+
{
|
| 1909 |
+
"grad_norm": 0.08609385043382645,
|
| 1910 |
+
"learning_rate": 1.4360539674192899e-05,
|
| 1911 |
+
"loss": 0.0033,
|
| 1912 |
+
"step": 3170
|
| 1913 |
+
},
|
| 1914 |
+
{
|
| 1915 |
+
"grad_norm": 0.08336781710386276,
|
| 1916 |
+
"learning_rate": 1.4279654109258472e-05,
|
| 1917 |
+
"loss": 0.003,
|
| 1918 |
+
"step": 3180
|
| 1919 |
+
},
|
| 1920 |
+
{
|
| 1921 |
+
"grad_norm": 0.12197946757078171,
|
| 1922 |
+
"learning_rate": 1.4198789533423466e-05,
|
| 1923 |
+
"loss": 0.0039,
|
| 1924 |
+
"step": 3190
|
| 1925 |
+
},
|
| 1926 |
+
{
|
| 1927 |
+
"grad_norm": 0.09763512015342712,
|
| 1928 |
+
"learning_rate": 1.4117948302881811e-05,
|
| 1929 |
+
"loss": 0.0038,
|
| 1930 |
+
"step": 3200
|
| 1931 |
+
},
|
| 1932 |
+
{
|
| 1933 |
+
"grad_norm": 0.07868342101573944,
|
| 1934 |
+
"learning_rate": 1.403713277314723e-05,
|
| 1935 |
+
"loss": 0.0038,
|
| 1936 |
+
"step": 3210
|
| 1937 |
+
},
|
| 1938 |
+
{
|
| 1939 |
+
"grad_norm": 0.051395829766988754,
|
| 1940 |
+
"learning_rate": 1.3956345298984574e-05,
|
| 1941 |
+
"loss": 0.0028,
|
| 1942 |
+
"step": 3220
|
| 1943 |
+
},
|
| 1944 |
+
{
|
| 1945 |
+
"grad_norm": 0.11672254651784897,
|
| 1946 |
+
"learning_rate": 1.3875588234341233e-05,
|
| 1947 |
+
"loss": 0.0028,
|
| 1948 |
+
"step": 3230
|
| 1949 |
+
},
|
| 1950 |
+
{
|
| 1951 |
+
"grad_norm": 0.11190174520015717,
|
| 1952 |
+
"learning_rate": 1.3794863932278527e-05,
|
| 1953 |
+
"loss": 0.0041,
|
| 1954 |
+
"step": 3240
|
| 1955 |
+
},
|
| 1956 |
+
{
|
| 1957 |
+
"grad_norm": 0.09065545350313187,
|
| 1958 |
+
"learning_rate": 1.3714174744903172e-05,
|
| 1959 |
+
"loss": 0.0032,
|
| 1960 |
+
"step": 3250
|
| 1961 |
+
},
|
| 1962 |
+
{
|
| 1963 |
+
"grad_norm": 0.11228013038635254,
|
| 1964 |
+
"learning_rate": 1.3633523023298717e-05,
|
| 1965 |
+
"loss": 0.0046,
|
| 1966 |
+
"step": 3260
|
| 1967 |
+
},
|
| 1968 |
+
{
|
| 1969 |
+
"grad_norm": 0.06375761330127716,
|
| 1970 |
+
"learning_rate": 1.3552911117457055e-05,
|
| 1971 |
+
"loss": 0.0044,
|
| 1972 |
+
"step": 3270
|
| 1973 |
+
},
|
| 1974 |
+
{
|
| 1975 |
+
"grad_norm": 0.07220669835805893,
|
| 1976 |
+
"learning_rate": 1.347234137620994e-05,
|
| 1977 |
+
"loss": 0.0032,
|
| 1978 |
+
"step": 3280
|
| 1979 |
+
},
|
| 1980 |
+
{
|
| 1981 |
+
"grad_norm": 0.09625758230686188,
|
| 1982 |
+
"learning_rate": 1.3391816147160568e-05,
|
| 1983 |
+
"loss": 0.0056,
|
| 1984 |
+
"step": 3290
|
| 1985 |
+
},
|
| 1986 |
+
{
|
| 1987 |
+
"grad_norm": 0.08487574756145477,
|
| 1988 |
+
"learning_rate": 1.3311337776615146e-05,
|
| 1989 |
+
"loss": 0.0037,
|
| 1990 |
+
"step": 3300
|
| 1991 |
+
},
|
| 1992 |
+
{
|
| 1993 |
+
"grad_norm": 0.07956377416849136,
|
| 1994 |
+
"learning_rate": 1.3230908609514544e-05,
|
| 1995 |
+
"loss": 0.0031,
|
| 1996 |
+
"step": 3310
|
| 1997 |
+
},
|
| 1998 |
+
{
|
| 1999 |
+
"grad_norm": 0.11078336089849472,
|
| 2000 |
+
"learning_rate": 1.3150530989365966e-05,
|
| 2001 |
+
"loss": 0.0044,
|
| 2002 |
+
"step": 3320
|
| 2003 |
+
},
|
| 2004 |
+
{
|
| 2005 |
+
"grad_norm": 0.08106371760368347,
|
| 2006 |
+
"learning_rate": 1.307020725817467e-05,
|
| 2007 |
+
"loss": 0.0033,
|
| 2008 |
+
"step": 3330
|
| 2009 |
+
},
|
| 2010 |
+
{
|
| 2011 |
+
"grad_norm": 0.05347970128059387,
|
| 2012 |
+
"learning_rate": 1.2989939756375711e-05,
|
| 2013 |
+
"loss": 0.004,
|
| 2014 |
+
"step": 3340
|
| 2015 |
+
},
|
| 2016 |
+
{
|
| 2017 |
+
"grad_norm": 0.08384165912866592,
|
| 2018 |
+
"learning_rate": 1.290973082276577e-05,
|
| 2019 |
+
"loss": 0.003,
|
| 2020 |
+
"step": 3350
|
| 2021 |
+
},
|
| 2022 |
+
{
|
| 2023 |
+
"grad_norm": 0.09720702469348907,
|
| 2024 |
+
"learning_rate": 1.2829582794434987e-05,
|
| 2025 |
+
"loss": 0.0031,
|
| 2026 |
+
"step": 3360
|
| 2027 |
+
},
|
| 2028 |
+
{
|
| 2029 |
+
"grad_norm": 0.10105139762163162,
|
| 2030 |
+
"learning_rate": 1.2749498006698879e-05,
|
| 2031 |
+
"loss": 0.0043,
|
| 2032 |
+
"step": 3370
|
| 2033 |
+
},
|
| 2034 |
+
{
|
| 2035 |
+
"grad_norm": 0.08893254399299622,
|
| 2036 |
+
"learning_rate": 1.2669478793030287e-05,
|
| 2037 |
+
"loss": 0.0029,
|
| 2038 |
+
"step": 3380
|
| 2039 |
+
},
|
| 2040 |
+
{
|
| 2041 |
+
"grad_norm": 0.09673310816287994,
|
| 2042 |
+
"learning_rate": 1.2589527484991387e-05,
|
| 2043 |
+
"loss": 0.0025,
|
| 2044 |
+
"step": 3390
|
| 2045 |
+
},
|
| 2046 |
+
{
|
| 2047 |
+
"grad_norm": 0.12181838601827621,
|
| 2048 |
+
"learning_rate": 1.2509646412165748e-05,
|
| 2049 |
+
"loss": 0.0039,
|
| 2050 |
+
"step": 3400
|
| 2051 |
+
},
|
| 2052 |
+
{
|
| 2053 |
+
"grad_norm": 0.11474212259054184,
|
| 2054 |
+
"learning_rate": 1.2429837902090462e-05,
|
| 2055 |
+
"loss": 0.0027,
|
| 2056 |
+
"step": 3410
|
| 2057 |
+
},
|
| 2058 |
+
{
|
| 2059 |
+
"grad_norm": 0.0729747787117958,
|
| 2060 |
+
"learning_rate": 1.2350104280188316e-05,
|
| 2061 |
+
"loss": 0.0028,
|
| 2062 |
+
"step": 3420
|
| 2063 |
+
},
|
| 2064 |
+
{
|
| 2065 |
+
"grad_norm": 0.09193936735391617,
|
| 2066 |
+
"learning_rate": 1.2270447869700051e-05,
|
| 2067 |
+
"loss": 0.0035,
|
| 2068 |
+
"step": 3430
|
| 2069 |
+
},
|
| 2070 |
+
{
|
| 2071 |
+
"grad_norm": 0.07126328349113464,
|
| 2072 |
+
"learning_rate": 1.219087099161665e-05,
|
| 2073 |
+
"loss": 0.0043,
|
| 2074 |
+
"step": 3440
|
| 2075 |
+
},
|
| 2076 |
+
{
|
| 2077 |
+
"grad_norm": 0.09029316902160645,
|
| 2078 |
+
"learning_rate": 1.2111375964611726e-05,
|
| 2079 |
+
"loss": 0.0032,
|
| 2080 |
+
"step": 3450
|
| 2081 |
+
},
|
| 2082 |
+
{
|
| 2083 |
+
"grad_norm": 0.10429364442825317,
|
| 2084 |
+
"learning_rate": 1.2031965104973935e-05,
|
| 2085 |
+
"loss": 0.0046,
|
| 2086 |
+
"step": 3460
