* modeld: RL driving model with 3-file split
Split the driving model into vision + off_policy + on_policy ONNX
files and wire up the RL policy:
- 3-file model split (vision / off_policy / on_policy), replacing the
combined big_driving_policy/vision models
- compiler updates for the split models
- actually consume the policy action in modeld
- add desire state to the driving model
- model iterations (smoothness, off/on-policy weight updates)
* modeld: update driving model
* 1e72cf5a-785f-45ea-888f-28cdb14785de/100
* tinygrad hack
* fix parsing
* looser timing
* big
* Remove unnecessary modeld rebase changes
* Tighten modeld split cleanup
---------
Co-authored-by: Comma Device <device@comma.ai>
Co-authored-by: Armandpl <adpl33@gmail.com>
* Revert "OP model 7 (#37760)"
This reverts commit 052692b25d.
* Revert "OP model (#37740)"
This reverts commit cb32793300.
* dead
* parse_model_outputs: drop extra space
* release model: 6f23a03f-486b-4d3e-a314-19d149644c7c/700
* old style model in tinygrad
* fix desire
* tg hack
* 20Hz
* no gas probs
* No gas here
* better indexing
---------
Co-authored-by: Yassine Yousfi <yyousfi1@binghamton.edu>
* 6f6e3749-1b7c-42e8-a33b-03929b7fc476/700
* oops deleted too much
* 1b4308b7-a659-4ebd-b4c6-c81c1c3890f8/700
* 1be192f3-f407-4217-9757-78b9ad92750a/700
* remove some todos
* more cleanup in lat planner
* vego > min_speed
* regen and update process replay refs
* update model replay ref
* update model replay ref commit again
* Revert "update model replay ref commit again"
This reverts commit 922cb796b8dfe264b0cce7a18206bb898b18bcb3.
* update again
* bump cereal
old-commit-hash: 4c2bd853e4410db7f0d80e258291408e94c6c068
* WIP try modeld all in python
* fix plan
* add lane lines stds
* fix lane lines prob
* add lead prob
* add meta
* simplify plan parsing
* add hard brake pred
* add confidence
* fix desire state and desire pred
* check this file for now
* rm prints
* rm debug
* add todos
* add plan_t_idxs
* same as cpp
* removed cython
* add wfd width - rm cpp code
* add new files rm old files
* get metadata at compile time
* forgot this file
* now uses more CPU
* not used
* update readme
* lint
* copy this too
* simplify disengage probs
* update model replay ref commit
* update again
* confidence: remove if statemens
* use publish_state.enqueue
* Revert "use publish_state.enqueue"
This reverts commit d8807c8348338a1f773a8de00fd796abb8181404.
* confidence: better shape defs
* use ModelConstants class
* fix confidence
* Parser
* slightly more power too
* no inline ifs :(
* confidence: just use if statements
old-commit-hash: cad17b125595c4654bfd8299b041b94ccb3faf73
Harald Schäfer