autodetect tg backend (#37778)

* pick fastest

* save config

* fix

* Ignore generated tg_compiled_flags file

* helper

* cleaner

* whitespace not needed

* no shebang

* whitespace

.gitignore

@@ -44,6 +44,7 @@ bin/
 config.json
 compile_commands.json
 compare_runtime*.html
+selfdrive/modeld/models/tg_compiled_flags.json
 
 # build artifacts
 selfdrive/pandad/pandad

selfdrive/modeld/SConscript

@@ -1,6 +1,9 @@
-import os
 import glob
+import json
+import os
+from SCons.Script import Value
 from openpilot.common.file_chunker import chunk_file, get_chunk_paths
+from tinygrad import Device
 
 Import('env', 'arch')
 chunker_file = File("#common/file_chunker.py")
@@ -13,31 +16,51 @@ tinygrad_files = ["#"+x for x in glob.glob(env.Dir("#tinygrad_repo").relpath + "
 def estimate_pickle_max_size(onnx_size):
   return 1.2 * onnx_size  + 10 * 1024 * 1024  # 20% + 10MB is plenty
 
-# compile warp
 # THREADS=0 is need to prevent bug: https://github.com/tinygrad/tinygrad/issues/14689
-tg_flags = {
-    'larch64': 'DEV=QCOM FLOAT16=1 NOLOCALS=1 JIT_BATCH_SIZE=0',
-    'Darwin': f'DEV=CPU THREADS=0 HOME={os.path.expanduser("~")}', # tinygrad calls brew which needs a $HOME in the env
-}.get(arch, 'DEV=CPU:LLVM THREADS=0')
+# get fastest TG config
+available = set(Device.get_available_devices())
+if 'CUDA' in available:
+  tg_backend = 'CUDA'
+  tg_flags = f'DEV={tg_backend}'
+elif 'QCOM' in available:
+  tg_backend = 'QCOM'
+  tg_flags = f'DEV={tg_backend} FLOAT16=1 NOLOCALS=1 JIT_BATCH_SIZE=0'
+else:
+  tg_backend = 'CPU'
+  tg_flags = f'DEV={tg_backend} THREADS=0'
+
+def write_tg_compiled_flags(target, source, env):
+  with open(str(target[0]), "w") as f:
+    json.dump({"DEV": tg_backend}, f)
+    f.write("\n")
+
+compiled_flags_node = lenv.Command(
+  File("models/tg_compiled_flags.json").abspath,
+  tinygrad_files + [Value(tg_backend)],
+  write_tg_compiled_flags,
+)
+
+# tinygrad calls brew which needs a $HOME in the env
+mac_brew_string = f'HOME={os.path.expanduser("~")}' if arch == 'Darwin' else ''
 
 # Get model metadata
 for model_name in ['driving_vision', 'driving_off_policy', 'driving_on_policy', 'dmonitoring_model']:
   fn = File(f"models/{model_name}").abspath
   script_files = [File(Dir("#selfdrive/modeld").File("get_model_metadata.py").abspath)]
-  cmd = f'{tg_flags} python3 {Dir("#selfdrive/modeld").abspath}/get_model_metadata.py {fn}.onnx'
-  lenv.Command(fn + "_metadata.pkl", [fn + ".onnx"] + tinygrad_files + script_files, cmd)
+  cmd = f'{tg_flags} {mac_brew_string} python3 {Dir("#selfdrive/modeld").abspath}/get_model_metadata.py {fn}.onnx'
+  lenv.Command(fn + "_metadata.pkl", [fn + ".onnx"] + tinygrad_files + script_files + [compiled_flags_node], cmd)
 
 image_flag = {
      'larch64': 'IMAGE=2',
 }.get(arch, 'IMAGE=0')
 script_files = [File(Dir("#selfdrive/modeld").File("compile_warp.py").abspath)]
-compile_warp_cmd = f'{tg_flags} python3 {Dir("#selfdrive/modeld").abspath}/compile_warp.py '
+compile_warp_cmd = f'{tg_flags} {mac_brew_string} python3 {Dir("#selfdrive/modeld").abspath}/compile_warp.py '
 from openpilot.common.transformations.camera import _ar_ox_fisheye, _os_fisheye
 warp_targets = []
 for cam in [_ar_ox_fisheye, _os_fisheye]:
   w, h = cam.width, cam.height
   warp_targets += [File(f"models/warp_{w}x{h}_tinygrad.pkl").abspath, File(f"models/dm_warp_{w}x{h}_tinygrad.pkl").abspath]
-lenv.Command(warp_targets, tinygrad_files + script_files, compile_warp_cmd)
+lenv.Command(warp_targets, tinygrad_files + script_files + [compiled_flags_node], compile_warp_cmd)
 