|
| 2087 |
+
},
|
| 2088 |
+
{
|
| 2089 |
+
"grad_norm": 0.08040885627269745,
|
| 2090 |
+
"learning_rate": 1.1952640726539533e-05,
|
| 2091 |
+
"loss": 0.0025,
|
| 2092 |
+
"step": 3470
|
| 2093 |
+
},
|
| 2094 |
+
{
|
| 2095 |
+
"grad_norm": 0.11755494773387909,
|
| 2096 |
+
"learning_rate": 1.1873405140624904e-05,
|
| 2097 |
+
"loss": 0.0039,
|
| 2098 |
+
"step": 3480
|
| 2099 |
+
},
|
| 2100 |
+
{
|
| 2101 |
+
"grad_norm": 0.12123305350542068,
|
| 2102 |
+
"learning_rate": 1.1794260655959247e-05,
|
| 2103 |
+
"loss": 0.0044,
|
| 2104 |
+
"step": 3490
|
| 2105 |
+
},
|
| 2106 |
+
{
|
| 2107 |
+
"grad_norm": 0.10822794586420059,
|
| 2108 |
+
"learning_rate": 1.1715209578617294e-05,
|
| 2109 |
+
"loss": 0.0037,
|
| 2110 |
+
"step": 3500
|
| 2111 |
+
},
|
| 2112 |
+
{
|
| 2113 |
+
"grad_norm": 0.065816231071949,
|
| 2114 |
+
"learning_rate": 1.1636254211952112e-05,
|
| 2115 |
+
"loss": 0.0028,
|
| 2116 |
+
"step": 3510
|
| 2117 |
+
},
|
| 2118 |
+
{
|
| 2119 |
+
"grad_norm": 0.06965353339910507,
|
| 2120 |
+
"learning_rate": 1.1557396856528013e-05,
|
| 2121 |
+
"loss": 0.0033,
|
| 2122 |
+
"step": 3520
|
| 2123 |
+
},
|
| 2124 |
+
{
|
| 2125 |
+
"grad_norm": 0.07491380721330643,
|
| 2126 |
+
"learning_rate": 1.1478639810053487e-05,
|
| 2127 |
+
"loss": 0.0044,
|
| 2128 |
+
"step": 3530
|
| 2129 |
+
},
|
| 2130 |
+
{
|
| 2131 |
+
"grad_norm": 0.08226829022169113,
|
| 2132 |
+
"learning_rate": 1.1399985367314274e-05,
|
| 2133 |
+
"loss": 0.0032,
|
| 2134 |
+
"step": 3540
|
| 2135 |
+
},
|
| 2136 |
+
{
|
| 2137 |
+
"grad_norm": 0.05845038592815399,
|
| 2138 |
+
"learning_rate": 1.1321435820106487e-05,
|
| 2139 |
+
"loss": 0.0026,
|
| 2140 |
+
"step": 3550
|
| 2141 |
+
},
|
| 2142 |
+
{
|
| 2143 |
+
"grad_norm": 0.11162237823009491,
|
| 2144 |
+
"learning_rate": 1.1242993457169867e-05,
|
| 2145 |
+
"loss": 0.003,
|
| 2146 |
+
"step": 3560
|
| 2147 |
+
},
|
| 2148 |
+
{
|
| 2149 |
+
"grad_norm": 0.13599853217601776,
|
| 2150 |
+
"learning_rate": 1.116466056412104e-05,
|
| 2151 |
+
"loss": 0.0033,
|
| 2152 |
+
"step": 3570
|
| 2153 |
+
},
|
| 2154 |
+
{
|
| 2155 |
+
"grad_norm": 0.09740614145994186,
|
| 2156 |
+
"learning_rate": 1.1086439423386972e-05,
|
| 2157 |
+
"loss": 0.0043,
|
| 2158 |
+
"step": 3580
|
| 2159 |
+
},
|
| 2160 |
+
{
|
| 2161 |
+
"grad_norm": 0.08986973017454147,
|
| 2162 |
+
"learning_rate": 1.1008332314138433e-05,
|
| 2163 |
+
"loss": 0.0026,
|
| 2164 |
+
"step": 3590
|
| 2165 |
+
},
|
| 2166 |
+
{
|
| 2167 |
+
"grad_norm": 0.09561492502689362,
|
| 2168 |
+
"learning_rate": 1.0930341512223596e-05,
|
| 2169 |
+
"loss": 0.0038,
|
| 2170 |
+
"step": 3600
|
| 2171 |
+
},
|
| 2172 |
+
{
|
| 2173 |
+
"grad_norm": 0.10061262547969818,
|
| 2174 |
+
"learning_rate": 1.0852469290101724e-05,
|
| 2175 |
+
"loss": 0.0034,
|
| 2176 |
+
"step": 3610
|
| 2177 |
+
},
|
| 2178 |
+
{
|
| 2179 |
+
"grad_norm": 0.07297629117965698,
|
| 2180 |
+
"learning_rate": 1.077471791677697e-05,
|
| 2181 |
+
"loss": 0.0025,
|
| 2182 |
+
"step": 3620
|
| 2183 |
+
},
|
| 2184 |
+
{
|
| 2185 |
+
"grad_norm": 0.10886093229055405,
|
| 2186 |
+
"learning_rate": 1.069708965773224e-05,
|
| 2187 |
+
"loss": 0.0026,
|
| 2188 |
+
"step": 3630
|
| 2189 |
+
},
|
| 2190 |
+
{
|
| 2191 |
+
"grad_norm": 0.102382592856884,
|
| 2192 |
+
"learning_rate": 1.0619586774863198e-05,
|
| 2193 |
+
"loss": 0.0031,
|
| 2194 |
+
"step": 3640
|
| 2195 |
+
},
|
| 2196 |
+
{
|
| 2197 |
+
"grad_norm": 0.08297941088676453,
|
| 2198 |
+
"learning_rate": 1.0542211526412351e-05,
|
| 2199 |
+
"loss": 0.0029,
|
| 2200 |
+
"step": 3650
|
| 2201 |
+
},
|
| 2202 |
+
{
|
| 2203 |
+
"grad_norm": 0.0679573193192482,
|
| 2204 |
+
"learning_rate": 1.0464966166903266e-05,
|
| 2205 |
+
"loss": 0.0035,
|
| 2206 |
+
"step": 3660
|
| 2207 |
+
},
|
| 2208 |
+
{
|
| 2209 |
+
"grad_norm": 0.0714852511882782,
|
| 2210 |
+
"learning_rate": 1.0387852947074858e-05,
|
| 2211 |
+
"loss": 0.0036,
|
| 2212 |
+
"step": 3670
|
| 2213 |
+
},
|
| 2214 |
+
{
|
| 2215 |
+
"grad_norm": 0.07032801955938339,
|
| 2216 |
+
"learning_rate": 1.0310874113815824e-05,
|
| 2217 |
+
"loss": 0.0026,
|
| 2218 |
+
"step": 3680
|
| 2219 |
+
},
|
| 2220 |
+
{
|
| 2221 |
+
"grad_norm": 0.0894472673535347,
|
| 2222 |
+
"learning_rate": 1.0234031910099153e-05,
|
| 2223 |
+
"loss": 0.0028,
|
| 2224 |
+
"step": 3690
|
| 2225 |
+
},
|
| 2226 |
+
{
|
| 2227 |
+
"grad_norm": 0.08256567269563675,
|
| 2228 |
+
"learning_rate": 1.0157328574916811e-05,
|
| 2229 |
+
"loss": 0.0047,
|
| 2230 |
+
"step": 3700
|
| 2231 |
+
},
|
| 2232 |
+
{
|
| 2233 |
+
"grad_norm": 0.08830638229846954,
|
| 2234 |
+
"learning_rate": 1.0080766343214462e-05,
|
| 2235 |
+
"loss": 0.0029,
|
| 2236 |
+
"step": 3710
|
| 2237 |
+
},
|
| 2238 |
+
{
|
| 2239 |
+
"grad_norm": 0.07382957637310028,
|
| 2240 |
+
"learning_rate": 1.0004347445826368e-05,
|
| 2241 |
+
"loss": 0.0034,
|
| 2242 |
+
"step": 3720
|
| 2243 |
+
},
|
| 2244 |
+
{
|
| 2245 |
+
"grad_norm": 0.0672360509634018,
|
| 2246 |
+
"learning_rate": 9.928074109410371e-06,
|
| 2247 |
+
"loss": 0.0036,
|
| 2248 |
+
"step": 3730
|
| 2249 |
+
},
|
| 2250 |
+
{
|
| 2251 |
+
"grad_norm": 0.062298908829689026,
|
| 2252 |
+
"learning_rate": 9.851948556383055e-06,
|
| 2253 |
+
"loss": 0.0034,
|
| 2254 |
+
"step": 3740
|
| 2255 |
+
},
|
| 2256 |
+
{
|
| 2257 |
+
"grad_norm": 0.054670773446559906,
|
| 2258 |
+
"learning_rate": 9.77597300485493e-06,
|
| 2259 |
+
"loss": 0.0031,
|
| 2260 |
+
"step": 3750
|
| 2261 |
+
},
|
| 2262 |
+
{
|
| 2263 |
+
"grad_norm": 0.0936974510550499,
|
| 2264 |
+
"learning_rate": 9.700149668565853e-06,
|
| 2265 |
+
"loss": 0.0037,
|
