kuroll
/

lingbot_vla_arena

Model card Files Files and versions

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

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

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support