admin/lerobot
1
0
Fork 0
mirror of https://github.com/huggingface/lerobot.git synced 2026-05-15 08:39:49 +00:00
Code Issues Packages Projects Releases Wiki Activity

reformat and clean up tutorial for multitask dit policy

Browse Source
This commit is contained in:
Bryson Jones
2025-12-11 09:33:30 -08:00
parent dd4ef1383f
commit 43c335d0d7
1 changed files with 19 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/source/multitask_dit.mdx
+19 -20
View File
@@ -40,15 +40,15 @@ Here's a complete training command for training Multi-Task DiT on your dataset:
```bash
lerobot-train \
--dataset.repo_id={{MY_DATASET_ID}} \
--output_dir={{MY_OUTPUT_DIR}} \
--policy.type=multi_task_dit \
--policy.device=cuda \
--policy.repo_id={{MY_REPO_ID}}
--dataset.repo_id=YOUR_DATASET \
--output_dir=./outputs/multitask_dit_training \
--batch_size=32 \
--steps=5000 \
--save_freq=500 \
--log_freq=100 \
--policy.type=multi_task_dit \
--policy.device=cuda \
--policy.repo_id="HF_USER/multitask-dit-your-robot" \
--wandb.enable=true
```
@@ -58,18 +58,18 @@ For reliable performance, start with these suggested default hyperparameters:
```bash
lerobot-train \
--dataset.repo_id={{MY_DATASET_ID}} \
--output_dir={{MY_OUTPUT_DIR}} \
--policy.type=multi_task_dit \
--policy.device=cuda \
--dataset.repo_id=YOUR_DATASET \
--output_dir=./outputs/mutitask_dit_training \
--batch_size=320 \
--steps=30000 \
--policy.type=multi_task_dit \
--policy.device=cuda \
--policy.horizon=32 \
--policy.n_action_steps=24 \
--policy.repo_id={{MY_REPO_ID}} \
--policy.objective=diffusion \
--policy.noise_scheduler_type=DDPM \
--policy.num_train_timesteps=100 \
--policy.repo_id="HF_USER/multitask-dit-your-robot" \
--wandb.enable=true
```
@@ -194,12 +194,11 @@ To resume training from a checkpoint:
```bash
lerobot-train \
--config_path=$OUTPUT_DIR/checkpoints/00001000/pretrained_model/train_config.json \
--resume=true \
--output_dir=$OUTPUT_DIR
--config_path=./outputs/mutitask_dit_training/checkpoints/last/pretrained_model/train_config.json \
--resume=true
```
The checkpoint directory should contain `model.safetensors` and `config.json` files (saved automatically during training).
The checkpoint directory should contain `model.safetensors` and `config.json` files (saved automatically during training). When resuming, the configuration is loaded from the checkpoint, so you don't need to specify other parameters.
## Common Failure Modes and Debugging
@@ -262,15 +261,15 @@ Here's a complete example training on a custom dataset:
```bash
lerobot-train \
--dataset.repo_id={{MY_DATASET_ID}} \
--output_dir={{MY_OUTPUT_DIR}} \
--policy.type=multi_task_dit \
--policy.device=cuda \
--dataset.repo_id=YOUR_DATASET \
--output_dir=./outputs/mutitask_dit_training \
--batch_size=320 \
--steps=30000 \
--save_freq=1000 \
--log_freq=100 \
--eval_freq=1000 \
--policy.type=multi_task_dit \
--policy.device=cuda \
--policy.horizon=32 \
--policy.n_action_steps=24 \
--policy.objective=diffusion \
@@ -280,9 +279,9 @@ lerobot-train \
--policy.vision_encoder_name=openai/clip-vit-base-patch16 \
--policy.image_resize_shape=[320,240] \
--policy.image_crop_shape=[224,224] \
--policy.repo_id="HF_USER/multitask-dit-your-robot" \
--wandb.enable=true \
--wandb.project=multitask_dit \
--policy.repo_id={{MY_REPO_ID}}
--wandb.project=multitask_dit
```
## References