| 2266 |
+
"step": 3760
|
| 2267 |
+
},
|
| 2268 |
+
{
|
| 2269 |
+
"grad_norm": 0.06597235053777695,
|
| 2270 |
+
"learning_rate": 9.624480756820495e-06,
|
| 2271 |
+
"loss": 0.0031,
|
| 2272 |
+
"step": 3770
|
| 2273 |
+
},
|
| 2274 |
+
{
|
| 2275 |
+
"grad_norm": 0.06676555424928665,
|
| 2276 |
+
"learning_rate": 9.548968474423984e-06,
|
| 2277 |
+
"loss": 0.0023,
|
| 2278 |
+
"step": 3780
|
| 2279 |
+
},
|
| 2280 |
+
{
|
| 2281 |
+
"grad_norm": 0.08981256186962128,
|
| 2282 |
+
"learning_rate": 9.473615021617646e-06,
|
| 2283 |
+
"loss": 0.004,
|
| 2284 |
+
"step": 3790
|
| 2285 |
+
},
|
| 2286 |
+
{
|
| 2287 |
+
"grad_norm": 0.07909717410802841,
|
| 2288 |
+
"learning_rate": 9.398422594014917e-06,
|
| 2289 |
+
"loss": 0.004,
|
| 2290 |
+
"step": 3800
|
| 2291 |
+
},
|
| 2292 |
+
{
|
| 2293 |
+
"grad_norm": 0.05415402725338936,
|
| 2294 |
+
"learning_rate": 9.32339338253735e-06,
|
| 2295 |
+
"loss": 0.0022,
|
| 2296 |
+
"step": 3810
|
| 2297 |
+
},
|
| 2298 |
+
{
|
| 2299 |
+
"grad_norm": 0.05480325594544411,
|
| 2300 |
+
"learning_rate": 9.248529573350782e-06,
|
| 2301 |
+
"loss": 0.0027,
|
| 2302 |
+
"step": 3820
|
| 2303 |
+
},
|
| 2304 |
+
{
|
| 2305 |
+
"grad_norm": 0.058878351002931595,
|
| 2306 |
+
"learning_rate": 9.173833347801639e-06,
|
| 2307 |
+
"loss": 0.0037,
|
| 2308 |
+
"step": 3830
|
| 2309 |
+
},
|
| 2310 |
+
{
|
| 2311 |
+
"grad_norm": 0.06265062093734741,
|
| 2312 |
+
"learning_rate": 9.099306882353377e-06,
|
| 2313 |
+
"loss": 0.0027,
|
| 2314 |
+
"step": 3840
|
| 2315 |
+
},
|
| 2316 |
+
{
|
| 2317 |
+
"grad_norm": 0.0681389793753624,
|
| 2318 |
+
"learning_rate": 9.024952348523053e-06,
|
| 2319 |
+
"loss": 0.0026,
|
| 2320 |
+
"step": 3850
|
| 2321 |
+
},
|
| 2322 |
+
{
|
| 2323 |
+
"grad_norm": 0.08739670366048813,
|
| 2324 |
+
"learning_rate": 8.95077191281807e-06,
|
| 2325 |
+
"loss": 0.0032,
|
| 2326 |
+
"step": 3860
|
| 2327 |
+
},
|
| 2328 |
+
{
|
| 2329 |
+
"grad_norm": 0.07649361342191696,
|
| 2330 |
+
"learning_rate": 8.876767736673037e-06,
|
| 2331 |
+
"loss": 0.003,
|
| 2332 |
+
"step": 3870
|
| 2333 |
+
},
|
| 2334 |
+
{
|
| 2335 |
+
"grad_norm": 0.08005782216787338,
|
| 2336 |
+
"learning_rate": 8.802941976386793e-06,
|
| 2337 |
+
"loss": 0.0023,
|
| 2338 |
+
"step": 3880
|
| 2339 |
+
},
|
| 2340 |
+
{
|
| 2341 |
+
"grad_norm": 0.0897689238190651,
|
| 2342 |
+
"learning_rate": 8.72929678305959e-06,
|
| 2343 |
+
"loss": 0.0026,
|
| 2344 |
+
"step": 3890
|
| 2345 |
+
},
|
| 2346 |
+
{
|
| 2347 |
+
"grad_norm": 0.07536846399307251,
|
| 2348 |
+
"learning_rate": 8.655834302530391e-06,
|
| 2349 |
+
"loss": 0.0029,
|
| 2350 |
+
"step": 3900
|
| 2351 |
+
},
|
| 2352 |
+
{
|
| 2353 |
+
"grad_norm": 0.07567661255598068,
|
| 2354 |
+
"learning_rate": 8.58255667531437e-06,
|
| 2355 |
+
"loss": 0.0035,
|
| 2356 |
+
"step": 3910
|
| 2357 |
+
},
|
| 2358 |
+
{
|
| 2359 |
+
"grad_norm": 0.12472916394472122,
|
| 2360 |
+
"learning_rate": 8.509466036540531e-06,
|
| 2361 |
+
"loss": 0.005,
|
| 2362 |
+
"step": 3920
|
| 2363 |
+
},
|
| 2364 |
+
{
|
| 2365 |
+
"grad_norm": 0.06811542809009552,
|
| 2366 |
+
"learning_rate": 8.436564515889497e-06,
|
| 2367 |
+
"loss": 0.0025,
|
| 2368 |
+
"step": 3930
|
| 2369 |
+
},
|
| 2370 |
+
{
|
| 2371 |
+
"grad_norm": 0.049239810556173325,
|
| 2372 |
+
"learning_rate": 8.363854237531442e-06,
|
| 2373 |
+
"loss": 0.0025,
|
| 2374 |
+
"step": 3940
|
| 2375 |
+
},
|
| 2376 |
+
{
|
| 2377 |
+
"grad_norm": 0.050166867673397064,
|
| 2378 |
+
"learning_rate": 8.291337320064241e-06,
|
| 2379 |
+
"loss": 0.0024,
|
| 2380 |
+
"step": 3950
|
| 2381 |
+
},
|
| 2382 |
+
{
|
| 2383 |
+
"grad_norm": 0.05276251584291458,
|
| 2384 |
+
"learning_rate": 8.219015876451675e-06,
|
| 2385 |
+
"loss": 0.0033,
|
| 2386 |
+
"step": 3960
|
| 2387 |
+
},
|
| 2388 |
+
{
|
| 2389 |
+
"grad_norm": 0.08960229158401489,
|
| 2390 |
+
"learning_rate": 8.146892013961928e-06,
|
| 2391 |
+
"loss": 0.0028,
|
| 2392 |
+
"step": 3970
|
| 2393 |
+
},
|
| 2394 |
+
{
|
| 2395 |
+
"grad_norm": 0.110183484852314,
|
| 2396 |
+
"learning_rate": 8.074967834106153e-06,
|
| 2397 |
+
"loss": 0.006,
|
| 2398 |
+
"step": 3980
|
| 2399 |
+
},
|
| 2400 |
+
{
|
| 2401 |
+
"grad_norm": 0.11177948862314224,
|
| 2402 |
+
"learning_rate": 8.003245432577229e-06,
|
| 2403 |
+
"loss": 0.0056,
|
| 2404 |
+
"step": 3990
|
| 2405 |
+
},
|
| 2406 |
+
{
|
| 2407 |
+
"grad_norm": 0.05508233234286308,
|
| 2408 |
+
"learning_rate": 7.931726899188727e-06,
|
| 2409 |
+
"loss": 0.0027,
|
| 2410 |
+
"step": 4000
|
| 2411 |
+
},
|
| 2412 |
+
{
|
| 2413 |
+
"grad_norm": 0.052221231162548065,
|
| 2414 |
+
"learning_rate": 7.860414317814007e-06,
|
| 2415 |
+
"loss": 0.0029,
|
| 2416 |
+
"step": 4010
|
| 2417 |
+
},
|
| 2418 |
+
{
|
| 2419 |
+
"grad_norm": 0.06484582275152206,
|
| 2420 |
+
"learning_rate": 7.789309766325483e-06,
|
| 2421 |
+
"loss": 0.003,
|
| 2422 |
+
"step": 4020
|
| 2423 |
+
},
|
| 2424 |
+
{
|
| 2425 |
+
"grad_norm": 0.0864526629447937,
|
| 2426 |
+
"learning_rate": 7.718415316534107e-06,
|
| 2427 |
+
"loss": 0.0026,
|
| 2428 |
+
"step": 4030
|
| 2429 |
+
},
|
| 2430 |
+
{
|
| 2431 |
+
"grad_norm": 0.05630047991871834,
|
| 2432 |
+
"learning_rate": 7.64773303412897e-06,
|
| 2433 |
+
"loss": 0.0025,
|
| 2434 |
+
"step": 4040
|
| 2435 |
+
},
|
| 2436 |
+
{
|
| 2437 |
+
"grad_norm": 0.07152245938777924,
|
| 2438 |
+
"learning_rate": 7.5772649786171415e-06,
|
| 2439 |
+
"loss": 0.0026,
|
| 2440 |
+
"step": 4050
|
| 2441 |
+
},
|
| 2442 |
+
{
|
| 2443 |
+
"grad_norm": 0.07800120860338211,
|
| 2444 |
+
"learning_rate": 7.507013203263653e-06,
|
| 2445 |
+
"loss": 0.0026,
|
| 2446 |
