This update replaces the generic `any` type with the more explicit `CLMem` type for better type safety and clarity. It ensures consistency across the `prepare_inputs` method implementations in derived classes, improving code readability and robustness.
Simplified the `run_model` method by removing the requirement to pass inputs as arguments, and instead leveraging an internal `inputs` state. Adjusted `prepare_inputs` methods across model runners to populate this internal state. This refactor improves code clarity and reduces redundancy in managing input data.
Updated the `run_model` method to explicitly convert tensor outputs to NumPy arrays using `.numpy()`. This ensures compatibility with downstream processes relying on NumPy array inputs.
This change cleans up the code by removing unused imports and redundant `pass` statements in abstract methods. It improves code readability and adheres to cleaner coding practices.
The `assign` operation was unnecessary as new tensors are realized for updated inputs. This simplifies the code and avoids redundant updates, improving clarity and maintainability.
Removed the `create_model_runner` factory function and replaced it with direct initialization of `TinyGradRunner` or `ONNXRunner`. Simplified the `__init__` methods by standardizing paths as constants within `model_runner.py` for cleaner and more maintainable code.
Consolidated multiline function declarations and calls into single lines where appropriate to improve code readability and maintainability. No changes were made to the functionality.
A significant refactoring of `modeld.py` was performed to enhance the handling of model logic. A new abstraction called `ModelRunner` has been introduced which encapsulates the model-running logic. This refactor simplifies the `modeld.py` script and provides easier management across different hardware configurations. Using this segregation, varying processing methods for models can be handled distinctly ensuring cleaner and more maintainable code. An instance of the appropriate model runner is now created during initialization based on whether a TICI hardware or a different type is used.
* needs cleanup
* only if tici
* bump tinygrad
* check width
* base modelframe
* .
* need to be args
* more cleanup
* no _frame in base
* tici only
* its DrivingModelFrame
* .6 is fair
---------
Co-authored-by: Comma Device <device@comma.ai>
* squash
* bump tg
* bump tg
* debump tinygrad
* bump tinygrad
* bump tg
* Skip init iteration
* fixes
* cleanups
* skip first test sample
* typos
* linter unhappy
* update cpu usage
* OPENCL just zeros for now
* imports
* Try printing
* Runs again, but slower
* unused import
* Allow more buffer with tg and all on gpu
* bump tinygrad
* seems ok
* stricter timings for driving looser for dm
* try llvm
* check nvidia
* More timeout for now
* make test pass
* Revert "try llvm"
This reverts commit ef136e478320101fea262bae3579e558da991902.
* small fixes
* whitespace
* revert test timeout
* No model runners
* Always CPU always fast
* No onnx runtime GPU
* more cores
* cleanup
* Is this faster
* Is this faster
* at least runs
* FP32 is faster than 16
* fix deps
* whitespace
* comment
---------
Co-authored-by: Adeeb Shihadeh <adeebshihadeh@gmail.com>
* Revert "Replace ThneedModel with TinygradModel (#33532)"
This reverts commit da952e9b64.
* Revert "camerad: move E + D cams image pipelines to the IFE (#33959)"
This reverts commit f2a1cce42b.
* Reapply "move car.capnp to opendbc" (#33725)
This reverts commit 9d52a5b485.
* why can't i repro?!
* Revert "why can't i repro?!"
This reverts commit 0435d218f790faf7b7aaed27d05ab9ee67b087e6.
* does this cause card to try and read it?
* better place
* wtf
* Reapply "why can't i repro?!"
This reverts commit d24fd5a0abf454f47d5591e3b39039fdc4d0251c.
* also here
DevTekVE