Downloads · 30 days
2
17% of all-time downloads
mimiminsoo/starvla-qwenpiv3-so101-marker-rt1warmstart
starvla-qwenpiv3-so101-marker-rt1warmstart is a machine learning model from mimiminsoo. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as mit.
QwenPIv3 (Qwen3-VL-4B backbone + layer-wise flow-matching DiT action head), fine-tuned on the marker100 SO-101 teleoperation dataset (97 episodes, 31,291 frames, singlearm(5)+gripper(1) absolute joint-position action).
Downloads · 30 days
2
17% of all-time downloads
All-time downloads
12
Public
Repo size
11 GB
Likes
0
Public
Click a slice to open those files.
.pt10.9 GB · 100%
From the Hugging Face model README
marker_100, RT-1 backbone warm-startQwenPI_v3 (Qwen3-VL-4B backbone + layer-wise flow-matching DiT action head),
fine-tuned on the marker_100 SO-101 teleoperation dataset (97 episodes,
31,291 frames, single_arm(5)+gripper(1) absolute joint-position action).
qwen_vl_interface warm-started from StarVLA/Qwen3VL-PI_v3-Bridge-RT_1
(checkpoints/steps_50000_pytorch_model.pt), then frozen for the whole run.
action_model + project_layers are freshly initialized (the RT-1 checkpoint's
action space is 7-dim WidowX/Google-robot EE-delta, incompatible with SO-101's
6-dim absolute joint action, so only the backbone transfers).marker_100), 49h03m wall-clock on a single RTX 3090 24GB.action_dit_loss): 0.049 at step 115,000 (down from 2.92 at step 10;
stabilized around 0.01-0.1 from ~step 4,000 onward).final_model/pytorch_model.pt only. Eleven intermediate
checkpoints (every 10,000 steps) exist locally but were not uploaded (112GB total).python deployment/model_server/server_policy_gr00t_zmq.py \
--ckpt_path final_model/pytorch_model.pt \
--port 5555 --use_bf16
Speaks the Isaac-GR00T N1.6 ZMQ protocol; compatible with leisaac's
Gr00t16ServicePolicyClient (--policy_type gr00tn1.6) out of the box — see
examples/SO101_Marker/eval_files/ in the StarVLA repo.