+
"step": 4060
|
| 2447 |
+
},
|
| 2448 |
+
{
|
| 2449 |
+
"grad_norm": 0.062226876616477966,
|
| 2450 |
+
"learning_rate": 7.436979755031649e-06,
|
| 2451 |
+
"loss": 0.0024,
|
| 2452 |
+
"step": 4070
|
| 2453 |
+
},
|
| 2454 |
+
{
|
| 2455 |
+
"grad_norm": 0.08144303411245346,
|
| 2456 |
+
"learning_rate": 7.367166674522775e-06,
|
| 2457 |
+
"loss": 0.0026,
|
| 2458 |
+
"step": 4080
|
| 2459 |
+
},
|
| 2460 |
+
{
|
| 2461 |
+
"grad_norm": 0.049730513244867325,
|
| 2462 |
+
"learning_rate": 7.297575995917708e-06,
|
| 2463 |
+
"loss": 0.003,
|
| 2464 |
+
"step": 4090
|
| 2465 |
+
},
|
| 2466 |
+
{
|
| 2467 |
+
"grad_norm": 0.07160552591085434,
|
| 2468 |
+
"learning_rate": 7.228209746916862e-06,
|
| 2469 |
+
"loss": 0.0026,
|
| 2470 |
+
"step": 4100
|
| 2471 |
+
},
|
| 2472 |
+
{
|
| 2473 |
+
"grad_norm": 0.06243562698364258,
|
| 2474 |
+
"learning_rate": 7.159069948681363e-06,
|
| 2475 |
+
"loss": 0.0029,
|
| 2476 |
+
"step": 4110
|
| 2477 |
+
},
|
| 2478 |
+
{
|
| 2479 |
+
"grad_norm": 0.0679006278514862,
|
| 2480 |
+
"learning_rate": 7.090158615774082e-06,
|
| 2481 |
+
"loss": 0.0028,
|
| 2482 |
+
"step": 4120
|
| 2483 |
+
},
|
| 2484 |
+
{
|
| 2485 |
+
"grad_norm": 0.08142750710248947,
|
| 2486 |
+
"learning_rate": 7.021477756101007e-06,
|
| 2487 |
+
"loss": 0.0024,
|
| 2488 |
+
"step": 4130
|
| 2489 |
+
},
|
| 2490 |
+
{
|
| 2491 |
+
"grad_norm": 0.05774086341261864,
|
| 2492 |
+
"learning_rate": 6.953029370852677e-06,
|
| 2493 |
+
"loss": 0.0036,
|
| 2494 |
+
"step": 4140
|
| 2495 |
+
},
|
| 2496 |
+
{
|
| 2497 |
+
"grad_norm": 0.0558801144361496,
|
| 2498 |
+
"learning_rate": 6.884815454445923e-06,
|
| 2499 |
+
"loss": 0.0027,
|
| 2500 |
+
"step": 4150
|
| 2501 |
+
},
|
| 2502 |
+
{
|
| 2503 |
+
"grad_norm": 0.061070751398801804,
|
| 2504 |
+
"learning_rate": 6.816837994465731e-06,
|
| 2505 |
+
"loss": 0.0045,
|
| 2506 |
+
"step": 4160
|
| 2507 |
+
},
|
| 2508 |
+
{
|
| 2509 |
+
"grad_norm": 0.06962040066719055,
|
| 2510 |
+
"learning_rate": 6.749098971607311e-06,
|
| 2511 |
+
"loss": 0.0023,
|
| 2512 |
+
"step": 4170
|
| 2513 |
+
},
|
| 2514 |
+
{
|
| 2515 |
+
"grad_norm": 0.07268668711185455,
|
| 2516 |
+
"learning_rate": 6.681600359618428e-06,
|
| 2517 |
+
"loss": 0.0026,
|
| 2518 |
+
"step": 4180
|
| 2519 |
+
},
|
| 2520 |
+
{
|
| 2521 |
+
"grad_norm": 0.05980467423796654,
|
| 2522 |
+
"learning_rate": 6.614344125241862e-06,
|
| 2523 |
+
"loss": 0.0024,
|
| 2524 |
+
"step": 4190
|
| 2525 |
+
},
|
| 2526 |
+
{
|
| 2527 |
+
"grad_norm": 0.09890970587730408,
|
| 2528 |
+
"learning_rate": 6.547332228158094e-06,
|
| 2529 |
+
"loss": 0.0026,
|
| 2530 |
+
"step": 4200
|
| 2531 |
+
},
|
| 2532 |
+
{
|
| 2533 |
+
"grad_norm": 0.047991473227739334,
|
| 2534 |
+
"learning_rate": 6.480566620928245e-06,
|
| 2535 |
+
"loss": 0.0027,
|
| 2536 |
+
"step": 4210
|
| 2537 |
+
},
|
| 2538 |
+
{
|
| 2539 |
+
"grad_norm": 0.09556986391544342,
|
| 2540 |
+
"learning_rate": 6.41404924893713e-06,
|
| 2541 |
+
"loss": 0.0024,
|
| 2542 |
+
"step": 4220
|
| 2543 |
+
},
|
| 2544 |
+
{
|
| 2545 |
+
"grad_norm": 0.08843955397605896,
|
| 2546 |
+
"learning_rate": 6.347782050336625e-06,
|
| 2547 |
+
"loss": 0.0022,
|
| 2548 |
+
"step": 4230
|
| 2549 |
+
},
|
| 2550 |
+
{
|
| 2551 |
+
"grad_norm": 0.06960665434598923,
|
| 2552 |
+
"learning_rate": 6.281766955989165e-06,
|
| 2553 |
+
"loss": 0.0048,
|
| 2554 |
+
"step": 4240
|
| 2555 |
+
},
|
| 2556 |
+
{
|
| 2557 |
+
"grad_norm": 0.08095689862966537,
|
| 2558 |
+
"learning_rate": 6.2160058894114835e-06,
|
| 2559 |
+
"loss": 0.0028,
|
| 2560 |
+
"step": 4250
|
| 2561 |
+
},
|
| 2562 |
+
{
|
| 2563 |
+
"grad_norm": 0.06031321734189987,
|
| 2564 |
+
"learning_rate": 6.150500766718582e-06,
|
| 2565 |
+
"loss": 0.0027,
|
| 2566 |
+
"step": 4260
|
| 2567 |
+
},
|
| 2568 |
+
{
|
| 2569 |
+
"grad_norm": 0.06310360133647919,
|
| 2570 |
+
"learning_rate": 6.085253496567878e-06,
|
| 2571 |
+
"loss": 0.0019,
|
| 2572 |
+
"step": 4270
|
| 2573 |
+
},
|
| 2574 |
+
{
|
| 2575 |
+
"grad_norm": 0.06618604063987732,
|
| 2576 |
+
"learning_rate": 6.020265980103609e-06,
|
| 2577 |
+
"loss": 0.0027,
|
| 2578 |
+
"step": 4280
|
| 2579 |
+
},
|
| 2580 |
+
{
|
| 2581 |
+
"grad_norm": 0.058316055685281754,
|
| 2582 |
+
"learning_rate": 5.955540110901447e-06,
|
| 2583 |
+
"loss": 0.0028,
|
| 2584 |
+
"step": 4290
|
| 2585 |
+
},
|
| 2586 |
+
{
|
| 2587 |
+
"grad_norm": 0.05983656644821167,
|
| 2588 |
+
"learning_rate": 5.891077774913282e-06,
|
| 2589 |
+
"loss": 0.0025,
|
| 2590 |
+
"step": 4300
|
| 2591 |
+
},
|
| 2592 |
+
{
|
| 2593 |
+
"grad_norm": 0.05791456997394562,
|
| 2594 |
+
"learning_rate": 5.8268808504123185e-06,
|
| 2595 |
+
"loss": 0.003,
|
| 2596 |
+
"step": 4310
|
| 2597 |
+
},
|
| 2598 |
+
{
|
| 2599 |
+
"grad_norm": 0.04390927404165268,
|
| 2600 |
+
"learning_rate": 5.762951207938309e-06,
|
| 2601 |
+
"loss": 0.0026,
|
| 2602 |
+
"step": 4320
|
| 2603 |
+
},
|
| 2604 |
+
{
|
| 2605 |
+
"grad_norm": 0.05586939677596092,
|
| 2606 |
+
"learning_rate": 5.699290710243082e-06,
|
| 2607 |
+
"loss": 0.003,
|
| 2608 |
+
"step": 4330
|
| 2609 |
+
},
|
| 2610 |
+
{
|
| 2611 |
+
"grad_norm": 0.10256128013134003,
|
| 2612 |
+
"learning_rate": 5.635901212236251e-06,
|
| 2613 |
+
"loss": 0.0032,
|
| 2614 |
+
"step": 4340
|
| 2615 |
+
},
|
| 2616 |
+
{
|
| 2617 |
+
"grad_norm": 0.04283665493130684,
|
| 2618 |
+
"learning_rate": 5.5727845609311545e-06,
|
| 2619 |
+
"loss": 0.0025,
|
| 2620 |
+
"step": 4350
|
| 2621 |
+
},
|
| 2622 |
+
{
|
| 2623 |
+
"grad_norm": 0.05426718667149544,
|
| 2624 |
+
"learning_rate": 5.509942595391073e-06,
|
| 2625 |
+
"loss": 0.0021,
|
