Commit Graph

196 Commits

Author SHA1 Message Date
Mitchell Goff
328b5d9d47 Nicki Minaj Model (#28218)
* New model: 2ff7490f-8a2f-4c0d-87a8-b0fa3a4a6a71

* Updated model_replay_ref_commit
2023-05-22 15:11:04 -07:00
YassineYousfi
bda0d6ac77 update dm model README (#28238)
add calib dm input in readme
2023-05-19 15:20:43 -07:00
YassineYousfi
2d4558d5cf update DM model README (#28211)
* update DM model readme

* rename + permalink + correct types and ranges
2023-05-16 13:40:49 -07:00
YassineYousfi
f80c1087fc new model: trained with augmented turn desire pulses (#28063)
* df733db3-fad2-4485-abe9-59a6a47f2233/449 85f20e13-dd4d-49a8-ba33-98e6afc161d9/700

* update ref commit
2023-05-01 16:03:16 -07:00
YassineYousfi
26541d9e7a new model: trained on new dataset (#27637)
* df733db3-fad2-4485-abe9-59a6a47f2233/449 946c0f55-247d-4790-a407-c07e4364b340/700

* fix reducel2 df733db3-fad2-4485-abe9-59a6a47f2233/449 946c0f55-247d-4790-a407-c07e4364b340/700

* update ref commit
2023-03-22 20:19:46 -07:00
YassineYousfi
e146bb1a96 update modeld/models readme (#27564)
* update readme with new model design

* typo

* space
2023-03-11 16:26:39 -08:00
Harald Schäfer
eb8bdc0026 MPC path in UI (#27380)
* 10s lat

* Full length MPC

* redfine N

* Leave controls the same for now

* Updates

* use long plan in lat plan

* interp plan

* simplergit add selfdrive/controls/plannerd.py selfdrive/controls/

* expand to 10s

* revert this

* fix linter

* vizualize

* fix long test

* typo

* cleanup

* compiles

* unused

* unused

* bump cereal

* bump cereal

* use model if no uiplanm

* update replay

* update ref commit

* bump cereal to master
2023-02-17 19:58:30 -08:00
YassineYousfi
1e49c54ffb new model: improved height estimation (#27234)
* 07a3a2b2-6dbe-43c3-9bb9-a8023932e054/449 c3d7d18a-6207-42e7-84d9-601ecb3c6f0e/700

* 07a3a2b2-6dbe-43c3-9bb9-a8023932e054/449 db92e7f3-3f4b-4f0f-95ff-76cb6ed6894a/700

* update ref commit
2023-02-08 22:21:31 -08:00
YassineYousfi
7316cd03fb new model: fixed weight decay logic (#27068)
* 07a3a2b2-6dbe-43c3-9bb9-a8023932e054/449 cdfa9e3c-2807-49fc-bdc7-a985f8ff1644/700

* update refs
2023-01-26 19:09:12 -08:00
Harald Schäfer
160c9ba1d2 Navmodel and driving style: update runner (#26762)
* Navmodel and driving style: update runner

* No driving style yet

* thneedrunner without extra inputs

* nav feature ref
2022-12-11 15:25:36 -08:00
Mitchell Goff
bb8a38a050 navmodeld (#26665)
* Added navmodeld

* New nav model: 7c306685-5476-4bd4-ab65-105b01b6bca8/300, feats only

* little cleanup

* Remove NAV flag

* Moved to_kj_array_ptr to commonmodel.h

* Switch from decimation to last_frame_id check

* add to release files

Co-authored-by: Adeeb Shihadeh <adeebshihadeh@gmail.com>
2022-12-02 13:14:30 -08:00
YassineYousfi
ceb94f1bcd Revert "Albert Einstein Model" (#26576)
Revert "Albert Einstein Model (#26456)"

This reverts commit 34c80a6fbd.
2022-11-22 11:41:22 -08:00
YassineYousfi
34c80a6fbd Albert Einstein Model (#26456)
* d0bc247d-9529-4e50-a5dd-95a8a5733874/440 38fb1ce9-18d7-4292-9ffd-678f8483ebf4/700

* update model replay ref commit
2022-11-18 17:50:24 -08:00
Harald Schäfer
f63f0de80a E2e long model: calibrate model speed to wheel speed (#26395)
* calibrate!

