Add files using upload-large-folder tool

config.yaml

@@ -0,0 +1,68 @@
+datasets:
+  vla_data:
+    CoT_prompt: Your task is {instruction}. To identify the key objects for your task.
+      Locate their bounding boxes in [x1,y1,x2,y2] format.
+    data_mix: libero_plus_all
+    data_root_dir: /home/Sylvest
+    dataset_py: lerobot_datasets
+    per_device_batch_size: 4
+    video_backend: torchvision_av
+framework:
+  action_model:
+    action_dim: 7
+    action_horizon: 10
+    action_model_type: DiT-B
+    add_pos_embed: true
+    diffusion_model_cfg:
+      cross_attention_dim: 2048
+      dropout: 0.2
+      final_dropout: true
+      interleave_self_attention: true
+      norm_type: ada_norm
+      num_layers: 16
+      output_dim: 1024
+      positional_embeddings: null
+    future_action_window_size: 9
+    hidden_size: 1024
+    max_action_dim: 32
+    max_seq_len: 1024
+    noise_beta_alpha: 1.5
+    noise_beta_beta: 1.0
+    noise_s: 0.999
+    num_inference_timesteps: 4
+    num_target_vision_tokens: 3
+    num_timestep_buckets: 1000
+    past_action_window_size: 0
+    state_dim: 7
+  name: QwenGR00T
+  qwenvl:
+    base_vlm: /root/.cache/huggingface/hub/models--StarVLA--Qwen2.5-VL-3B-Instruct-Action/snapshots/01f25c3e057346fda67bdd17fba264d8e31d88b0
+output_dir: ./results/QwenGR00T/exp_c4
+run_id: exp_c4
+run_root_dir: ./results/QwenGR00T
+seed: 42
+trainer:
+  eval_interval: 100
+  freeze_modules: true
+  gradient_accumulation_steps: 4
+  gradient_clipping: 1.0
+  is_resume: false
+  learning_rate:
+    action_model: 0.0001
+    base: 2.5e-05
+    qwen_vl_interface: 1.0e-05
+  logging_frequency: 100
+  lr_scheduler_type: cosine_with_min_lr
+  max_train_steps: 40000
+  num_warmup_steps: 5000
+  optimizer:
+    betas:
+    - 0.9
+    - 0.95
+    eps: 1.0e-08
+    weight_decay: 1.0e-08
+  save_interval: 10000
+  scheduler_specific_kwargs:
+    min_lr: 1.0e-06
+wandb_entity: xxx
+wandb_project: starVLA_Libero

dataset_statistics.json

@@ -0,0 +1,133 @@
+{
+  "franka": {
+    "action": {
+      "mean": [
+        0.07236806908622384,
+        0.09041699254885316,
+        -0.1012995345517993,
+        -0.000578228005906567,
+        0.006175444577820599,
+        -0.002866805996745825,
+        -0.043408477678895
+      ],
+      "std": [
+        0.34934414734580516,
+        0.37898891540933966,
+        0.46305894119710334,
+        0.04004100987887957,
+        0.06589697614744162,
+        0.076847854462235,
+        1.000541573605827
+      ],
+      "max": [
+        0.9375,
+        0.9375,
+        0.9375,
+        0.3557142913341522,
+        0.375,
+        0.375,
+        1.0
+      ],
+      "min": [
+        -0.9375,
+        -0.9375,
+        -0.9375,
+        -0.2582142949104309,
+        -0.375,
+        -0.3642857074737549,
+        -1.0
+      ],
+      "q01": [
+        -0.8732143044471741,
+        -0.8758928775787354,
+        -0.9375,
+        -0.15214285254478455,
+        -0.20571428537368774,
+        -0.26249998807907104,
+        -1.0
+      ],
+      "q99": [
+        0.9375,
+        0.9133928418159485,
+        0.9375,
+        0.20250000059604645,
+        0.26249998807907104,
+        0.375,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        false
+      ]
+    },
+    "state": {
+      "mean": [
+        -0.047402460128068924,
+        0.03764284774661064,
+        0.7824649065732956,
+        2.980945110321045,
+        -0.0917410310357809,
+        -0.10754720494151115,
+        0.026381169445812702,
+        -0.026589241810142994
+      ],
+      "std": [
+        0.10643359323201027,
+        0.1512195563913168,
+        0.3829725662785133,
+        0.34897153380459534,
+        0.7877290237854367,
+        0.3158265355416683,
+        0.014528665058790028,
+        0.01448397257314164
+      ],
+      "max": [
+        0.21031762659549713,
+        0.39128610491752625,
+        1.3660105466842651,
+        3.7743608951568604,
+        3.5752453804016113,
+        1.3747878074645996,
+        0.042380549013614655,
+        0.0010137537028640509
+      ],
+      "min": [
+        -0.4856736361980438,
+        -0.3313629925251007,
+        0.008128180168569088,
+        0.8415665030479431,
+        -3.616978406906128,
+        -1.842738389968872,
+        -0.002542249858379364,
+        -0.042119648307561874
+      ],
+      "q01": [
+        -0.424397349357605,
+        -0.27999457716941833,
+        0.009928198531270027,
+        1.3091226816177368,
+        -2.817831039428711,
+        -1.1050233840942383,
+        0.0014720855979248881,
+        -0.04026537388563156
+      ],
+      "q99": [
+        0.15359212458133698,
+        0.36184924840927124,
+        1.2928963899612427,
+        3.2899329662323,
+        2.7776882648468018,
+        0.7240961790084839,
+        0.04062278941273689,
+        -0.00104702974203974
+      ]
+    },
+    "num_transitions": 2448544,
+    "num_trajectories": 15874
+  }
+}

