Merge branch 'master' into ccnc-port-better-gpu

This reverts commit 75b4e61913.

openpilot/common/params_keys.h

@@ -132,6 +132,7 @@ inline static std::unordered_map<std::string, ParamKeyAttributes> keys = {
     {"UptimeOnroad", {PERSISTENT, FLOAT, "0.0"}},
     {"UsbGpuActive", {CLEAR_ON_MANAGER_START | CLEAR_ON_OFFROAD_TRANSITION | CLEAR_ON_IGNITION_ON, BOOL}},
     {"UsbGpuLoading", {CLEAR_ON_MANAGER_START | CLEAR_ON_OFFROAD_TRANSITION | CLEAR_ON_IGNITION_ON, BOOL}},
+    {"UsbGpuLoadProgress", {CLEAR_ON_MANAGER_START | CLEAR_ON_OFFROAD_TRANSITION | CLEAR_ON_IGNITION_ON, INT, "0"}},
     {"Version", {PERSISTENT, STRING}},
 
     // --- sunnypilot params --- //

openpilot/selfdrive/modeld/load_progress.py

@@ -0,0 +1,44 @@
+import os
+from openpilot.common.file_chunker import open_file_chunked, get_existing_chunks
+from openpilot.common.params import Params
+
+PARAM = "UsbGpuLoadProgress"
+
+
+class ProgressReader:
+  def __init__(self, inner, total):
+    self._inner = inner
+    self._total = total
+    self._params = Params()
+    self._read = 0
+    self._pct = -1
+    self._step = max(64 * 1024, total // 100)
+
+  def _bump(self, n):
+    self._read += n
+    if self._total:
+      pct = min(100, self._read * 100 // self._total)
+      if pct != self._pct:
+        self._pct = pct
+        self._params.put(PARAM, pct)
+
+  def read(self, size=-1):
+    data = self._inner.read(size)
+    self._bump(len(data))
+    return data
+
+  def readinto(self, b):
+    view = memoryview(b)
+    done = 0
+    while done < len(view):
+      n = self._inner.readinto(view[done:done + self._step])
+      if not n:
+        break
+      done += n
+      self._bump(n)
+    return done
+
+
+def open_with_progress(pkl_path):
+  total = sum(os.path.getsize(p) for p in get_existing_chunks(pkl_path))
+  return ProgressReader(open_file_chunked(pkl_path), total)

openpilot/selfdrive/modeld/modeld.py

@@ -29,6 +29,7 @@ from openpilot.selfdrive.modeld.parse_model_outputs import Parser
 from openpilot.selfdrive.modeld.compile_modeld import make_input_queues, WARP_INPUTS, POLICY_INPUTS
 from openpilot.selfdrive.modeld.fill_model_msg import fill_model_msg, fill_driving_model_data, fill_pose_msg, PublishState
 from openpilot.common.file_chunker import open_file_chunked
+from openpilot.selfdrive.modeld.load_progress import open_with_progress
 from openpilot.selfdrive.modeld.constants import ModelConstants, Plan
 from openpilot.selfdrive.modeld.helpers import usbgpu_present, usbgpu_compiled, modeld_pkl_path, get_tg_input_devices, load_oob
 
@@ -145,7 +146,7 @@ class ModelState(ModelStateBase):
     ModelStateBase.__init__(self)
     input_devices = get_tg_input_devices(PROCESS_NAME, usbgpu)
     self.WARP_DEV, self.QUEUE_DEV = input_devices['WARP_DEV'], input_devices['QUEUE_DEV']
-    jits = load_oob(open_file_chunked(modeld_pkl_path(usbgpu)))
+    jits = load_oob(open_with_progress(modeld_pkl_path(usbgpu)) if usbgpu else open_file_chunked(modeld_pkl_path(usbgpu)))
     metadata = jits['metadata']
     self.input_shapes = metadata['input_shapes']
     self.vision_input_names = [k for k in self.input_shapes if 'img' in k]

openpilot/selfdrive/ui/mici/onroad/hud_renderer.py

@@ -223,7 +223,7 @@ class HudRenderer(Widget):
     if icon is not self._egpu_icon:
       self._egpu_fade_time = rl.get_time()
       self._egpu_icon = icon
-    alpha = self._egpu_alpha_filter.update(loading or 0 < rl.get_time() - self._egpu_fade_time < SET_SPEED_PERSISTENCE)
+    alpha = self._egpu_alpha_filter.update(True)
     if alpha < 1e-2:
       return
 
@@ -231,6 +231,18 @@ class HudRenderer(Widget):
                      rect.y + rect.height - 14 - (self._txt_wheel.height + icon.height) / 2)
     rl.draw_texture_ex(icon, pos, 0.0, 1.0, rl.Color(255, 255, 255, int(255 * opacity * alpha)))
 
