KerryGoldModel, AGNOS12.3, ButtonMode3, autoDetectLFA2, (#181)

* fix.. speed_limit error... * draw tpms settings. * fix.. traffic light stopping only.. * fix.. waze cam * fix.. waze... * add setting (Enable comma connect ) * auto detect LFA2 * fix.. cruisespeed1 * vff2 driving model. * fix.. * agnos 12.3 * fix.. * ff * ff * test * ff * fix.. drawTurnInfo.. * Update drive_helpers.py * fix.. support eng voice eng sounds fix settings... english fix.. mph.. fix.. roadlimit speed bug.. * new vff model.. 250608 * fix soundd.. * fix safe exit speed.. * fix.. sounds. * fix.. radar timeStep.. * KerryGold model * Update drive_helpers.py * fix.. model. * fix.. * fix.. * Revert "fix.." This reverts commit b09ec459afb855c533d47fd7e8a1a6b1a09466e7. * Revert "fix.." This reverts commit 290bec6b83a4554ca232d531a911edccf94a2156. * fix esim * add more acc table. 10kph * kg update.. * fix cruisebutton mode3 * test atc..cond. * fix.. canfd * fix.. angle control limit
2025-06-13 15:59:36 +09:00 · 2025-06-13 15:59:36 +09:00 · efee1712aa
commit efee1712aa
parent 15e6ed3e3b
432 changed files with 25034 additions and 1096533 deletions
--- a/cereal/init.py
+++ b/cereal/init.py
@ -1,9 +1,11 @@
 import os
 import capnp
+from importlib.resources import as_file, files

-CEREAL_PATH = os.path.dirname(os.path.abspath(__file__))
 capnp.remove_import_hook()

+with as_file(files("cereal")) as fspath:
+  CEREAL_PATH = fspath.as_posix()
  log = capnp.load(os.path.join(CEREAL_PATH, "log.capnp"))
  car = capnp.load(os.path.join(CEREAL_PATH, "car.capnp"))
  custom = capnp.load(os.path.join(CEREAL_PATH, "custom.capnp"))
--- a/cereal/log.capnp
+++ b/cereal/log.capnp
@ -48,6 +48,7 @@ struct OnroadEvent @0xc4fa6047f024e718 {
    preEnableStandstill @12;  # added during pre-enable state with brake
    gasPressedOverride @13;  # added when user is pressing gas with no disengage on gas
    steerOverride @14;
+    steerDisengage @94;  # exits active state
    cruiseDisabled @15;
    speedTooLow @16;
    outOfSpace @17;
@ -126,10 +127,11 @@ struct OnroadEvent @0xc4fa6047f024e718 {
    espActive @90;
    personalityChanged @91;
    aeb @92;
+    userFlag @95;

    softHold @115;
-    trafficStopping @94;
-    audioPrompt @95;
+    trafficStopping @116;
+    audioPrompt @117;
    audioRefuse @96;
    stopStop @97;
    audioLaneChange @98;
@ -512,6 +514,7 @@ struct DeviceState @0xa4d8b5af2aa492eb {
  # device thermals
  cpuTempC @26 :List(Float32);
  gpuTempC @27 :List(Float32);
+  dspTempC @49 :Float32;
  memoryTempC @28 :Float32;
  nvmeTempC @35 :List(Float32);
  modemTempC @36 :List(Float32);
--- a/common/params_keys.h
+++ b/common/params_keys.h
@ -140,6 +140,7 @@ inline static std::unordered_map<std::string, uint32_t> keys = {
    {"SupportedCars", PERSISTENT},
    {"SupportedCars_gm", PERSISTENT},
    {"ShowDebugUI", PERSISTENT},
+    {"ShowTpms", PERSISTENT},
    {"ShowDateTime", PERSISTENT},
    {"ShowPathEnd", PERSISTENT},
    {"ShowCustomBrightness", PERSISTENT},
@ -203,6 +204,7 @@ inline static std::unordered_map<std::string, uint32_t> keys = {
    {"TrafficLightDetectMode", PERSISTENT},
    {"SteerActuatorDelay", PERSISTENT},
    {"CruiseOnDist", PERSISTENT},
+    {"CruiseMaxVals0", PERSISTENT},
    {"CruiseMaxVals1", PERSISTENT},
    {"CruiseMaxVals2", PERSISTENT},
    {"CruiseMaxVals3", PERSISTENT},
@ -262,6 +264,7 @@ inline static std::unordered_map<std::string, uint32_t> keys = {
    {"SpeedFromPCM", PERSISTENT},
    {"MaxTimeOffroadMin", PERSISTENT},
    {"DisableDM", PERSISTENT},
+    {"EnableConnect", PERSISTENT},
    {"MuteDoor", PERSISTENT},
    {"MuteSeatbelt", PERSISTENT},
    {"CarrotException", CLEAR_ON_MANAGER_START},
--- a/common/util.py
+++ b/common/util.py
@ -1,3 +1,25 @@
+import os
+import subprocess
+
+def sudo_write(val: str, path: str) -> None:
+  try:
+    with open(path, 'w') as f:
+      f.write(str(val))
+  except PermissionError:
+    os.system(f"sudo chmod a+w {path}")
+    try:
+      with open(path, 'w') as f:
+        f.write(str(val))
+    except PermissionError:
+      # fallback for debugfs files
+      os.system(f"sudo su -c 'echo {val} > {path}'")
+
+def sudo_read(path: str) -> str:
+  try:
+    return subprocess.check_output(f"sudo cat {path}", shell=True, encoding='utf8').strip()
+  except Exception:
+    return ""
+
 class MovingAverage:
  def __init__(self, window_size: int):
    self.window_size: int = window_size
--- a/launch_env.sh
+++ b/launch_env.sh
@ -7,7 +7,7 @@ export OPENBLAS_NUM_THREADS=1
 export VECLIB_MAXIMUM_THREADS=1

 if [ -z "$AGNOS_VERSION" ]; then
-  export AGNOS_VERSION="12.1"
+  export AGNOS_VERSION="12.3"
 fi

 export STAGING_ROOT="/data/safe_staging"
--- a/opendbc_repo/opendbc/car/car.capnp
+++ b/opendbc_repo/opendbc/car/car.capnp
@ -798,5 +798,5 @@ struct CarParams {
  maxSteeringAngleDegDEPRECATED @54 :Float32;
  longitudinalActuatorDelayLowerBoundDEPRECATED @61 :Float32;
  stoppingControlDEPRECATED @31 :Bool; # Does the car allow full control even at lows speeds when stopping
-  radarTimeStepDEPRECATED @45: Float32 = 0.05;  # time delta between radar updates, 20Hz is very standard
+  radarTimeStep @45: Float32;  # time delta between radar updates, 20Hz is very standard
 }
--- a/opendbc_repo/opendbc/car/hyundai/carstate.py
+++ b/opendbc_repo/opendbc/car/hyundai/carstate.py
@ -450,7 +450,7 @@ class CarState(CarStateBase):
        self.hda_info_4a3 = copy.copy(cp.vl.get("HDA_INFO_4A3", {}))
        speedLimit = self.hda_info_4a3["SPEED_LIMIT"]
        if not self.is_metric:
-          speedLimit = CV.MPH_TO_KPH
+          speedLimit *= CV.MPH_TO_KPH
        ret.speedLimit = speedLimit if speedLimit < 255 else 0
        if int(self.hda_info_4a3["NEW_SIGNAL_4"]) == 17:
          speed_limit_cam = True
--- a/opendbc_repo/opendbc/car/hyundai/interface.py
+++ b/opendbc_repo/opendbc/car/hyundai/interface.py
@ -57,6 +57,10 @@ class CarInterface(CarInterfaceBase):
      if 0x4a3 in fingerprint[CAN.ECAN]:
        ret.extFlags |= HyundaiExtFlags.CANFD_4A3.value

+      if 203 in fingerprint[CAN.CAM]: # LFA_ALT
+        print("##### Anglecontrol detected (LFA_ALT)")
+        ret.flags |= HyundaiFlags.ANGLE_CONTROL.value
+
      # detect HDA2 with ADAS Driving ECU
      if hda2:
        print("$$$CANFD HDA2")
@ -188,6 +192,8 @@ class CarInterface(CarInterfaceBase):

    #ret.radarUnavailable = False  # TODO: canfd... carrot, hyundai cars have radar

+    ret.radarTimeStep = 0.05 if params.get_int("EnableRadarTracks") > 0 else 0.02
+
    ret.pcmCruise = not ret.openpilotLongitudinalControl
    ret.startingState = False # True  # carrot
    ret.vEgoStarting = 0.1
--- a/opendbc_repo/opendbc/car/hyundai/values.py
+++ b/opendbc_repo/opendbc/car/hyundai/values.py
@ -16,7 +16,7 @@ class CarControllerParams:
  ACCEL_MAX = 2.5 # m/s
  ANGLE_LIMITS: AngleSteeringLimits = AngleSteeringLimits(
    # LKAS angle command is unlimited, but LFA is limited to 176.7 deg (but does not fault if requesting above)
-    180, # deg
+    175, # deg
    # stock comma's
    #([0, 9, 16, 25], [1.4, 0.6, 0.4, 0.1]),
    #([0, 9, 16, 25], [1.4, 0.7, 0.5, 0.1]),
--- a/opendbc_repo/opendbc/car/interfaces.py
+++ b/opendbc_repo/opendbc/car/interfaces.py
@ -154,8 +154,12 @@ class RadarInterfaceBase(ABC):
    self.init_done = False

  def estimate_dt(self, rcv_time):
-    if len(self.init_samples) > 50:
-      estimated_dt = np.mean(np.diff(self.init_samples[20:]))
+    if self.CP.radarTimeStep > 0.0:
+      self.dt = self.CP.radarTimeStep
+      self.init_done = True
+      print(f"Using radar dt: {self.dt} sec")
+    elif len(self.init_samples) > 100:
+      estimated_dt = np.mean(np.diff(self.init_samples[50:]))
      self.dt = estimated_dt
      self.init_done = True
      print(f"Estimated radar dt: {self.dt} sec")
--- a/opendbc_repo/opendbc/dbc/hyundai_canfd.dbc
+++ b/opendbc_repo/opendbc/dbc/hyundai_canfd.dbc
@ -260,6 +260,7 @@ BO_ 352 ADRV_0x160: 16 ADRV
 SG_ SET_ME_FC : 72|8@1+ (1,0) [0|255] "" XXX
 SG_ SET_ME_9 : 80|8@1+ (1,0) [0|255] "" XXX
 SG_ NEW_SIGNAL_3 : 95|8@0+ (1,0) [0|255] "" XXX
+ SG_ NEW_SIGNAL_7 : 102|1@0+ (1,0) [0|1] "" XXX

 BO_ 353 ADRV_0x161: 32 ADRV
 SG_ CHECKSUM : 0|16@1+ (1,0) [0|65535] "" XXX
--- a/selfdrive/assets/sounds_eng/audio_1.wav
+++ b/selfdrive/assets/sounds_eng/audio_1.wav
--- a/selfdrive/assets/sounds_eng/audio_10.wav
+++ b/selfdrive/assets/sounds_eng/audio_10.wav
--- a/selfdrive/assets/sounds_eng/audio_2.wav
+++ b/selfdrive/assets/sounds_eng/audio_2.wav
--- a/selfdrive/assets/sounds_eng/audio_3.wav
+++ b/selfdrive/assets/sounds_eng/audio_3.wav
--- a/selfdrive/assets/sounds_eng/audio_4.wav
+++ b/selfdrive/assets/sounds_eng/audio_4.wav
--- a/selfdrive/assets/sounds_eng/audio_5.wav
+++ b/selfdrive/assets/sounds_eng/audio_5.wav
--- a/selfdrive/assets/sounds_eng/audio_6.wav
+++ b/selfdrive/assets/sounds_eng/audio_6.wav
--- a/selfdrive/assets/sounds_eng/audio_7.wav
+++ b/selfdrive/assets/sounds_eng/audio_7.wav
--- a/selfdrive/assets/sounds_eng/audio_8.wav
+++ b/selfdrive/assets/sounds_eng/audio_8.wav
--- a/selfdrive/assets/sounds_eng/audio_9.wav
+++ b/selfdrive/assets/sounds_eng/audio_9.wav
--- a/selfdrive/assets/sounds_eng/audio_auto_hold.wav
+++ b/selfdrive/assets/sounds_eng/audio_auto_hold.wav
--- a/selfdrive/assets/sounds_eng/audio_car_watchout.wav
+++ b/selfdrive/assets/sounds_eng/audio_car_watchout.wav
--- a/selfdrive/assets/sounds_eng/audio_disengage.wav
+++ b/selfdrive/assets/sounds_eng/audio_disengage.wav
--- a/selfdrive/assets/sounds_eng/audio_engage.wav
+++ b/selfdrive/assets/sounds_eng/audio_engage.wav
--- a/selfdrive/assets/sounds_eng/audio_lane_change.wav
+++ b/selfdrive/assets/sounds_eng/audio_lane_change.wav
--- a/selfdrive/assets/sounds_eng/audio_lanechange.wav
+++ b/selfdrive/assets/sounds_eng/audio_lanechange.wav
--- a/selfdrive/assets/sounds_eng/audio_speed_down.wav
+++ b/selfdrive/assets/sounds_eng/audio_speed_down.wav
--- a/selfdrive/assets/sounds_eng/audio_stopping.wav
+++ b/selfdrive/assets/sounds_eng/audio_stopping.wav
--- a/selfdrive/assets/sounds_eng/audio_stopstop.wav
+++ b/selfdrive/assets/sounds_eng/audio_stopstop.wav
--- a/selfdrive/assets/sounds_eng/audio_traffic_error.wav
+++ b/selfdrive/assets/sounds_eng/audio_traffic_error.wav
--- a/selfdrive/assets/sounds_eng/audio_turn.wav
+++ b/selfdrive/assets/sounds_eng/audio_turn.wav
--- a/selfdrive/assets/sounds_eng/audio_turn2.wav
+++ b/selfdrive/assets/sounds_eng/audio_turn2.wav
--- a/selfdrive/assets/sounds_eng/disengage.wav
+++ b/selfdrive/assets/sounds_eng/disengage.wav
--- a/selfdrive/assets/sounds_eng/engage.wav
+++ b/selfdrive/assets/sounds_eng/engage.wav
--- a/selfdrive/assets/sounds_eng/nnff.wav
+++ b/selfdrive/assets/sounds_eng/nnff.wav
--- a/selfdrive/assets/sounds_eng/prompt.wav
+++ b/selfdrive/assets/sounds_eng/prompt.wav
--- a/selfdrive/assets/sounds_eng/prompt_distracted.wav
+++ b/selfdrive/assets/sounds_eng/prompt_distracted.wav
--- a/selfdrive/assets/sounds_eng/refuse.wav
+++ b/selfdrive/assets/sounds_eng/refuse.wav
--- a/selfdrive/assets/sounds_eng/reverse_gear.wav
+++ b/selfdrive/assets/sounds_eng/reverse_gear.wav
--- a/selfdrive/assets/sounds_eng/tici_disengaged.wav
+++ b/selfdrive/assets/sounds_eng/tici_disengaged.wav
--- a/selfdrive/assets/sounds_eng/tici_engaged.wav
+++ b/selfdrive/assets/sounds_eng/tici_engaged.wav
--- a/selfdrive/assets/sounds_eng/traffic_sign_changed.wav
+++ b/selfdrive/assets/sounds_eng/traffic_sign_changed.wav
--- a/selfdrive/assets/sounds_eng/traffic_sign_green.wav
+++ b/selfdrive/assets/sounds_eng/traffic_sign_green.wav
--- a/selfdrive/assets/sounds_eng/warning_immediate.wav
+++ b/selfdrive/assets/sounds_eng/warning_immediate.wav
--- a/selfdrive/assets/sounds_eng/warning_soft.wav
+++ b/selfdrive/assets/sounds_eng/warning_soft.wav
--- a/selfdrive/car/cruise.py
+++ b/selfdrive/car/cruise.py
@ -236,7 +236,7 @@ class VCruiseCarrot:
      #self.event = event
      self._log_timer = self._log_timeout

-  def update_params(self):
+  def update_params(self, is_metric):
    if self.frame % 10 == 0:
      self.autoCruiseControl = self.params.get_int("AutoCruiseControl")
      self.autoGasTokSpeed = self.params.get_int("AutoGasTokSpeed")
@ -252,17 +252,25 @@ class VCruiseCarrot:
      self._paddle_mode = self.params.get_int("PaddleMode")
      self._cruise_button_mode = self.params.get_int("CruiseButtonMode")
      self._lfa_button_mode = self.params.get_int("LfaButtonMode")
+      self.autoRoadSpeedLimitOffset = self.params.get_int("AutoRoadSpeedLimitOffset")
+      self.autoNaviSpeedSafetyFactor = self.params.get_float("AutoNaviSpeedSafetyFactor") * 0.01
      self.cruiseOnDist = self.params.get_float("CruiseOnDist") * 0.01
-      cruiseSpeed1 = self.params.get_float("CruiseSpeed1")
-      cruiseSpeed2 = self.params.get_float("CruiseSpeed2")
-      cruiseSpeed3 = self.params.get_float("CruiseSpeed3")
-      cruiseSpeed4 = self.params.get_float("CruiseSpeed4")
-      cruiseSpeed5 = self.params.get_float("CruiseSpeed5")
+      unit_factor = 1.0 if is_metric else CV.MPH_TO_KPH
+      cruiseSpeed1 = self.params.get_float("CruiseSpeed1") * unit_factor
+      cruiseSpeed2 = self.params.get_float("CruiseSpeed2") * unit_factor
+      cruiseSpeed3 = self.params.get_float("CruiseSpeed3") * unit_factor
+      cruiseSpeed4 = self.params.get_float("CruiseSpeed4") * unit_factor
+      cruiseSpeed5 = self.params.get_float("CruiseSpeed5") * unit_factor
+      if cruiseSpeed1 <= 0:
+        if self.autoRoadSpeedLimitOffset < 0:
+          cruiseSpeed1 = self.nRoadLimitSpeed * self.autoNaviSpeedSafetyFactor
+        else:
+          cruiseSpeed1 = self.nRoadLimitSpeed + self.autoRoadSpeedLimitOffset
      self._cruise_speed_table = [cruiseSpeed1, cruiseSpeed2, cruiseSpeed3, cruiseSpeed4, cruiseSpeed5]

  def update_v_cruise(self, CS, sm, is_metric):
    self._add_log("")
-    self.update_params()
+    self.update_params(is_metric)
    self.frame += 1
    if CS.gearShifter != GearShifter.drive:
      self.autoCruiseControl_cancel_timer = 20 * 100  # 20 sec
--- a/selfdrive/carrot/carrot_functions.py
+++ b/selfdrive/carrot/carrot_functions.py
@ -39,7 +39,7 @@ class TrafficState(Enum):
  def __str__(self):
    return self.name

-A_CRUISE_MAX_BP_CARROT = [0., 40 * CV.KPH_TO_MS, 60 * CV.KPH_TO_MS, 80 * CV.KPH_TO_MS, 110 * CV.KPH_TO_MS, 140 * CV.KPH_TO_MS]
+A_CRUISE_MAX_BP_CARROT = [0., 10 * CV.KPH_TO_MS, 40 * CV.KPH_TO_MS, 60 * CV.KPH_TO_MS, 80 * CV.KPH_TO_MS, 110 * CV.KPH_TO_MS, 140 * CV.KPH_TO_MS]

 class CarrotPlanner:
  def __init__(self):
@ -102,6 +102,7 @@ class CarrotPlanner:
    self.dynamicTFollow = 0.0
    self.dynamicTFollowLC = 0.0

+    self.cruiseMaxVals0 = 1.6
    self.cruiseMaxVals1 = 1.6
    self.cruiseMaxVals2 = 1.2
    self.cruiseMaxVals3 = 1.0
@ -162,6 +163,7 @@ class CarrotPlanner:
      self.dynamicTFollow = self.params.get_float("DynamicTFollow") / 100.
      self.dynamicTFollowLC = self.params.get_float("DynamicTFollowLC") / 100.
    elif self.params_count == 30:
+      self.cruiseMaxVals0 = self.params.get_float("CruiseMaxVals0") / 100.
      self.cruiseMaxVals1 = self.params.get_float("CruiseMaxVals1") / 100.
      self.cruiseMaxVals2 = self.params.get_float("CruiseMaxVals2") / 100.
      self.cruiseMaxVals3 = self.params.get_float("CruiseMaxVals3") / 100.
@ -179,7 +181,7 @@ class CarrotPlanner:
      self.params_count = 0

  def get_carrot_accel(self, v_ego):
-    cruiseMaxVals = [self.cruiseMaxVals1, self.cruiseMaxVals2, self.cruiseMaxVals3, self.cruiseMaxVals4, self.cruiseMaxVals5, self.cruiseMaxVals6]
+    cruiseMaxVals = [self.cruiseMaxVals0, self.cruiseMaxVals1, self.cruiseMaxVals2, self.cruiseMaxVals3, self.cruiseMaxVals4, self.cruiseMaxVals5, self.cruiseMaxVals6]
    factor = self.myHighModeFactor if self.myDrivingMode == DrivingMode.High else self.mySafeFactor
    return np.interp(v_ego, A_CRUISE_MAX_BP_CARROT, cruiseMaxVals) * factor

@ -387,8 +389,6 @@ class CarrotPlanner:

    if self.myDrivingMode == DrivingMode.High or self.trafficLightDetectMode == 0:
      self.trafficState = TrafficState.off
-    if self.trafficState == TrafficState.green and self.trafficLightDetectMode == 1:  # Stopping only
-      self.trafficState = TrafficState.off
    if abs(carstate.steeringAngleDeg) > 20:
      self.trafficState = TrafficState.off