 def tg_compile(flags, model_name):
   pythonpath_string = 'PYTHONPATH="${PYTHONPATH}:' + env.Dir("#tinygrad_repo").abspath + '"'
@@ -47,7 +70,7 @@ def tg_compile(flags, model_name):
   chunk_targets = get_chunk_paths(pkl, estimate_pickle_max_size(os.path.getsize(onnx_path)))
   compile_node = lenv.Command(
     pkl,
-    [onnx_path] + tinygrad_files + [chunker_file],
+    [onnx_path] + tinygrad_files + [chunker_file, compiled_flags_node],
     f'{pythonpath_string} {flags} {image_flag} python3 {Dir("#tinygrad_repo").abspath}/examples/openpilot/compile3.py {fn}.onnx {pkl}',
   )
   def do_chunk(target, source, env):

selfdrive/modeld/dmonitoringmodeld.py

@@ -1,12 +1,12 @@
 #!/usr/bin/env python3
 import os
-from openpilot.system.hardware import TICI
-os.environ['DEV'] = 'QCOM' if TICI else 'CPU'
+from openpilot.selfdrive.modeld.tinygrad_helpers import MODELS_DIR, set_tinygrad_backend_from_compiled_flags
+set_tinygrad_backend_from_compiled_flags()
+
 from tinygrad.tensor import Tensor
 import time
 import pickle
 import numpy as np
-from pathlib import Path
 
 from cereal import messaging
 from cereal.messaging import PubMaster, SubMaster
@@ -21,9 +21,8 @@ from openpilot.selfdrive.modeld.parse_model_outputs import sigmoid, safe_exp
 
 PROCESS_NAME = "selfdrive.modeld.dmonitoringmodeld"
 SEND_RAW_PRED = os.getenv('SEND_RAW_PRED')
-MODEL_PKL_PATH = Path(__file__).parent / 'models/dmonitoring_model_tinygrad.pkl'
-METADATA_PATH = Path(__file__).parent / 'models/dmonitoring_model_metadata.pkl'
-MODELS_DIR = Path(__file__).parent / 'models'
+MODEL_PKL_PATH = MODELS_DIR / 'dmonitoring_model_tinygrad.pkl'
+METADATA_PATH = MODELS_DIR / 'dmonitoring_model_metadata.pkl'
 
 class ModelState:
   inputs: dict[str, np.ndarray]

selfdrive/modeld/modeld.py

@@ -1,7 +1,8 @@
 #!/usr/bin/env python3
 import os
-from openpilot.system.hardware import TICI
-os.environ['DEV'] = 'QCOM' if TICI else 'CPU'
+from openpilot.selfdrive.modeld.tinygrad_helpers import MODELS_DIR, set_tinygrad_backend_from_compiled_flags
+set_tinygrad_backend_from_compiled_flags()
+
 USBGPU = "USBGPU" in os.environ
 if USBGPU:
   os.environ['DEV'] = 'AMD'
@@ -12,7 +13,6 @@ import pickle
 import numpy as np
 import cereal.messaging as messaging
 from cereal import car, log
-from pathlib import Path
 from cereal.messaging import PubMaster, SubMaster
 from msgq.visionipc import VisionIpcClient, VisionStreamType, VisionBuf
 from opendbc.car.car_helpers import get_demo_car_params
@@ -34,7 +34,6 @@ from openpilot.selfdrive.modeld.constants import ModelConstants, Plan
 PROCESS_NAME = "selfdrive.modeld.modeld"
 SEND_RAW_PRED = os.getenv('SEND_RAW_PRED')
 
-MODELS_DIR = Path(__file__).parent / 'models'
 VISION_PKL_PATH = MODELS_DIR / 'driving_vision_tinygrad.pkl'
 VISION_METADATA_PATH = MODELS_DIR / 'driving_vision_metadata.pkl'
 ON_POLICY_PKL_PATH = MODELS_DIR / 'driving_on_policy_tinygrad.pkl'

selfdrive/modeld/tinygrad_helpers.py

@@ -0,0 +1,11 @@
+import json
+import os
+from pathlib import Path
+
+MODELS_DIR = Path(__file__).parent / 'models'
+COMPILED_FLAGS_PATH = MODELS_DIR / 'tg_compiled_flags.json'
+
+
+def set_tinygrad_backend_from_compiled_flags() -> None:
+  with open(COMPILED_FLAGS_PATH) as f:
+    os.environ['DEV'] = str(json.load(f)['DEV'])