| 2626 |
+
"step": 4360
|
| 2627 |
+
},
|
| 2628 |
+
{
|
| 2629 |
+
"grad_norm": 0.08707068860530853,
|
| 2630 |
+
"learning_rate": 5.447377146675602e-06,
|
| 2631 |
+
"loss": 0.0029,
|
| 2632 |
+
"step": 4370
|
| 2633 |
+
},
|
| 2634 |
+
{
|
| 2635 |
+
"grad_norm": 0.10243280977010727,
|
| 2636 |
+
"learning_rate": 5.385090037787334e-06,
|
| 2637 |
+
"loss": 0.0029,
|
| 2638 |
+
"step": 4380
|
| 2639 |
+
},
|
| 2640 |
+
{
|
| 2641 |
+
"grad_norm": 0.05103703960776329,
|
| 2642 |
+
"learning_rate": 5.323083083618727e-06,
|
| 2643 |
+
"loss": 0.0033,
|
| 2644 |
+
"step": 4390
|
| 2645 |
+
},
|
| 2646 |
+
{
|
| 2647 |
+
"grad_norm": 0.060124531388282776,
|
| 2648 |
+
"learning_rate": 5.261358090899206e-06,
|
| 2649 |
+
"loss": 0.0033,
|
| 2650 |
+
"step": 4400
|
| 2651 |
+
},
|
| 2652 |
+
{
|
| 2653 |
+
"grad_norm": 0.10158316045999527,
|
| 2654 |
+
"learning_rate": 5.199916858142556e-06,
|
| 2655 |
+
"loss": 0.0033,
|
| 2656 |
+
"step": 4410
|
| 2657 |
+
},
|
| 2658 |
+
{
|
| 2659 |
+
"grad_norm": 0.07515929639339447,
|
| 2660 |
+
"learning_rate": 5.138761175594495e-06,
|
| 2661 |
+
"loss": 0.0025,
|
| 2662 |
+
"step": 4420
|
| 2663 |
+
},
|
| 2664 |
+
{
|
| 2665 |
+
"grad_norm": 0.07259644567966461,
|
| 2666 |
+
"learning_rate": 5.077892825180497e-06,
|
| 2667 |
+
"loss": 0.0029,
|
| 2668 |
+
"step": 4430
|
| 2669 |
+
},
|
| 2670 |
+
{
|
| 2671 |
+
"grad_norm": 0.07611123472452164,
|
| 2672 |
+
"learning_rate": 5.017313580453905e-06,
|
| 2673 |
+
"loss": 0.0025,
|
| 2674 |
+
"step": 4440
|
| 2675 |
+
},
|
| 2676 |
+
{
|
| 2677 |
+
"grad_norm": 0.08785836398601532,
|
| 2678 |
+
"learning_rate": 4.9570252065442274e-06,
|
| 2679 |
+
"loss": 0.0034,
|
| 2680 |
+
"step": 4450
|
| 2681 |
+
},
|
| 2682 |
+
{
|
| 2683 |
+
"grad_norm": 0.039934225380420685,
|
| 2684 |
+
"learning_rate": 4.89702946010572e-06,
|
| 2685 |
+
"loss": 0.002,
|
| 2686 |
+
"step": 4460
|
| 2687 |
+
},
|
| 2688 |
+
{
|
| 2689 |
+
"grad_norm": 0.07307283580303192,
|
| 2690 |
+
"learning_rate": 4.837328089266195e-06,
|
| 2691 |
+
"loss": 0.0024,
|
| 2692 |
+
"step": 4470
|
| 2693 |
+
},
|
| 2694 |
+
{
|
| 2695 |
+
"grad_norm": 0.0927189365029335,
|
| 2696 |
+
"learning_rate": 4.777922833576088e-06,
|
| 2697 |
+
"loss": 0.0034,
|
| 2698 |
+
"step": 4480
|
| 2699 |
+
},
|
| 2700 |
+
{
|
| 2701 |
+
"grad_norm": 0.04589124023914337,
|
| 2702 |
+
"learning_rate": 4.718815423957778e-06,
|
| 2703 |
+
"loss": 0.0039,
|
| 2704 |
+
"step": 4490
|
| 2705 |
+
},
|
| 2706 |
+
{
|
| 2707 |
+
"grad_norm": 0.0727127268910408,
|
| 2708 |
+
"learning_rate": 4.660007582655125e-06,
|
| 2709 |
+
"loss": 0.0031,
|
| 2710 |
+
"step": 4500
|
| 2711 |
+
},
|
| 2712 |
+
{
|
| 2713 |
+
"grad_norm": 0.07311692833900452,
|
| 2714 |
+
"learning_rate": 4.6015010231833275e-06,
|
| 2715 |
+
"loss": 0.0025,
|
| 2716 |
+
"step": 4510
|
| 2717 |
+
},
|
| 2718 |
+
{
|
| 2719 |
+
"grad_norm": 0.06421023607254028,
|
| 2720 |
+
"learning_rate": 4.5432974502789754e-06,
|
| 2721 |
+
"loss": 0.0025,
|
| 2722 |
+
"step": 4520
|
| 2723 |
+
},
|
| 2724 |
+
{
|
| 2725 |
+
"grad_norm": 0.04857269302010536,
|
| 2726 |
+
"learning_rate": 4.4853985598503625e-06,
|
| 2727 |
+
"loss": 0.0031,
|
| 2728 |
+
"step": 4530
|
| 2729 |
+
},
|
| 2730 |
+
{
|
| 2731 |
+
"grad_norm": 0.059340883046388626,
|
| 2732 |
+
"learning_rate": 4.427806038928109e-06,
|
| 2733 |
+
"loss": 0.0032,
|
| 2734 |
+
"step": 4540
|
| 2735 |
+
},
|
| 2736 |
+
{
|
| 2737 |
+
"grad_norm": 0.08296361565589905,
|
| 2738 |
+
"learning_rate": 4.370521565615962e-06,
|
| 2739 |
+
"loss": 0.0029,
|
| 2740 |
+
"step": 4550
|
| 2741 |
+
},
|
| 2742 |
+
{
|
| 2743 |
+
"grad_norm": 0.07902832329273224,
|
| 2744 |
+
"learning_rate": 4.313546809041942e-06,
|
| 2745 |
+
"loss": 0.0027,
|
| 2746 |
+
"step": 4560
|
| 2747 |
+
},
|
| 2748 |
+
{
|
| 2749 |
+
"grad_norm": 0.1092028021812439,
|
| 2750 |
+
"learning_rate": 4.256883429309685e-06,
|
| 2751 |
+
"loss": 0.0022,
|
| 2752 |
+
"step": 4570
|
| 2753 |
+
},
|
| 2754 |
+
{
|
| 2755 |
+
"grad_norm": 0.052551236003637314,
|
| 2756 |
+
"learning_rate": 4.2005330774500645e-06,
|
| 2757 |
+
"loss": 0.0035,
|
| 2758 |
+
"step": 4580
|
| 2759 |
+
},
|
| 2760 |
+
{
|
| 2761 |
+
"grad_norm": 0.05813279747962952,
|
| 2762 |
+
"learning_rate": 4.144497395373106e-06,
|
| 2763 |
+
"loss": 0.0031,
|
| 2764 |
+
"step": 4590
|
| 2765 |
+
},
|
| 2766 |
+
{
|
| 2767 |
+
"grad_norm": 0.0455247201025486,
|
| 2768 |
+
"learning_rate": 4.088778015820135e-06,
|
| 2769 |
+
"loss": 0.0031,
|
| 2770 |
+
"step": 4600
|
| 2771 |
+
},
|
| 2772 |
+
{
|
| 2773 |
+
"grad_norm": 0.0658392682671547,
|
| 2774 |
+
"learning_rate": 4.033376562316191e-06,
|
| 2775 |
+
"loss": 0.0033,
|
| 2776 |
+
"step": 4610
|
| 2777 |
+
},
|
| 2778 |
+
{
|
| 2779 |
+
"grad_norm": 0.06841374188661575,
|
| 2780 |
+
"learning_rate": 3.978294649122752e-06,
|
| 2781 |
+
"loss": 0.0026,
|
| 2782 |
+
"step": 4620
|
| 2783 |
+
},
|
| 2784 |
+
{
|
| 2785 |
+
"grad_norm": 0.06853021681308746,
|
| 2786 |
+
"learning_rate": 3.923533881190662e-06,
|
| 2787 |
+
"loss": 0.003,
|
| 2788 |
+
"step": 4630
|
| 2789 |
+
},
|
| 2790 |
+
{
|
| 2791 |
+
"grad_norm": 0.07420346140861511,
|
| 2792 |
+
"learning_rate": 3.8690958541134e-06,
|
| 2793 |
+
"loss": 0.003,
|
| 2794 |
+
"step": 4640
|
| 2795 |
+
},
|
| 2796 |
+
{
|
| 2797 |
+
"grad_norm": 0.06575186550617218,
|
| 2798 |
+
"learning_rate": 3.814982154080569e-06,
|
| 2799 |
+
"loss": 0.0034,
|
| 2800 |
+
"step": 4650
|
| 2801 |
+
},
|
| 2802 |
+
{
|
| 2803 |
+
"grad_norm": 0.07205885648727417,
|
| 2804 |
+
"learning_rate": 3.7611943578316824e-06,
|
| 2805 |
+
"loss": 0.0027,
