Fix: Adaptive gripper threshold with debug logging
40494ef
aw1app
commited on
Spaces:
Runtime error
Runtime error
Commit History
Force output_hidden_states=True in lang_encoder forward call
7cc7ee4
Debug: Check if hidden_states are actually enabled
44a0f72
Final fix: Exact checkpoint dimensions (action=6, not 7)
adf1b6d
Fix: Use correct LSTM dimensions (4 layers, hidden=1024)
ed14d35
Fix: Map checkpoint policy weights correctly and enable hidden states
10f4702
Add LSTM policy head for robot action prediction
bc9ccfd
Cache checkpoint in /data + diagnose missing policy head
8052e7f
Add debug output to see model output format
c37e94b
Remove simulation fallback - real model only
ba08b6e
Fix: Convert attention_mask to boolean before passing to model
1164ea7
Patch OpenFlamingo to remove 'labels' before calling MosaicGPT
0130f3f
Fix: Remove 'labels' argument from model forward pass
92e2be9
Use patched factory to bypass BCFlamingo init bug
a0d16a3
Add einops-exts==0.0.4 dependency
acd072f
Remove large eval files
418dfe1
Deploy real interactive model
c6cb161
Load checkpoint directly in app.py
1b585c8
Load trained RoboFlamingo checkpoint from HuggingFace (robovlms/RoboFlamingo)
e041848
Add custom factory wrapper to avoid BCFlamingo argument issues
6e5893d
Remove problematic arguments from create_model_and_transforms
77dd63f
Add zarr and RoboFlamingo dependencies
184d102
Use robot_flamingo.models.factory for REAL RoboFlamingo
bc2b1f8
Fix: Use actual RoboFlamingo structure
0f64440
Deploy REAL RoboFlamingo with LSTM policy head and trained checkpoint
f71030e
Complete deployment: Real interactive RoboFlamingo model with correct shapes
ed17192
Complete deployment: Real interactive RoboFlamingo model with correct shapes
b66336a
Remove decoder_type parameter for vanilla OpenFlamingo
6073035
Add einops-exts dependency
b01ee14
Fix imports: use open_flamingo.src.factory
16708b1
Deploy with src only (no large eval data)
ff95b80
Deploy real interactive RoboFlamingo model
15c4e01
Colab User
commited on