* Fix test

* Fix proc replay

* check len

* get v_ego from model 8501d20-bb59-4193-aa82-82b2737dedd6/449 609d90f3-65e6-4617-a60c-d6d99eead408/700

* bump cereal

* initialize v_model_error

* typo

* better names

* cleanup

* bump cereal

* update model replay ref commit

* bump to cereal master

Co-authored-by: Yassine Yousfi <yyousfi1@binghamton.edu>
2022-11-07 17:49:36 -08:00
YassineYousfi
c3388d216c Model trained with headlight augmentation (#26284)
* d8501d20-bb59-4193-aa82-82b2737dedd6/449 ddfe2a22-fb0e-4ff0-a97f-0157b95eb44d/700

* update ref
2022-10-30 23:55:48 -07:00
HaraldSchafer
2b7290a142 Model trained with more long noise (#26241)
* d8501d20-bb59-4193-aa82-82b2737dedd6/449 d37d320d-7d78-479a-9be2-d58c75284307/700

* Update model ref
2022-10-26 00:20:30 -07:00
HaraldSchafer
0edbbdeaf9 Modeld: Add comment (#26188)
Rename and add comment
2022-10-21 11:20:52 -07:00
HaraldSchafer
d25ab3e366 Wide transform model (#26165)
* d8501d20-bb59-4193-aa82-82b2737dedd6/449 f78009d1-2184-4136-9fbc-1c5de55d460d/700

* Change model outs

* Add cereal

* Publish transform

* d8501d20-bb59-4193-aa82-82b2737dedd6/449 f78009d1-2184-4136-9fbc-1c5de55d460d/700

* Bump cereal to master

* Bump cereal to master

* Update model ref
2022-10-21 01:29:58 -07:00
Bruce Wayne
3a50d5fd85 Revert "d8501d20-bb59-4193-aa82-82b2737dedd6/449 f78009d1-2184-4136-9fbc-1c5de55d460d/700"
This reverts commit 0ac12b82c7.
2022-10-21 00:32:34 -07:00
Bruce Wayne
0ac12b82c7 d8501d20-bb59-4193-aa82-82b2737dedd6/449 f78009d1-2184-4136-9fbc-1c5de55d460d/700 2022-10-21 00:24:11 -07:00
YassineYousfi
6b36f73870 desire: give the pulse at t=0 (#26180)
input desire at t0
2022-10-20 15:13:58 -07:00
ZwX1616
d522492ba0 DM: add use of e2e preds (#26078)
* try ml

* de56

* j914ef75a

* jd1124586

* jd1124586

* d112

* oops

* set

* update ref

* use offset

* bump DM power usage

* new ref
2022-10-17 19:40:06 -07:00
HaraldSchafer
b654ebdd25 Refactor model: no klblock (#26035)
* ff138dc0-d097-4818-b40e-dba5ba89d5d6/449 13274b7d-b546-4b91-a587-33b4af7dec6a/700

* b1bb39be-c6ce-4744-8e63-92969fda6bfc/449 f3ebfba1-f686-448f-be9b-b4d5010be91c/700

* model ref

Co-authored-by: Yassine Yousfi <yyousfi1@binghamton.edu>
2022-10-15 20:04:35 -07:00
YassineYousfi
6c5693e965 faster rocket launcher model (#26009)
* cache tokens 1456d261-d232-4654-8885-4d9fde883894/440 e63ab895-2222-4abd-a9a5-af86bb70e260/700

* udpate ref commit

* bump tinygrad to master
2022-10-07 17:22:42 -07:00
HaraldSchafer
cb0b7375b7 Rocket Launcher Model (#25963)
* 1456d261-d232-4654-8885-4d9fde883894/440 6b7d7cec-ead8-40f3-86cc-86d52c9b03fe/300

