Closed pgimenes closed 1 month ago
added tensorrt backbone, tensorrt quantize.py tests
Added requirements and enabled tensorrt as action in cmd line
onnx transform working
minor changes
added calibrate and calib-quant-test
Setup non-terminal client
Added training for quantization
improved train calib
calib changes
Added utlils
adding fake quant
Fixed toml issue
minor change
Fixed toml issues and now calibrate working in py
Improved file system for quantize
Calibrate and Quantize working
Added jsc-trt as test model
Added analysis pass
Added lots of metrics and trt support. need to test
data_loader change
Squashing bugs
Quant inference working, need to clean
unexpected regression model
Almost finsihed analysis
TensorRT improves latency!!
need to fix int8calib
added training. need to test
added int8calibrator and onnx summary
added onnxruntime setup & placeholders
devised onnxruntime & transform pass structure
modify requirement from ort to ort-gpu
ort-gpu in setup.py
devised onnx_runtime_transform_pass, test_performances to define
added ort inference session
added execution provider; todo: add input type to toml and according onnx processing
Minor changes
FP16 working
INT8 calibrator implemented but takes ages
Minor change
Revert "Minor change"
This reverts commit 745723bfca8e3d3659641485a705b9419f3526d2.
Minor tidy up
fixed INT8 and FP16
Added fine_tune transform pass but getting circular import for chop
Added scheduler to Chop CLI and train for Cosine Annealing LR capability
Fine Tuning QAT working
Improved fine tune
improvements and bug fixes
inference consumption measurement
Improved table
JSC behaving itself!
reformatted tensorrt tutorial
Adding documentation
more documentation
added defaults to fine tune params
improved directory
changed folders to lowercase according to conventions
onnxruntime for jsc-tiny behaves and improves latency :D
onnx ambiguously slower on vgg7 cpu
small fixes
Added summarize quant
Improved calibration and docs
Bug fix
fixed fine tune bug
vision models functioning on mnist (added input channel pre-processing)
transform int8 changes
INT8 Float16 comparison complete
added checkpoitns for tensorrt demos
VGG7 test not behaving
opt125 config
added opt125 to notebook
onnxruntime works and improves latency on vgg7
adding opt125
adjusted onnxruntime batching mismatch
changed opt toml
fixes
tidy up
Opt 125 toml
Improved mixed precision
Section1 notebook working
section 1of tutorial complete
added pooling and other conv support
modified module support
test commit
transfer to another gpu
Added lstm support
added lightning logs to gitignore
MaseRT docs
Documentation improvements
MaseRT documentation
Improvements to docs
Runtime analysis refractor
updated requirements for onnx
ONNXRT implementation
Bug fixes and OnnxRT dynamic quant
dynamic quantization working
static quantization working
added cpu / gpu inference options
Little improvements on jsc_toy onnx performance
onnxruntime fixed and toml file structure change
created trt mixed precision search space
minors
pre-tensorrt runner for search
fixed mixed precision int8
search space changes
mixed precision onnx
Onnxrt tutorial improvements
still fixin mixed precision
Mixed precision fixed
Fixed minor static quant bug. Still large VGG latency
fixex formatting to keep up with expected coding style
opt not working for any transform action, investigating cause
coding spacing adjustments
Documentation only change
Fixed onnxruntime large latency - onnxruntime package issue
quantization onnxruntime debugging for vgg model
standardized batch sizes for experiments
mobilnet experiments
Starting sphinx documentation for RT
Transform interface refractor
Sphinx transforms
fixed import errors for transform interface passes
Sphinx documentation 5/6 passes
deleted old tensorrtdev playground folder
transform analysis pass fix
masert readme improvement
tutorials added to sphinx
added docstring comments
adjust mnist dummy_inputs for vision models, tutorial fixes, toml fixes
minor style changes
readme docs improvment
Docstring and readme improvemnets
Added open source contribution section
Section 1 and 2 TensorRT tutorial complete
added open source contribution to masert readme
updated masert onnxrt overview
updated jsc toy checkpoint load dir
minor toml changes
Reformatted using Black
Fixed sphinx formatting issue
Updated tomls to support new transfrom style config
minor pr readme fix
tensorRT_tutorial ready
onnx quantization tutorial ready
Added MASERT tutorials to sphinx docs
Final formatting
Onnxrt tutorial finalized
added tensorrt backbone, tensorrt quantize.py tests
Added requirements and enabled tensorrt as action in cmd line
onnx transform working
minor changes
added calibrate and calib-quant-test
minor changes
Setup non-terminal client
Added training for quantization
improved train calib
calib changes
Added utlils
adding fake quant
Fixed toml issue
minor change
Fixed toml issues and now calibrate working in py
Improved file system for quantize
Calibrate and Quantize working
Added jsc-trt as test model
Added analysis pass
Added lots of metrics and trt support. need to test
data_loader change
Squashing bugs
Quant inference working, need to clean
unexpected regression model
Almost finsihed analysis
TensorRT improves latency!!
