Downloads · 30 days
13
32% of all-time downloads
zhoumiaosen/groot-n1p5-lora-libero-2ep-stresstest
groot-n1p5-lora-libero-2ep-stresstest is a robotics model from zhoumiaosen. Use it for the robotics task on the model card, and read the license before you ship it in a product. It is set up for lerobot. The card lists the license as apache-2.0.
This is not a usable trained policy. It's a GPU/software stress test: NVIDIA GR00T-N1.5-3B fine-tuned via LeRobot's groot policy wrapper on only 2 episodes of the HuggingFaceVLA/libero dataset, for 25,000 steps (out o…
Downloads · 30 days
13
32% of all-time downloads
All-time downloads
41
Public
Parameters
2.4B
7 GB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors7 GB · 100%
How the weights are stored.
BF161.3B · 56%
From the Hugging Face model README
This is not a usable trained policy. It's a GPU/software stress test: NVIDIA
GR00T-N1.5-3B fine-tuned via LeRobot's
groot policy wrapper on only 2 episodes of the HuggingFaceVLA/libero
dataset, for 25,000 steps (out of an intended 30,000 — training crashed twice
with a transformers/accelerate compatibility bug inside PyTorch's module
iteration and was stopped at this checkpoint).
tune_llm=False, tune_visual=False, tune_projector=False,
tune_diffusion_model=True, batch size 1 — provided the GPU isn't shared
with another active process at the same time.nvidia/GR00T-N1.5-3BHuggingFaceVLA/libero, episodes [0, 1] onlylora_rank=8 was passed but did not reduce the trainable parameter count in
this LeRobot integration (751M / 2.4B total params were fully trainable,
not adapted through actual low-rank layers) — worth knowing if comparing
against a "real" LoRA run.