@ -407,7 +407,7 @@ class CarrotPlanner:
      elif lead_detected and (radarstate.leadOne.dRel - stop_model_x) < 2.0:
        self.xState = XState.lead
      elif self.stopping_count == 0:
-        if self.trafficState == TrafficState.green and not self.carrot_stay_stop and not carstate.leftBlinker:
+        if self.trafficState == TrafficState.green and not self.carrot_stay_stop and not carstate.leftBlinker and self.trafficLightDetectMode != 1:
          self.xState = XState.e2ePrepare
          self.events.add(EventName.trafficSignGreen)
      self.stopping_count = max(0, self.stopping_count - 1)
@ -501,7 +501,7 @@ class DrivingModeDetector:
        self.speed_threshold = 2  # (km/h)
        self.accel_threshold = 1.5  # (m/s^2)
        self.distance_threshold = 12  # (m)
-        self.lead_speed_exit_threshold = 20  # (km/h)
+        self.lead_speed_exit_threshold = 35  # (km/h)

    def update_data(self, my_speed, lead_speed, my_accel, lead_accel, distance):
        # 1. 정체 조건: 앞차가 가까이 있고 정지된 상황
--- a/selfdrive/carrot/carrot_man.py
+++ b/selfdrive/carrot/carrot_man.py
@ -853,6 +853,7 @@ class CarrotMan:
    host = '0.0.0.0'  # 혹은 다른 호스트 주소
    port = 7709  # 포트 번호

+    try:
      with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
        s.bind((host, port))
        s.listen()
@ -915,7 +916,9 @@ class CarrotMan:

            except Exception as e:
              print(e)
-
+    except Exception as e:
+      print("################# CarrotMan route server error #####################")
+      print(e)

  def carrot_curve_speed_params(self):
    self.autoCurveSpeedFactor = self.params.get_int("AutoCurveSpeedFactor")*0.01
@ -1553,7 +1556,7 @@ class CarrotServ:
      print(f"{id_str}: {distance} m")
      xSpdType = -1
      if 'camera' in id_str:
-        xSpdType = 1
+        xSpdType = 101    # 101: waze speed cam, 100: police
      elif 'police' in id_str:
        xSpdType = 100

@ -1616,7 +1619,7 @@ class CarrotServ:
      self.nGoPosDist = 0
      self.update_nav_instruction(sm)

-    if self.xSpdType < 0 or (self.xSpdType != 100 and self.xSpdDist <= 0) or (self.xSpdType == 100 and self.xSpdDist < -250):
+    if self.xSpdType < 0 or (self.xSpdType not in [100,101] and self.xSpdDist <= 0) or (self.xSpdType in [100,101] and self.xSpdDist < -250):
      self.xSpdType = -1
      self.xSpdDist = self.xSpdLimit = 0
    if self.xTurnInfo < 0 or self.xDistToTurn < -50:
@ -1628,12 +1631,12 @@ class CarrotServ:
    sdi_speed = 250
    hda_active = False
    ### 과속카메라, 사고방지턱
-    if (self.xSpdDist > 0 or self.xSpdType == 100) and self.active_carrot > 0:
+    if (self.xSpdDist > 0 or self.xSpdType in [100, 101]) and self.active_carrot > 0:
      safe_sec = self.autoNaviSpeedBumpTime if self.xSpdType == 22 else self.autoNaviSpeedCtrlEnd
      decel = self.autoNaviSpeedDecelRate
      sdi_speed = min(sdi_speed, self.calculate_current_speed(self.xSpdDist, self.xSpdLimit, safe_sec, decel))
      self.active_carrot = 5 if self.xSpdType == 22 else 3
-      if self.xSpdType == 4 or (self.xSpdType == 100 and self.xSpdDist <= 0):
+      if self.xSpdType == 4 or (self.xSpdType in [100, 101] and self.xSpdDist <= 0):
        sdi_speed = self.xSpdLimit
        self.active_carrot = 4
    elif CS is not None and CS.speedLimit > 0 and CS.speedLimitDistance > 0:
--- a/selfdrive/carrot_settings.json
+++ b/selfdrive/carrot_settings.json
@ -302,11 +302,24 @@
    },
    {
      "group": "가속설정",
-      "name": "CruiseMaxVals1",
-      "title": "가속설정1:0km/h(160)",
+      "name": "CruiseMaxVals0",
+      "title": "가속설정0:0km/h(160)",
      "descr": "",
      "egroup": "ACCEL",
-      "etitle": "AccelSet1:0km/h(160)",
+      "etitle": "AccelSet0:0km/h(160)",
+      "edescr": "",
+      "min": 1,
+      "max": 250,
+      "default": 160,
+      "unit": 5
+    },
+    {
+      "group": "가속설정",
+      "name": "CruiseMaxVals1",
+      "title": "가속설정1:10km/h(160)",
+      "descr": "",
+      "egroup": "ACCEL",
+      "etitle": "AccelSet1:10km/h(160)",
      "edescr": "",
      "min": 1,
      "max": 250,
@ -642,11 +655,11 @@
      "group": "버튼설정",
      "name": "CruiseSpeed1",
      "title": "크루즈속도1(30)",
-      "descr": "버튼모드3, 단계별속도",
+      "descr": "버튼모드3, 단계별속도, 0:RoadLimitSpeed+offset",
      "egroup": "BUTN",
      "etitle": "Cruise Speed1(30)",
-      "edescr": "CruiseButtonMode2, prefer speed",
-      "min": 1,
+      "edescr": "CruiseButtonMode3, prefer speed, 0:RoadLimitSpeed+offset",
+      "min": 0,
      "max": 160,
      "default": 10,
      "unit": 10
@ -658,7 +671,7 @@
      "descr": "버튼모드3, 단계별속도",
      "egroup": "BUTN",
      "etitle": "Cruise Speed2(50)",
-      "edescr": "CruiseButtonMode2, prefer speed",
+      "edescr": "CruiseButtonMode3, prefer speed",
      "min": 1,
      "max": 160,
      "default": 10,
@ -671,7 +684,7 @@
      "descr": "버튼모드3, 단계별속도",
      "egroup": "BUTN",
      "etitle": "Cruise Speed3(80)",
-      "edescr": "CruiseButtonMode2, prefer speed",
+      "edescr": "CruiseButtonMode3, prefer speed",
      "min": 1,
      "max": 160,
      "default": 10,
@ -684,7 +697,7 @@
      "descr": "버튼모드3, 단계별속도",
      "egroup": "BUTN",
      "etitle": "Cruise Speed4(110)",
-      "edescr": "CruiseButtonMode2, prefer speed",
+      "edescr": "CruiseButtonMode3, prefer speed",
      "min": 1,
      "max": 160,
      "default": 10,
@ -697,7 +710,7 @@
      "descr": "버튼모드3, 단계별속도",
      "egroup": "BUTN",
      "etitle": "Cruise Speed5(130)",
-      "edescr": "CruiseButtonMode2, prefer speed",
+      "edescr": "CruiseButtonMode3, prefer speed",
      "min": 1,
      "max": 160,
      "default": 10,
@ -837,10 +850,23 @@
      "group": "시작",
      "name": "DisableDM",
      "title": "DisableDM",
-      "descr": "",
+      "descr": "Reboot필요",
      "egroup": "START",
      "etitle": "DisableDM",
-      "edescr": "",
+      "edescr": "Reboot required",
+      "min": 0,
+      "max": 1,
+      "default": 0,
+      "unit": 1
+    },
+    {
+      "group": "시작",
+      "name": "EnableConnect",
+      "title": "EnableConnect",
+      "descr": "Reboot, 콤마 Connect에 연결합니다. 상황에 따라 BAN될수 있습니다.",
+      "egroup": "START",
+      "etitle": "EnableConnect",
+      "edescr": "Reboot required, Enable comma connect, Your device may be banned by Comma",
      "min": 0,
      "max": 1,
      "default": 0,
@ -925,6 +951,19 @@
      "default": 1,
      "unit": 1
    },
+    {
+      "group": "화면",
+      "name": "ShowTpms",
+      "title": "TPMS정보 표시",
+      "descr": "0:None, 1: 위, 2: 아래, 3: 둘다",
+      "egroup": "DISP",
+      "etitle": "Show TPMS Info",
+      "edescr": "0:None, 1: Upper, 2:Lower, 3: Both",
+      "min": 0,
+      "max": 3,
+      "default": 1,
+      "unit": 1
+    },
    {
      "group": "화면",
      "name": "ShowDateTime",
--- a/selfdrive/controls/lib/desire_helper.py
+++ b/selfdrive/controls/lib/desire_helper.py
@ -149,7 +149,7 @@ class DesireHelper:
                                                                                                    modeldata.laneLines[2], modeldata.roadEdges[1])
    self.lane_exist_left_count.update(lane_prob_left)
    self.lane_exist_right_count.update(lane_prob_right)
-    min_lane_width = 2.0
+    min_lane_width = 2.8
    self.lane_width_left_count.update(self.lane_width_left > min_lane_width)
    self.lane_width_right_count.update(self.lane_width_right > min_lane_width)
    self.road_edge_left_count.update(self.distance_to_road_edge_left > min_lane_width)
@ -269,7 +269,7 @@ class DesireHelper:
      auto_lane_change_available = lane_available
    else:
      auto_lane_change_blocked = ((atc_blinker_state == BLINKER_LEFT) and (driver_blinker_state != BLINKER_LEFT))
-      auto_lane_change_available = not auto_lane_change_blocked and lane_availabled and edge_availabled and not side_object_detected
+      auto_lane_change_available = not auto_lane_change_blocked and (lane_availabled or edge_availabled or lane_appeared) and not side_object_detected

    if not lateral_active or self.lane_change_timer > LANE_CHANGE_TIME_MAX:
      #print("Desire canceled")
--- a/selfdrive/controls/lib/drive_helpers.py
+++ b/selfdrive/controls/lib/drive_helpers.py
@ -49,12 +49,12 @@ def get_lag_adjusted_curvature(CP, v_ego, psis, curvatures, steer_actuator_delay
  #average_curvature_desired = psi / (v_ego * delay)

  # curve -> straight or reverse curve
-  if abs(current_curvature_desired) > 0.002 and \
+  if v_ego > 5 and abs(current_curvature_desired) > 0.002 and \
     (abs(future_curvature_desired) < 0.001 or np.sign(current_curvature_desired) != np.sign(future_curvature_desired)):
    psis_damping = 0.2
  else:
    psis_damping = 1.0
-  psi *= psis_damping
+  #psi *= psis_damping


  average_curvature_desired = psi / distance
--- a/selfdrive/modeld/SConscript
+++ b/selfdrive/modeld/SConscript
@ -1,3 +1,4 @@
+import os
 import glob

 Import('env', 'envCython', 'arch', 'cereal', 'messaging', 'common', 'gpucommon', 'visionipc', 'transformations')
@ -13,7 +14,6 @@ common_src = [
  "transforms/transform.cc",
 ]

-
 # OpenCL is a framework on Mac
 if arch == "Darwin":
  frameworks += ['OpenCL']
@ -38,17 +38,35 @@ for model_name in ['driving_vision', 'driving_policy']:
  cmd = f'python3 {Dir("#selfdrive/modeld").abspath}/get_model_metadata.py {fn}.onnx'
  lenv.Command(fn + "_metadata.pkl", [fn + ".onnx"] + tinygrad_files + script_files, cmd)

-# Compile tinygrad model
+def tg_compile(flags, model_name):
  pythonpath_string = 'PYTHONPATH="${PYTHONPATH}:' + env.Dir("#tinygrad_repo").abspath + '"'
-if arch == 'larch64':
-  device_string = 'QCOM=1'
-elif arch == 'Darwin':
-  device_string = 'CLANG=1 IMAGE=0'
-else:
-  device_string = 'LLVM=1 LLVMOPT=1 BEAM=0 IMAGE=0'
-
-for model_name in ['driving_vision', 'driving_policy', 'dmonitoring_model']:
  fn = File(f"models/{model_name}").abspath
-  cmd = f'{pythonpath_string} {device_string} python3 {Dir("#tinygrad_repo").abspath}/examples/openpilot/compile3.py {fn}.onnx {fn}_tinygrad.pkl'
-  lenv.Command(fn + "_tinygrad.pkl", [fn + ".onnx"] + tinygrad_files, cmd)
+  return lenv.Command(
+    fn + "_tinygrad.pkl",
+    [fn + ".onnx"] + tinygrad_files,
+    f'{pythonpath_string} {flags} python3 {Dir("#tinygrad_repo").abspath}/examples/openpilot/compile3.py {fn}.onnx {fn}_tinygrad.pkl'
+  )

+# Compile small models
+for model_name in ['driving_vision', 'driving_policy', 'dmonitoring_model']:
+  flags = {
+    'larch64': 'QCOM=1',
+    'Darwin': 'CPU=1 IMAGE=0 JIT=2',
+  }.get(arch, 'LLVM=1 LLVMOPT=1 BEAM=0 IMAGE=0 JIT=2')
+  tg_compile(flags, model_name)
+
+# Compile BIG model if USB GPU is available
+import subprocess
+from tinygrad import Device
+
+# because tg doesn't support multi-process
+devs = subprocess.check_output('python3 -c "from tinygrad import Device; print(list(Device.get_available_devices()))"', shell=True)
+if b"AMD" in devs:
+  del Device
+  print("USB GPU detected... building")
+  flags = "AMD=1 AMD_IFACE=USB AMD_LLVM=1 NOLOCALS=0 IMAGE=0"
+  bp = tg_compile(flags, "big_driving_policy")
+  bv = tg_compile(flags, "big_driving_vision")
+  lenv.SideEffect('lock', [bp, bv])  # tg doesn't support multi-process so build serially
+else:
+  print("USB GPU not detected... skipping")
--- a/selfdrive/modeld/dmonitoringmodeld.py
+++ b/selfdrive/modeld/dmonitoringmodeld.py
@ -173,9 +173,6 @@ def main():
    model_output, gpu_execution_time = model.run(buf, calib, model_transform)
    t2 = time.perf_counter()

-    # run one more time, just for the load
-    model.run(buf, calib, model_transform)
-
    pm.send("driverStateV2", get_driverstate_packet(model_output, vipc_client.frame_id, vipc_client.timestamp_sof, t2 - t1, gpu_execution_time))


--- a/selfdrive/modeld/modeld.py
+++ b/selfdrive/modeld/modeld.py
@ -1,13 +1,18 @@
 #!/usr/bin/env python3
 import os
 from openpilot.system.hardware import TICI
-from tinygrad.tensor import Tensor
-from tinygrad.dtype import dtypes
-if TICI:
+USBGPU = "USBGPU" in os.environ
+if USBGPU:
+  os.environ['AMD'] = '1'
+  os.environ['AMD_IFACE'] = 'USB'
+elif TICI:
  from openpilot.selfdrive.modeld.runners.tinygrad_helpers import qcom_tensor_from_opencl_address
  os.environ['QCOM'] = '1'
 else:
  os.environ['LLVM'] = '1'
+  os.environ['JIT'] = '2'
+from tinygrad.tensor import Tensor
+from tinygrad.dtype import dtypes
 import time
 import pickle
 import numpy as np
@ -33,8 +38,6 @@ from openpilot.selfdrive.modeld.constants import ModelConstants, Plan
 from openpilot.selfdrive.modeld.models.commonmodel_pyx import DrivingModelFrame, CLContext


-
-
 PROCESS_NAME = "selfdrive.modeld.modeld"
 SEND_RAW_PRED = os.getenv('SEND_RAW_PRED')

@ -45,10 +48,11 @@ POLICY_METADATA_PATH = Path(__file__).parent / 'models/driving_policy_metadata.p

 LAT_SMOOTH_SECONDS = 0.13
 LONG_SMOOTH_SECONDS = 0.3
+MIN_LAT_CONTROL_SPEED = 0.3


 def get_action_from_model(model_output: dict[str, np.ndarray], prev_action: log.ModelDataV2.Action,
-                          lat_action_t: float, long_action_t: float, lat_smooth_seconds: float) -> log.ModelDataV2.Action:
+                          lat_action_t: float, long_action_t: float, v_ego: float, lat_smooth_seconds: float) -> log.ModelDataV2.Action:
    plan = model_output['plan'][0]
    desired_accel, should_stop = get_accel_from_plan(plan[:,Plan.VELOCITY][:,0],
                                                     plan[:,Plan.ACCELERATION][:,0],
@ -57,7 +61,10 @@ def get_action_from_model(model_output: dict[str, np.ndarray], prev_action: log.
    desired_accel = smooth_value(desired_accel, prev_action.desiredAcceleration, LONG_SMOOTH_SECONDS)

    desired_curvature = model_output['desired_curvature'][0, 0]
+    if v_ego > MIN_LAT_CONTROL_SPEED:
      desired_curvature = smooth_value(desired_curvature, prev_action.desiredCurvature, lat_smooth_seconds)
+    else:
+      desired_curvature = prev_action.desiredCurvature

    return log.ModelDataV2.Action(desiredCurvature=float(desired_curvature),
                                  desiredAcceleration=float(desired_accel),
@ -79,10 +86,20 @@ class ModelState:
  prev_desire: np.ndarray  # for tracking the rising edge of the pulse

  def __init__(self, context: CLContext):
-    self.frames = {
-      'input_imgs': DrivingModelFrame(context, ModelConstants.TEMPORAL_SKIP),
-      'big_input_imgs': DrivingModelFrame(context, ModelConstants.TEMPORAL_SKIP)
-    }
+    with open(VISION_METADATA_PATH, 'rb') as f:
+      vision_metadata = pickle.load(f)
+      self.vision_input_shapes =  vision_metadata['input_shapes']
+      self.vision_input_names = list(self.vision_input_shapes.keys())
+      self.vision_output_slices = vision_metadata['output_slices']
+      vision_output_size = vision_metadata['output_shapes']['outputs'][1]
+
+    with open(POLICY_METADATA_PATH, 'rb') as f:
+      policy_metadata = pickle.load(f)
+      self.policy_input_shapes =  policy_metadata['input_shapes']
+      self.policy_output_slices = policy_metadata['output_slices']
+      policy_output_size = policy_metadata['output_shapes']['outputs'][1]
+
+    self.frames = {name: DrivingModelFrame(context, ModelConstants.TEMPORAL_SKIP) for name in self.vision_input_names}
    self.prev_desire = np.zeros(ModelConstants.DESIRE_LEN, dtype=np.float32)

    self.full_features_buffer = np.zeros((1, ModelConstants.FULL_HISTORY_BUFFER_LEN,  ModelConstants.FEATURE_LEN), dtype=np.float32)
@ -99,18 +116,6 @@ class ModelState:
      'features_buffer': np.zeros((1, ModelConstants.INPUT_HISTORY_BUFFER_LEN,  ModelConstants.FEATURE_LEN), dtype=np.float32),
    }

-    with open(VISION_METADATA_PATH, 'rb') as f:
-      vision_metadata = pickle.load(f)
-      self.vision_input_shapes =  vision_metadata['input_shapes']
-      self.vision_output_slices = vision_metadata['output_slices']
-      vision_output_size = vision_metadata['output_shapes']['outputs'][1]
-
-    with open(POLICY_METADATA_PATH, 'rb') as f:
-      policy_metadata = pickle.load(f)
-      self.policy_input_shapes =  policy_metadata['input_shapes']
-      self.policy_output_slices = policy_metadata['output_slices']
-      policy_output_size = policy_metadata['output_shapes']['outputs'][1]
-
    # img buffers are managed in openCL transform code
    self.vision_inputs: dict[str, Tensor] = {}
    self.vision_output = np.zeros(vision_output_size, dtype=np.float32)
@ -128,7 +133,7 @@ class ModelState:
    parsed_model_outputs = {k: model_outputs[np.newaxis, v] for k,v in output_slices.items()}
    return parsed_model_outputs

-  def run(self, buf: VisionBuf, wbuf: VisionBuf, transform: np.ndarray, transform_wide: np.ndarray,
+  def run(self, bufs: dict[str, VisionBuf], transforms: dict[str, np.ndarray],
                inputs: dict[str, np.ndarray], prepare_only: bool) -> dict[str, np.ndarray] | None:
    # Model decides when action is completed, so desire input is just a pulse triggered on rising edge
    inputs['desire'][0] = 0
@ -141,10 +146,9 @@ class ModelState:

    self.numpy_inputs['traffic_convention'][:] = inputs['traffic_convention']
    self.numpy_inputs['lateral_control_params'][:] = inputs['lateral_control_params']
-    imgs_cl = {'input_imgs': self.frames['input_imgs'].prepare(buf, transform.flatten()),
-               'big_input_imgs': self.frames['big_input_imgs'].prepare(wbuf, transform_wide.flatten())}
+    imgs_cl = {name: self.frames[name].prepare(bufs[name], transforms[name].flatten()) for name in self.vision_input_names}

-    if TICI:
+    if TICI and not USBGPU:
      # The imgs tensors are backed by opencl memory, only need init once
      for key in imgs_cl:
        if key not in self.vision_inputs:
@ -185,13 +189,17 @@ def main(demo=False):
  sentry.set_tag("daemon", PROCESS_NAME)
  cloudlog.bind(daemon=PROCESS_NAME)
  setproctitle(PROCESS_NAME)
+  if not USBGPU:
+    # USB GPU currently saturates a core so can't do this yet,
+    # also need to move the aux USB interrupts for good timings
    config_realtime_process(7, 54)

+  st = time.monotonic()
  cloudlog.warning("setting up CL context")
  cl_context = CLContext()
  cloudlog.warning("CL context ready; loading model")
  model = ModelState(cl_context)
-  cloudlog.warning("models loaded, modeld starting")
+  cloudlog.warning(f"models loaded in {time.monotonic() - st:.1f}s, modeld starting")

  # visionipc clients
  while True:
@ -330,6 +338,8 @@ def main(demo=False):
    if prepare_only:
      cloudlog.error(f"skipping model eval. Dropped {vipc_dropped_frames} frames")

+    bufs = {name: buf_extra if 'big' in name else buf_main for name in model.vision_input_names}
+    transforms = {name: model_transform_extra if 'big' in name else model_transform_main for name in model.vision_input_names}
    inputs:dict[str, np.ndarray] = {
      'desire': vec_desire,
      'traffic_convention': traffic_convention,
@ -337,7 +347,7 @@ def main(demo=False):
    }

    mt1 = time.perf_counter()
-    model_output = model.run(buf_main, buf_extra, model_transform_main, model_transform_extra, inputs, prepare_only)
+    model_output = model.run(bufs, transforms, inputs, prepare_only)
    mt2 = time.perf_counter()
    model_execution_time = mt2 - mt1