* compute only 9 tokens: 1456d261-d232-4654-8885-4d9fde883894/440 6b7d7cec-ead8-40f3-86cc-86d52c9b03fe/300

* tinygrad: cleanup gather

* 1456d261-d232-4654-8885-4d9fde883894/440 6b7d7cec-ead8-40f3-86cc-86d52c9b03fe/700

* empty commit for tests

* bump tinygrad

* dont use tinygrad matmul for now

* bump tinygrad

* 1456d261-d232-4654-8885-4d9fde883894/440 e63ab895-2222-4abd-a9a5-af86bb70e260/700

* float16 1456d261-d232-4654-8885-4d9fde883894/440 e63ab895-2222-4abd-a9a5-af86bb70e260/700

* increase steer rate cost

* Revert "increase steer rate cost"

This reverts commit 74ce9ab9be7ef17ecfec931f96851b12f37f2336.

* fork tinygrad

* empty commit for tests

* basics

* Kinda works

* new lat

* new tuning

* Move LATMPCN so scons compiles

* Update long weights

* Add tinygrad optim

* Update model ref

* update weights

* Update ref

* Try

* Error message for field ignore

* update model regf

* ref commit

* Fix onnx test

Co-authored-by: Yassine Yousfi <yyousfi1@binghamton.edu>
2022-10-05 21:43:38 -07:00
HaraldSchafer
e1b7a37a1f Support e2e long in longitudinal planner (#25636)
* refactor

* Add planer modes to support offline, acc, and blended

* add acceleration

* Fix index

* Update model ref

* Read in model outputs

* Add model msg

* Add e2e logic

* Add source
2022-09-01 15:40:25 -07:00
George Hotz
40d6f4b65c modeld: Move from SNPE to tinygrad (#25207)
* compiling, won't work yet

* running with inputs and outputs

* there's some magic chance this works

* no more dlc, include onnx

* yolo tests plz

* bump tinygrad

* files_common + delete dlc

* tinygrad_repo -> tinygrad

* pre commit config

* llops needed

* extra in files_common

* bump tinygrad

* fix indent

* tinygrad/nn/__init__

* tinygrad_repo

* bump tinygrad repo

* bump tinygrad

* bump with native_exp, match maybe

* native_explog is argument

* pyopencl no cache

* 5% chance this matches

* work in float32?

* bump tinygrad

* fix build

* no __init__

* fix recip

* dumb hack

* adding thneed PC support

* fix pc segfault

* pc thneed is working

* to_image

* prints stuff with debug=2

* it sort of works

* copy host ptr is simpler

* bug fix

* build on c3

* this correct?

* reenable float16

* fix private, fixup copy_inputs internal

* bump tinygrad and update ref commit

* fix OPTWG on PC

* maybe fix non determinism

* revert model replay ref commit

* comments, init zeroed out buffers

* upd ref commit

* bump tinygrad to fix initial image

* try this ref

Co-authored-by: Comma Device <device@comma.ai>
2022-09-01 10:31:14 -07:00
George Hotz
b6e355a933 modeld: PC Thneed prereqs (#25615)
* pc thneed prereqs

* ugh, out of date

* that can stay private

* memcpy here is fine in SNPE variant

* release files

* thneed docs don't work anymore. they didn't look too useful

Co-authored-by: Comma Device <device@comma.ai>
2022-08-30 17:57:14 -07:00
Dean Lee
68b1dbc0ea modeld: delete wide_frame in model_free (#25562) 2022-08-26 11:19:05 -07:00
HaraldSchafer
2f46fe5d85 Nuclear Grade Model: less memory, more accuracy (#25524)
* c9d10c64-bea4-41ec-8ca3-d8c886fda172/440 6d1c8a6b-4070-4780-80f1-6f08f234275e/900

* update ref
2022-08-25 00:01:21 -07:00
HaraldSchafer
041458f632 Falcon Punch Model: turn cutting improvements (#25413)
* simplified change to mpc dynamics