+    if loading:
+      pct_text = f"{ui_state.usbgpu_load_progress}%"
+      size = FONT_SIZES.max_speed
+      cell = measure_text_cached(self._font_bold, "0", size)
+      widths = [cell.x if c.isdigit() else measure_text_cached(self._font_bold, c, size).x for c in pct_text]
+      x = pos.x - 8 - sum(widths)
+      y = pos.y + (icon.height - cell.y) / 2
+      for c, w in zip(pct_text, widths):
+        glyph = measure_text_cached(self._font_bold, c, size).x
+        rl.draw_text_ex(self._font_bold, c, rl.Vector2(x + (w - glyph) / 2, y), size, 0, rl.WHITE)
+        x += w
+
   def _draw_steering_wheel(self, rect: rl.Rectangle) -> None:
     wheel_txt = self._txt_wheel_critical if self._show_wheel_critical else self._txt_wheel
 

openpilot/selfdrive/ui/ui_state.py

@@ -86,6 +86,7 @@ class UIState(UIStateSP):
     self.usbgpu_compiled: bool = usbgpu_compiled()
     self.usbgpu_active: bool | None = self.params.get("UsbGpuActive")
     self.usbgpu_loading: bool = self.params.get_bool("UsbGpuLoading")
+    self.usbgpu_load_progress: int = self.params.get("UsbGpuLoadProgress", return_default=True)
     self.started: bool = False
     self.ignition: bool = False
     self.recording_audio: bool = False
@@ -164,6 +165,9 @@ class UIState(UIStateSP):
     # Update started state
     self.started = self.sm["deviceState"].started and self.ignition
 
+    if self.usbgpu_loading:
+      self.usbgpu_load_progress = self.params.get("UsbGpuLoadProgress", return_default=True)
+
     # Update body state
     if self.CP is not None and self.is_body != self.CP.notCar:
       self.is_body = self.CP.notCar

openpilot/sunnypilot/modeld_v2/compile_modeld.py

@@ -7,7 +7,6 @@ See the LICENSE.md file in the root directory for more details.
 """
 
 import argparse
-import math
 import os
 import tempfile
 import time
@@ -67,15 +66,14 @@ def get_policy_npy_shapes(input_shapes: dict, is_supercombo: bool = False) -> tu
   if desire_key:
     shapes['desire'] = (input_shapes[desire_key][2],)
 
+  if is_supercombo and 'features_buffer' in input_shapes:
+    fb = input_shapes['features_buffer']
+    shapes['prev_feat'] = (fb[0], fb[2])
+
   for key, shape in input_shapes.items():
     if key not in (desire_key, 'features_buffer') and 'img' not in key:
       shapes[key] = tuple(shape)
 
-  if is_supercombo and 'features_buffer' in input_shapes:
-    fb = input_shapes['features_buffer']
-    feat_dim = math.prod(fb[2:])
-    shapes['prev_feat'] = (fb[0], feat_dim)
-
   sizes = [int(np.prod(size)) for size in shapes.values()]
   return shapes, sizes
 
@@ -119,9 +117,8 @@ def generate_queues_and_npy(input_shapes: dict, frame_skip: int, device: str = D
   }
 
   if features_buffer:
-    feat_dim = math.prod(features_buffer[2:])
     feat_q_len = frame_skip * features_buffer[1] if is_supercombo else frame_skip * (features_buffer[1] - 1) + 1
-    queues['feat_q'] = Tensor(np.zeros((feat_q_len, features_buffer[0], feat_dim),
+    queues['feat_q'] = Tensor(np.zeros((feat_q_len, features_buffer[0], features_buffer[2]),
                        dtype=np.float32), device=device).contiguous().realize()
 
   queues.update({key: Tensor(value, device='NPY').realize() for key, value in npy_arrays.items() if key in ('tfm', 'big_tfm')})
@@ -186,14 +183,14 @@ def make_run_policy(vision_runner, policy_runners: list, features_slice: slice,
     warped_dev = warped.to(Device.DEFAULT)
     Tensor.realize(packed_npy_inputs_dev, warped_dev)
 