@ -346,7 +356,7 @@ def main(demo=False):
      drivingdata_send = messaging.new_message('drivingModelData')
      posenet_send = messaging.new_message('cameraOdometry')

-      action = get_action_from_model(model_output, prev_action, lat_delay + DT_MDL, long_delay + DT_MDL, lat_smooth_seconds)
+      action = get_action_from_model(model_output, prev_action, lat_delay + DT_MDL, long_delay + DT_MDL, v_ego, lat_smooth_seconds)
      prev_action = action
      fill_model_msg(drivingdata_send, modelv2_send, model_output, action,
                     publish_state, meta_main.frame_id, meta_extra.frame_id, frame_id,
--- a/selfdrive/modeld/models/big_driving_policy.onnx
+++ b/selfdrive/modeld/models/big_driving_policy.onnx
@ -0,0 +1 @@
+driving_policy.onnx
--- a/selfdrive/modeld/models/big_driving_vision.onnx
+++ b/selfdrive/modeld/models/big_driving_vision.onnx
@ -0,0 +1 @@
+driving_vision.onnx
--- a/selfdrive/modeld/models/driving_policy.onnx
+++ b/selfdrive/modeld/models/driving_policy.onnx
--- a/selfdrive/modeld/models/driving_vision.onnx
+++ b/selfdrive/modeld/models/driving_vision.onnx
--- a/selfdrive/ui/carrot.cc
+++ b/selfdrive/ui/carrot.cc
@ -2677,14 +2677,22 @@ void ui_draw(UIState *s, ModelRenderer* model_renderer, int w, int h) {
  drawCarrot.drawDateTime(s);
  //drawCarrot.drawConnInfo(s);
  drawCarrot.drawDeviceInfo(s);
+  int show_tpms = params.getInt("ShowTpms");
+  switch (show_tpms) {
+  case 0: break;
+  case 1:
    drawCarrot.drawTpms2(s);
-
-  int draw_turn_info = drawTurnInfo.draw(s);
-
-  if (draw_turn_info < 0) {
+    break;
+  case 2:
    drawCarrot.drawTpms3(s);
+    break;
+  case 3:
+    drawCarrot.drawTpms2(s);
+    drawCarrot.drawTpms3(s);
+    break;
  }

+  drawTurnInfo.draw(s);

  ui_draw_text_a2(s);
  ui_draw_alert(s);
--- a/selfdrive/ui/qt/offroad/settings.cc
+++ b/selfdrive/ui/qt/offroad/settings.cc
@ -702,14 +702,15 @@ CarrotPanel::CarrotPanel(QWidget* parent) : QWidget(parent) {
  latLongToggles->addItem(new CValueControl("StopDistanceCarrot", "LONG: StopDistance (600)cm", "", "../assets/offroad/icon_logic.png", 300, 1000, 10));
  //latLongToggles->addItem(new CValueControl("TraffStopDistanceAdjust", "LONG: TrafficStopDistance adjust(150)cm", "", "../assets/offroad/icon_road.png", -1000, 1000, 10));
  latLongToggles->addItem(new CValueControl("JLeadFactor3", "LONG: Jerk Lead Factor (0)", "x0.01", "../assets/offroad/icon_logic.png", 0, 100, 5));
-  latLongToggles->addItem(new CValueControl("CruiseMaxVals1", "ACCEL:0km/h(160)", "속도별 가속도를 지정합니다.(x0.01m/s^2)", "../assets/offroad/icon_logic.png", 1, 250, 5));
-  latLongToggles->addItem(new CValueControl("CruiseMaxVals2", "ACCEL:40km/h(120)", "속도별 가속도를 지정합니다.(x0.01m/s^2)", "../assets/offroad/icon_logic.png", 1, 250, 5));
-  latLongToggles->addItem(new CValueControl("CruiseMaxVals3", "ACCEL:60km/h(100)", "속도별 가속도를 지정합니다.(x0.01m/s^2)", "../assets/offroad/icon_logic.png", 1, 250, 5));
-  latLongToggles->addItem(new CValueControl("CruiseMaxVals4", "ACCEL:80km/h(80)", "속도별 가속도를 지정합니다.(x0.01m/s^2)", "../assets/offroad/icon_logic.png", 1, 250, 5));
-  latLongToggles->addItem(new CValueControl("CruiseMaxVals5", "ACCEL:110km/h(70)", "속도별 가속도를 지정합니다.(x0.01m/s^2)", "../assets/offroad/icon_logic.png", 1, 250, 5));
-  latLongToggles->addItem(new CValueControl("CruiseMaxVals6", "ACCEL:140km/h(60)", "속도별 가속도를 지정합니다.(x0.01m/s^2)", "../assets/offroad/icon_logic.png", 1, 250, 5));
-  //latLongToggles->addItem(new CValueControl("CruiseMinVals", "DECEL:(120)", "감속도를 설정합니다.(x0.01m/s^2)", "../assets/offroad/icon_road.png", 50, 250, 5));
-  latLongToggles->addItem(new CValueControl("MaxAngleFrames", "MaxAngleFrames(89)", "89:기본, 스티어계기판에러시 85~87", "../assets/offroad/icon_logic.png", 80, 100, 1));
+  latLongToggles->addItem(new CValueControl("CruiseMaxVals0", "ACCEL:0km/h(160)", "Acceleration needed at specified speed.(x0.01m/s^2)", "../assets/offroad/icon_logic.png", 1, 250, 5));
+  latLongToggles->addItem(new CValueControl("CruiseMaxVals1", "ACCEL:10km/h(160)", "Acceleration needed at specified speed.(x0.01m/s^2)", "../assets/offroad/icon_logic.png", 1, 250, 5));
+  latLongToggles->addItem(new CValueControl("CruiseMaxVals2", "ACCEL:40km/h(120)", "Acceleration needed at specified speed.(x0.01m/s^2)", "../assets/offroad/icon_logic.png", 1, 250, 5));
+  latLongToggles->addItem(new CValueControl("CruiseMaxVals3", "ACCEL:60km/h(100)", "Acceleration needed at specified speed.(x0.01m/s^2)", "../assets/offroad/icon_logic.png", 1, 250, 5));
+  latLongToggles->addItem(new CValueControl("CruiseMaxVals4", "ACCEL:80km/h(80)", "Acceleration needed at specified speed.(x0.01m/s^2)", "../assets/offroad/icon_logic.png", 1, 250, 5));
+  latLongToggles->addItem(new CValueControl("CruiseMaxVals5", "ACCEL:110km/h(70)", "Acceleration needed at specified speed.(x0.01m/s^2)", "../assets/offroad/icon_logic.png", 1, 250, 5));
+  latLongToggles->addItem(new CValueControl("CruiseMaxVals6", "ACCEL:140km/h(60)", "Acceleration needed at specified speed.(x0.01m/s^2)", "../assets/offroad/icon_logic.png", 1, 250, 5));
+  //latLongToggles->addItem(new CValueControl("CruiseMinVals", "DECEL:(120)", "Sets the deceleration rate.(x0.01m/s^2)", "../assets/offroad/icon_road.png", 50, 250, 5));
+  latLongToggles->addItem(new CValueControl("MaxAngleFrames", "MaxAngleFrames(89)", "89:Basic, steering instrument panel error 85~87", "../assets/offroad/icon_logic.png", 80, 100, 1));
  latLongToggles->addItem(new CValueControl("SteerActuatorDelay", "LAT:SteerActuatorDelay(30)", "x0.01, 0:LiveDelay", "../assets/offroad/icon_logic.png", 0, 100, 1));
  latLongToggles->addItem(new CValueControl("LateralTorqueCustom", "LAT: TorqueCustom(0)", "", "../assets/offroad/icon_logic.png", 0, 2, 1));
  latLongToggles->addItem(new CValueControl("LateralTorqueAccelFactor", "LAT: TorqueAccelFactor(2500)", "", "../assets/offroad/icon_logic.png", 1000, 6000, 10));
@ -721,6 +722,7 @@ CarrotPanel::CarrotPanel(QWidget* parent) : QWidget(parent) {
  dispToggles = new ListWidget(this);
  //dispToggles->addItem(new CValueControl("ShowHudMode", "DISP:Display Mode", "0:Frog,1:APilot,2:Bottom,3:Top,4:Left,5:Left-Bottom", "../assets/offroad/icon_shell.png", 0, 5, 1));
  dispToggles->addItem(new CValueControl("ShowDebugUI", "DISP:Debug Info", "", "../assets/offroad/icon_shell.png", 0, 2, 1));
+  dispToggles->addItem(new CValueControl("ShowTpms", "DISP:Tpms Info", "", "../assets/offroad/icon_shell.png", 0, 3, 1));
  dispToggles->addItem(new CValueControl("ShowDateTime", "DISP:Time Info", "0:None,1:Time/Date,2:Time,3:Date", "../assets/offroad/icon_calendar.png", 0, 3, 1));
  //dispToggles->addItem(new CValueControl("ShowSteerRotate", "DISP:Handle rotate", "0:None,1:Rotate", "../assets/offroad/icon_shell.png", 0, 1, 1));
  dispToggles->addItem(new CValueControl("ShowPathEnd", "DISP:Path End", "0:None,1:Display", "../assets/offroad/icon_shell.png", 0, 1, 1));
@ -809,7 +811,7 @@ CarrotPanel::CarrotPanel(QWidget* parent) : QWidget(parent) {
  startToggles->addItem(new CValueControl("AutoCruiseControl", "Auto Cruise control", "Softhold, Auto Cruise ON/OFF control", "../assets/offroad/icon_road.png", 0, 3, 1));
  startToggles->addItem(new CValueControl("CruiseOnDist", "CRUISE: Auto ON distance(0cm)", "When GAS/Brake is OFF, Cruise ON when the lead car gets closer.", "../assets/offroad/icon_road.png", 0, 2500, 50));
  startToggles->addItem(new CValueControl("AutoEngage", "Auto Engage control on start", "1:SteerEnable, 2:Steer/Cruise Engage", "../assets/offroad/icon_road.png", 0, 2, 1));
-  startToggles->addItem(new ParamControl("DisableMinSteerSpeed", "Disable Min.SteerSpeed", "", "../assets/offroad/icon_road.png", this));
+  startToggles->addItem(new ParamControl("DisableMinSteerSpeed", "Disable Min.SteerSpeed (Eg. SMDPS", "", "../assets/offroad/icon_road.png", this));
  startToggles->addItem(new CValueControl("AutoGasTokSpeed", "Auto AccelTok speed", "Gas(Accel)Tok enable speed", "../assets/offroad/icon_road.png", 0, 200, 5));
  startToggles->addItem(new ParamControl("AutoGasSyncSpeed", "Auto update Cruise speed", "", "../assets/offroad/icon_road.png", this));
  startToggles->addItem(new CValueControl("SpeedFromPCM", "Read Cruise Speed from PCM", "Toyota must set to 1, Honda 3", "../assets/offroad/icon_road.png", 0, 3, 1));
@ -817,6 +819,7 @@ CarrotPanel::CarrotPanel(QWidget* parent) : QWidget(parent) {
  startToggles->addItem(new CValueControl("SoundVolumeAdjustEngage", "Sound Volume, Engage(10%)", "", "../assets/offroad/icon_sound.png", 5, 200, 5));
  startToggles->addItem(new CValueControl("MaxTimeOffroadMin", "Power off time (min)", "", "../assets/offroad/icon_sandtimer.png", 1, 600, 10));
  startToggles->addItem(new ParamControl("DisableDM", "Disable DM", "", "../assets/img_driver_face_static_x.png", this));
+  startToggles->addItem(new CValueControl("EnableConnect", "EnableConnect", "Your device may be banned by Comma", "../assets/offroad/icon_sandtimer.png", 0, 1, 1));
  //startToggles->addItem(new CValueControl("CarrotCountDownSpeed", "NaviCountDown Speed(10)", "", "../assets/offroad/icon_shell.png", 0, 200, 5));
  startToggles->addItem(new CValueControl("MapboxStyle", "Mapbox Style(0)", "", "../assets/offroad/icon_shell.png", 0, 2, 1));
  startToggles->addItem(new CValueControl("RecordRoadCam", "Record Road camera(0)", "1:RoadCam, 2:RoadCam+WideRoadCam", "../assets/offroad/icon_shell.png", 0, 2, 1));
@ -828,12 +831,12 @@ CarrotPanel::CarrotPanel(QWidget* parent) : QWidget(parent) {
  //startToggles->addItem(new CValueControl("LaneChangeLaneCheck", "LaneChange: Check lane exist", "(0:No,1:Lane,2:+Edge)", "../assets/offroad/icon_shell.png", 0, 2, 1));

  speedToggles = new ListWidget(this);
-  speedToggles->addItem(new CValueControl("AutoCurveSpeedLowerLimit", "CURVE: Lower limit speed(30)", "곡선도로를 만나면 속도를 줄여줍니다. 최저속도", "../assets/offroad/icon_road.png", 30, 200, 5));
+  speedToggles->addItem(new CValueControl("AutoCurveSpeedLowerLimit", "CURVE: Lower limit speed(30)", "When you approach a curve, reduce your speed. Minimum speed", "../assets/offroad/icon_road.png", 30, 200, 5));
  speedToggles->addItem(new CValueControl("AutoCurveSpeedFactor", "CURVE: Auto Control ratio(100%)", "", "../assets/offroad/icon_road.png", 50, 300, 1));
  speedToggles->addItem(new CValueControl("AutoCurveSpeedAggressiveness", "CURVE: Aggressiveness (100%)", "", "../assets/offroad/icon_road.png", 50, 300, 1));
-  speedToggles->addItem(new CValueControl("AutoNaviSpeedCtrlEnd", "SpeedCameraDecelEnd(6s)", "감속완료시점을 설정합니다.값이 크면 카메라에서 멀리 감속 완료", "../assets/offroad/icon_road.png", 3, 20, 1));
-  speedToggles->addItem(new CValueControl("AutoNaviSpeedCtrlMode", "NaviSpeedControlMode(2)", "0:감속안함,1:과속카메라,2:+사고방지턱,3:+이동식카메라", "../assets/offroad/icon_road.png", 0, 3, 1));
-  speedToggles->addItem(new CValueControl("AutoNaviSpeedDecelRate", "SpeedCameraDecelRatex0.01m/s^2(80)", "낮으면 멀리서부터 감속함", "../assets/offroad/icon_road.png", 10, 200, 10));
+  speedToggles->addItem(new CValueControl("AutoNaviSpeedCtrlEnd", "SpeedCameraDecelEnd(6s)", "Sets the deceleration completion point. A larger value completes deceleration farther away from the camera.", "../assets/offroad/icon_road.png", 3, 20, 1));
+  speedToggles->addItem(new CValueControl("AutoNaviSpeedCtrlMode", "NaviSpeedControlMode(2)", "0:No slowdown, 1: speed camera, 2: + accident prevention bump, 3: + mobile camera", "../assets/offroad/icon_road.png", 0, 3, 1));
+  speedToggles->addItem(new CValueControl("AutoNaviSpeedDecelRate", "SpeedCameraDecelRatex0.01m/s^2(80)", "Lower number, slows down from a greater distance", "../assets/offroad/icon_road.png", 10, 200, 10));
  speedToggles->addItem(new CValueControl("AutoNaviSpeedSafetyFactor", "SpeedCameraSafetyFactor(105%)", "", "../assets/offroad/icon_road.png", 80, 120, 1));
  speedToggles->addItem(new CValueControl("AutoNaviSpeedBumpTime", "SpeedBumpTimeDistance(1s)", "", "../assets/offroad/icon_road.png", 1, 50, 1));
  speedToggles->addItem(new CValueControl("AutoNaviSpeedBumpSpeed", "SpeedBumpSpeed(35Km/h)", "", "../assets/offroad/icon_road.png", 10, 100, 5));
@ -841,8 +844,8 @@ CarrotPanel::CarrotPanel(QWidget* parent) : QWidget(parent) {
  speedToggles->addItem(new CValueControl("AutoNaviCountDownMode", "NaviCountDown mode(2)", "0: off, 1:tbt+camera, 2:tbt+camera+bump", "../assets/offroad/icon_road.png", 0, 2, 1));
  speedToggles->addItem(new CValueControl("TurnSpeedControlMode", "Turn Speed control mode(1)", "0: off, 1:vision, 2:vision+route, 3: route", "../assets/offroad/icon_road.png", 0, 3, 1));
  speedToggles->addItem(new CValueControl("MapTurnSpeedFactor", "Map TurnSpeed Factor(100)", "", "../assets/offroad/icon_map.png", 50, 300, 5));
-  speedToggles->addItem(new CValueControl("AutoTurnControl", "ATC: Auto turn control(0)", "0:없음,1:차선변경,2:차선변경+속도,3:속도", "../assets/offroad/icon_road.png", 0, 3, 1));
-  speedToggles->addItem(new CValueControl("AutoTurnControlSpeedTurn", "ATC: Turn Speed (20)", "0:없음, 턴속도", "../assets/offroad/icon_road.png", 0, 100, 5));
+  speedToggles->addItem(new CValueControl("AutoTurnControl", "ATC: Auto turn control(0)", "0:None, 1: lane change, 2: lane change + speed, 3: speed", "../assets/offroad/icon_road.png", 0, 3, 1));
+  speedToggles->addItem(new CValueControl("AutoTurnControlSpeedTurn", "ATC: Turn Speed (20)", "0:None, turn speed", "../assets/offroad/icon_road.png", 0, 100, 5));
  speedToggles->addItem(new CValueControl("AutoTurnControlTurnEnd", "ATC: Turn CtrlDistTime (6)", "dist=speed*time", "../assets/offroad/icon_road.png", 0, 30, 1));
  speedToggles->addItem(new CValueControl("AutoRoadSpeedAdjust", "Auto Roadlimit Speed adjust (50%)", "", "../assets/offroad/icon_road.png", 0, 100, 10));
  speedToggles->addItem(new CValueControl("AutoTurnMapChange", "ATC Auto Map Change(0)", "", "../assets/offroad/icon_road.png", 0, 1, 1));
--- a/selfdrive/ui/soundd.py
+++ b/selfdrive/ui/soundd.py
@ -108,6 +108,7 @@ class Soundd:
    self.soundVolumeAdjust = 1.0
    self.carrot_count_down = 0

+    self.lang = self.params.get('LanguageSetting', encoding='utf8')
    self.load_sounds()

    self.current_alert = AudibleAlert.none
@ -118,6 +119,7 @@ class Soundd:

    self.spl_filter_weighted = FirstOrderFilter(0, 2.5, FILTER_DT, initialized=False)

+
  def load_sounds(self):
    self.loaded_sounds: dict[int, np.ndarray] = {}

@ -125,7 +127,10 @@ class Soundd:
    for sound in sound_list:
      filename, play_count, volume = sound_list[sound]

+      if self.lang == "main_ko":
        wavefile = wave.open(BASEDIR + "/selfdrive/assets/sounds/" + filename, 'r')
+      else:
+        wavefile = wave.open(BASEDIR + "/selfdrive/assets/sounds_eng/" + filename, 'r')

      #assert wavefile.getnchannels() == 1
      assert wavefile.getsampwidth() == 2
--- a/system/hardware/base.py
+++ b/system/hardware/base.py
@ -6,6 +6,19 @@ from cereal import log

 NetworkType = log.DeviceState.NetworkType

+class LPAError(RuntimeError):
+  pass
+
+class LPAProfileNotFoundError(LPAError):
+  pass
+
+@dataclass
+class Profile:
+  iccid: str
+  nickname: str
+  enabled: bool
+  provider: str
+
@dataclass
 class ThermalZone:
  # a zone from /sys/class/thermal/thermal_zone*
@ -33,6 +46,7 @@ class ThermalZone:
 class ThermalConfig:
  cpu: list[ThermalZone] | None = None
  gpu: list[ThermalZone] | None = None
+  dsp: ThermalZone | None = None
  pmic: list[ThermalZone] | None = None
  memory: ThermalZone | None = None
  intake: ThermalZone | None = None
@ -50,6 +64,31 @@ class ThermalConfig:
          ret[f.name + "TempC"] = v.read()
    return ret

+class LPABase(ABC):
+  @abstractmethod
+  def list_profiles(self) -> list[Profile]:
+    pass
+
+  @abstractmethod
+  def get_active_profile(self) -> Profile | None:
+    pass
+
+  @abstractmethod
+  def delete_profile(self, iccid: str) -> None:
+    pass
+
+  @abstractmethod
+  def download_profile(self, qr: str, nickname: str | None = None) -> None:
+    pass
+
+  @abstractmethod
+  def nickname_profile(self, iccid: str, nickname: str) -> None:
+    pass
+
+  @abstractmethod
+  def switch_profile(self, iccid: str) -> None:
+    pass
+
 class HardwareBase(ABC):
  @staticmethod
  def get_cmdline() -> dict[str, str]:
@ -104,6 +143,10 @@ class HardwareBase(ABC):
  def get_sim_info(self):
    pass

+  @abstractmethod
+  def get_sim_lpa(self) -> LPABase:
+    pass
+
  @abstractmethod
  def get_network_strength(self, network_type):
    pass
@ -130,6 +173,9 @@ class HardwareBase(ABC):
  def get_thermal_config(self):
    return ThermalConfig()

+  def set_display_power(self, on: bool):
+    pass
+
  @abstractmethod
  def set_screen_brightness(self, percentage):
    pass
@ -164,6 +210,9 @@ class HardwareBase(ABC):
  def configure_modem(self):
    pass