|
| 2806 |
+
"step": 4660
|
| 2807 |
+
},
|
| 2808 |
+
{
|
| 2809 |
+
"grad_norm": 0.04615858942270279,
|
| 2810 |
+
"learning_rate": 3.707734032610228e-06,
|
| 2811 |
+
"loss": 0.003,
|
| 2812 |
+
"step": 4670
|
| 2813 |
+
},
|
| 2814 |
+
{
|
| 2815 |
+
"grad_norm": 0.07501405477523804,
|
| 2816 |
+
"learning_rate": 3.654602736117982e-06,
|
| 2817 |
+
"loss": 0.0035,
|
| 2818 |
+
"step": 4680
|
| 2819 |
+
},
|
| 2820 |
+
{
|
| 2821 |
+
"grad_norm": 0.09194950014352798,
|
| 2822 |
+
"learning_rate": 3.601802016469653e-06,
|
| 2823 |
+
"loss": 0.0042,
|
| 2824 |
+
"step": 4690
|
| 2825 |
+
},
|
| 2826 |
+
{
|
| 2827 |
+
"grad_norm": 0.04328564181923866,
|
| 2828 |
+
"learning_rate": 3.5493334121477544e-06,
|
| 2829 |
+
"loss": 0.0028,
|
| 2830 |
+
"step": 4700
|
| 2831 |
+
},
|
| 2832 |
+
{
|
| 2833 |
+
"grad_norm": 0.05529750511050224,
|
| 2834 |
+
"learning_rate": 3.49719845195777e-06,
|
| 2835 |
+
"loss": 0.003,
|
| 2836 |
+
"step": 4710
|
| 2837 |
+
},
|
| 2838 |
+
{
|
| 2839 |
+
"grad_norm": 0.10001294314861298,
|
| 2840 |
+
"learning_rate": 3.445398654983628e-06,
|
| 2841 |
+
"loss": 0.0034,
|
| 2842 |
+
"step": 4720
|
| 2843 |
+
},
|
| 2844 |
+
{
|
| 2845 |
+
"grad_norm": 0.06471127271652222,
|
| 2846 |
+
"learning_rate": 3.393935530543418e-06,
|
| 2847 |
+
"loss": 0.0027,
|
| 2848 |
+
"step": 4730
|
| 2849 |
+
},
|
| 2850 |
+
{
|
| 2851 |
+
"grad_norm": 0.05960933491587639,
|
| 2852 |
+
"learning_rate": 3.3428105781454364e-06,
|
| 2853 |
+
"loss": 0.0025,
|
| 2854 |
+
"step": 4740
|
| 2855 |
+
},
|
| 2856 |
+
{
|
| 2857 |
+
"grad_norm": 0.08707603067159653,
|
| 2858 |
+
"learning_rate": 3.2920252874444788e-06,
|
| 2859 |
+
"loss": 0.0043,
|
| 2860 |
+
"step": 4750
|
| 2861 |
+
},
|
| 2862 |
+
{
|
| 2863 |
+
"grad_norm": 0.08540203422307968,
|
| 2864 |
+
"learning_rate": 3.2415811381984264e-06,
|
| 2865 |
+
"loss": 0.0037,
|
| 2866 |
+
"step": 4760
|
| 2867 |
+
},
|
| 2868 |
+
{
|
| 2869 |
+
"grad_norm": 0.0713149905204773,
|
| 2870 |
+
"learning_rate": 3.191479600225159e-06,
|
| 2871 |
+
"loss": 0.0033,
|
| 2872 |
+
"step": 4770
|
| 2873 |
+
},
|
| 2874 |
+
{
|
| 2875 |
+
"grad_norm": 0.06586753576993942,
|
| 2876 |
+
"learning_rate": 3.1417221333596918e-06,
|
| 2877 |
+
"loss": 0.0032,
|
| 2878 |
+
"step": 4780
|
| 2879 |
+
},
|
| 2880 |
+
{
|
| 2881 |
+
"grad_norm": 0.07737500220537186,
|
| 2882 |
+
"learning_rate": 3.0923101874116756e-06,
|
| 2883 |
+
"loss": 0.0031,
|
| 2884 |
+
"step": 4790
|
| 2885 |
+
},
|
| 2886 |
+
{
|
| 2887 |
+
"grad_norm": 0.07109493762254715,
|
| 2888 |
+
"learning_rate": 3.0432452021231287e-06,
|
| 2889 |
+
"loss": 0.0022,
|
| 2890 |
+
"step": 4800
|
| 2891 |
+
},
|
| 2892 |
+
{
|
| 2893 |
+
"grad_norm": 0.07665184885263443,
|
| 2894 |
+
"learning_rate": 2.9945286071264845e-06,
|
| 2895 |
+
"loss": 0.0038,
|
| 2896 |
+
"step": 4810
|
| 2897 |
+
},
|
| 2898 |
+
{
|
| 2899 |
+
"grad_norm": 0.05713554471731186,
|
| 2900 |
+
"learning_rate": 2.9461618219029563e-06,
|
| 2901 |
+
"loss": 0.0028,
|
| 2902 |
+
"step": 4820
|
| 2903 |
+
},
|
| 2904 |
+
{
|
| 2905 |
+
"grad_norm": 0.06393369287252426,
|
| 2906 |
+
"learning_rate": 2.89814625574116e-06,
|
| 2907 |
+
"loss": 0.0029,
|
| 2908 |
+
"step": 4830
|
| 2909 |
+
},
|
| 2910 |
+
{
|
| 2911 |
+
"grad_norm": 0.06203538551926613,
|
| 2912 |
+
"learning_rate": 2.8504833076960567e-06,
|
| 2913 |
+
"loss": 0.0027,
|
| 2914 |
+
"step": 4840
|
| 2915 |
+
},
|
| 2916 |
+
{
|
| 2917 |
+
"grad_norm": 0.04352898150682449,
|
| 2918 |
+
"learning_rate": 2.803174366548188e-06,
|
| 2919 |
+
"loss": 0.0021,
|
| 2920 |
+
"step": 4850
|
| 2921 |
+
},
|
| 2922 |
+
{
|
| 2923 |
+
"grad_norm": 0.07385428994894028,
|
| 2924 |
+
"learning_rate": 2.756220810763202e-06,
|
| 2925 |
+
"loss": 0.0033,
|
| 2926 |
+
"step": 4860
|
| 2927 |
+
},
|
| 2928 |
+
{
|
| 2929 |
+
"grad_norm": 0.06604447215795517,
|
| 2930 |
+
"learning_rate": 2.709624008451707e-06,
|
| 2931 |
+
"loss": 0.0028,
|
| 2932 |
+
"step": 4870
|
| 2933 |
+
},
|
| 2934 |
+
{
|
| 2935 |
+
"grad_norm": 0.05847585201263428,
|
| 2936 |
+
"learning_rate": 2.6633853173293834e-06,
|
| 2937 |
+
"loss": 0.002,
|
| 2938 |
+
"step": 4880
|
| 2939 |
+
},
|
| 2940 |
+
{
|
| 2941 |
+
"grad_norm": 0.0321679562330246,
|
| 2942 |
+
"learning_rate": 2.6175060846774485e-06,
|
| 2943 |
+
"loss": 0.0024,
|
| 2944 |
+
"step": 4890
|
| 2945 |
+
},
|
| 2946 |
+
{
|
| 2947 |
+
"grad_norm": 0.035010624676942825,
|
| 2948 |
+
"learning_rate": 2.5719876473033825e-06,
|
| 2949 |
+
"loss": 0.0025,
|
| 2950 |
+
"step": 4900
|
| 2951 |
+
},
|
| 2952 |
+
{
|
| 2953 |
+
"grad_norm": 0.06074056029319763,
|
| 2954 |
+
"learning_rate": 2.5268313315019796e-06,
|
| 2955 |
+
"loss": 0.0024,
|
| 2956 |
+
"step": 4910
|
| 2957 |
+
},
|
| 2958 |
+
{
|
| 2959 |
+
"grad_norm": 0.04323519021272659,
|
| 2960 |
+
"learning_rate": 2.482038453016712e-06,
|
| 2961 |
+
"loss": 0.0038,
|
| 2962 |
+
"step": 4920
|
| 2963 |
+
},
|
| 2964 |
+
{
|
| 2965 |
+
"grad_norm": 0.06676605343818665,
|
| 2966 |
+
"learning_rate": 2.437610317001387e-06,
|
| 2967 |
+
"loss": 0.0044,
|
| 2968 |
+
"step": 4930
|
| 2969 |
+
},
|
| 2970 |
+
{
|
| 2971 |
+
"grad_norm": 0.05648193508386612,
|
| 2972 |
+
"learning_rate": 2.3935482179821067e-06,
|
| 2973 |
+
"loss": 0.0028,
|
| 2974 |
+
"step": 4940
|
| 2975 |
+
},
|
| 2976 |
+
{
|
| 2977 |
+
"grad_norm": 0.05948924645781517,
|
| 2978 |
+
"learning_rate": 2.349853439819576e-06,
|
| 2979 |
+
"loss": 0.0032,
|
| 2980 |
+
"step": 4950
|
| 2981 |
+
},
|
| 2982 |
+
{
|
| 2983 |
+
"grad_norm": 0.05742521211504936,
|
| 2984 |
+
"learning_rate": 2.3065272556716604e-06,
|
| 2985 |