* add jerk pts

* increase jerk cost

* increase jerk pts multipler to master value

* Add final commit

* 1456d261-d232-4654-8885-4d9fde883894/440 ac1a6744-85b0-4ec6-8ba7-608d0717b8f1/750

* some copies are useful

* update model replay ref

* less frames in model replay onnx cpu

* 1456d261-d232-4654-8885-4d9fde883894/440 264b67f5-3f52-4b58-b11f-58dd8aaf08bf/950

* 1456d261-d232-4654-8885-4d9fde883894/440 236fc556-fba3-4255-8ccf-684b22637160/950

* c9d10c64-bea4-41ec-8ca3-d8c886fda172/440 26d73dd2-862a-44ae-bbdd-32cc4f397ad7/900

* Fix couple tests

* Update ref

* Unused for now

* Add lateral factor comment

* Unused variable

Co-authored-by: nuwandavek <vivekaithal44@gmail.com>
Co-authored-by: Bruce Wayne <yassine@comma.ai>
Co-authored-by: Yassine Yousfi <yyousfi1@binghamton.edu>
Co-authored-by: Bruce Wayne <batman@gpu06.internal>
2022-08-12 00:47:59 -07:00
YassineYousfi
3afe7464df Model file too big: Revert "New model: minor fixes" (#25390)
Revert "New model: minor fixes (#25368)"

This reverts commit 55bf938504.
2022-08-08 12:42:46 -07:00
HaraldSchafer
55bf938504 New model: minor fixes (#25368)
* 1456d261-d232-4654-8885-4d9fde883894/440 ac1a6744-85b0-4ec6-8ba7-608d0717b8f1/750

* some copies are useful

* update model replay ref

* less frames in model replay onnx cpu

Co-authored-by: Bruce Wayne <yassine@comma.ai>
Co-authored-by: Yassine Yousfi <yyousfi1@binghamton.edu>
2022-08-05 17:28:15 -07:00
ZwX1616
22408781f7 Remove RHD toggle: determine by driver side instead (#24802)
* driverview

* auto choose

* useless

* remove

* modeld not use toggle

* remove from params

* should have commissue anyway

* delete icon file

* deal with init more elegantly and prohibit switching when engaged

* nonblock save when not unstable

* filter very low speeds

* a few more secs dont harm

* draw icon

* match

* radius

* use const

* ref commit

* new transs

* put_bool_nonblocking

* 5k route tune

* ref comm

Co-authored-by: Comma Device <device@comma.ai>
2022-07-29 13:52:24 -07:00
ZwX1616
a0c114b8b0 DM: track RHD predictions (#24947)
* driverview

* auto choose

* useless

* remove

* modeld not use toggle

* remove from params

* Revert "remove from params"

This reverts commit a08df0b4921e03deac24f4da2c0f1e9e9255a717.

* Revert "modeld not use toggle"

This reverts commit 2730bf8f57c8b057db2e4a76541e92880506cedd.

* Revert "remove"

This reverts commit 21f7cfaaee5452e53ee719762078cb153b3cc766.

* Revert "driverview"

This reverts commit 222d129711e6aa34c0468294b94f60ebbd1bb126.

* semi revert
2022-06-22 13:51:37 -07:00
HaraldSchafer
0e0b5c4e24 Revert "Rocket league model (#24869)" (#24936)
* Revert rocket league

* revert ref commit

* New model ref commit
2022-06-21 12:07:00 -07:00
ZwX1616
1f2f9ea9c9 fullframe DM model (#24860)
* Revert "put cereal on master"

This reverts commit a8ccd8f838.

* Revert "Revert fullframe DM model (#24812)"

This reverts commit c646eeee0a.

