lingbot_vla_arena

kuroll/lingbot_vla_arena

lingbot_vla_arena is a machine learning model from kuroll. Use it for the machine learning task on the model card, and read the license before you ship it in a product.

Downloads · 30 days

Access

Public

Updated Apr 27, 2026

Repo size

66.7 GB

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.distcp49.9 GB · 75%

At a glance

Access: Public
Created: Apr 27, 2026
Updated: Apr 27, 2026
SHA: 7fa9de15

Model card

From the Hugging Face model README

LingBot-VLA (VLA-Arena) Training Hyperparameters

1. Model

Key	Value
`post_training`	`true`
`adanorm_time`	`true`
`old_adanorm`	`true`

2. Data

Key	Value
`datasets_type`	`vla`
`data_name`	`vla_arena`
`num_workers`	`8`
`norm_type`	`bounds_99_woclip`

3. Training

Key	Value
`loss_type`	`L1_fm`
`data_parallel_mode`	`fsdp2`
`enable_full_shard`	`false`
`module_fsdp_enable`	`true`
`use_compile`	`true`
`use_wandb`	`true`
`rmpad`	`false`
`rmpad_with_pos_ids`	`false`
`ulysses_parallel_size`	`1`
`freeze_vision_encoder`	`false`
`tokenizer_max_length`	`48`
`action_dim`	`7`
`max_action_dim`	`75`
`max_state_dim`	`75`
`lr`	`2.0e-5`
`lr_warmup_ratio`	`0.05`
`lr_decay_style`	`cosine`
`lr_decay_ratio`	`1.0`
`num_train_epochs`	`10`
`micro_batch_size`	`16`
`global_batch_size`	`256`
`max_steps`	`4000`
`ckpt_manager`	`dcp`
`enable_fp32`	`true`
`enable_resume`	`true`

采用8卡h100

Created: Apr 27, 2026
Updated: Apr 27, 2026