+  def reboot_modem(self):
+    pass
+
  @abstractmethod
  def get_networks(self):
    pass
--- a/system/hardware/esim.py
+++ b/system/hardware/esim.py
@ -0,0 +1,45 @@
+#!/usr/bin/env python3
+
+import argparse
+import time
+from openpilot.system.hardware import HARDWARE
+
+
+if __name__ == '__main__':
+  parser = argparse.ArgumentParser(prog='esim.py', description='manage eSIM profiles on your comma device', epilog='comma.ai')
+  parser.add_argument('--backend', choices=['qmi', 'at'], default='qmi', help='use the specified backend, defaults to qmi')
+  parser.add_argument('--switch', metavar='iccid', help='switch to profile')
+  parser.add_argument('--delete', metavar='iccid', help='delete profile (warning: this cannot be undone)')
+  parser.add_argument('--download', nargs=2, metavar=('qr', 'name'), help='download a profile using QR code (format: LPA:1$rsp.truphone.com$QRF-SPEEDTEST)')
+  parser.add_argument('--nickname', nargs=2, metavar=('iccid', 'name'), help='update the nickname for a profile')
+  args = parser.parse_args()
+
+  mutated = False
+  lpa = HARDWARE.get_sim_lpa()
+  if args.switch:
+    lpa.switch_profile(args.switch)
+    mutated = True
+  elif args.delete:
+    confirm = input('are you sure you want to delete this profile? (y/N) ')
+    if confirm == 'y':
+      lpa.delete_profile(args.delete)
+      mutated = True
+    else:
+      print('cancelled')
+      exit(0)
+  elif args.download:
+    lpa.download_profile(args.download[0], args.download[1])
+  elif args.nickname:
+    lpa.nickname_profile(args.nickname[0], args.nickname[1])
+  else:
+    parser.print_help()
+
+  if mutated:
+    HARDWARE.reboot_modem()
+    # eUICC needs a small delay post-reboot before querying profiles
+    time.sleep(.5)
+
+  profiles = lpa.list_profiles()
+  print(f'\n{len(profiles)} profile{"s" if len(profiles) > 1 else ""}:')
+  for p in profiles:
+    print(f'- {p.iccid} (nickname: {p.nickname or "<none provided>"}) (provider: {p.provider}) - {"enabled" if p.enabled else "disabled"}')
--- a/system/hardware/pc/hardware.py
+++ b/system/hardware/pc/hardware.py
@ -1,12 +1,11 @@
 import random

 from cereal import log
-from openpilot.system.hardware.base import HardwareBase
+from openpilot.system.hardware.base import HardwareBase, LPABase

 NetworkType = log.DeviceState.NetworkType
 NetworkStrength = log.DeviceState.NetworkStrength

-
 class Pc(HardwareBase):
  def get_os_version(self):
    return None
@ -41,6 +40,9 @@ class Pc(HardwareBase):
      'data_connected': False
    }

+  def get_sim_lpa(self) -> LPABase:
+    raise NotImplementedError("SIM LPA not implemented for PC")
+
  def get_network_strength(self, network_type):
    return NetworkStrength.unknown

--- a/system/hardware/tici/agnos.json
+++ b/system/hardware/tici/agnos.json
@ -1,25 +1,25 @@
 [
  {
    "name": "xbl",
-    "url": "https://commadist.azureedge.net/agnosupdate/xbl-468f1ad6ab55e198647ff9191f91bd2918db9c0a3e27bae5673b4c5575c1254c.img.xz",
-    "hash": "468f1ad6ab55e198647ff9191f91bd2918db9c0a3e27bae5673b4c5575c1254c",
-    "hash_raw": "468f1ad6ab55e198647ff9191f91bd2918db9c0a3e27bae5673b4c5575c1254c",
+    "url": "https://commadist.azureedge.net/agnosupdate/xbl-6710967ca9701f205d7ab19c3a9b0dd2f547e65b3d96048b7c2b03755aafa0f1.img.xz",
+    "hash": "6710967ca9701f205d7ab19c3a9b0dd2f547e65b3d96048b7c2b03755aafa0f1",
+    "hash_raw": "6710967ca9701f205d7ab19c3a9b0dd2f547e65b3d96048b7c2b03755aafa0f1",
    "size": 3282256,
    "sparse": false,
    "full_check": true,
    "has_ab": true,
-    "ondevice_hash": "d35a86e7b8ddd9279b513a6f27da1521aa0f89fb93987ea74d57d0f0bbbbd247"
+    "ondevice_hash": "003a17ab1be68a696f7efe4c9938e8be511d4aacfc2f3211fc896bdc1681d089"
  },
  {
    "name": "xbl_config",
-    "url": "https://commadist.azureedge.net/agnosupdate/xbl_config-92b675dc2862ed15c732d91d9eb307d7e852e349217db8bee8f8829db543686b.img.xz",
-    "hash": "92b675dc2862ed15c732d91d9eb307d7e852e349217db8bee8f8829db543686b",
-    "hash_raw": "92b675dc2862ed15c732d91d9eb307d7e852e349217db8bee8f8829db543686b",
+    "url": "https://commadist.azureedge.net/agnosupdate/xbl_config-63922cfbfdf4ab87986c4ba8f3a4df5bf28414b3f71a29ec5947336722215535.img.xz",
+    "hash": "63922cfbfdf4ab87986c4ba8f3a4df5bf28414b3f71a29ec5947336722215535",
+    "hash_raw": "63922cfbfdf4ab87986c4ba8f3a4df5bf28414b3f71a29ec5947336722215535",
    "size": 98124,
    "sparse": false,
    "full_check": true,
    "has_ab": true,
-    "ondevice_hash": "623f1568072ee2d687ba8449a3d894c1c83dc4131b2e79eff35696885f70a419"
+    "ondevice_hash": "2a855dd636cc94718b64bea83a44d0a31741ecaa8f72a63613ff348ec7404091"
  },
  {
    "name": "abl",
@ -34,50 +34,50 @@
  },
  {
    "name": "aop",
-    "url": "https://commadist.azureedge.net/agnosupdate/aop-f0fcf7611d0890a72984f15a516dd37fa532dfcb70d428a8406838cf74ce23d5.img.xz",
-    "hash": "f0fcf7611d0890a72984f15a516dd37fa532dfcb70d428a8406838cf74ce23d5",
-    "hash_raw": "f0fcf7611d0890a72984f15a516dd37fa532dfcb70d428a8406838cf74ce23d5",
+    "url": "https://commadist.azureedge.net/agnosupdate/aop-21370172e590bd4ea907a558bcd6df20dc7a6c7d38b8e62fdde18f4a512ba9e9.img.xz",
+    "hash": "21370172e590bd4ea907a558bcd6df20dc7a6c7d38b8e62fdde18f4a512ba9e9",
+    "hash_raw": "21370172e590bd4ea907a558bcd6df20dc7a6c7d38b8e62fdde18f4a512ba9e9",
    "size": 184364,
    "sparse": false,
    "full_check": true,
    "has_ab": true,
-    "ondevice_hash": "bf74feca486f650589f6b7c90eab73274e35a68b5e00bfc1de0ed5f5484d4b3d"
+    "ondevice_hash": "c1be2f4aac5b3af49b904b027faec418d05efd7bd5144eb4fdfcba602bcf2180"
  },
  {
    "name": "devcfg",
-    "url": "https://commadist.azureedge.net/agnosupdate/devcfg-225b24ea7b1d2fee7f7d2da21386920ddacac2e33e9e938168436292f4eae180.img.xz",
-    "hash": "225b24ea7b1d2fee7f7d2da21386920ddacac2e33e9e938168436292f4eae180",
-    "hash_raw": "225b24ea7b1d2fee7f7d2da21386920ddacac2e33e9e938168436292f4eae180",
+    "url": "https://commadist.azureedge.net/agnosupdate/devcfg-d7d7e52963bbedbbf8a7e66847579ca106a0a729ce2cf60f4b8d8ea4b535d620.img.xz",
+    "hash": "d7d7e52963bbedbbf8a7e66847579ca106a0a729ce2cf60f4b8d8ea4b535d620",
+    "hash_raw": "d7d7e52963bbedbbf8a7e66847579ca106a0a729ce2cf60f4b8d8ea4b535d620",
    "size": 40336,
    "sparse": false,
    "full_check": true,
    "has_ab": true,
-    "ondevice_hash": "70f682b59ca0fe2f197d1486bd8be7b9b7e560798ad40ddef83b9f0a2f497938"
+    "ondevice_hash": "17b229668b20305ff8fa3cd5f94716a3aaa1e5bf9d1c24117eff7f2f81ae719f"
  },
  {
    "name": "boot",
-    "url": "https://commadist.azureedge.net/agnosupdate/boot-3d8e848796924081f5a6b3d745808b1117ae2ec41c03f2d41ee2e75633bd6425.img.xz",
-    "hash": "3d8e848796924081f5a6b3d745808b1117ae2ec41c03f2d41ee2e75633bd6425",
-    "hash_raw": "3d8e848796924081f5a6b3d745808b1117ae2ec41c03f2d41ee2e75633bd6425",
+    "url": "https://commadist.azureedge.net/agnosupdate/boot-4143170bad94968fd9be870b1498b4100bf273ed0aec2a2601c9017991d4bd42.img.xz",
+    "hash": "4143170bad94968fd9be870b1498b4100bf273ed0aec2a2601c9017991d4bd42",
+    "hash_raw": "4143170bad94968fd9be870b1498b4100bf273ed0aec2a2601c9017991d4bd42",
    "size": 18479104,
    "sparse": false,
    "full_check": true,
    "has_ab": true,
-    "ondevice_hash": "2075104847d1c96a06f07e85efb9f48d0e792d75a059047eae7ba4b463ffeadf"
+    "ondevice_hash": "6b7b3371100ad36d8a5a9ff19a1663b9b9e2d5e99cbe3cf9255e9c3017291ce3"
  },
  {
    "name": "system",
-    "url": "https://commadist.azureedge.net/agnosupdate/system-b22bd239c6f9527596fd49b98b7d521a563f99b90953ce021ee36567498e99c7.img.xz",
-    "hash": "cb9bfde1e995b97f728f5d5ad8d7a0f7a01544db5d138ead9b2350f222640939",
-    "hash_raw": "b22bd239c6f9527596fd49b98b7d521a563f99b90953ce021ee36567498e99c7",
+    "url": "https://commadist.azureedge.net/agnosupdate/system-c51bb5841011728f7cf108a9138ba68228ffb4232dfd91d6e082a6d8a6a8deaa.img.xz",
+    "hash": "993d6a1cd2b684e2b1cf6ff840f8996f02a529011372d9c1471e4c80719e7da9",
+    "hash_raw": "c51bb5841011728f7cf108a9138ba68228ffb4232dfd91d6e082a6d8a6a8deaa",
    "size": 5368709120,
    "sparse": true,
    "full_check": false,
    "has_ab": true,
-    "ondevice_hash": "e92a1f34158c60364c8d47b8ebbb6e59edf8d4865cd5edfeb2355d6f54f617fc",
+    "ondevice_hash": "59db25651da977eeb16a1af741fd01fc3d6b50d21544b1a7428b7c86b2cdef2d",
    "alt": {
-      "hash": "b22bd239c6f9527596fd49b98b7d521a563f99b90953ce021ee36567498e99c7",
-      "url": "https://commadist.azureedge.net/agnosupdate/system-b22bd239c6f9527596fd49b98b7d521a563f99b90953ce021ee36567498e99c7.img",
+      "hash": "c51bb5841011728f7cf108a9138ba68228ffb4232dfd91d6e082a6d8a6a8deaa",
+      "url": "https://commadist.azureedge.net/agnosupdate/system-c51bb5841011728f7cf108a9138ba68228ffb4232dfd91d6e082a6d8a6a8deaa.img",
      "size": 5368709120
    }
  }
--- a/system/hardware/tici/all-partitions.json
+++ b/system/hardware/tici/all-partitions.json
@ -130,25 +130,25 @@
  },
  {
    "name": "xbl",
-    "url": "https://commadist.azureedge.net/agnosupdate/xbl-468f1ad6ab55e198647ff9191f91bd2918db9c0a3e27bae5673b4c5575c1254c.img.xz",
-    "hash": "468f1ad6ab55e198647ff9191f91bd2918db9c0a3e27bae5673b4c5575c1254c",
-    "hash_raw": "468f1ad6ab55e198647ff9191f91bd2918db9c0a3e27bae5673b4c5575c1254c",
+    "url": "https://commadist.azureedge.net/agnosupdate/xbl-6710967ca9701f205d7ab19c3a9b0dd2f547e65b3d96048b7c2b03755aafa0f1.img.xz",
+    "hash": "6710967ca9701f205d7ab19c3a9b0dd2f547e65b3d96048b7c2b03755aafa0f1",
+    "hash_raw": "6710967ca9701f205d7ab19c3a9b0dd2f547e65b3d96048b7c2b03755aafa0f1",
    "size": 3282256,
    "sparse": false,
    "full_check": true,
    "has_ab": true,
-    "ondevice_hash": "d35a86e7b8ddd9279b513a6f27da1521aa0f89fb93987ea74d57d0f0bbbbd247"
+    "ondevice_hash": "003a17ab1be68a696f7efe4c9938e8be511d4aacfc2f3211fc896bdc1681d089"
  },
  {
    "name": "xbl_config",
-    "url": "https://commadist.azureedge.net/agnosupdate/xbl_config-92b675dc2862ed15c732d91d9eb307d7e852e349217db8bee8f8829db543686b.img.xz",
-    "hash": "92b675dc2862ed15c732d91d9eb307d7e852e349217db8bee8f8829db543686b",
-    "hash_raw": "92b675dc2862ed15c732d91d9eb307d7e852e349217db8bee8f8829db543686b",
+    "url": "https://commadist.azureedge.net/agnosupdate/xbl_config-63922cfbfdf4ab87986c4ba8f3a4df5bf28414b3f71a29ec5947336722215535.img.xz",
+    "hash": "63922cfbfdf4ab87986c4ba8f3a4df5bf28414b3f71a29ec5947336722215535",
+    "hash_raw": "63922cfbfdf4ab87986c4ba8f3a4df5bf28414b3f71a29ec5947336722215535",
    "size": 98124,
    "sparse": false,
    "full_check": true,
    "has_ab": true,
-    "ondevice_hash": "623f1568072ee2d687ba8449a3d894c1c83dc4131b2e79eff35696885f70a419"
+    "ondevice_hash": "2a855dd636cc94718b64bea83a44d0a31741ecaa8f72a63613ff348ec7404091"
  },
  {
    "name": "abl",
@ -163,14 +163,14 @@
  },
  {
    "name": "aop",
-    "url": "https://commadist.azureedge.net/agnosupdate/aop-f0fcf7611d0890a72984f15a516dd37fa532dfcb70d428a8406838cf74ce23d5.img.xz",
-    "hash": "f0fcf7611d0890a72984f15a516dd37fa532dfcb70d428a8406838cf74ce23d5",
-    "hash_raw": "f0fcf7611d0890a72984f15a516dd37fa532dfcb70d428a8406838cf74ce23d5",
+    "url": "https://commadist.azureedge.net/agnosupdate/aop-21370172e590bd4ea907a558bcd6df20dc7a6c7d38b8e62fdde18f4a512ba9e9.img.xz",
+    "hash": "21370172e590bd4ea907a558bcd6df20dc7a6c7d38b8e62fdde18f4a512ba9e9",
+    "hash_raw": "21370172e590bd4ea907a558bcd6df20dc7a6c7d38b8e62fdde18f4a512ba9e9",
    "size": 184364,
    "sparse": false,
    "full_check": true,
    "has_ab": true,
-    "ondevice_hash": "bf74feca486f650589f6b7c90eab73274e35a68b5e00bfc1de0ed5f5484d4b3d"
+    "ondevice_hash": "c1be2f4aac5b3af49b904b027faec418d05efd7bd5144eb4fdfcba602bcf2180"
  },
  {
    "name": "bluetooth",
@ -207,14 +207,14 @@
  },
  {
    "name": "devcfg",
-    "url": "https://commadist.azureedge.net/agnosupdate/devcfg-225b24ea7b1d2fee7f7d2da21386920ddacac2e33e9e938168436292f4eae180.img.xz",
-    "hash": "225b24ea7b1d2fee7f7d2da21386920ddacac2e33e9e938168436292f4eae180",
-    "hash_raw": "225b24ea7b1d2fee7f7d2da21386920ddacac2e33e9e938168436292f4eae180",
+    "url": "https://commadist.azureedge.net/agnosupdate/devcfg-d7d7e52963bbedbbf8a7e66847579ca106a0a729ce2cf60f4b8d8ea4b535d620.img.xz",
+    "hash": "d7d7e52963bbedbbf8a7e66847579ca106a0a729ce2cf60f4b8d8ea4b535d620",
+    "hash_raw": "d7d7e52963bbedbbf8a7e66847579ca106a0a729ce2cf60f4b8d8ea4b535d620",
    "size": 40336,
    "sparse": false,
    "full_check": true,
    "has_ab": true,
-    "ondevice_hash": "70f682b59ca0fe2f197d1486bd8be7b9b7e560798ad40ddef83b9f0a2f497938"
+    "ondevice_hash": "17b229668b20305ff8fa3cd5f94716a3aaa1e5bf9d1c24117eff7f2f81ae719f"
  },
  {
    "name": "devinfo",
@ -339,62 +339,62 @@
  },
  {
    "name": "boot",
-    "url": "https://commadist.azureedge.net/agnosupdate/boot-3d8e848796924081f5a6b3d745808b1117ae2ec41c03f2d41ee2e75633bd6425.img.xz",
-    "hash": "3d8e848796924081f5a6b3d745808b1117ae2ec41c03f2d41ee2e75633bd6425",
-    "hash_raw": "3d8e848796924081f5a6b3d745808b1117ae2ec41c03f2d41ee2e75633bd6425",
+    "url": "https://commadist.azureedge.net/agnosupdate/boot-4143170bad94968fd9be870b1498b4100bf273ed0aec2a2601c9017991d4bd42.img.xz",
+    "hash": "4143170bad94968fd9be870b1498b4100bf273ed0aec2a2601c9017991d4bd42",
+    "hash_raw": "4143170bad94968fd9be870b1498b4100bf273ed0aec2a2601c9017991d4bd42",
    "size": 18479104,
    "sparse": false,
    "full_check": true,
    "has_ab": true,
-    "ondevice_hash": "2075104847d1c96a06f07e85efb9f48d0e792d75a059047eae7ba4b463ffeadf"
+    "ondevice_hash": "6b7b3371100ad36d8a5a9ff19a1663b9b9e2d5e99cbe3cf9255e9c3017291ce3"
  },
  {
    "name": "system",
-    "url": "https://commadist.azureedge.net/agnosupdate/system-b22bd239c6f9527596fd49b98b7d521a563f99b90953ce021ee36567498e99c7.img.xz",
-    "hash": "cb9bfde1e995b97f728f5d5ad8d7a0f7a01544db5d138ead9b2350f222640939",
-    "hash_raw": "b22bd239c6f9527596fd49b98b7d521a563f99b90953ce021ee36567498e99c7",
+    "url": "https://commadist.azureedge.net/agnosupdate/system-c51bb5841011728f7cf108a9138ba68228ffb4232dfd91d6e082a6d8a6a8deaa.img.xz",
+    "hash": "993d6a1cd2b684e2b1cf6ff840f8996f02a529011372d9c1471e4c80719e7da9",
+    "hash_raw": "c51bb5841011728f7cf108a9138ba68228ffb4232dfd91d6e082a6d8a6a8deaa",
    "size": 5368709120,
    "sparse": true,
    "full_check": false,
    "has_ab": true,
-    "ondevice_hash": "e92a1f34158c60364c8d47b8ebbb6e59edf8d4865cd5edfeb2355d6f54f617fc",
+    "ondevice_hash": "59db25651da977eeb16a1af741fd01fc3d6b50d21544b1a7428b7c86b2cdef2d",
    "alt": {
-      "hash": "b22bd239c6f9527596fd49b98b7d521a563f99b90953ce021ee36567498e99c7",
-      "url": "https://commadist.azureedge.net/agnosupdate/system-b22bd239c6f9527596fd49b98b7d521a563f99b90953ce021ee36567498e99c7.img",
+      "hash": "c51bb5841011728f7cf108a9138ba68228ffb4232dfd91d6e082a6d8a6a8deaa",
+      "url": "https://commadist.azureedge.net/agnosupdate/system-c51bb5841011728f7cf108a9138ba68228ffb4232dfd91d6e082a6d8a6a8deaa.img",
      "size": 5368709120
    }
  },
  {
    "name": "userdata_90",
-    "url": "https://commadist.azureedge.net/agnosupdate/userdata_90-4bb7239f7e82c846e4d2584c0c433f03c582a80950de4094e6c190563d6d84ac.img.xz",
-    "hash": "b18001a2a87caa070fabf6321f8215ac353d6444564e3f86329b4dccc039ce54",
-    "hash_raw": "4bb7239f7e82c846e4d2584c0c433f03c582a80950de4094e6c190563d6d84ac",
+    "url": "https://commadist.azureedge.net/agnosupdate/userdata_90-89a161f17b86637413fe10a641550110b626b699382f5138c02267b7866a8494.img.xz",
+    "hash": "99d9e6cf6755581c6879bbf442bd62212beb8a04116e965ab987135b8842188b",
+    "hash_raw": "89a161f17b86637413fe10a641550110b626b699382f5138c02267b7866a8494",
    "size": 96636764160,
    "sparse": true,
    "full_check": true,
    "has_ab": false,
-    "ondevice_hash": "15ce16f2349d5b4d5fec6ad1e36222b1ae744ed10b8930bc9af75bd244dccb3c"
+    "ondevice_hash": "24ea29ab9c4ecec0568a4aa83e38790fedfce694060e90f4bde725931386ff41"
  },
  {
    "name": "userdata_89",
-    "url": "https://commadist.azureedge.net/agnosupdate/userdata_89-e36b59bf9ff755b6ca488df2ba1e20da8f7dab6b8843129f3fdcccd7ff2ff7d8.img.xz",
-    "hash": "12682cf54596ab1bd1c2464c4ca85888e4e06b47af5ff7d0432399e9907e2f64",
-    "hash_raw": "e36b59bf9ff755b6ca488df2ba1e20da8f7dab6b8843129f3fdcccd7ff2ff7d8",
+    "url": "https://commadist.azureedge.net/agnosupdate/userdata_89-cdd3401168819987c4840765bba1aa2217641b1a6a4165c412f44cac14ccfcbf.img.xz",
+    "hash": "5fbfa008a7f6b58ab01d4d171f3185924d4c9db69b54f4bfc0f214c6f17c2435",
+    "hash_raw": "cdd3401168819987c4840765bba1aa2217641b1a6a4165c412f44cac14ccfcbf",
    "size": 95563022336,
    "sparse": true,
    "full_check": true,
    "has_ab": false,
-    "ondevice_hash": "e4df9dea47ff04967d971263d50c17460ef240457e8d814e7c4f409f7493eb8a"
+    "ondevice_hash": "c07dc2e883a23d4a24d976cdf53a767a2fd699c8eeb476d60cdf18e84b417a52"
  },
  {
    "name": "userdata_30",
-    "url": "https://commadist.azureedge.net/agnosupdate/userdata_30-fe1d86f5322c675c58b3ae9753a4670abf44a25746bf6ac822aed108bb577282.img.xz",
-    "hash": "fa471703be0f0647617d183312d5209d23407f1628e4ab0934e6ec54b1a6b263",
-    "hash_raw": "fe1d86f5322c675c58b3ae9753a4670abf44a25746bf6ac822aed108bb577282",
+    "url": "https://commadist.azureedge.net/agnosupdate/userdata_30-2a8e8278b3bb545e6d7292c2417ccebdca9b47507eb5924f7c1e068737a7edfd.img.xz",
+    "hash": "b3bc293c9c5e0480ef663e980c8ccb2fb83ffd230c85f8797830fb61b8f59360",
+    "hash_raw": "2a8e8278b3bb545e6d7292c2417ccebdca9b47507eb5924f7c1e068737a7edfd",
    "size": 32212254720,
    "sparse": true,
    "full_check": true,
    "has_ab": false,
-    "ondevice_hash": "0b5b2402c9caa1ed7b832818e66580c974251e735bda91f2f226c41499d5616e"
+    "ondevice_hash": "8dae1cda089828c750d1d646337774ccd9432f567ecefde19a06dc7feeda9cd3"
  }
 ]
--- a/system/hardware/tici/esim.py
+++ b/system/hardware/tici/esim.py
@ -1,115 +1,106 @@
-#!/usr/bin/env python3
+import json
 import os
-import math
-import time
-import binascii
-import requests
-import serial
+import shutil
 import subprocess
+from typing import Literal