+
"loss": 0.0043,
|
| 2986 |
+
"step": 4960
|
| 2987 |
+
},
|
| 2988 |
+
{
|
| 2989 |
+
"grad_norm": 0.07226058095693588,
|
| 2990 |
+
"learning_rate": 2.2635709279563204e-06,
|
| 2991 |
+
"loss": 0.0027,
|
| 2992 |
+
"step": 4970
|
| 2993 |
+
},
|
| 2994 |
+
{
|
| 2995 |
+
"grad_norm": 0.04860719293355942,
|
| 2996 |
+
"learning_rate": 2.2209857083148098e-06,
|
| 2997 |
+
"loss": 0.0033,
|
| 2998 |
+
"step": 4980
|
| 2999 |
+
},
|
| 3000 |
+
{
|
| 3001 |
+
"grad_norm": 0.06960118561983109,
|
| 3002 |
+
"learning_rate": 2.1787728375752042e-06,
|
| 3003 |
+
"loss": 0.0024,
|
| 3004 |
+
"step": 4990
|
| 3005 |
+
},
|
| 3006 |
+
{
|
| 3007 |
+
"grad_norm": 0.0939183309674263,
|
| 3008 |
+
"learning_rate": 2.136933545716257e-06,
|
| 3009 |
+
"loss": 0.0037,
|
| 3010 |
+
"step": 5000
|
| 3011 |
+
},
|
| 3012 |
+
{
|
| 3013 |
+
"grad_norm": 0.0371067188680172,
|
| 3014 |
+
"learning_rate": 2.095469051831561e-06,
|
| 3015 |
+
"loss": 0.0029,
|
| 3016 |
+
"step": 5010
|
| 3017 |
+
},
|
| 3018 |
+
{
|
| 3019 |
+
"grad_norm": 0.10542552173137665,
|
| 3020 |
+
"learning_rate": 2.0543805640940117e-06,
|
| 3021 |
+
"loss": 0.0034,
|
| 3022 |
+
"step": 5020
|
| 3023 |
+
},
|
| 3024 |
+
{
|
| 3025 |
+
"grad_norm": 0.08033856749534607,
|
| 3026 |
+
"learning_rate": 2.0136692797206276e-06,
|
| 3027 |
+
"loss": 0.0029,
|
| 3028 |
+
"step": 5030
|
| 3029 |
+
},
|
| 3030 |
+
{
|
| 3031 |
+
"grad_norm": 0.048041995614767075,
|
| 3032 |
+
"learning_rate": 1.973336384937634e-06,
|
| 3033 |
+
"loss": 0.0028,
|
| 3034 |
+
"step": 5040
|
| 3035 |
+
},
|
| 3036 |
+
{
|
| 3037 |
+
"grad_norm": 0.07258376479148865,
|
| 3038 |
+
"learning_rate": 1.9333830549459415e-06,
|
| 3039 |
+
"loss": 0.0021,
|
| 3040 |
+
"step": 5050
|
| 3041 |
+
},
|
| 3042 |
+
{
|
| 3043 |
+
"grad_norm": 0.07235071808099747,
|
| 3044 |
+
"learning_rate": 1.8938104538868595e-06,
|
| 3045 |
+
"loss": 0.0025,
|
| 3046 |
+
"step": 5060
|
| 3047 |
+
},
|
| 3048 |
+
{
|
| 3049 |
+
"grad_norm": 0.04660137742757797,
|
| 3050 |
+
"learning_rate": 1.854619734808209e-06,
|
| 3051 |
+
"loss": 0.002,
|
| 3052 |
+
"step": 5070
|
| 3053 |
+
},
|
| 3054 |
+
{
|
| 3055 |
+
"grad_norm": 0.06338319182395935,
|
| 3056 |
+
"learning_rate": 1.815812039630716e-06,
|
| 3057 |
+
"loss": 0.0022,
|
| 3058 |
+
"step": 5080
|
| 3059 |
+
},
|
| 3060 |
+
{
|
| 3061 |
+
"grad_norm": 0.06109411269426346,
|
| 3062 |
+
"learning_rate": 1.7773884991147243e-06,
|
| 3063 |
+
"loss": 0.0028,
|
| 3064 |
+
"step": 5090
|
| 3065 |
+
},
|
| 3066 |
+
{
|
| 3067 |
+
"grad_norm": 0.04851362481713295,
|
| 3068 |
+
"learning_rate": 1.7393502328272699e-06,
|
| 3069 |
+
"loss": 0.0022,
|
| 3070 |
+
"step": 5100
|
| 3071 |
+
},
|
| 3072 |
+
{
|
| 3073 |
+
"grad_norm": 0.08846496045589447,
|
| 3074 |
+
"learning_rate": 1.7016983491094513e-06,
|
| 3075 |
+
"loss": 0.0024,
|
| 3076 |
+
"step": 5110
|
| 3077 |
+
},
|
| 3078 |
+
{
|
| 3079 |
+
"grad_norm": 0.04170979931950569,
|
| 3080 |
+
"learning_rate": 1.6644339450441214e-06,
|
| 3081 |
+
"loss": 0.0024,
|
| 3082 |
+
"step": 5120
|
| 3083 |
+
},
|
| 3084 |
+
{
|
| 3085 |
+
"grad_norm": 0.04883277043700218,
|
| 3086 |
+
"learning_rate": 1.6275581064239497e-06,
|
| 3087 |
+
"loss": 0.0027,
|
| 3088 |
+
"step": 5130
|
| 3089 |
+
},
|
| 3090 |
+
{
|
| 3091 |
+
"grad_norm": 0.09936371445655823,
|
| 3092 |
+
"learning_rate": 1.5910719077197516e-06,
|
| 3093 |
+
"loss": 0.0032,
|
| 3094 |
+
"step": 5140
|
| 3095 |
+
},
|
| 3096 |
+
{
|
| 3097 |
+
"grad_norm": 0.028741255402565002,
|
| 3098 |
+
"learning_rate": 1.554976412049211e-06,
|
| 3099 |
+
"loss": 0.0025,
|
| 3100 |
+
"step": 5150
|
| 3101 |
+
},
|
| 3102 |
+
{
|
| 3103 |
+
"grad_norm": 0.03936922922730446,
|
| 3104 |
+
"learning_rate": 1.5192726711458899e-06,
|
| 3105 |
+
"loss": 0.0032,
|
| 3106 |
+
"step": 5160
|
| 3107 |
+
},
|
| 3108 |
+
{
|
| 3109 |
+
"grad_norm": 0.03335015848278999,
|
| 3110 |
+
"learning_rate": 1.4839617253285758e-06,
|
| 3111 |
+
"loss": 0.0019,
|
| 3112 |
+
"step": 5170
|
| 3113 |
+
},
|
| 3114 |
+
{
|
| 3115 |
+
"grad_norm": 0.03196994587779045,
|
| 3116 |
+
"learning_rate": 1.449044603470987e-06,
|
| 3117 |
+
"loss": 0.0018,
|
| 3118 |
+
"step": 5180
|
| 3119 |
+
},
|
| 3120 |
+
{
|
| 3121 |
+
"grad_norm": 0.0717940479516983,
|
| 3122 |
+
"learning_rate": 1.414522322971772e-06,
|
| 3123 |
+
"loss": 0.0026,
|
| 3124 |
+
"step": 5190
|
| 3125 |
+
},
|
| 3126 |
+
{
|
| 3127 |
+
"grad_norm": 0.025321830064058304,
|
| 3128 |
+
"learning_rate": 1.3803958897249003e-06,
|
| 3129 |
+
"loss": 0.0025,
|
| 3130 |
+
"step": 5200
|
| 3131 |
+
},
|
| 3132 |
+
{
|
| 3133 |
+
"grad_norm": 0.054642051458358765,
|
| 3134 |
+
"learning_rate": 1.3466662980903065e-06,
|
| 3135 |
+
"loss": 0.0026,
|
| 3136 |
+
"step": 5210
|
| 3137 |
+
},
|
| 3138 |
+
{
|
| 3139 |
+
"grad_norm": 0.03861148655414581,
|
| 3140 |
+
"learning_rate": 1.313334530864953e-06,
|
| 3141 |
+
"loss": 0.0027,
|
| 3142 |
+
"step": 5220
|
| 3143 |
+
},
|
| 3144 |
+
{
|
| 3145 |
+
"grad_norm": 0.07443683594465256,
|
| 3146 |
+
"learning_rate": 1.2804015592541856e-06,
|
| 3147 |
+
"loss": 0.0019,
|
| 3148 |
+
"step": 5230
|
| 3149 |
+
},
|
| 3150 |
+
{
|
| 3151 |
+
"grad_norm": 0.04590648040175438,
|
| 3152 |
+
"learning_rate": 1.247868342843419e-06,
|
| 3153 |
+
"loss": 0.0028,
|
| 3154 |
+
"step": 5240
|
| 3155 |
+
},
|
| 3156 |
+
{
|
| 3157 |
+
"grad_norm": 0.02897215634584427,
|
| 3158 |
+
"learning_rate": 1.2157358295701981e-06,
|
| 3159 |
+
"loss": 0.0026,
|
| 3160 |
+
"step": 5250
|
| 3161 |
+
},
|
| 3162 |
+
{
|
| 3163 |
+
"grad_norm": 0.043594203889369965,
|
| 3164 |
+
"learning_rate": 1.1840049556965693e-06,