* revert revert cereal

* clip6

* 0.8 is fair

* Fiction compensation should be based on error

* Update refs

* Add deadzone

* not that

* good mg

* ref

* ref

* ee8f

* minor tweak

* ref

* recompile

* ref

* cereal

* match driverstatus

* new ref

* new ref

* pass token through jenkins credentials

* quote

* fix snpe dead weights

* final ref

Co-authored-by: Harald Schafer <harald.the.engineer@gmail.com>
Co-authored-by: Adeeb Shihadeh <adeebshihadeh@gmail.com>
2022-06-20 16:24:51 -07:00
HaraldSchafer
9283040d84 Rocket league model (#24869)
* dd9a502d-c8e2-4831-b365-804b0ae0739d/600 80041070-d276-4fed-bdb9-0075e5442908/420

* no elementwise op

* 9dabf0fe-2e60-44bf-8d3a-d20a74aca072/600 ae746590-0bb5-4a16-80db-15f02d314f03/300 c4663a12-b499-4c9b-90dd-b169e3948cb1/60

* explicit slice

* some copies are useful

* 1456d261-d232-4654-8885-4d9fde883894/440 c06eba55-1931-4e00-9d63-acad00161be0/700 af2eb6ba-1935-4318-aaf8-868db81a4932/425

* 154f663e-d3e9-4020-ad49-0e640588ebbe/399 badb5e69-504f-4544-a99e-ba75ed204b74/800 08330327-7663-4874-af7a-dcbd2c994ba7/800

* set steer rate cost to 1.0

* smaller temporal size

* Update model reg

* update model ref again

* This did upload somehow

* Update steer rate cost

Co-authored-by: Yassine Yousfi <yyousfi1@binghamton.edu>
2022-06-15 15:29:42 -07:00
Adeeb Shihadeh
1139fe507b Move selfdrive/hardware/ to system/ (#24725)
* move hardware to system/

* fix mypy
2022-06-11 16:38:24 -07:00
ZwX1616
c646eeee0a Revert fullframe DM model (#24812)
* Revert "fullframe DM: flip RHD yaw to use matching thresholds"

This reverts commit 2ac6931003.

* Revert "fullframe DM model (#24762)"

This reverts commit d6c07a6b15.

* revert cereal
2022-06-10 16:16:46 -07:00
ZwX1616
d6c07a6b15 fullframe DM model (#24762)
* get log

* simplify two nonsense

* not needed

* libyuv is a joke

* clean up

* try small

* fast but not bad

* working

* clean up driverview

* simplified

* thats mirrored

* smol

* tweak

* ref is screen

* w/ ee

* update camera model

* no if TICI

* start

* update pose thresh

* less cpu more dsp

* new libyuv

* new snpe

* add files

* test

* should be fast

* update out len

* trigger test

* use master snpe

* add cereal

* update cereal

* refactor parsing

* missing ;

* get

* wrong type

* test model

* use driver data

* 10829278-72fe-4283-a118-2cef959ce174/1550

* no pf

* adapt driverview

* ;

* rhd learner

* update libyuv buildi x64

* ad4337ea

* remove blink slack

* test

* no

* use toggle

* b16

* fix for nv12

* 5b02cff5 both

* update test

* update cereal

* update cereal

* update cereal

* v2 packets

* revert libyuv

* no /

* update snpemodel

* ;

* memcpy

* fix test

* use toggle in driverview

* update power

* update replay

* Revert "update replay"

This reverts commit 1d0979ca59dbc89bc5890656e9501e83f0556d50.

* update model ref

* halve cpu

* fake 8bit onnx runner

* same thresh as report

* cereal master

Co-authored-by: Comma Device <device@comma.ai>
2022-06-08 20:13:46 -07:00
HaraldSchafer
5b0f660a0e Update README.md 2022-06-02 14:19:29 -07:00
HaraldSchafer
f07f7cbc8e Modeld readme replaced with driving.h 2022-06-02 14:19:19 -07:00
Willem Melching
f49a9c9fd2 less TICI when not needed (#24698)
* less TICI when not needed