+from openpilot.system.hardware.base import LPABase, LPAError, LPAProfileNotFoundError, Profile

-def post(url, payload):
-  print()
-  print("POST to", url)
-  r = requests.post(
-    url,
-    data=payload,
-    verify=False,
-    headers={
-      "Content-Type": "application/json",
-      "X-Admin-Protocol": "gsma/rsp/v2.2.0",
-      "charset": "utf-8",
-      "User-Agent": "gsma-rsp-lpad",
-    },
-  )
-  print("resp", r)
-  print("resp text", repr(r.text))
-  print()
-  r.raise_for_status()
+class TiciLPA(LPABase):
+  def __init__(self, interface: Literal['qmi', 'at'] = 'qmi'):
+    self.env = os.environ.copy()
+    self.env['LPAC_APDU'] = interface
+    self.env['QMI_DEVICE'] = '/dev/cdc-wdm0'
+    self.env['AT_DEVICE'] = '/dev/ttyUSB2'

-  ret = f"HTTP/1.1 {r.status_code}"
-  ret += ''.join(f"{k}: {v}" for k, v in r.headers.items() if k != 'Connection')
-  return ret.encode() + r.content
+    self.timeout_sec = 45

+    if shutil.which('lpac') is None:
+      raise LPAError('lpac not found, must be installed!')

-class LPA:
-  def __init__(self):
-    self.dev = serial.Serial('/dev/ttyUSB2', baudrate=57600, timeout=1, bytesize=8)
-    self.dev.reset_input_buffer()
-    self.dev.reset_output_buffer()
-    assert "OK" in self.at("AT")
+  def list_profiles(self) -> list[Profile]:
+    msgs = self._invoke('profile', 'list')
+    self._validate_successful(msgs)
+    return [Profile(
+      iccid=p['iccid'],
+      nickname=p['profileNickname'],
+      enabled=p['profileState'] == 'enabled',
+      provider=p['serviceProviderName']
+    ) for p in msgs[-1]['payload']['data']]

-  def at(self, cmd):
-    print(f"==> {cmd}")
-    self.dev.write(cmd.encode() + b'\r\n')
+  def get_active_profile(self) -> Profile | None:
+    return next((p for p in self.list_profiles() if p.enabled), None)

-    r = b""
-    cnt = 0
-    while b"OK" not in r and b"ERROR" not in r and cnt < 20:
-      r += self.dev.read(8192).strip()
-      cnt += 1
-    r = r.decode()
-    print(f"<== {repr(r)}")
-    return r
+  def delete_profile(self, iccid: str) -> None:
+    self._validate_profile_exists(iccid)
+    latest = self.get_active_profile()
+    if latest is not None and latest.iccid == iccid:
+      raise LPAError('cannot delete active profile, switch to another profile first')
+    self._validate_successful(self._invoke('profile', 'delete', iccid))
+    self._process_notifications()

-  def download_ota(self, qr):
-    return self.at(f'AT+QESIM="ota","{qr}"')
+  def download_profile(self, qr: str, nickname: str | None = None) -> None:
+    msgs = self._invoke('profile', 'download', '-a', qr)
+    self._validate_successful(msgs)
+    new_profile = next((m for m in msgs if m['payload']['message'] == 'es8p_meatadata_parse'), None)
+    if new_profile is None:
+      raise LPAError('no new profile found')
+    if nickname:
+      self.nickname_profile(new_profile['payload']['data']['iccid'], nickname)
+    self._process_notifications()

-  def download(self, qr):
-    smdp = qr.split('$')[1]
-    out = self.at(f'AT+QESIM="download","{qr}"')
-    for _ in range(5):
-      line = out.split("+QESIM: ")[1].split("\r\n\r\nOK")[0]
+  def nickname_profile(self, iccid: str, nickname: str) -> None:
+    self._validate_profile_exists(iccid)
+    self._validate_successful(self._invoke('profile', 'nickname', iccid, nickname))

-      parts = [x.strip().strip('"') for x in line.split(',', maxsplit=4)]
-      print(repr(parts))
-      trans, ret, url, payloadlen, payload = parts
-      assert trans == "trans" and ret == "0"
-      assert len(payload) == int(payloadlen)
+  def switch_profile(self, iccid: str) -> None:
+    self._validate_profile_exists(iccid)
+    latest = self.get_active_profile()
+    if latest and latest.iccid == iccid:
+      return
+    self._validate_successful(self._invoke('profile', 'enable', iccid))
+    self._process_notifications()

-      r = post(f"https://{smdp}/{url}", payload)
-      to_send = binascii.hexlify(r).decode()
+  def _invoke(self, *cmd: str):
+    proc = subprocess.Popen(['sudo', '-E', 'lpac'] + list(cmd), stdout=subprocess.PIPE, stderr=subprocess.PIPE, env=self.env)
+    try:
+      out, err = proc.communicate(timeout=self.timeout_sec)
+    except subprocess.TimeoutExpired as e:
+      proc.kill()
+      raise LPAError(f"lpac {cmd} timed out after {self.timeout_sec} seconds") from e

-      chunk_len = 1400
-      for i in range(math.ceil(len(to_send) / chunk_len)):
-        state = 1 if (i+1)*chunk_len < len(to_send) else 0
-        data = to_send[i * chunk_len : (i+1)*chunk_len]
-        out = self.at(f'AT+QESIM="trans",{len(to_send)},{state},{i},{len(data)},"{data}"')
-        assert "OK" in out
+    messages = []
+    for line in out.decode().strip().splitlines():
+      if line.startswith('{'):
+        message = json.loads(line)

-      if '+QESIM:"download",1' in out:
-        raise Exception("profile install failed")
-      elif '+QESIM:"download",0' in out:
-        print("done, successfully loaded")
-        break
+        # lpac response format validations
+        assert 'type' in message, 'expected type in message'
+        assert message['type'] == 'lpa' or message['type'] == 'progress', 'expected lpa or progress message type'
+        assert 'payload' in message, 'expected payload in message'
+        assert 'code' in message['payload'], 'expected code in message payload'
+        assert 'data' in message['payload'], 'expected data in message payload'

-  def enable(self, iccid):
-    self.at(f'AT+QESIM="enable","{iccid}"')
+        msg_ret_code = message['payload']['code']
+        if msg_ret_code != 0:
+          raise LPAError(f"lpac {' '.join(cmd)} failed with code {msg_ret_code}: <{message['payload']['message']}> {message['payload']['data']}")

-  def disable(self, iccid):
-    self.at(f'AT+QESIM="disable","{iccid}"')
+        messages.append(message)

-  def delete(self, iccid):
-    self.at(f'AT+QESIM="delete","{iccid}"')
+    if len(messages) == 0:
+      raise LPAError(f"lpac {cmd} returned no messages")

-  def list_profiles(self):
-    out = self.at('AT+QESIM="list"')
-    return out.strip().splitlines()[1:]
+    return messages

+  def _process_notifications(self) -> None:
+    """
+    Process notifications stored on the eUICC, typically to activate/deactivate the profile with the carrier.
+    """
+    self._validate_successful(self._invoke('notification', 'process', '-a', '-r'))

-if __name__ == "__main__":
-  import sys
+  def _validate_profile_exists(self, iccid: str) -> None:
+    if not any(p.iccid == iccid for p in self.list_profiles()):
+      raise LPAProfileNotFoundError(f'profile {iccid} does not exist')

-  if "RESTART" in os.environ:
-    subprocess.check_call("sudo systemctl stop ModemManager", shell=True)
-    subprocess.check_call("/usr/comma/lte/lte.sh stop_blocking", shell=True)
-    subprocess.check_call("/usr/comma/lte/lte.sh start", shell=True)
-    while not os.path.exists('/dev/ttyUSB2'):
-      time.sleep(1)
-    time.sleep(3)
-
-  lpa = LPA()
-  print(lpa.list_profiles())
-  if len(sys.argv) > 1:
-    lpa.download(sys.argv[1])
-    print(lpa.list_profiles())
+  def _validate_successful(self, msgs: list[dict]) -> None:
+    assert msgs[-1]['payload']['message'] == 'success', 'expected success notification'
--- a/system/hardware/tici/hardware.py
+++ b/system/hardware/tici/hardware.py
@ -9,9 +9,11 @@ from functools import cached_property, lru_cache
 from pathlib import Path

 from cereal import log
+from openpilot.common.util import sudo_read, sudo_write
 from openpilot.common.gpio import gpio_set, gpio_init, get_irqs_for_action
-from openpilot.system.hardware.base import HardwareBase, ThermalConfig, ThermalZone
+from openpilot.system.hardware.base import HardwareBase, LPABase, ThermalConfig, ThermalZone
 from openpilot.system.hardware.tici import iwlist
+from openpilot.system.hardware.tici.esim import TiciLPA
 from openpilot.system.hardware.tici.pins import GPIO
 from openpilot.system.hardware.tici.amplifier import Amplifier

@ -61,25 +63,6 @@ MM_MODEM_ACCESS_TECHNOLOGY_UMTS = 1 << 5
 MM_MODEM_ACCESS_TECHNOLOGY_LTE = 1 << 14


-def sudo_write(val, path):
-  try:
-    with open(path, 'w') as f:
-      f.write(str(val))
-  except PermissionError:
-    os.system(f"sudo chmod a+w {path}")
-    try:
-      with open(path, 'w') as f:
-        f.write(str(val))
-    except PermissionError:
-      # fallback for debugfs files
-      os.system(f"sudo su -c 'echo {val} > {path}'")
-
-def sudo_read(path: str) -> str:
-  try:
-    return subprocess.check_output(f"sudo cat {path}", shell=True, encoding='utf8').strip()
-  except Exception:
-    return ""
-
 def affine_irq(val, action):
  irqs = get_irqs_for_action(action)
  if len(irqs) == 0:
@ -198,6 +181,9 @@ class Tici(HardwareBase):
        'data_connected': modem.Get(MM_MODEM, 'State', dbus_interface=DBUS_PROPS, timeout=TIMEOUT) == MM_MODEM_STATE.CONNECTED,
      }

+  def get_sim_lpa(self) -> LPABase:
+    return TiciLPA()
+
  def get_imei(self, slot):
    if slot != 0:
      return ""
@ -297,13 +283,11 @@ class Tici(HardwareBase):
      return None

  def get_modem_temperatures(self):
-    if self.get_device_type() == "mici":
-      return []
    timeout = 0.2  # Default timeout is too short
    try:
      modem = self.get_modem()
      temps = modem.Command("AT+QTEMP", math.ceil(timeout), dbus_interface=MM_MODEM, timeout=timeout)
-      return list(map(int, temps.split(' ')[1].split(',')))
+      return list(filter(lambda t: t != 255, map(int, temps.split(' ')[1].split(','))))
    except Exception:
      return []

@ -335,12 +319,20 @@ class Tici(HardwareBase):
    return ThermalConfig(cpu=[ThermalZone(f"cpu{i}-silver-usr") for i in range(4)] +
                             [ThermalZone(f"cpu{i}-gold-usr") for i in range(4)],
                         gpu=[ThermalZone("gpu0-usr"), ThermalZone("gpu1-usr")],
+                         dsp=ThermalZone("compute-hvx-usr"),
                         memory=ThermalZone("ddr-usr"),
                         pmic=[ThermalZone("pm8998_tz"), ThermalZone("pm8005_tz")],
                         intake=intake,
                         exhaust=exhaust,
                         case=case)

+  def set_display_power(self, on):
+    try:
+      with open("/sys/class/backlight/panel0-backlight/bl_power", "w") as f:
+        f.write("0" if on else "4")
+    except Exception:
+      pass
+
  def set_screen_brightness(self, percentage):
    try:
      with open("/sys/class/backlight/panel0-backlight/max_brightness") as f:
--- a/system/loggerd/uploader.py
+++ b/system/loggerd/uploader.py
@ -89,7 +89,7 @@ class Uploader:

  def list_upload_files(self, metered: bool) -> Iterator[tuple[str, str, str]]:
    r = self.params.get("AthenadRecentlyViewedRoutes", encoding="utf8")
-    requested_routes = [] if r is None else r.split(",")
+    requested_routes = [] if r is None else [route for route in r.split(",") if route]

    for logdir in listdir_by_creation(self.root):
      path = os.path.join(self.root, logdir)
--- a/system/loggerd/xattr_cache.py
+++ b/system/loggerd/xattr_cache.py
@ -1,16 +1,17 @@
-import os
 import errno

+import xattr
+
 _cached_attributes: dict[tuple, bytes | None] = {}

 def getxattr(path: str, attr_name: str) -> bytes | None:
  key = (path, attr_name)
  if key not in _cached_attributes:
    try:
-      response = os.getxattr(path, attr_name)
+      response = xattr.getxattr(path, attr_name)
    except OSError as e:
-      # ENODATA means attribute hasn't been set
-      if e.errno == errno.ENODATA:
+      # ENODATA (Linux) or ENOATTR (macOS) means attribute hasn't been set
+      if e.errno == errno.ENODATA or (hasattr(errno, 'ENOATTR') and e.errno == errno.ENOATTR):
        response = None
      else:
        raise
@ -19,4 +20,4 @@ def getxattr(path: str, attr_name: str) -> bytes | None:

 def setxattr(path: str, attr_name: str, attr_value: bytes) -> None:
  _cached_attributes.pop((path, attr_name), None)
-  return os.setxattr(path, attr_name, attr_value)
+  xattr.setxattr(path, attr_name, attr_value)
--- a/system/manager/manager.py
+++ b/system/manager/manager.py
@ -40,6 +40,7 @@ def get_default_params():

    ("LongitudinalPersonalityMax", "3"),
    ("ShowDebugUI", "0"),
+    ("ShowTpms", "1"),
    ("ShowDateTime", "1"),
    ("ShowPathEnd", "1"),
    ("ShowCustomBrightness", "100"),
@ -101,6 +102,7 @@ def get_default_params():
    ("MyDrivingMode", "3"),
    ("MyDrivingModeAuto", "0"),
    ("TrafficLightDetectMode", "2"),
+    ("CruiseMaxVals0", "160"),
    ("CruiseMaxVals1", "200"),
    ("CruiseMaxVals2", "160"),
    ("CruiseMaxVals3", "130"),
@ -156,6 +158,7 @@ def get_default_params():
    ("SteerActuatorDelay", "0"),
    ("MaxTimeOffroadMin", "60"),
    ("DisableDM", "0"),
+    ("EnableConnect", "0"),
    ("MuteDoor", "0"),
    ("MuteSeatbelt", "0"),
    ("RecordRoadCam", "0"),
--- a/system/manager/process_config.py
+++ b/system/manager/process_config.py
@ -72,6 +72,9 @@ def and_(*fns):
 def enable_dm(started, params, CP: car.CarParams) -> bool:
  return (started or params.get_bool("IsDriverViewEnabled")) and params.get_int("DisableDM") == 0

+def enable_connect(started, params, CP: car.CarParams) -> bool:
+  return params.get_int("EnableConnect") >= 0
+
 procs = [
  DaemonProcess("manage_athenad", "system.athena.manage_athenad", "AthenadPid"),

@ -121,7 +124,7 @@ procs = [
  PythonProcess("hardwared", "system.hardware.hardwared", always_run),
  PythonProcess("tombstoned", "system.tombstoned", always_run, enabled=not PC),
  PythonProcess("updated", "system.updated.updated", enable_updated, enabled=not PC),
-  #PythonProcess("uploader", "system.loggerd.uploader", always_run),
+  PythonProcess("uploader", "system.loggerd.uploader", enable_connect),
  PythonProcess("statsd", "system.statsd", always_run),

  # debug procs
--- a/system/micd.py
+++ b/system/micd.py
@ -1,6 +1,7 @@
 #!/usr/bin/env python3
 import numpy as np
 from functools import cache
+import threading

 from cereal import messaging
 from openpilot.common.realtime import Ratekeeper
@ -52,12 +53,18 @@ class Mic:
    self.sound_pressure_weighted = 0
    self.sound_pressure_level_weighted = 0

-  def update(self):
-    msg = messaging.new_message('microphone', valid=True)
-    msg.microphone.soundPressure = float(self.sound_pressure)
-    msg.microphone.soundPressureWeighted = float(self.sound_pressure_weighted)
+    self.lock = threading.Lock()

-    msg.microphone.soundPressureWeightedDb = float(self.sound_pressure_level_weighted)
+  def update(self):
+    with self.lock:
+      sound_pressure = self.sound_pressure
+      sound_pressure_weighted = self.sound_pressure_weighted
+      sound_pressure_level_weighted = self.sound_pressure_level_weighted
+
+    msg = messaging.new_message('microphone', valid=True)
+    msg.microphone.soundPressure = float(sound_pressure)
+    msg.microphone.soundPressureWeighted = float(sound_pressure_weighted)
+    msg.microphone.soundPressureWeightedDb = float(sound_pressure_level_weighted)

    self.pm.send('microphone', msg)
    self.rk.keep_time()
@ -69,7 +76,7 @@ class Mic:

    Logged A-weighted equivalents are rough approximations of the human-perceived loudness.
    """
-
+    with self.lock:
      self.measurements = np.concatenate((self.measurements, indata[:, 0]))

      while self.measurements.size >= FFT_SAMPLES:
--- a/system/updated/updated.py
+++ b/system/updated/updated.py
@ -9,7 +9,6 @@ import signal
 import fcntl
 import time
 import threading
-import gc
 from collections import defaultdict
 from pathlib import Path

@ -72,12 +71,8 @@ def read_time_from_param(params, param) -> datetime.datetime | None:
    pass
  return None

-def run_org(cmd: list[str], cwd: str = None) -> str:
-  return subprocess.check_output(cmd, cwd=cwd, stderr=subprocess.STDOUT, encoding='utf8')
-
 def run(cmd: list[str], cwd: str = None) -> str:
-  proc = subprocess.run(cmd, cwd=cwd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, encoding='utf8', check=True)
-  return proc.stdout
+  return subprocess.check_output(cmd, cwd=cwd, stderr=subprocess.STDOUT, encoding='utf8')


 def set_consistent_flag(consistent: bool) -> None:
@ -497,9 +492,6 @@ def main() -> None:
        cloudlog.exception("uncaught updated exception, shouldn't happen")
        exception = str(e)
        OVERLAY_INIT.unlink(missing_ok=True)
-      finally:
-        gc.collect()
-

      try:
        params.put("UpdaterState", "idle")
--- a/tinygrad_repo/README.md
+++ b/tinygrad_repo/README.md
@ -21,11 +21,11 @@ tinygrad: For something between [PyTorch](https://github.com/pytorch/pytorch) an

 ---

-This may not be the best deep learning framework, but it is a deep learning framework.
+Despite tinygrad's size, it is a fully featured deep learning framework.

-Due to its extreme simplicity, it aims to be the easiest framework to add new accelerators to, with support for both inference and training. If XLA is CISC, tinygrad is RISC.
+Due to its extreme simplicity, it is the easiest framework to add new accelerators to, with support for both inference and training. If XLA is CISC, tinygrad is RISC.

-tinygrad is still alpha software, but we [raised some money](https://geohot.github.io/blog/jekyll/update/2023/05/24/the-tiny-corp-raised-5M.html) to make it good. Someday, we will tape out chips.
+tinygrad is now beta software, we [raised some money](https://geohot.github.io/blog/jekyll/update/2023/05/24/the-tiny-corp-raised-5M.html) to make it good. Someday, we will tape out chips.

