James Vecellio-Grant 7c6d887187 modeld_v2: infer model shapes from inputs (#1162) ...

* modeld_v2: dynamify temporal buffer management.

* skip redundant reshaping and flattening.

* simplify MHP checks for lead and plan

* modeld_v2: add unit tests for buffer logic and refactor index mapping

* Let’s possibly fail a test :)

* Update test_buffer_logic_inspect.py

* Update test_buffer_logic_inspect.py

* modeld_v2: better temporal mapping for non-split

* Bump to 10 I guess

* Downgrade CURRENT_SELECTOR_VERSION to 9

* red diff ya know?

* add dynamic buffer update tests and compare against legacy logic.

 Cover modelState.init and modelState.run

* send

* Revert "send"

This reverts commit 9e6c95fbfde134eeba952da6eef012baa0396fa0.

* format

---------

Co-authored-by: Jason Wen <haibin.wen3@gmail.com>

2025-08-24 22:08:55 -04:00

..

onnx

model: Refactor modeld with modular runners and split model support (#877)

2025-05-25 17:10:16 +02:00

tinygrad

modeld_v2: infer model shapes from inputs (#1162)

2025-08-24 22:08:55 -04:00

constants.py

model: Refactor modeld with modular runners and split model support (#877)

2025-05-25 17:10:16 +02:00

helpers.py

model: Refactor modeld with modular runners and split model support (#877)

2025-05-25 17:10:16 +02:00

model_runner.py

Model: Sync modeld with upstream arbitrary vision inputs (#1004)

2025-06-19 07:24:20 -07:00