Update lingbotvla_cli.yaml

lingbotvla_cli.yaml

@@ -4,24 +4,22 @@ data:
   - camera_wrist_left
   - camera_wrist_right
   chat_template: default
-  data_name: multi
-  data_root: null
-  data_tag: default
+  data_name: robotwin
   data_type: conversation
   dataloader_type: native
   datasets_type: vla
   drop_last: true
-  image_keys: images
+  img_size: 224
   joints:
   - '{''arm.position'': 14}'
   - '{''effector.position'': 2}'
-  max_seq_len: 64
-  norm_type: bounds_99_woclip
-  num_workers: 20
+  max_seq_len: 2048
+  norm_stats_file: assets/norm_stats/robotwin_50.json
+  norm_type: bounds_99
+  num_workers: 8
   pin_memory: true
   prefetch_factor: 4
-  robot_config_root: ./configs/robot_configs
-  source_name: null
+  robot_config_root: configs/robot_configs
   text_keys: messages
   train_path: ./assets/training_data/robotwin_all_tasks_with_aug.txt
   train_size: 8000000
@@ -30,17 +28,12 @@ model:
   attn_implementation: flash_attention_2
   basic_modules: []
   decoders: {}
-  depth_incremental_training: false
-  encode_target: false
   encoders: {}
   final_norm_adanorm: false
   force_use_huggingface: false
-  incremental_training: false
-  input_encoder: encoder
-  moe_implementation: null
+  moge_path: null
+  morgbd_path: null
   nosplit_gate_liner: false
-  old_adanorm: true
-  output_encoder: decoder
   post_training: true
   separate_time_proj: false
   split_gate_liner: false
@@ -48,8 +41,9 @@ model:
   vlm_repo_id: null
   vocab_size: 0
 train:
-  action_dim: 14
+  action_dim: null
   activation_gpu_limit: 0.0
+  adapt_to_pi_aloha: false
   align_params: {}
   bsz_warmup_init_mbtoken: 200
   bsz_warmup_ratio: 0
@@ -63,8 +57,6 @@ train:
   dyn_bsz: true
   dyn_bsz_buffer_size: 200
   dyn_bsz_margin: 0
-  ema_rate: 0.9999
-  embodiment_name: null
   empty_cache_steps: 500
   enable_activation_offload: false
   enable_expert_vision: false
@@ -74,7 +66,6 @@ train:
   enable_full_determinism: false
   enable_full_shard: false
   enable_gradient_checkpointing: true
-  enable_manual_eager: false
   enable_mixed_precision: true
   enable_profiling: false
   enable_reentrant: false
@@ -85,11 +76,9 @@ train:
   freeze_vision_encoder: false
   freeze_vit: false
   global_batch_size: 256
-  ignore_depth: false
   init_device: cuda
-  interact_layer_end: null
-  interact_layer_start: null
   load_checkpoint_path: null
+  loss_type: L1_fm
   lr: 0.0001
   lr_decay_ratio: 1.0
   lr_decay_style: constant
@@ -102,42 +91,39 @@ train:
   max_steps: 220000
   micro_batch_size: 32
   module_fsdp_enable: true
-  my_tokenizer_max_length: 72
   norm_qkv: false
-  num_train_epochs: 29000
+  num_steps: 10
+  num_train_epochs: 69
   optimizer: adamw
   pipeline_parallel_size: 1
-  pre_train: false
   profile_end_step: 2
   profile_profile_memory: true
   profile_record_shapes: true
   profile_start_step: 1
   profile_trace_dir: ./trace
   profile_with_stack: true
-  qwenvl_bos: true
+  resize_imgs_with_padding:
+  - 224
+  - 224
   resume_dataloader_state: true
   rmpad: false
   rmpad_with_pos_ids: false
-  save_epochs: 29000
+  save_epochs: 23
   save_hf_weights: true
   save_steps: 10000
   seed: 42
-  skip_max_norm: false
   stable_train_steps: 100000
   tensor_parallel_size: 1
-  tokenizer_max_length: 24
+  tokenizer_max_length: 72
+  train_expert_only: false
+  train_state_proj: true
   ulysses_parallel_size: 1
+  use_cache: true
   use_compile: true
-  use_doptim: false
-  use_ema: false
-  use_fast_action: false
-  use_ki: false
-  use_prompt: false
-  use_state: false
-  use_subtask: false
+  use_delta_joint_actions_aloha: false
   use_wandb: false
   vit_lr: 1.0e-06
   vlm_causal: false
-  wandb_name: pi0_libero
-  wandb_project: pi0_libero
-  weight_decay: 0
+  wandb_name: null
+  wandb_project: LingBotVLA
+  weight_decay: 0