 ## Features

--- a/tinygrad_repo/autogen_stubs.sh
+++ b/tinygrad_repo/autogen_stubs.sh
@ -35,7 +35,7 @@ def _try_dlopen_$name():
  for candidate in PATHS_TO_TRY:
    try: return ctypes.CDLL(candidate)
    except OSError: pass
-  raise RuntimeError("library $name not found")
+  return None
 EOF
 }

@ -78,11 +78,11 @@ generate_kfd() {
  clang2py /usr/include/linux/kfd_ioctl.h -o $BASE/kfd.py -k cdefstum

  fixup $BASE/kfd.py
-  sed -i "s\import ctypes\import ctypes, os\g" $BASE/kfd.py
-  sed -i "s\import fcntl, functools\import functools" $BASE/kfd.py
-  sed -i "s\import ctypes,os\a from tinygrad.runtime.support import HWInterface\g" $BASE/kfd.py
-  sed -i "s\def _do_ioctl(__idir, __base, __nr, __user_struct, __fd, **kwargs):\def _do_ioctl(__idir, __base, __nr, __user_struct, __fd:HWInterface, **kwargs):\g" $BASE/kfd.py
-  sed -i "s\fcntl.ioctl(__fd, (__idir<<30)\__fd.ioctl((__idir<<30)\g" $BASE/kfd.py
+  sed -i "s/import ctypes/import ctypes, os/g" $BASE/kfd.py
+  sed -i "s/import fcntl, functools/import functools/g" $BASE/kfd.py
+  sed -i "/import functools/a from tinygrad.runtime.support.hcq import FileIOInterface" $BASE/kfd.py
+  sed -i "s/def _do_ioctl(__idir, __base, __nr, __user_struct, __fd, \*\*kwargs):/def _do_ioctl(__idir, __base, __nr, __user_struct, __fd:FileIOInterface, \*\*kwargs):/g" $BASE/kfd.py
+  sed -i "s/fcntl.ioctl(__fd, (__idir<<30)/__fd.ioctl((__idir<<30)/g" $BASE/kfd.py
  python3 -c "import tinygrad.runtime.autogen.kfd"
 }

@ -104,10 +104,10 @@ generate_nvrtc() {
 }

 generate_nv() {
-  NVKERN_COMMIT_HASH=d6b75a34094b0f56c2ccadf14e5d0bd515ed1ab6
+  NVKERN_COMMIT_HASH=81fe4fb417c8ac3b9bdcc1d56827d116743892a5
  NVKERN_SRC=/tmp/open-gpu-kernel-modules-$NVKERN_COMMIT_HASH
  if [ ! -d "$NVKERN_SRC" ]; then
-    git clone https://github.com/tinygrad/open-gpu-kernel-modules $NVKERN_SRC
+    git clone https://github.com/NVIDIA/open-gpu-kernel-modules $NVKERN_SRC
    pushd .
    cd $NVKERN_SRC
    git reset --hard $NVKERN_COMMIT_HASH
@ -116,15 +116,19 @@ generate_nv() {

  clang2py -k cdefstum \
    extra/nv_gpu_driver/clc6c0qmd.h \
+    extra/nv_gpu_driver/clcec0qmd.h \
    $NVKERN_SRC/src/common/sdk/nvidia/inc/class/cl0080.h \
    $NVKERN_SRC/src/common/sdk/nvidia/inc/class/cl2080_notification.h \
    $NVKERN_SRC/src/common/sdk/nvidia/inc/class/clc56f.h \
-    $NVKERN_SRC/src/common/sdk/nvidia/inc/class/clc56f.h \
-    $NVKERN_SRC/src/common/sdk/nvidia/inc/class/clc56f.h \
+    $NVKERN_SRC/src/common/sdk/nvidia/inc/class/clc86f.h \
+    $NVKERN_SRC/src/common/sdk/nvidia/inc/class/clc96f.h \
+    $NVKERN_SRC/src/common/sdk/nvidia/inc/class/clc761.h \
    $NVKERN_SRC/src/common/sdk/nvidia/inc/class/cl83de.h \
    $NVKERN_SRC/src/nvidia/generated/g_allclasses.h \
    $NVKERN_SRC/src/common/sdk/nvidia/inc/class/clc6c0.h \
+    $NVKERN_SRC/src/common/sdk/nvidia/inc/class/clcdc0.h \
    $NVKERN_SRC/kernel-open/nvidia-uvm/clc6b5.h \
+    $NVKERN_SRC/kernel-open/nvidia-uvm/clc9b5.h \
    $NVKERN_SRC/kernel-open/nvidia-uvm/uvm_ioctl.h \
    $NVKERN_SRC/kernel-open/nvidia-uvm/uvm_linux_ioctl.h \
    $NVKERN_SRC/kernel-open/nvidia-uvm/hwref/ampere/ga100/dev_fault.h \
@ -149,6 +153,7 @@ generate_nv() {
  sed -i "s\import ctypes\import ctypes, os\g" $BASE/nv_gpu.py
  sed -i 's/#\?\s\([A-Za-z0-9_]\+\) = MW ( \([0-9]\+\) : \([0-9]\+\) )/\1 = (\2 , \3)/' $BASE/nv_gpu.py # NVC6C0_QMDV03_00 processing
  sed -i 's/#\sdef NVC6C0_QMD\([A-Za-z0-9_()]\+\):/def NVC6C0_QMD\1:/' $BASE/nv_gpu.py
+  sed -i 's/#\sdef NVCEC0_QMD\([A-Za-z0-9_()]\+\):/def NVCEC0_QMD\1:/' $BASE/nv_gpu.py
  sed -i 's/#\s*return MW(\([0-9i()*+]\+\):\([0-9i()*+]\+\))/    return (\1 , \2)/' $BASE/nv_gpu.py
  sed -i 's/#\?\s*\(.*\)\s*=\s*\(NV\)\?BIT\(32\)\?\s*(\s*\([0-9]\+\)\s*)/\1 = (1 << \4)/' $BASE/nv_gpu.py # name = BIT(x) -> name = (1 << x)
  sed -i "s/UVM_\([A-Za-z0-9_]\+\) = \['i', '(', '\([0-9]\+\)', ')'\]/UVM_\1 = \2/" $BASE/nv_gpu.py # UVM_name = ['i', '(', '<num>', ')'] -> UVM_name = <num>
@ -287,7 +292,7 @@ generate_vfio() {
  fixup $BASE/vfio.py
  sed -i "s\import ctypes\import ctypes, os\g" $BASE/vfio.py
  sed -i "s\import fcntl, functools\import functools" $BASE/vfio.py
-  sed -i "s\import ctypes,os\a from tinygrad.runtime.support import HWInterface\g" $BASE/vfio.py
+  sed -i "s\import ctypes,os\a from tinygrad.runtime.support import FileIOInterface\g" $BASE/vfio.py
  sed -i "s\fcntl.ioctl(__fd, (__idir<<30)\return __fd.ioctl((__idir<<30)\g" $BASE/vfio.py
 }

@ -349,50 +354,6 @@ generate_am() {
    -o $BASE/am/soc24.py
  fixup $BASE/am/soc24.py

-  clang2py -k cdefstum \
-    $AMKERN_INC/asic_reg/mp/mp_13_0_0_offset.h \
-    $AMKERN_INC/asic_reg/mp/mp_13_0_0_sh_mask.h \
-    -o $BASE/am/mp_13_0_0.py
-  fixup $BASE/am/mp_13_0_0.py
-
-  # 14_0_3 reuses 14_0_2
-  clang2py -k cdefstum \
-    $AMKERN_INC/asic_reg/mp/mp_14_0_2_offset.h \
-    $AMKERN_INC/asic_reg/mp/mp_14_0_2_sh_mask.h \
-    -o $BASE/am/mp_14_0_3.py
-  fixup $BASE/am/mp_14_0_3.py
-
-  clang2py -k cdefstum \
-    $AMKERN_INC/asic_reg/mp/mp_11_0_offset.h \
-    $AMKERN_INC/asic_reg/mp/mp_11_0_sh_mask.h \
-    -o $BASE/am/mp_11_0.py
-  fixup $BASE/am/mp_11_0.py
-
-  clang2py -k cdefstum \
-    $AMKERN_INC/asic_reg/gc/gc_9_4_3_offset.h \
-    $AMKERN_INC/asic_reg/gc/gc_9_4_3_sh_mask.h \
-    extra/amdpci/overlay/gc_9_4_3.h \
-    -o $BASE/am/gc_9_4_3.py
-  fixup $BASE/am/gc_9_4_3.py
-
-  clang2py -k cdefstum \
-    $AMKERN_INC/asic_reg/gc/gc_10_3_0_offset.h \
-    $AMKERN_INC/asic_reg/gc/gc_10_3_0_sh_mask.h \
-    -o $BASE/am/gc_10_3_0.py
-  fixup $BASE/am/gc_10_3_0.py
-
-  clang2py -k cdefstum \
-    $AMKERN_INC/asic_reg/gc/gc_11_0_0_offset.h \
-    $AMKERN_INC/asic_reg/gc/gc_11_0_0_sh_mask.h \
-    -o $BASE/am/gc_11_0_0.py
-  fixup $BASE/am/gc_11_0_0.py
-
-  clang2py -k cdefstum \
-    $AMKERN_INC/asic_reg/gc/gc_12_0_0_offset.h \
-    $AMKERN_INC/asic_reg/gc/gc_12_0_0_sh_mask.h \
-    -o $BASE/am/gc_12_0_0.py
-  fixup $BASE/am/gc_12_0_0.py
-
  clang2py -k cdefstum \
    extra/hip_gpu_driver/sdma_registers.h \
    $AMKERN_AMD/amdgpu/vega10_sdma_pkt_open.h \
@ -414,66 +375,6 @@ generate_am() {
    -o $BASE/am/sdma_6_0_0.py
  fixup $BASE/am/sdma_6_0_0.py

-  clang2py -k cdefstum \
-    $AMKERN_INC/asic_reg/mmhub/mmhub_3_0_0_offset.h \
-    $AMKERN_INC/asic_reg/mmhub/mmhub_3_0_0_sh_mask.h \
-    -o $BASE/am/mmhub_3_0_0.py
-  fixup $BASE/am/mmhub_3_0_0.py
-
-  clang2py -k cdefstum \
-    $AMKERN_INC/asic_reg/mmhub/mmhub_3_0_2_offset.h \
-    $AMKERN_INC/asic_reg/mmhub/mmhub_3_0_2_sh_mask.h \
-    -o $BASE/am/mmhub_3_0_2.py
-  fixup $BASE/am/mmhub_3_0_2.py
-
-  clang2py -k cdefstum \
-    $AMKERN_INC/asic_reg/nbio/nbio_2_3_offset.h \
-    $AMKERN_INC/asic_reg/nbio/nbio_2_3_sh_mask.h \
-    -o $BASE/am/nbio_2_3_0.py
-  fixup $BASE/am/nbio_2_3_0.py
-
-  clang2py -k cdefstum \
-    $AMKERN_INC/asic_reg/mmhub/mmhub_4_1_0_offset.h \
-    $AMKERN_INC/asic_reg/mmhub/mmhub_4_1_0_sh_mask.h \
-    -o $BASE/am/mmhub_4_1_0.py
-  fixup $BASE/am/mmhub_4_1_0.py
-
-  clang2py -k cdefstum \
-    $AMKERN_INC/asic_reg/nbio/nbio_4_3_0_offset.h \
-    $AMKERN_INC/asic_reg/nbio/nbio_4_3_0_sh_mask.h \
-    -o $BASE/am/nbio_4_3_0.py
-  fixup $BASE/am/nbio_4_3_0.py
-
-  clang2py -k cdefstum \
-    $AMKERN_INC/asic_reg/nbif/nbif_6_3_1_offset.h \
-    $AMKERN_INC/asic_reg/nbif/nbif_6_3_1_sh_mask.h \
-    -o $BASE/am/nbif_6_3_1.py
-  fixup $BASE/am/nbif_6_3_1.py
-
-  clang2py -k cdefstum \
-    $AMKERN_INC/asic_reg/nbio/nbio_7_9_0_offset.h \
-    $AMKERN_INC/asic_reg/nbio/nbio_7_9_0_sh_mask.h \
-    -o $BASE/am/nbio_7_9_0.py
-  fixup $BASE/am/nbio_7_9_0.py
-
-  clang2py -k cdefstum \
-    $AMKERN_INC/asic_reg/nbio/nbio_7_11_0_offset.h \
-    $AMKERN_INC/asic_reg/nbio/nbio_7_11_0_sh_mask.h \
-    -o $BASE/am/nbio_7_11_0.py
-  fixup $BASE/am/nbio_7_11_0.py
-
-  clang2py -k cdefstum \
-    $AMKERN_INC/asic_reg/oss/osssys_6_0_0_offset.h \
-    $AMKERN_INC/asic_reg/oss/osssys_6_0_0_sh_mask.h \
-    -o $BASE/am/osssys_6_0_0.py
-  fixup $BASE/am/osssys_6_0_0.py
-
-  clang2py -k cdefstum \
-    $AMKERN_INC/asic_reg/oss/osssys_7_0_0_offset.h \
-    $AMKERN_INC/asic_reg/oss/osssys_7_0_0_sh_mask.h \
-    -o $BASE/am/osssys_7_0_0.py
-  fixup $BASE/am/osssys_7_0_0.py
-
  clang2py -k cdefstum \
    $AMKERN_AMD/pm/swsmu/inc/pmfw_if/smu_v13_0_0_ppsmc.h \
    $AMKERN_AMD/pm/swsmu/inc/pmfw_if/smu13_driver_if_v13_0_0.h \
@ -484,24 +385,11 @@ generate_am() {
  clang2py -k cdefstum \
    $AMKERN_AMD/pm/swsmu/inc/pmfw_if/smu_v14_0_0_pmfw.h \
    $AMKERN_AMD/pm/swsmu/inc/pmfw_if/smu_v14_0_2_ppsmc.h \
-    $AMKERN_AMD/pm/swsmu/inc/pmfw_if/smu14_driver_if_v14_0_0.h \
    $AMKERN_AMD/pm/swsmu/inc/pmfw_if/smu14_driver_if_v14_0.h \
    extra/amdpci/headers/amdgpu_smu.h \
    --clang-args="-include stdint.h" \
    -o $BASE/am/smu_v14_0_3.py
  fixup $BASE/am/smu_v14_0_3.py
-
-  clang2py -k cdefstum \
-    $AMKERN_INC/asic_reg/hdp/hdp_6_0_0_offset.h \
-    $AMKERN_INC/asic_reg/hdp/hdp_6_0_0_sh_mask.h \
-    -o $BASE/am/hdp_6_0_0.py
-  fixup $BASE/am/hdp_6_0_0.py
-
-  clang2py -k cdefstum \
-    $AMKERN_INC/asic_reg/hdp/hdp_7_0_0_offset.h \
-    $AMKERN_INC/asic_reg/hdp/hdp_7_0_0_sh_mask.h \
-    -o $BASE/am/hdp_7_0_0.py
-  fixup $BASE/am/hdp_7_0_0.py
 }

 generate_sqtt() {
@ -523,6 +411,19 @@ generate_webgpu() {
  python3 -c "import tinygrad.runtime.autogen.webgpu"
 }

+generate_libusb() {
+  clang2py -k cdefstum \
+    /usr/include/libusb-1.0/libusb.h \
+    -o $BASE/libusb.py
+
+  fixup $BASE/libusb.py
+  sed -i "s\import ctypes\import ctypes, os\g" $BASE/libusb.py
+  sed -i "s/FIXME_STUB/libusb/g" "$BASE/libusb.py"
+  sed -i "s/libusb_le16_to_cpu = libusb_cpu_to_le16//g" "$BASE/libusb.py"
+  sed -i "s/FunctionFactoryStub()/None if (lib_path:=os.getenv('LIBUSB_PATH', ctypes.util.find_library('usb-1.0'))) is None else ctypes.CDLL(lib_path)/g" "$BASE/libusb.py"
+  python3 -c "import tinygrad.runtime.autogen.libusb"
+}
+
 if [ "$1" == "opencl" ]; then generate_opencl
 elif [ "$1" == "hip" ]; then generate_hip
 elif [ "$1" == "comgr" ]; then generate_comgr
@ -543,6 +444,7 @@ elif [ "$1" == "adreno" ]; then generate_adreno
 elif [ "$1" == "pci" ]; then generate_pci
 elif [ "$1" == "vfio" ]; then generate_vfio
 elif [ "$1" == "webgpu" ]; then generate_webgpu
+elif [ "$1" == "libusb" ]; then generate_libusb
 elif [ "$1" == "all" ]; then generate_opencl; generate_hip; generate_comgr; generate_cuda; generate_nvrtc; generate_hsa; generate_kfd; generate_nv; generate_amd; generate_io_uring; generate_libc; generate_am; generate_webgpu
 else echo "usage: $0 <type>"
 fi
--- a/tinygrad_repo/docs/abstractions2.py
+++ b/tinygrad_repo/docs/abstractions2.py
@ -1,7 +1,7 @@
 # tinygrad is a tensor library, and as a tensor library it has multiple parts
 # 1. a "runtime". this allows buffer management, compilation, and running programs
 # 2. a "Device" that uses the runtime but specifies compute in an abstract way for all
-# 3. a "LazyBuffer" that fuses the compute into kernels, using memory only when needed
+# 3. a "UOp" that fuses the compute into kernels, using memory only when needed
 # 4. a "Tensor" that provides an easy to use frontend with autograd ".backward()"


@ -39,7 +39,7 @@ DEVICE = "CPU"   # NOTE: you can change this!
 import struct
 from tinygrad.dtype import dtypes
 from tinygrad.device import Buffer, Device
-from tinygrad.ops import UOp, Ops
+from tinygrad.uop.ops import UOp, Ops
 from tinygrad.shape.shapetracker import ShapeTracker

 # allocate some buffers + load in values
@ -74,30 +74,47 @@ fxn.exec([out, a, b])
 assert out.as_buffer().cast('I')[0] == 5


-print("******** third, the LazyBuffer ***********")
+print("******** third, the UOp ***********")

 from tinygrad.engine.realize import run_schedule
 from tinygrad.engine.schedule import create_schedule_with_vars
+from tinygrad.engine.grouper import get_kernelize_map

 # allocate some values + load in values
-a = UOp.metaop(Ops.EMPTY, (1,), dtypes.int32, DEVICE)
-b = UOp.metaop(Ops.EMPTY, (1,), dtypes.int32, DEVICE)
+a = UOp.new_buffer(DEVICE, 1, dtypes.int32)
+b = UOp.new_buffer(DEVICE, 1, dtypes.int32)
 a.buffer.allocate().copyin(memoryview(bytearray(struct.pack("I", 2))))
 b.buffer.allocate().copyin(memoryview(bytearray(struct.pack("I", 3))))

 # describe the computation
-out = a.alu(Ops.ADD, b)
+out = a + b
+s = UOp(Ops.SINK, dtypes.void, (out,))

-# schedule the computation as a list of kernels
-sched, _, becomes_map = create_schedule_with_vars(out.sink())
-for si in sched: print(si.ast.op)  # NOTE: the first two convert it to CPU
-# NOTE: UOps are no longer mutable, the scheduler gives you a map to lookup which BUFFER the result was written to
-out = becomes_map[out]
+# group the computation into kernels
+becomes_map = get_kernelize_map(s)
+
+# the compute maps to an assign
+assign = becomes_map[a+b]
+
+# the first source is the output buffer (data)
+assert assign.src[0].op is Ops.BUFFER
+# the second source is the kernel (compute)
+assert assign.src[1].op is Ops.KERNEL
+
+# schedule the kernel graph in a linear list
+s = UOp(Ops.SINK, dtypes.void, (assign,))
+sched, _ = create_schedule_with_vars(s)
+assert len(sched) == 1

 # DEBUGGING: print the compute ast
 print(sched[-1].ast)
 # NOTE: sched[-1].ast is the same as st_0 above

+# the output will be stored in a new buffer
+out = assign.buf_uop
+assert out.op is Ops.BUFFER and not out.buffer.is_allocated()
+print(out)
+
 # run that schedule
 run_schedule(sched)

--- a/tinygrad_repo/docs/developer/am.md
+++ b/tinygrad_repo/docs/developer/am.md
@ -1,6 +1,6 @@
 # AM Driver

-AM driver is a userspace driver targeting AMD's 7900XTX. You only need tinygrad to send compute tasks to your GPU!
+AM driver is a userspace driver targeting AMD's RDNA3/RDNA4. You only need tinygrad to send compute tasks to your GPU!

