503 MB

Ctrl+K

3 contributors

History: 38 commits

Luigi

Support Inference over batch with TensorRT Engine Model

0cdc9a7 almost 2 years ago

video
Add a sample video about 2 years ago
.gitattributes

1.66 kB
Add a sample video about 2 years ago
README.md

1.47 kB
Update README.md about 2 years ago
constant_folding.sh

337 Bytes
Apply constant folding about 2 years ago
convert_to_fp16.py

1.25 kB
Re-quantize models in FP16 in keep positional encoding in FP32 to avoid bad accuracy about 2 years ago
convert_to_fp16_all.py

1.36 kB
Bugfix: correct batch dimension fixation then regenerate ONNX models about 2 years ago
convert_to_mixed.py

5.26 kB
Improve Debug Message Layout in Validate Funciton about 2 years ago
demo.sh

56 Bytes
Remove models with fixed batch size as dynamic batch is generally supported about 2 years ago
demo_batch.sh

64 Bytes
Remove models with fixed batch size as dynamic batch is generally supported about 2 years ago
onnx_to_engine.py

7.05 kB
Use dynamic batch size by default almost 2 years ago
rtmo-l.fp16.onnx

88 MB
xet

Re-quantize models in FP16 in keep positional encoding in FP32 to avoid bad accuracy about 2 years ago
rtmo-l.onnx

176 MB
xet

Apply constant folding about 2 years ago
rtmo-m.fp16.onnx

44.7 MB
xet

Re-quantize models in FP16 in keep positional encoding in FP32 to avoid bad accuracy about 2 years ago
rtmo-m.onnx

89.3 MB
xet

Apply constant folding about 2 years ago
rtmo-s.fp16.onnx

19.9 MB
xet

Re-quantize models in FP16 in keep positional encoding in FP32 to avoid bad accuracy about 2 years ago
rtmo-s.onnx

39.6 MB
xet

Apply constant folding about 2 years ago
rtmo-t.fp16.onnx

13.7 MB
xet

Re-quantize models in FP16 in keep positional encoding in FP32 to avoid bad accuracy about 2 years ago
rtmo-t.onnx

27.4 MB
xet

Apply constant folding about 2 years ago
rtmo_demo.py

1.99 kB
Resize in keeping aspect ratio in visualization almost 2 years ago
rtmo_demo_batch.py

3.06 kB
Resize in keeping aspect ratio in visualization almost 2 years ago
rtmo_gpu.py

21.8 kB
Support Inference over batch with TensorRT Engine Model almost 2 years ago
symbolic_shape_infer.py

138 kB
Make ONNX models compatible with ONNXruntime's TensorrtExecutionProvider about 2 years ago
symbolic_shape_infer_all.sh

135 Bytes
Make ONNX models compatible with ONNXruntime's TensorrtExecutionProvider about 2 years ago