need to fix int8calib
added training. need to test
added int8calibrator and onnx summary
added onnxruntime setup & placeholders
devised onnxruntime & transform pass structure
modify requirement from ort to ort-gpu
ort-gpu in setup.py
devised onnx_runtime_transform_pass, test_performances to define
added ort inference session
added execution provider; todo: add input type to toml and according onnx processing
Minor changes
minor change
minor changes
FP16 working
INT8 calibrator implemented but takes ages
Minor change
Revert "Minor change"
This reverts commit 745723bfca8e3d3659641485a705b9419f3526d2.
Minor tidy up
fixed INT8 and FP16
Added fine_tune transform pass but getting circular import for chop
Added scheduler to Chop CLI and train for Cosine Annealing LR capability
Fine Tuning QAT working
Improved fine tune
improvements and bug fixes
inference consumption measurement
Improved table
minor changes
Minor change
Minor changes
JSC behaving itself!
reformatted tensorrt tutorial
Adding documentation
more documentation
added defaults to fine tune params
improved directory
changed folders to lowercase according to conventions
onnxruntime for jsc-tiny behaves and improves latency :D
onnx ambiguously slower on vgg7 cpu
small fixes
Added summarize quant
Improved calibration and docs
Bug fix
fixed fine tune bug
vision models functioning on mnist (added input channel pre-processing)
Minor changes
minor changes
transform int8 changes
INT8 Float16 comparison complete
added checkpoitns for tensorrt demos
VGG7 test not behaving
opt125 config
added opt125 to notebook
onnxruntime works and improves latency on vgg7
adding opt125
adjusted onnxruntime batching mismatch
changed opt toml
fixes
tidy up
Opt 125 toml
Improved mixed precision
Section1 notebook working
minor changes
section 1of tutorial complete
added pooling and other conv support
modified module support
test commit
test commit
test commit
transfer to another gpu
Added lstm support
added lightning logs to gitignore
MaseRT docs
Documentation improvements
MaseRT documentation
Improvements to docs
Runtime analysis refractor
updated requirements for onnx
ONNXRT implementation
Bug fixes and OnnxRT dynamic quant
dynamic quantization working
static quantization working
added cpu / gpu inference options
Little improvements on jsc_toy onnx performance
onnxruntime fixed and toml file structure change
minor changes
minor changes
created trt mixed precision search space
minor changes
minors
pre-tensorrt runner for search
fixed mixed precision int8
search space changes
mixed precision onnx
Onnxrt tutorial improvements
minor changes
still fixin mixed precision
Mixed precision fixed
Fixed minor static quant bug. Still large VGG latency
fixex formatting to keep up with expected coding style
opt not working for any transform action, investigating cause
coding spacing adjustments
Documentation only change
Fixed onnxruntime large latency - onnxruntime package issue
quantization onnxruntime debugging for vgg model
standardized batch sizes for experiments
mobilnet experiments
Starting sphinx documentation for RT
Transform interface refractor
Sphinx transforms
fixed import errors for transform interface passes
Sphinx documentation 5/6 passes
deleted old tensorrtdev playground folder
transform analysis pass fix
masert readme improvement
tutorials added to sphinx
added docstring comments
adjust mnist dummy_inputs for vision models, tutorial fixes, toml fixes
minor style changes
readme docs improvment
Docstring and readme improvemnets
Added open source contribution section
Section 1 and 2 TensorRT tutorial complete
added open source contribution to masert readme
minor changes
updated masert onnxrt overview
minor changes
updated jsc toy checkpoint load dir
minor toml changes
minor toml changes
Reformatted using Black
Fixed sphinx formatting issue
Updated tomls to support new transfrom style config
minor changes
minor pr readme fix
tensorRT_tutorial ready
onnx quantization tutorial ready
Added MASERT tutorials to sphinx docs
Final formatting
Onnxrt tutorial finalized