 ## How to run?
 Make sure that amdgpu module is unloaded and just run tinygrad with `AMD=1`!
--- a/tinygrad_repo/docs/developer/kernelize.md
+++ b/tinygrad_repo/docs/developer/kernelize.md
@ -0,0 +1,109 @@
+# Kernel Creation
+
+Tinygrad lazily builds up a graph of Tensor operations. The Tensor graph includes a mix of:
+
+- Buffer and Assignment Ops: `BUFFER`, `BUFFER_VIEW`, `COPY`, `ASSIGN`
+- Movement Ops: `RESHAPE`, `EXPAND`, `PERMUTE`, `PAD`, `SHRINK`, `FLIP`
+- Compute Ops: `ADD`, `MUL`, `REDUCE_AXIS`, ...
+
+`Tensor.kernelize` creates the kernels and buffers needed to realize the output Tensor(s).
+
+## Kernelize flow
+
+Let's see how a multiply add Tensor graph becomes a fused elementwise kernel.
+
+```py
+# initialize 3 input buffers on the device
+a = Tensor([1]).realize()
+b = Tensor([2]).realize()
+c = Tensor([3]).realize()
+
+# create the Tensor graph
+mul = a*b
+out = mul+c
+
+print(mul) # <Tensor <UOp METAL (1,) int (<Ops.MUL: 48>, None)> on METAL with grad None>
+print(out) # <Tensor <UOp METAL (1,) int (<Ops.ADD: 52>, None)> on METAL with grad None>
+
+out.kernelize()
+
+print(mul) # <Tensor <UOp METAL (1,) int (<Ops.MUL: 48>, None)> on METAL with grad None>
+print(out) # <Tensor <UOp METAL (1,) int (<Ops.ASSIGN: 66>, None)> on METAL with grad None>
+```
+
+The multiply Tensor stays the same because it is fused. The output Tensor's UOp becomes a new ASSIGN UOp:
+
+```py
+print(out.lazydata)
+```
+
+The first source is the output BUFFER:
+
+```
+UOp(Ops.BUFFER, dtypes.int, arg=1, src=(
+  UOp(Ops.DEVICE, dtypes.void, arg='METAL', src=()),
+  UOp(Ops.UNIQUE, dtypes.void, arg=6, src=()),))
+```
+
+And the second source is the KERNEL and its 4 buffer edges (output_buffer, a, b, c):
+
+```
+UOp(Ops.KERNEL, dtypes.void, arg=<Kernel 12 SINK(<Ops.STORE: 45>,) (__add__, __mul__)>, src=(
+  UOp(Ops.BUFFER, dtypes.int, arg=1, src=(
+    x1:=UOp(Ops.DEVICE, dtypes.void, arg='METAL', src=()),
+    UOp(Ops.UNIQUE, dtypes.void, arg=6, src=()),)),
+  UOp(Ops.BUFFER, dtypes.int, arg=1, src=(
+     x1,
+    UOp(Ops.UNIQUE, dtypes.void, arg=1, src=()),)),
+  UOp(Ops.BUFFER, dtypes.int, arg=1, src=(
+     x1,
+    UOp(Ops.UNIQUE, dtypes.void, arg=3, src=()),)),
+  UOp(Ops.BUFFER, dtypes.int, arg=1, src=(
+     x1,
+    UOp(Ops.UNIQUE, dtypes.void, arg=5, src=()),)),))
+```
+
+KERNEL describes the compute AST, metadata and memory dependencies.
+
+BUFFER holds a reference to the device memory where the output will be stored.
+
+Once a Tensor is kernelized, all children will LOAD its BUFFER, instead of fusing it:
+
+```py
+child = out+2
+child.kernelize()
+print(child.lazydata.src[1].arg.ast)
+```
+
+```
+UOp(Ops.SINK, dtypes.void, arg=None, src=(
+  UOp(Ops.STORE, dtypes.void, arg=None, src=(
+    UOp(Ops.DEFINE_GLOBAL, dtypes.int.ptr(1), arg=0, src=()),
+    x2:=UOp(Ops.VIEW, dtypes.void, arg=ShapeTracker(views=(View(shape=(1,), strides=(0,), offset=0, mask=None, contiguous=True),)), src=()),
+    UOp(Ops.ADD, dtypes.int, arg=None, src=(
+      UOp(Ops.LOAD, dtypes.int, arg=None, src=(
+        UOp(Ops.DEFINE_GLOBAL, dtypes.int.ptr(1), arg=1, src=()),
+         x2,)),
+      UOp(Ops.CONST, dtypes.int, arg=2, src=(
+         x2,)),)),)),))
+```
+
+`Tensor.realize` will execute the kernels and write outputs to memory:
+
+```py
+Tensor.realize(out)
+print(out)        # <Tensor <UOp METAL (1,) int (<Ops.BUFFER: 23>, <buf real:True device:METAL size:1 dtype:dtypes.int offset:0>)> on METAL with grad None>
+print(out.item()) # 5
+```
+
+<hr />
+
+**Summary**
+
+- The large Tensor graph is built from a mix of data, compute and movement Ops.
+
+- `Tensor.kernelize` splits the Tensor graph into data (BUFFER), compute (KERNEL) and links dependencies with ASSIGN.
+
+- `Tensor.realize` executes KERNELs on device and replaces the Tensor graph with just a BUFFER.
+
+- Kernelize can be called multiple times on a Tensor. This allows for incrementally building the kernel fusion layout of a large Tensor graph, without having to call `realize` or `schedule`.
--- a/tinygrad_repo/docs/developer/speed.md
+++ b/tinygrad_repo/docs/developer/speed.md
@ -68,4 +68,4 @@ We have a simple framework in tinygrad for adding these ALU blocks and achieving

 ### Indexing

-Indexing determines the address of the memory we need to load. GPUs often have less integer math resources than floating point math, so this can sometimes be the bottleneck. We have a symbolic math engine in our rewrite rules to simplifiy indexing before it's emitted to the kernel. Newer NVIDIA GPUs have a "Tensor Memory Accelerator" to assist with fast indexing, however, this is not supported in tinygrad yet.
+Indexing determines the address of the memory we need to load. GPUs often have less integer math resources than floating point math, so this can sometimes be the bottleneck. We have a symbolic math engine in our rewrite rules to simplify indexing before it's emitted to the kernel. Newer NVIDIA GPUs have a "Tensor Memory Accelerator" to assist with fast indexing, however, this is not supported in tinygrad yet.
--- a/tinygrad_repo/docs/developer/uop.md
+++ b/tinygrad_repo/docs/developer/uop.md
@ -1,10 +1,10 @@
-::: tinygrad.ops.UOp
+::: tinygrad.uop.ops.UOp
    options:
        members: false
        members_order: source
        show_labels: false

-::: tinygrad.ops.Ops
+::: tinygrad.uop.ops.Ops
    options:
        members: true
        members_order: source
--- a/tinygrad_repo/docs/env_vars.md
+++ b/tinygrad_repo/docs/env_vars.md
@ -30,7 +30,7 @@ These control the behavior of core tinygrad even when used as a library.

 Variable | Possible Value(s) | Description
 ---|---|---
-DEBUG               | [1-6]      | enable debugging output, with 4 you get operations, timings, speed, generated code and more
+DEBUG               | [1-7]      | enable debugging output (operations, timings, speed, generated code and more)
 GPU                 | [1]        | enable the GPU (OpenCL) backend
 CUDA                | [1]        | enable CUDA backend
 AMD                 | [1]        | enable AMD backend
@ -50,3 +50,16 @@ JIT                 | [0-2]      | 0=disabled, 1=[jit enabled](quickstart.md#jit
 VIZ                 | [1]        | 0=disabled, 1=[viz enabled](https://github.com/tinygrad/tinygrad/tree/master/tinygrad/viz)
 ALLOW_TF32          | [1]        | enable TensorFloat-32 tensor cores on Ampere or newer GPUs.
 WEBGPU_BACKEND      | [WGPUBackendType_Metal, ...]          | Force select a backend for WebGPU (Metal, DirectX, OpenGL, Vulkan...)
+CUDA_PATH           | str        | Use `CUDA_PATH/include` for CUDA headers for CUDA and NV backends. If not set, TinyGrad will use `/usr/local/cuda/include`, `/usr/include` and `/opt/cuda/include`.
+
+## Debug breakdown
+
+Variable | Value | Description
+---|---|---
+DEBUG               | >= 1       | Enables debugging and lists devices being used
+DEBUG               | >= 2       | Provides performance metrics for operations, including timing, memory usage, bandwidth for each kernel execution
+DEBUG               | >= 3       | Outputs buffers used for each kernel (shape, dtype and strides) and the applied optimizations at a kernel level
+DEBUG               | >= 4       | Outputs the generated kernel code
+DEBUG               | >= 5       | Displays the intermediate representation of the computation UOps (AST)
+DEBUG               | >= 6       | Displays the intermediate representation of the computation UOps in a linearized manner, detailing the operation sequence
+DEBUG               | >= 7       | Outputs the assembly code generated for the target hardware
--- a/tinygrad_repo/docs/index.md
+++ b/tinygrad_repo/docs/index.md
@ -42,7 +42,7 @@ There's nothing special about a "Module" class in tinygrad, it's just a normal c

 ### tinygrad is functional

-In tinygrad, you can do [`x.conv2d(w, b)`](tensor/ops.md/#tinygrad.Tensor.conv2d) or [`x.sparse_categorical_cross_entropy(y)`](tensor/ops.md/#tinygrad.Tensor.sparse_categorical_crossentropy). We do also have a [`Conv2D`](nn.md/#tinygrad.nn.Conv2d) class like PyTorch if you want a place to keep the state, but all stateless operations don't have classes.
+In tinygrad, you can do [`x.conv2d(w, b)`](tensor/ops.md/#tinygrad.Tensor.conv2d) or [`x.sparse_categorical_crossentropy(y)`](tensor/ops.md/#tinygrad.Tensor.sparse_categorical_crossentropy). We do also have a [`Conv2D`](nn.md/#tinygrad.nn.Conv2d) class like PyTorch if you want a place to keep the state, but all stateless operations don't have classes.