* fix process replay

* move reading voltages into hw abstraction layer

* Update selfdrive/hardware/tici/hardware.h

Co-authored-by: Adeeb Shihadeh <adeebshihadeh@gmail.com>

* Update selfdrive/hardware/hw.h

Co-authored-by: Adeeb Shihadeh <adeebshihadeh@gmail.com>

* Update selfdrive/hardware/base.h

Co-authored-by: Adeeb Shihadeh <adeebshihadeh@gmail.com>

* rename init function

* Update selfdrive/athena/athenad.py

Co-authored-by: Robbe Derks <robbe.derks@gmail.com>

* Update selfdrive/boardd/boardd.cc

* Apply suggestions from code review

* Update selfdrive/thermald/thermald.py

* update ref

* fix alert width if all cameras are bad

* add ecam to test_loggerd

* bump cereal

* bump cereal

Co-authored-by: Adeeb Shihadeh <adeebshihadeh@gmail.com>
Co-authored-by: Robbe Derks <robbe.derks@gmail.com>
2022-06-02 15:20:51 +02:00
George Hotz
ea5b8cdfb1 nv12: encoderd avoids a full frame copy (#24519)
* rgb to nv12

* nv12 works (w memcpy)

* correct now

* no copy

* fix nv12 with fast debayer

* reverts of unused stuff

* ui use nv12

* comment out thumbnails for now

* rebase fix

* dm read nv12

* model read nv12

* fix ffmpeg encoder

* thumbnails from nv12

* replay to nv12

* python framereader support nv12

* remove hardcoded frame/buffer sizes

* fix build

* ffmpeg encoder fix buffers

* small cleanup

* reduce power usage test

* fix cpu usage test

* fix snapshot

* fix loggerd test

* bump cereal

Co-authored-by: Comma Device <device@comma.ai>
Co-authored-by: Joost Wooning <jwooning@gmail.com>
2022-06-01 17:18:28 +02:00
HaraldSchafer
7da48ebdba New model: better laneless (#24605)
* Fix ping pong

* Fix hugging: e6fd3a62-aa86-46d1-a414-19137ed1babc/800

* Remove lateral toggle

* a6da6265-1952-48dd-986b-2b5996dd2642/950

* Model replay commit

* Revert "Remove lateral toggle"

This reverts commit 69e57eb7e1a72e4bad5e12bcfded6ff431a21a0e.
2022-05-24 09:11:31 -07:00
Adeeb Shihadeh
cb8885cffb Merge common/ and selfdrive/common (#24556)
* Merge common/ and selfdrive/common

* fix that

* fix version

* fix unit tests
2022-05-18 14:11:57 -07:00
HaraldSchafer
e2e5378dd3 Revert "Joint model (#24521)" (#24581)
This reverts commit 7996fd37ef.
2022-05-18 10:56:45 -07:00
HaraldSchafer
7996fd37ef Joint model (#24521)
* update models

* wip lanelines mhp parsing

* typos

* newer models

* 95995a49-db0c-4261-8776-b90780dc2a8c/600

* Use laneful policy to prevent planner aggression

* c8c3ab69-bd2c-4d1c-a7f8-9e8457e4827b/950

* ede8f75a-df3e-41fb-bc56-63e2f33858e5/950

* 7205c5af-7532-4215-ad9b-678789054db7/995

* bfb34fa3-1d72-4987-854f-43d42ee01015/950

* change cost

* 8ae2477a-b452-4601-b04c-d3af66ea5e98/950

* Update heading cost

* 9b991314-94f5-4f0b-872e-58d95266e4d0/333

* 9b991314-94f5-4f0b-872e-58d95266e4d0/950

* ede8f75a-df3e-41fb-bc56-63e2f33858e5/950

* undo lateral planner changes

* rm whitespace

* update ref

* c0acc19d-f2f1-47ae-8838-5382c6c6c7f5/950

* c0acc19d-f2f1-47ae-8838-5382c6c6c7f5/950

* update model refg

* Update model ref

Co-authored-by: Yassine Yousfi <yyousfi1@binghamton.edu>
2022-05-16 17:26:23 -07:00
Adeeb Shihadeh
319f3d1475 move models to selfdrive/modeld/ (#24539)
* move models to selfdrive/modeld/

* two more refs
2022-05-15 17:29:09 -07:00