final_model/5791.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d69f446f48fad2e0e9f571627893828dd2eab2fe2229d836d21b2015cb28b324
+size 66

final_model/5793.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d69f446f48fad2e0e9f571627893828dd2eab2fe2229d836d21b2015cb28b324
+size 66

final_model/5795.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d69f446f48fad2e0e9f571627893828dd2eab2fe2229d836d21b2015cb28b324
+size 66

final_model/5797.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d69f446f48fad2e0e9f571627893828dd2eab2fe2229d836d21b2015cb28b324
+size 66

final_model/pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec8d4476f81138ae307f9974d31b124dac06300f3d459767dd53bec141ee2a74
+size 8459026659

run_libero_train_cloud.sh

@@ -0,0 +1,59 @@
+source activate /home/starvla
+export HF_ENDPOINT=https://hf-mirror.com
+# export CUDA_VISIBLE_DEVICES=1
+export NCCL_SOCKET_IFNAME=lo
+export NCCL_IB_HCA=mlx5_2,mlx5_3
+
+# used for check save when communication
+export NCCL_BLOCKING_WAIT=1
+export NCCL_ASYNC_ERROR_HANDLING=1
+export NCCL_TIMEOUT=10000  # timeout set to 1 hour (unit: seconds)
+export NCCL_SOCKET_TIMEOUT_MS=360000
+###########################################################################################
+# === Please modify the following paths according to your environment ===
+Framework_name=QwenGR00T
+freeze_module_list=''
+
+base_vlm=/root/.cache/huggingface/hub/models--StarVLA--Qwen2.5-VL-3B-Instruct-Action/snapshots/01f25c3e057346fda67bdd17fba264d8e31d88b0
+config_yaml=./examples/LIBERO/train_files/starvla_cotrain_libero.yaml
+
+libero_data_root=/home/Sylvest
+data_mix=libero_plus_all
+run_root_dir=./results/${Framework_name}
+# === End of environment variable configuration ===
+###########################################################################################
+
+
+export WANDB_MODE=disabled
+export WANDB_DISABLED=true
+export WANDB_SILENT=true
+
+output_dir=${run_root_dir}/${run_id}
+mkdir -p ${output_dir}
+# mv this script to the output dir
+cp $0 ${output_dir}/
+
+accelerate launch \
+  --config_file starVLA/config/deepseeds/deepspeed_zero2.yaml \
+  --num_processes 4 \
+  --main_process_port 29501 \
+  starVLA/training/train_starvla.py \
+  --config_yaml ${config_yaml} \
+  --framework.name ${Framework_name} \
+  --framework.qwenvl.base_vlm ${base_vlm} \
+  --framework.action_model.num_target_vision_tokens 3 \
+  --datasets.vla_data.data_root_dir ${libero_data_root} \
+  --datasets.vla_data.data_mix ${data_mix} \
+  --datasets.vla_data.per_device_batch_size 4 \
+  --trainer.vla_data.video_backend torchvision_av \
+  --trainer.freeze_modules ${freeze_module_list} \
+  --trainer.gradient_accumulation_steps 4 \
+  --trainer.max_train_steps 40000 \
+  --trainer.save_interval 10000 \
+  --trainer.logging_frequency 100 \
+  --trainer.eval_interval 100 \
+  --run_root_dir ${run_root_dir} \
+  --run_id ${run_id} \
+  --wandb_project starVLA_Libero \
+  --wandb_entity xxx \
+  # --is_debug True

summary.jsonl

@@ -0,0 +1,4 @@
+{"steps": 10000}
+{"steps": 20000}
+{"steps": 30000}
+{"steps": 40000}