 ### tinygrad is lazy

--- a/tinygrad_repo/docs/runtime.md
+++ b/tinygrad_repo/docs/runtime.md
@ -5,7 +5,7 @@ tinygrad supports various runtimes, enabling your code to scale across a wide ra
 | Runtime | Description | Requirements |
 |---------|-------------|--------------|
 | [NV](https://github.com/tinygrad/tinygrad/tree/master/tinygrad/runtime/ops_nv.py) | Provides acceleration for NVIDIA GPUs | Ampere/Ada series GPUs |
-| [AMD](https://github.com/tinygrad/tinygrad/tree/master/tinygrad/runtime/ops_amd.py) | Provides acceleration for AMD GPUs | RDNA2/RDNA3 series GPUs |
+| [AMD](https://github.com/tinygrad/tinygrad/tree/master/tinygrad/runtime/ops_amd.py) | Provides acceleration for AMD GPUs | RDNA2/RDNA3/RDNA4 series GPUs. You can select one of the interfaces for communication by setting `AMD_IFACE=(KFD|PCI)`. See [AMD interfaces](#amd-interfaces) for more details. |
 | [QCOM](https://github.com/tinygrad/tinygrad/tree/master/tinygrad/runtime/ops_qcom.py) | Provides acceleration for QCOM GPUs | 6xx series GPUs |
 | [METAL](https://github.com/tinygrad/tinygrad/tree/master/tinygrad/runtime/ops_metal.py) | Utilizes Metal for acceleration on Apple devices | M1+ Macs; Metal 3.0+ for `bfloat` support |
 | [CUDA](https://github.com/tinygrad/tinygrad/tree/master/tinygrad/runtime/ops_cuda.py) | Utilizes CUDA for acceleration on NVIDIA GPUs | NVIDIA GPU with CUDA support |
@ -64,3 +64,11 @@ rawbuf_ptr = to_mv(cl_buf_desc_ptr, 0x100).cast('Q')[20] # offset 0xA0 is a raw
 # create tiny tensor
 tiny = Tensor.from_blob(rawbuf_ptr, (h*w*4,), dtype=dtypes.imagef((h,w)), device='QCOM')
 ```
+
+## AMD Interfaces
+AMD backend supports several interfaces for communicating with devices:
+
+* `KFD`: uses the amdgpu driver
+* `PCI`: uses the [AM driver](developer/am.md)
+
+You can force an interface by setting `AMD_IFACE` to one of these values. In the case of `AMD_IFACE=PCI`, this may unbind your GPU from the amdgpu driver.
--- a/tinygrad_repo/docs/tensor/creation.md
+++ b/tinygrad_repo/docs/tensor/creation.md
@ -22,6 +22,7 @@
 ::: tinygrad.Tensor.rand
 ::: tinygrad.Tensor.rand_like
 ::: tinygrad.Tensor.randn
+::: tinygrad.Tensor.randn_like
 ::: tinygrad.Tensor.randint
 ::: tinygrad.Tensor.normal
 ::: tinygrad.Tensor.uniform
--- a/tinygrad_repo/docs/tensor/movement.md
+++ b/tinygrad_repo/docs/tensor/movement.md
@ -18,6 +18,7 @@
 ::: tinygrad.Tensor.repeat_interleave
 ::: tinygrad.Tensor.split
 ::: tinygrad.Tensor.chunk
+::: tinygrad.Tensor.unfold
 ::: tinygrad.Tensor.meshgrid
 ::: tinygrad.Tensor.squeeze
 ::: tinygrad.Tensor.unsqueeze
--- a/tinygrad_repo/examples/beautiful_mnist.py
+++ b/tinygrad_repo/examples/beautiful_mnist.py
@ -1,4 +1,4 @@
-# model based off https://towardsdatascience.com/going-beyond-99-mnist-handwritten-digits-recognition-cfff96337392
+# model based off https://medium.com/data-science/going-beyond-99-mnist-handwritten-digits-recognition-cfff96337392
 from typing import List, Callable
 from tinygrad import Tensor, TinyJit, nn, GlobalCounters
 from tinygrad.helpers import getenv, colored, trange
--- a/tinygrad_repo/examples/benchmark_onnx.py
+++ b/tinygrad_repo/examples/benchmark_onnx.py
@ -1,12 +1,12 @@
-import sys, onnx, time
-from tinygrad import TinyJit, Device, GlobalCounters, fetch, getenv
+import sys, onnx, time, pickle
+from tinygrad import TinyJit, GlobalCounters, fetch, getenv
 from tinygrad.frontend.onnx import OnnxRunner
 from extra.onnx_helpers import get_example_inputs, validate

 def load_onnx_model(onnx_file):
  onnx_model = onnx.load(onnx_file)
  run_onnx = OnnxRunner(onnx_model)
-  run_onnx_jit = TinyJit(lambda **kwargs: next(iter(run_onnx({k:v.to(Device.DEFAULT) for k,v in kwargs.items()}).values())), prune=True)
+  run_onnx_jit = TinyJit(lambda **kwargs: next(iter(run_onnx({k:v.to(None) for k,v in kwargs.items()}).values())), prune=True, optimize=True)
  return run_onnx_jit, run_onnx.graph_inputs

 if __name__ == "__main__":
--- a/tinygrad_repo/examples/coder.py
+++ b/tinygrad_repo/examples/coder.py
@ -34,8 +34,8 @@ if __name__ == "__main__":
    part2 = nn.state.torch_load(fetch("https://huggingface.co/teknium/OpenHermes-2.5-Mistral-7B/resolve/main/pytorch_model-00002-of-00002.bin?download=true"))

  with Timing("weights -> model: "):
-    nn.state.load_state_dict(model, fix_bf16(convert_from_huggingface(part1, model, 32, 8)), strict=False)
-    nn.state.load_state_dict(model, fix_bf16(convert_from_huggingface(part2, model, 32, 8)), strict=False)
+    nn.state.load_state_dict(model, fix_bf16(convert_from_huggingface(part1, 32, 32, 8)), strict=False)
+    nn.state.load_state_dict(model, fix_bf16(convert_from_huggingface(part2, 32, 32, 8)), strict=False)

  if not os.path.isfile("/tmp/tokenizer.model"): create_fixed_tokenizer("/tmp/tokenizer.model")
  spp = SentencePieceProcessor(model_file="/tmp/tokenizer.model")
--- a/tinygrad_repo/examples/gpt2.py
+++ b/tinygrad_repo/examples/gpt2.py
@ -3,10 +3,11 @@ import os, argparse, contextlib
 from typing import Optional, Union
 with contextlib.suppress(ImportError): import tiktoken
 from tinygrad import Tensor, TinyJit, Device, GlobalCounters, Variable, dtypes
-from tinygrad.ops import UOp
+from tinygrad.uop.ops import UOp
 from tinygrad.helpers import Timing, DEBUG, JIT, getenv, fetch, colored, trange
 from tinygrad.nn import Embedding, Linear, LayerNorm
 from tinygrad.nn.state import gguf_load, torch_load, load_state_dict, get_state_dict
+from extra.bench_log import BenchEvent, WallTimeEvent

 MAX_CONTEXT = getenv("MAX_CONTEXT", 128)
 HALF = getenv("HALF")
@ -134,6 +135,7 @@ class GPT2:
    # lm head and wte are tied
    weights['lm_head.weight'] = weights['wte.weight']

+    with WallTimeEvent(BenchEvent.LOAD_WEIGHTS):
      load_state_dict(model, weights)

      if HALF:
@ -167,6 +169,7 @@ class GPT2:
      return key
    state_dict = { _remap_gguf_key(k): v for k, v in state_dict.items() }
    model = Transformer(**gpt2_params)
+    with WallTimeEvent(BenchEvent.LOAD_WEIGHTS):
      load_state_dict(model, state_dict)
    return GPT2(model, tiktoken.get_encoding("gpt2"))

@ -185,11 +188,12 @@ class GPT2:
      with Timing("ran model in ", on_exit=(lambda et: (f", {(GlobalCounters.time_sum_s-st)*1e3:.2f} ms on GPU" if DEBUG>=2 else "")+
                  f", {GlobalCounters.global_ops*1e-9:.2f} GOPS, {GlobalCounters.global_mem*1e-9:.2f} GB"+
                  (f", {GlobalCounters.global_mem*1e-9/(GlobalCounters.time_sum_s-st):.2f} GB/s" if DEBUG>=2 else "")) if DEBUG else None, enabled=timing):
+        with WallTimeEvent(BenchEvent.STEP):
          if batch_size == 1 and len(toks[0][start_pos:]) == 1:
            tokens = Variable("tokens", 0, VOCAB_SIZE).bind(toks[0][start_pos])
          else:
            tokens = Tensor([x[start_pos:] for x in toks])
-        tok = self.model(tokens, Variable("start_pos", 1 if start_pos else 0, MAX_CONTEXT).bind(start_pos), temperature).tolist()
+          tok = self.model(tokens, Variable("start_pos", 1 if start_pos else 0, MAX_CONTEXT-1).bind(start_pos), temperature).tolist()
      start_pos = len(toks[0])
      for i,t in enumerate(tok): toks[i].append(t)
    return [self.tokenizer.decode(x) for x in toks]
--- a/tinygrad_repo/examples/handcode_opt.py
+++ b/tinygrad_repo/examples/handcode_opt.py
@ -3,7 +3,8 @@ from extra.mcts_search import mcts_search
 from examples.mlperf.helpers import get_mlperf_bert_model
 from tinygrad import Tensor, Device, dtypes, nn
 from tinygrad.codegen.kernel import Kernel
-from tinygrad.ops import Ops, sym_infer
+from tinygrad.codegen.heuristic import hand_coded_optimizations
+from tinygrad.uop.ops import Ops, sym_infer
 from tinygrad.device import Compiled
 from tinygrad.engine.search import beam_search, bufs_from_lin
 from tinygrad.helpers import DEBUG, ansilen, getenv, colored, TRACEMETA
@ -83,7 +84,7 @@ if __name__ == "__main__":

    # always try hand coded opt
    lin = Kernel(si.ast, opts=device.renderer)
-    lin.hand_coded_optimizations()
+    lin.apply_opts(hand_coded_optimizations(lin))
    lins.append((lin, "HC"))

    # maybe try tensor cores
--- a/tinygrad_repo/examples/hlb_cifar10.py
+++ b/tinygrad_repo/examples/hlb_cifar10.py
@ -11,6 +11,7 @@ from tinygrad import nn, dtypes, Tensor, Device, GlobalCounters, TinyJit
 from tinygrad.nn.state import get_state_dict, get_parameters
 from tinygrad.nn import optim
 from tinygrad.helpers import Context, BEAM, WINO, getenv, colored, prod
+from extra.bench_log import BenchEvent, WallTimeEvent

 cifar_mean = [0.4913997551666284, 0.48215855929893703, 0.4465309133731618]
 cifar_std = [0.24703225141799082, 0.24348516474564, 0.26158783926049628]
@ -395,6 +396,8 @@ def train_cifar():
      if STEPS == 0 or i == STEPS: break

      GlobalCounters.reset()
+
+      with WallTimeEvent(BenchEvent.STEP):
        X, Y = next(batcher)
        if len(GPUS) > 1:
          X.shard_(GPUS, axis=0)
@ -409,6 +412,7 @@ def train_cifar():
          if model_ema is None:
            model_ema = modelEMA(W, model)
          model_ema.update(model, Tensor([projected_ema_decay_val*(i/STEPS)**hyp['ema']['decay_pow']]))
+
      cl = time.monotonic()
      device_str = loss.device if isinstance(loss.device, str) else f"{loss.device[0]} * {len(loss.device)}"
      #  53  221.74 ms run,    2.22 ms python,  219.52 ms CL,  803.39 loss, 0.000807 LR, 4.66 GB used,   3042.49 GFLOPS,    674.65 GOPS
@ -424,4 +428,5 @@ def train_cifar():
      raise ValueError(colored(f"{eval_acc_pct=} < {target}", "red"))

 if __name__ == "__main__":
+  with WallTimeEvent(BenchEvent.FULL):
    train_cifar()
--- a/tinygrad_repo/examples/llama.py
+++ b/tinygrad_repo/examples/llama.py
@ -13,6 +13,7 @@ from extra.models.llama import Transformer, convert_from_huggingface, fix_bf16
 from sentencepiece import SentencePieceProcessor
 import tiktoken, sys
 from tiktoken.load import load_tiktoken_bpe
+from extra.bench_log import BenchEvent, WallTimeEvent

 MAX_CONTEXT = getenv("MAX_CONTEXT", 4096)

@ -206,16 +207,19 @@ class LLaMa:

    model = Transformer(**params["args"], linear=linear, max_context=MAX_CONTEXT, jit=bool(JIT))

+    with WallTimeEvent(BenchEvent.LOAD_WEIGHTS):
      if model_path.is_dir():
        weights = concat_weights([load(filename) for filename in [f"{model_path}/consolidated.{i:02d}.pth" for i in range(params["files"])]], device[0] if isinstance(device, tuple) else device)
      else:
        weights = load(str(model_path))
      if "model.embed_tokens.weight" in weights:
-      weights = convert_from_huggingface(weights, model, params["args"]["n_heads"], params["args"].get("n_kv_heads", params["args"]["n_heads"]))
+        weights = convert_from_huggingface(weights, params["args"]["n_layers"], params["args"]["n_heads"], params["args"].get("n_kv_heads", params["args"]["n_heads"]))

      weights = fix_bf16(weights)

-    with Context(BEAM=0):
+      # prevent tracking model weights
+      # this is a part of a larger problem with BUFFER UOps and gc in TRACK_MATCH_STATS=2
+      with Context(BEAM=0, TRACK_MATCH_STATS=0):
        # quantize
        if quantize is not None:
          weights = linear.quantize(weights, device)
@ -475,6 +479,7 @@ After you are done speaking, output [EOS]. You are not Chad.
      next_tok = Tensor([toks[start_pos:]], device=device) if tok_tensor is None or (len(toks)-start_pos) > 1 else tok_tensor.reshape(1, 1)
      with Profiling(enabled=args.profile):
        with Timing("total ", enabled=args.timing, on_exit=lambda x: f", {1e9/x:.2f} tok/s, {GlobalCounters.global_mem/x:.2f} GB/s, param {param_bytes/x:.2f} GB/s"):
+          with WallTimeEvent(BenchEvent.STEP):
            with Timing("enqueue in ", on_exit=(lambda et: (f", {(GlobalCounters.time_sum_s-st)*1e3:.2f} ms on GPU" if DEBUG>=2 else "")+
                        f", {GlobalCounters.global_ops*1e-9:.2f} GOPS, {GlobalCounters.global_mem*1e-9:.2f} GB"+
                        (f", {GlobalCounters.global_mem*1e-9/(GlobalCounters.time_sum_s-st):.2f} GB/s, param {param_bytes*1e-9/(GlobalCounters.time_sum_s-st):.2f} GB/s" if DEBUG>=2 else "")) if DEBUG else None, enabled=args.timing):
--- a/tinygrad_repo/examples/llama3.py
+++ b/tinygrad_repo/examples/llama3.py
@ -7,6 +7,7 @@ from extra.models.llama import Transformer, convert_from_huggingface, convert_fr
 from tinygrad.nn.state import safe_load, torch_load, load_state_dict, get_parameters, gguf_load
 from tinygrad import Tensor, dtypes, nn, Context, Device, GlobalCounters
 from tinygrad.helpers import Profiling, Timing, DEBUG, colored, fetch, tqdm
+from extra.bench_log import BenchEvent, WallTimeEvent

 class Tokenizer:
  pat_str = r"(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\r\n\p{L}\p{N}]?\p{L}+|\p{N}{1,3}| ?[^\s\p{L}\p{N}]+[\r\n]*|\s*[\r\n]+|\s+(?!\S)|\s+"
@ -126,7 +127,8 @@ def NF4Linear(block_size):
      return x.linear(unscaled.reshape(self.out_features, self.in_features).T)

    @staticmethod
-    def quantize(state_dict: dict[str, Tensor], device, scale_dtype=dtypes.float16) -> dict[str, Tensor]:
+    def quantize(state_dict: dict[str, Tensor], device, scale_dtype=dtypes.float16, quantize_embeds=False) -> dict[str, Tensor]:
+      assert not quantize_embeds  # TODO: support this?
      new_state_dict = {}
      for k, v in state_dict.items():
        if "feed_forward" in k or "attention.w" in k:
@ -165,7 +167,9 @@ def build_transformer(model_path: Path, model_size="8B", quantize=None, scale_dt
  model = Transformer(**MODEL_PARAMS[model_size]["args"], linear=linear, embedding=embedding, max_context=max_context, jit=True)

  if not load_weights: return model
+
  # load weights
+  with WallTimeEvent(BenchEvent.LOAD_WEIGHTS):
    if model_path.is_dir():
      if (model_path / "model.safetensors.index.json").exists(): weights = load(str(model_path / "model.safetensors.index.json"))
      elif (model_path / "model.safetensors").exists(): weights = load(str(model_path / "model.safetensors"))
@ -173,9 +177,9 @@ def build_transformer(model_path: Path, model_size="8B", quantize=None, scale_dt
    else:
      weights = load(str(model_path))
    if "model.embed_tokens.weight" in weights:
-    weights = convert_from_huggingface(weights, model, MODEL_PARAMS[model_size]["args"]["n_heads"], MODEL_PARAMS[model_size]["args"]["n_kv_heads"])
+      weights = convert_from_huggingface(weights, MODEL_PARAMS[model_size]["args"]["n_layers"], MODEL_PARAMS[model_size]["args"]["n_heads"], MODEL_PARAMS[model_size]["args"]["n_kv_heads"])
    elif "token_embd.weight" in weights:
-    weights = convert_from_gguf(weights, model)
+      weights = convert_from_gguf(weights, MODEL_PARAMS[model_size]["args"]["n_layers"])
    weights = fix_bf16(weights)

    with Context(BEAM=0):
@ -434,6 +438,7 @@ if __name__ == "__main__":
      st = GlobalCounters.time_sum_s
      with Profiling(enabled=args.profile):
        with Timing("total ", on_exit=lambda x: f", {1e9/x:.2f} tok/s, {GlobalCounters.global_mem/x:.2f} GB/s, param {param_bytes/x:.2f} GB/s"):
+          with WallTimeEvent(BenchEvent.STEP):
            with Timing("enqueue in ", on_exit=(lambda et: (f", {(GlobalCounters.time_sum_s-st)*1e3:.2f} ms on GPU" if DEBUG>=2 else "")+
                        f", {GlobalCounters.global_ops*1e-9:.2f} GOPS, {GlobalCounters.global_mem*1e-9:.2f} GB"+
                        (f", {GlobalCounters.global_mem*1e-9/(GlobalCounters.time_sum_s-st):.2f} GB/s, param {param_bytes*1e-9/(GlobalCounters.time_sum_s-st):.2f} GB/s" if DEBUG>=2 else "")) if DEBUG else None):
--- a/tinygrad_repo/examples/llm.c/export.py
+++ b/tinygrad_repo/examples/llm.c/export.py
@ -7,7 +7,7 @@ from train_gpt2 import GPT, GPTConfig
 from tinygrad.helpers import dedup, to_function_name, flatten, getenv, GlobalCounters, ansilen, to_function_name
 from tinygrad.engine.realize import get_kernel, run_schedule
 from tinygrad.engine.memory import memory_planner
-from tinygrad.ops import Ops
+from tinygrad.uop.ops import Ops

 TIMING = getenv("TIMING")

@ -25,7 +25,7 @@ if __name__ == "__main__":
  Tensor.training = True
  optimizer = nn.optim.Adam(nn.state.get_parameters(model), lr=1e-4)
  warmup_count = getenv("WARMUP", 3)
-  for i in range(warmup_count):  # TODO: why does it take three and not two to stablize
+  for i in range(warmup_count):  # TODO: why does it take three and not two to stabilize
    GlobalCounters.reset()
    X = Tensor.empty(4, 64, dtype=dtypes.int).reshape(B, T)
    Y = Tensor.empty(4, 64, dtype=dtypes.int).reshape(B, T)
--- a/tinygrad_repo/examples/llm.c/train_gpt2.py
+++ b/tinygrad_repo/examples/llm.c/train_gpt2.py
@ -99,7 +99,7 @@ class GPT:

  def __call__(self, idx:Tensor, targets=None):
    b, t = idx.shape
-    pos = Tensor.arange(0, t)
+    pos = Tensor.arange(0, t, device=idx.device)

    tok_emb = self.wte(idx) # token embeddings of shape (b, t, n_embd)
    pos_emb = self.wpe(pos) # position embeddings of shape (t, n_embd)
@ -124,6 +124,7 @@ if __name__ == "__main__":
  parser.add_argument("--batch_size", type=int, default=4, help="batch size")
  parser.add_argument("--sequence_length", type=int, default=64, help="sequence length")
  parser.add_argument("--skip_test", action="store_true", help="skip test")
+  parser.add_argument("--gpus", type=int, default=1, help="sequence length")
  args = parser.parse_args()
  B, T = args.batch_size, args.sequence_length
  assert 1 <= T <= 1024
@ -131,6 +132,10 @@ if __name__ == "__main__":
  model = GPT(GPTConfig(n_layer=12, n_head=12, n_embd=768))
  model.load_pretrained()

+  if args.gpus > 1:
+    GPUS = tuple(f'{Device.DEFAULT}:{i}' for i in range(args.gpus))
+    for x in nn.state.get_parameters(model): x.to_(GPUS)  # we put a copy of the model on every GPU
+
  # init the tokenizer
  enc = tiktoken.get_encoding("gpt2")
  encode = lambda s: enc.encode(s, allowed_special={"<|endoftext|>"})
@ -165,23 +170,32 @@ if __name__ == "__main__":
  x, y = next(data_iter) # we'll overfit this batch below
  optimizer = nn.optim.AdamW(nn.state.get_parameters(model), lr=1e-4, weight_decay=0)

+  print(f"model state:     {sum(x.nbytes() for x in nn.state.get_parameters(model))/1e9:.2f} GB")
+  print(f"optimizer state: {sum(x.nbytes() for x in nn.state.get_parameters(optimizer))/1e9:.2f} GB")
+
+  # shard the data on axis 0
+  if args.gpus > 1: x, y = x.shard(GPUS, axis=0), y.shard(GPUS, axis=0)
+
  @TinyJit
-  def step(x, y):
+  @Tensor.train()
+  def step(x:Tensor, y:Tensor) -> Tensor:
    _, loss = model(x, y)
    optimizer.zero_grad()
    loss.backward()
    return loss.realize(*optimizer.schedule_step())

-  with Tensor.train():
  for i in range(args.num_iterations):
    GlobalCounters.reset()
-      t0 = time.time()
+    t0 = time.perf_counter()
    loss = step(x.contiguous(), y.contiguous())
    Device[Device.DEFAULT].synchronize()
-      t1 = time.time()
-      print(f"iteration {i}, loss: {loss.item():.6f}, time: {(t1-t0)*1000:.3f}ms, {int(B*T/(t1-t0))} tok/s")
+    t1 = time.perf_counter()
+    print(f"iteration {i}, loss: {loss.item():.6f}, time: {(t1-t0)*1000:.3f}ms, {int(B*T/(t1-t0))} tok/s, {GlobalCounters.global_mem/1e9:.2f} GB")

  if not args.skip_test:
+    # copy back to single gpu for test
+    if args.gpus > 1:
+      for x in nn.state.get_parameters(model): x.to_(Device.DEFAULT)
    start = "<|endoftext|>"
    start_ids = encode(start)
    x = (Tensor(start_ids)[None, ...])
--- a/tinygrad_repo/examples/minrf.py
+++ b/tinygrad_repo/examples/minrf.py
@ -0,0 +1,157 @@
+# much taken from https://github.com/cloneofsimo/minRF
+from tinygrad import Tensor, nn, GlobalCounters, TinyJit
+from tinygrad.helpers import getenv, trange
+from extra.models.llama import Attention, FeedForward, precompute_freqs_cis
+
+def modulate(x:Tensor, shift:Tensor, scale:Tensor) -> Tensor: return x * (1 + scale.unsqueeze(1)) + shift.unsqueeze(1)
+
+# TODO: why doesn't the TimestepEmbedder from minRF work?
+class TimestepEmbedder:
+  def __init__(self, hidden_size): self.mlp = [nn.Linear(1, hidden_size), Tensor.silu, nn.Linear(hidden_size, hidden_size)]
+  def __call__(self, t:Tensor): return t.reshape(-1, 1).sequential(self.mlp)
+
+class TransformerBlock:
+  def __init__(self, dim, n_heads, norm_eps=1e-5):
+    self.attention = Attention(dim, n_heads)
+    self.feed_forward = FeedForward(dim, 4*dim)
+    self.attention_norm = nn.LayerNorm(dim, eps=norm_eps)
+    self.ffn_norm = nn.LayerNorm(dim, eps=norm_eps)
+    self.adaLN_modulation = nn.Linear(dim, 6 * dim, bias=True)
+
+  def __call__(self, x:Tensor, freqs_cis:Tensor, adaln_input:Tensor):
+    shift_msa, scale_msa, gate_msa, shift_mlp, scale_mlp, gate_mlp = self.adaLN_modulation(adaln_input.silu()).chunk(6, dim=1)
+    x = x + gate_msa.unsqueeze(1) * self.attention(modulate(self.attention_norm(x), shift_msa, scale_msa), 0, freqs_cis)
+    x = x + gate_mlp.unsqueeze(1) * self.feed_forward(modulate(self.ffn_norm(x), shift_mlp, scale_mlp))
+    return x.contiguous().contiguous_backward()
+
+class FinalLayer:
+  def __init__(self, dim, patch_size, out_channels):
+    self.norm_final = nn.LayerNorm(dim, elementwise_affine=False, eps=1e-6)
+    self.linear = nn.Linear(dim, patch_size*patch_size*out_channels, bias=True)
+    self.adaLN_modulation = nn.Linear(dim, 2 * dim, bias=True)
+
+    # init weights/bias to 0
+    self.linear.weight.replace(self.linear.weight.zeros_like().contiguous())
+    self.linear.bias.replace(self.linear.bias.zeros_like().contiguous())
+
+  def __call__(self, x:Tensor, c:Tensor):
+    shift, scale = self.adaLN_modulation(c.silu()).chunk(2, dim=1)
+    x = modulate(self.norm_final(x), shift, scale)
+    return self.linear(x)
+
+# channels=1, input_size=32, dim=64, n_layers=6, n_heads=4, num_classes=10
+class DiT_Llama:
+  def __init__(self, in_channels=1, dim=64, n_layers=6, n_heads=4, num_classes=10, patch_size=2):
+    self.patch_size = patch_size
+    self.out_channels = in_channels
+    self.num_classes = num_classes
+
+    self.init_conv_seq = [
+      nn.Conv2d(in_channels, dim // 2, kernel_size=5, padding=2, stride=1), Tensor.silu, nn.GroupNorm(32, dim//2),
+      nn.Conv2d(dim //2, dim // 2, kernel_size=5, padding=2, stride=1), Tensor.silu, nn.GroupNorm(32, dim//2),
+    ]
+
+    self.x_embedder = nn.Linear(self.patch_size * self.patch_size * dim // 2, dim, bias=True)
+    self.t_embedder = TimestepEmbedder(dim)
+    self.y_embedder = nn.Embedding(num_classes+1, dim)
+    self.final_layer = FinalLayer(dim, self.patch_size, self.out_channels)
+
+    self.freqs_cis = precompute_freqs_cis(dim // n_heads, 4096)
+    self.layers = [TransformerBlock(dim, n_heads) for _ in range(n_layers)]
+
+  def unpatchify(self, x:Tensor):
+    c, p = self.out_channels, self.patch_size
+    h = w = int(x.shape[1] ** 0.5)
+    x = x.reshape(shape=(x.shape[0], h, w, p, p, c))
+    x = x.rearrange("n h w p q c -> n c h p w q")
+    return x.reshape(shape=(x.shape[0], c, h * p, h * p))
+
+  def patchify(self, x:Tensor):
+    B, C, H, W = x.shape
+    x = x.reshape(B, C, H // self.patch_size, self.patch_size, W // self.patch_size, self.patch_size)
+    x = x.permute(0, 2, 4, 1, 3, 5).flatten(-3).flatten(1, 2)
+    return x  # B <H*W ish> <C*patch_size*patch_size>
+
+  def __call__(self, x:Tensor, t:Tensor, y:Tensor) -> Tensor:
+    x = x.sequential(self.init_conv_seq)
+    x = self.patchify(x)
+    x = self.x_embedder(x)
+    adaln_input = self.t_embedder(t) + self.y_embedder(y)
+    adaln_input = adaln_input.contiguous()
+    for layer in self.layers:
+      x = layer(x, self.freqs_cis[:, :x.size(1)], adaln_input=adaln_input)
+    x = self.final_layer(x, adaln_input)
+    return self.unpatchify(x)
+
+  def rf(self, x:Tensor, cond:Tensor):
+    b = x.shape[0]
+    # self.ln is True
+    t = Tensor.randn((b,)).sigmoid()
+    texp = t.view([b, *([1] * len(x.shape[1:]))])
+
+    # conditional dropout
+    dropout_prob = 0.1
+    cond = (Tensor.rand(cond.shape[0]) < dropout_prob).where(cond.full_like(self.num_classes), cond)
+
+    # this is rectified flow
+    z1 = x.randn_like()
+    zt = (1 - texp) * x + texp * z1
+    vtheta = self(zt, t, cond)
+
+    # MSE loss
+    return ((z1 - x) - vtheta).square().mean()
+
+  def sample(self, z, cond, null_cond, sample_steps=50, cfg=2.0):
+    b = z.size(0)
+    dt = Tensor.full((b,)+(1,)*len(z.shape[1:]), fill_value=1.0/sample_steps).contiguous()
+    images = [z]
+    for i in range(sample_steps, 0, -1):
+      t = Tensor.full((b,), fill_value=i/sample_steps).contiguous()
+      vc = self(z, t, cond)
+      vu = self(z, t, null_cond)
+      vc = vu + cfg * (vc - vu)
+      z = z - dt * vc
+      z = z.contiguous()
+      images.append(z)
+    return images
+
+def mviz(t:Tensor):
+  assert len(t.shape) == 4 and t.shape[1] == 1
+  ft = t.permute(1,2,0,3).reshape(32, -1)
+  assert ft.shape[-1]%32 == 0
+  print("")
+  for y in ((ft+1)/2).clamp(0,1).tolist():
+    ln = [f"\033[38;5;{232+int(x*23)}m██" for x in y]
+    print(''.join(ln) + "\033[0m")
+
+if __name__ == "__main__":
+  X_train, Y_train, X_test, Y_test = nn.datasets.mnist()
+  X_train = X_train.pad((2,2,2,2))
+  X_train = ((X_train.float()/255)-0.5)/0.5
+  Y_train = Y_train.int()
+
+  model = DiT_Llama(patch_size=getenv("PATCH_SIZE", 2))
+  for r in nn.state.get_parameters(model): r.realize()
+  optimizer = nn.optim.Adam(nn.state.get_parameters(model), lr=5e-4)
+
+  @TinyJit
+  @Tensor.train()
+  def train_step():
+    if getenv("OVERFIT"): samples = Tensor.zeros(getenv("BS", 256), dtype='int')
+    else: samples = Tensor.randint(getenv("BS", 256), high=X_train.shape[0])
+    optimizer.zero_grad()
+    loss = model.rf(X_train[samples], Y_train[samples])
+    loss.backward()
+    optimizer.step()
+    return loss
+
+  @TinyJit
+  @Tensor.test()
+  def sample(z:Tensor, cond:Tensor) -> Tensor:
+    return model.sample(z, cond, Tensor.full_like(cond, 10), sample_steps=getenv("SAMPLE_STEPS", 20))[-1]
+
+  for steps in (t:=trange(getenv("STEPS", 5000))):
+    if steps%10 == 0: mviz(sample(Tensor.randn(3, 1, 32, 32), Tensor([5,0,4], dtype='int')))
+    GlobalCounters.reset()
+    loss = train_step()
+    t.set_description(f"loss: {loss.item():9.2f}")
--- a/tinygrad_repo/examples/mixtral.py
+++ b/tinygrad_repo/examples/mixtral.py
@ -3,6 +3,7 @@ from tinygrad import Tensor, nn, Device, GlobalCounters, Variable
 from tinygrad.helpers import Timing, Profiling, CI, tqdm
 from tinygrad.nn.state import torch_load, get_state_dict
 from extra.models.llama import FeedForward, Transformer
+from extra.bench_log import BenchEvent, WallTimeEvent

 class MixtureFeedForward:
  def __init__(self, num_experts:int, dim:int, hidden_dim:int, linear=nn.Linear):
@ -30,6 +31,7 @@ if __name__ == "__main__":
                      help="Path to the downloaded weights")
  args = parser.parse_args()

+  with WallTimeEvent(BenchEvent.LOAD_WEIGHTS):
    state = torch_load(args.weights + "/consolidated.00.pth.b")
    model = Transformer(n_layers=32, dim=4096, hidden_dim=14336, n_heads=32, n_kv_heads=8, norm_eps=1e-5, vocab_size=32000, feed_forward=functools.partial(MixtureFeedForward, 8), jit=False)
    model_state_dict = get_state_dict(model)
@ -53,6 +55,7 @@ if __name__ == "__main__":
    GlobalCounters.reset()
    with Profiling(sort="time", frac=0.1, enabled=args.profile):
      with Timing("total ", enabled=args.timing, on_exit=lambda x: f", {1e9/x:.2f} tok/sec"):
+        with WallTimeEvent(BenchEvent.STEP):
          tok = model(Tensor([toks[start_pos:]]), 0 if start_pos == 0 else Variable("start_pos", 1, 1024).bind(start_pos), args.temperature).item()
    toks.append(tok)
    start_pos += 1
--- a/tinygrad_repo/examples/mlperf/dataloader.py
+++ b/tinygrad_repo/examples/mlperf/dataloader.py
@ -5,7 +5,7 @@ from multiprocessing import Queue, Process, shared_memory, connection, Lock, cpu

 import numpy as np
 from tinygrad import dtypes, Tensor
-from tinygrad.helpers import getenv, prod, Context, round_up, tqdm
+from tinygrad.helpers import getenv, prod, Context, round_up, tqdm, OSX

 ### ResNet

@ -129,14 +129,15 @@ def batch_load_resnet(batch_size=64, val=False, shuffle=True, seed=None, pad_fir
  q_in, q_out = Queue(), Queue()

  sz = (batch_size*BATCH_COUNT, 224, 224, 3)
-  if os.path.exists("/dev/shm/resnet_X"): os.unlink("/dev/shm/resnet_X")
-  shm = shared_memory.SharedMemory(name="resnet_X", create=True, size=prod(sz))
+  shm_name = "resnet_X_val" if val else "resnet_X_train"
+  if not OSX and os.path.exists(f"/dev/shm/{shm_name}"): os.unlink(f"/dev/shm/{shm_name}")
+  shm = shared_memory.SharedMemory(name=shm_name, create=True, size=prod(sz))
  procs = []

  try:
    # disk:shm is slower
-    #X = Tensor.empty(*sz, dtype=dtypes.uint8, device=f"disk:shm:{shm.name}")
-    X = Tensor.empty(*sz, dtype=dtypes.uint8, device=f"disk:/dev/shm/resnet_X")
+    if OSX: X = Tensor.empty(*sz, dtype=dtypes.uint8, device=f"disk:shm:{shm.name}")
+    else: X = Tensor.empty(*sz, dtype=dtypes.uint8, device=f"disk:/dev/shm/{shm_name}")
    Y = [None] * (batch_size*BATCH_COUNT)

    for _ in range(cpu_count()):
@ -532,7 +533,7 @@ if __name__ == "__main__":
  def load_retinanet(val):
    from extra.datasets.openimages import BASEDIR, download_dataset
    from pycocotools.coco import COCO
-    dataset = COCO(download_dataset(base_dir:=getenv("BASE_DIR", BASEDIR), "validation" if val else "train"))
+    dataset = COCO(download_dataset(base_dir:=getenv("BASEDIR", BASEDIR), "validation" if val else "train"))
    with tqdm(total=len(dataset.imgs.keys())) as pbar:
      for x in batch_load_retinanet(dataset, val, base_dir):
        pbar.update(x[0].shape[0])
--- a/Show More
+++ b/Show More