|
| 3165 |
+
"loss": 0.0043,
|
| 3166 |
+
"step": 5260
|
| 3167 |
+
},
|
| 3168 |
+
{
|
| 3169 |
+
"grad_norm": 0.0517614483833313,
|
| 3170 |
+
"learning_rate": 1.1526766457817916e-06,
|
| 3171 |
+
"loss": 0.0025,
|
| 3172 |
+
"step": 5270
|
| 3173 |
+
},
|
| 3174 |
+
{
|
| 3175 |
+
"grad_norm": 0.07484445720911026,
|
| 3176 |
+
"learning_rate": 1.1217518126554156e-06,
|
| 3177 |
+
"loss": 0.0022,
|
| 3178 |
+
"step": 5280
|
| 3179 |
+
},
|
| 3180 |
+
{
|
| 3181 |
+
"grad_norm": 0.03570960462093353,
|
| 3182 |
+
"learning_rate": 1.0912313573906617e-06,
|
| 3183 |
+
"loss": 0.0019,
|
| 3184 |
+
"step": 5290
|
| 3185 |
+
},
|
| 3186 |
+
{
|
| 3187 |
+
"grad_norm": 0.05601223558187485,
|
| 3188 |
+
"learning_rate": 1.0611161692781885e-06,
|
| 3189 |
+
"loss": 0.0023,
|
| 3190 |
+
"step": 5300
|
| 3191 |
+
},
|
| 3192 |
+
{
|
| 3193 |
+
"grad_norm": 0.021517347544431686,
|
| 3194 |
+
"learning_rate": 1.0314071258001722e-06,
|
| 3195 |
+
"loss": 0.0021,
|
| 3196 |
+
"step": 5310
|
| 3197 |
+
},
|
| 3198 |
+
{
|
| 3199 |
+
"grad_norm": 0.03928108140826225,
|
| 3200 |
+
"learning_rate": 1.0021050926047315e-06,
|
| 3201 |
+
"loss": 0.002,
|
| 3202 |
+
"step": 5320
|
| 3203 |
+
},
|
| 3204 |
+
{
|
| 3205 |
+
"grad_norm": 0.032030925154685974,
|
| 3206 |
+
"learning_rate": 9.732109234807124e-07,
|
| 3207 |
+
"loss": 0.0025,
|
| 3208 |
+
"step": 5330
|
| 3209 |
+
},
|
| 3210 |
+
{
|
| 3211 |
+
"grad_norm": 0.050179291516542435,
|
| 3212 |
+
"learning_rate": 9.447254603328142e-07,
|
| 3213 |
+
"loss": 0.0029,
|
| 3214 |
+
"step": 5340
|
| 3215 |
+
},
|
| 3216 |
+
{
|
| 3217 |
+
"grad_norm": 0.02549734152853489,
|
| 3218 |
+
"learning_rate": 9.166495331570491e-07,
|
| 3219 |
+
"loss": 0.0018,
|
| 3220 |
+
"step": 5350
|
| 3221 |
+
},
|
| 3222 |
+
{
|
| 3223 |
+
"grad_norm": 0.04321710392832756,
|
| 3224 |
+
"learning_rate": 8.889839600165656e-07,
|
| 3225 |
+
"loss": 0.0025,
|
| 3226 |
+
"step": 5360
|
| 3227 |
+
},
|
| 3228 |
+
{
|
| 3229 |
+
"grad_norm": 0.06186482310295105,
|
| 3230 |
+
"learning_rate": 8.617295470178039e-07,
|
| 3231 |
+
"loss": 0.0023,
|
| 3232 |
+
"step": 5370
|
| 3233 |
+
},
|
| 3234 |
+
{
|
| 3235 |
+
"grad_norm": 0.04183085635304451,
|
| 3236 |
+
"learning_rate": 8.348870882870269e-07,
|
| 3237 |
+
"loss": 0.002,
|
| 3238 |
+
"step": 5380
|
| 3239 |
+
},
|
| 3240 |
+
{
|
| 3241 |
+
"grad_norm": 0.019996939226984978,
|
| 3242 |
+
"learning_rate": 8.084573659471512e-07,
|
| 3243 |
+
"loss": 0.0019,
|
| 3244 |
+
"step": 5390
|
| 3245 |
+
},
|
| 3246 |
+
{
|
| 3247 |
+
"grad_norm": 0.06442452222108841,
|
| 3248 |
+
"learning_rate": 7.82441150094988e-07,
|
| 3249 |
+
"loss": 0.0037,
|
| 3250 |
+
"step": 5400
|
| 3251 |
+
},
|
| 3252 |
+
{
|
| 3253 |
+
"grad_norm": 0.077437624335289,
|
| 3254 |
+
"learning_rate": 7.568391987787849e-07,
|
| 3255 |
+
"loss": 0.0023,
|
| 3256 |
+
"step": 5410
|
| 3257 |
+
},
|
| 3258 |
+
{
|
| 3259 |
+
"grad_norm": 0.027246888726949692,
|
| 3260 |
+
"learning_rate": 7.316522579761431e-07,
|
| 3261 |
+
"loss": 0.0021,
|
| 3262 |
+
"step": 5420
|
| 3263 |
+
},
|
| 3264 |
+
{
|
| 3265 |
+
"grad_norm": 0.0590815469622612,
|
| 3266 |
+
"learning_rate": 7.068810615722887e-07,
|
| 3267 |
+
"loss": 0.0024,
|
| 3268 |
+
"step": 5430
|
| 3269 |
+
},
|
| 3270 |
+
{
|
| 3271 |
+
"grad_norm": 0.03910129517316818,
|
| 3272 |
+
"learning_rate": 6.825263313386793e-07,
|
| 3273 |
+
"loss": 0.0018,
|
| 3274 |
+
"step": 5440
|
| 3275 |
+
},
|
| 3276 |
+
{
|
| 3277 |
+
"grad_norm": 0.04055808112025261,
|
| 3278 |
+
"learning_rate": 6.585887769119758e-07,
|
| 3279 |
+
"loss": 0.0022,
|
| 3280 |
+
"step": 5450
|
| 3281 |
+
},
|
| 3282 |
+
{
|
| 3283 |
+
"grad_norm": 0.032348066568374634,
|
| 3284 |
+
"learning_rate": 6.350690957733712e-07,
|
| 3285 |
+
"loss": 0.0017,
|
| 3286 |
+
"step": 5460
|
| 3287 |
+
},
|
| 3288 |
+
{
|
| 3289 |
+
"grad_norm": 0.028897378593683243,
|
| 3290 |
+
"learning_rate": 6.119679732282596e-07,
|
| 3291 |
+
"loss": 0.0022,
|
| 3292 |
+
"step": 5470
|
| 3293 |
+
},
|
| 3294 |
+
{
|
| 3295 |
+
"grad_norm": 0.03753242269158363,
|
| 3296 |
+
"learning_rate": 5.89286082386274e-07,
|
| 3297 |
+
"loss": 0.002,
|
| 3298 |
+
"step": 5480
|
| 3299 |
+
},
|
| 3300 |
+
{
|
| 3301 |
+
"grad_norm": 0.037839531898498535,
|
| 3302 |
+
"learning_rate": 5.670240841416718e-07,
|
| 3303 |
+
"loss": 0.0022,
|
| 3304 |
+
"step": 5490
|
| 3305 |
+
},
|
| 3306 |
+
{
|
| 3307 |
+
"grad_norm": 0.05074967443943024,
|
| 3308 |
+
"learning_rate": 5.451826271540744e-07,
|
| 3309 |
+
"loss": 0.0032,
|
| 3310 |
+
"step": 5500
|
| 3311 |
+
}
|
| 3312 |
+
],
|
| 3313 |
+
"logging_steps": 10,
|
| 3314 |
+
"max_steps": 6000,
|
| 3315 |
+
"num_input_tokens_seen": 0,
|
| 3316 |
+
"num_train_epochs": 9223372036854775807,
|
| 3317 |
+
"save_steps": 250,
|
| 3318 |
+
"stateful_callbacks": {
|
| 3319 |
+
"TrainerControl": {
|
| 3320 |
+
"args": {
|
| 3321 |
+
"should_epoch_stop": false,
|
| 3322 |
+
"should_evaluate": false,
|
| 3323 |
+
"should_log": false,
|
| 3324 |
+
"should_save": true,
|
| 3325 |
+
"should_training_stop": false
|
| 3326 |
+
},
|
| 3327 |
+
"attributes": {}
|
| 3328 |
+
}
|
| 3329 |
+
},
|
| 3330 |
+
"total_flos": 0.0,
|
| 3331 |
+
"train_batch_size": 64,
|
| 3332 |
+
"trial_name": null,
|
| 3333 |
+
"trial_params": null
|
| 3334 |
+
}
|
checkpoint-5500/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0801163b8198b3ceaf7a15d9a0802ae791a344f1974e4b729b15632016a2cb11
|
| 3 |
+
size 7825
|
checkpoint-5500/wandb_config.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"project": "screwing_hil", "run_id": "groot_dagger_it1"}
|