-    img = shift_and_sample(img_q, warped_dev[0:1], sample_skip_fn)
+    img = shift_and_sample(img_q, warped_dev[0:1], sample_skip_fn).realize()
-    big_img = shift_and_sample(big_img_q, warped_dev[1:2], sample_skip_fn)
+    big_img = shift_and_sample(big_img_q, warped_dev[1:2], sample_skip_fn).realize()
 
     unpacked_tensors = [tensor.reshape(shape) for tensor, shape in zip(packed_npy_inputs_dev.split(npy_sizes), npy_shapes.values(), strict=True)]
     unpacked_dict = dict(zip(npy_shapes.keys(), unpacked_tensors, strict=True))
 
     desire_dev = unpacked_dict['desire']
-    desire_buf = shift_and_sample(desire_q, desire_dev.reshape(1, 1, -1), sample_desire_fn)
+    desire_buf = shift_and_sample(desire_q, desire_dev.reshape(1, 1, -1), sample_desire_fn).realize()
 
     inputs = {desire_key: desire_buf}
     for key, tensor_val in unpacked_dict.items():
@@ -202,22 +199,19 @@ def make_run_policy(vision_runner, policy_runners: list, features_slice: slice,
 
     if 'prev_feat' in unpacked_dict:
       prev_feat_dev = unpacked_dict['prev_feat']
-      feat_buf = shift_and_sample(feat_q, prev_feat_dev.reshape(1, 1, -1), sample_skip_fn)
+      inputs['features_buffer'] = shift_and_sample(feat_q, prev_feat_dev.reshape(1, 1, -1), sample_skip_fn).realize()
-      inputs['features_buffer'] = feat_buf if len(fb := input_shapes['features_buffer']) <= 3 else feat_buf.reshape(fb)
 
     if vision_runner:
       vision_out_cast = next(iter(vision_runner({road_key: img, wide_key: big_img}).values())).cast('float32').realize()
       if 'features_buffer' not in inputs:
         new_feat = vision_out_cast[:, features_slice].reshape(1, -1).unsqueeze(0)
-        feat_buf = shift_and_sample(feat_q, new_feat, sample_skip_fn).realize()
+        inputs['features_buffer'] = shift_and_sample(feat_q, new_feat, sample_skip_fn).realize()
-        inputs['features_buffer'] = feat_buf if len(fb := input_shapes['features_buffer']) <= 3 else feat_buf.reshape(fb)
       policy_outs = [next(iter(pol_runner(inputs).values())).cast('float32').realize() for pol_runner in policy_runners]
       return (vision_out_cast, *policy_outs) if len(policy_outs) > 1 else (vision_out_cast, policy_outs[0])
 
     inputs.update({road_key: img, wide_key: big_img})
     if 'features_buffer' not in inputs:
-      feat_buf = sample_skip_fn(feat_q)
+      inputs['features_buffer'] = sample_skip_fn(feat_q)
-      inputs['features_buffer'] = feat_buf if len(fb := input_shapes['features_buffer']) <= 3 else feat_buf.reshape(fb)
 
     policy_out = next(iter(policy_runners[0](inputs).values())).cast('float32').realize()
     if 'features_buffer' not in inputs and features_slice is not None:

openpilot/sunnypilot/modeld_v2/modeld.py

@@ -25,6 +25,7 @@ from opendbc.car.car_helpers import get_demo_car_params
 from tinygrad.tensor import Tensor
 
 from openpilot.common.file_chunker import open_file_chunked
+from openpilot.selfdrive.modeld.load_progress import open_with_progress
 from openpilot.common.swaglog import cloudlog
 from openpilot.common.params import Params
 from openpilot.common.filter_simple import FirstOrderFilter
@@ -107,7 +108,7 @@ class ModelState(ModelStateBase):
 
   def _init_combined(self, pkl_path, cam_w, cam_h, bundle):
     cloudlog.warning(f"loading combined pkl: {pkl_path}")
-    jits = load_oob(open_file_chunked(pkl_path))
+    jits = load_oob(open_with_progress(pkl_path) if self.usbgpu else open_file_chunked(pkl_path))
 
     self.WARP_DEV = 'QCOM' if COMMA_HARDWARE else 'CPU'
     self.DEV = 'AMD' if self.usbgpu else self.WARP_DEV

openpilot/sunnypilot/modeld_v2/tests/test_compile_modeld.py

@@ -195,85 +195,3 @@ class TestReadFileChunkedToDisk(OpenpilotTestCase):
 
       assert out.parent == Path(d)
       assert out.read_bytes() == payload
-
-
-class Test4DFeaturesBuffer(OpenpilotTestCase):
-  def test_get_policy_npy_shapes_4d(self):
-    from openpilot.sunnypilot.modeld_v2.compile_modeld import get_policy_npy_shapes
-    input_shapes = {
-      'desire_pulse': (1, 25, 8),
-      'features_buffer': (1, 24, 32, 512),  # compare 4d to 3d for regression
-      'traffic_convention': (1, 2),
-      'action_t': (1, 2)
-    }
-    shapes, sizes = get_policy_npy_shapes(input_shapes, is_supercombo=True)
-    assert shapes['prev_feat'] == (1, 16384)
-    assert sizes == [8, 2, 2, 16384]
-
-  def test_get_policy_npy_shapes_3d(self):
-    from openpilot.sunnypilot.modeld_v2.compile_modeld import get_policy_npy_shapes
-    input_shapes = {
-      'desire_pulse': (1, 25, 8),
-      'features_buffer': (1, 24, 512),
-      'traffic_convention': (1, 2),
-      'action_t': (1, 2)
-    }
-    shapes, sizes = get_policy_npy_shapes(input_shapes, is_supercombo=True)
-    assert shapes['prev_feat'] == (1, 512)
-    assert sizes == [8, 2, 2, 512]
-
-
-class TestStockCompileModeldEquivalence(OpenpilotTestCase):
-  def test_get_policy_npy_shapes_matches_stock(self):
-    from openpilot.selfdrive.modeld.compile_modeld import get_policy_npy_shapes as stock_get_policy_npy_shapes
-    from openpilot.sunnypilot.modeld_v2.compile_modeld import get_policy_npy_shapes as sunny_get_policy_npy_shapes
-
-    stock_input_shapes = {
-      'desire_pulse': (1, 25, 8),
-      'features_buffer': (1, 24, 512), # see below comment
-      'traffic_convention': (1, 2),
-      'action_t': (1, 2),
-    }
-
-    stock_shapes, stock_sizes = stock_get_policy_npy_shapes(stock_input_shapes)
-    sunny_shapes, sunny_sizes = sunny_get_policy_npy_shapes(stock_input_shapes, is_supercombo=True)
-
-    assert sunny_shapes == stock_shapes
-    assert sunny_sizes == stock_sizes
-    assert sunny_shapes['prev_feat'] == (1, 512)
-
-  def test_make_input_queues_full_stock_equivalence(self):
-    from openpilot.selfdrive.modeld.compile_modeld import make_input_queues as stock_make_input_queues
-    from openpilot.sunnypilot.modeld_v2.compile_modeld import make_supercombo_input_queues as sunny_make_supercombo_input_queues
-    input_shapes = {
-      'img': (1, 12, 128, 256),
-      'desire_pulse': (1, 25, 8),
-      'features_buffer': (1, 24, 512),  # when https://github.com/commaai/openpilot/pull/38681 merges, update to 1,24,32,512
-      'traffic_convention': (1, 2),
-      'action_t': (1, 2),
-    }
-    frame_skip = 4
-
-    stock_queues, stock_npy = stock_make_input_queues(input_shapes, frame_skip, device='NPY')
-    sunny_queues, sunny_npy = sunny_make_supercombo_input_queues(input_shapes, frame_skip, device='NPY')
-    assert set(sunny_queues.keys()) == set(stock_queues.keys())
-    for key in stock_queues:
-      assert sunny_queues[key].shape == stock_queues[key].shape, \
-        f"Queue shape mismatch for {key}: sunny {sunny_queues[key].shape} != stock {stock_queues[key].shape}"
-    assert set(sunny_npy.keys()) == set(stock_npy.keys())
-    for key in stock_npy:
-      assert sunny_npy[key].shape == stock_npy[key].shape, \
-        f"Numpy array shape mismatch for {key}: sunny {sunny_npy[key].shape} != stock {stock_npy[key].shape}"
-
-  def test_make_warp_queues_stock_equivalence(self):
-    from openpilot.selfdrive.modeld.compile_modeld import make_warp_input_queues as stock_make_warp_queues
-    from openpilot.sunnypilot.modeld_v2.compile_modeld import make_warp_queues as sunny_make_warp_queues
-    stock_vision_shapes = {'img': (1, 12, 128, 256)}  # for now?
-    stock_queues, stock_npy = stock_make_warp_queues(stock_vision_shapes, frame_skip=4, device='NPY')
-    sunny_queues, sunny_npy = sunny_make_warp_queues(device='NPY')
-
-    assert set(sunny_npy.keys()) == set(stock_npy.keys()) == {'tfm', 'big_tfm'}
-    for key in sunny_npy:
-      assert sunny_npy[key].shape == stock_npy[key].shape == (3, 3)
-
-

openpilot/sunnypilot/models/fetcher.py

@@ -141,7 +141,7 @@ class ModelCache:
 class ModelFetcher:
   """Handles fetching and caching of model data from remote source"""
   MODEL_URL = "https://raw.githubusercontent.com/sunnypilot/sunnypilot-models/refs/heads/gh-pages/docs/driving_models_v20.json"
-  MODEL_URL_USBGPU = "https://raw.githubusercontent.com/sunnypilot/sunnypilot-models/refs/heads/gh-pages/docs/driving_models_usbgpu_v22.json"
+  MODEL_URL_USBGPU = "https://raw.githubusercontent.com/sunnypilot/sunnypilot-models/refs/heads/gh-pages/docs/driving_models_usbgpu_v21.json"
 
   def __init__(self, params: Params):
     self.params = params