letify 2.4.5__tar.gz → 2.4.6__tar.gz

{letify-2.4.5 → letify-2.4.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: letify
-Version: 2.4.5
+Version: 2.4.6
 Summary: Declarations that become infrastructure.
 Project-URL: Homepage, https://github.com/DarkPyonix/letify
 Project-URL: Documentation, https://github.com/DarkPyonix/letify/tree/main/docs

{letify-2.4.5 → letify-2.4.6}/letify/__init__.py

@@ -64,7 +64,7 @@ distributed: Final = Host.distributed
 # The enum class stays importable from letify.declare.instance for letify's own use.
 del Host
 
-__version__ = "2.4.5"
+__version__ = "2.4.6"
 
 
 def fetch(tensor):  # type: ignore[no-untyped-def]

{letify-2.4.5 → letify-2.4.6}/letify/runtime/bootstrap.py

@@ -146,9 +146,20 @@ CARD_SOURCE = (
     "        pass\n"
     # /proc can list the right cards on a session whose driver cannot be used, and the
     # first .to('cuda') then raises minutes into the call. Spec "Kaggle".
+    # cuInit is the call CUDA initialization makes, so it answers whether the driver
+    # works rather than standing in for it: a live session had the cards and the control
+    # device and still could not start CUDA. ctypes needs no torch, which the session does
+    # not have yet at this point. Spec "Kaggle".
+    "__letify_cuda = -1\n"
+    "try:\n"
+    "    import ctypes as _letify_ctypes\n"
+    "    __letify_cuda = _letify_ctypes.CDLL('libcuda.so.1').cuInit(0)\n"
+    "except Exception:\n"
+    "    pass\n"
     "__letify_value__ = {\n"
     "    'cards': __letify_cards,\n"
     "    'driver': os.path.exists('/dev/nvidiactl'),\n"
+    "    'cuinit': __letify_cuda,\n"
     "}\n"
 )
 

{letify-2.4.5 → letify-2.4.6}/letify/runtime/session.py

@@ -720,11 +720,14 @@ class Runtime:
         except Exception:
             # Reading /proc is not worth failing a session over when it cannot be read.
             return
-        found = found if isinstance(found, dict) else {"cards": found, "driver": True}
+        if not isinstance(found, dict):
+            found = {"cards": found, "driver": True, "cuinit": 0}
         cards = [str(card) for card in found.get("cards") or ()]
         driver = bool(found.get("driver"))
+        cuinit = found.get("cuinit", 0)
+        started = cuinit == 0
         matched = [card for card in cards if fragment.casefold() in card.casefold()]
-        if len(matched) == count and len(cards) == count and driver:
+        if len(matched) == count and len(cards) == count and driver and started:
             return
         asked = getattr(instance, "accelerator", None) or fragment
         arrived = ", ".join(cards) if cards else "no GPU at all"
@@ -735,6 +738,12 @@ class Runtime:
                 f"{arrived}, and no usable driver: /dev/nvidiactl is not there, so CUDA "
                 f"cannot start"
             )
+        elif cards and not started:
+            arrived = (
+                f"{arrived}, and a driver that does not start: cuInit returned {cuinit}, "
+                f"which is the call CUDA initialization makes, so nothing on this session "
+                f"can use those cards"
+            )
         message = (
             f"{self.name}: {self.provider.alias} was asked for {count} x {asked} and the "
             f"session has {arrived}. A declaration says where a function runs, so the "

{letify-2.4.5 → letify-2.4.6}/letify-core/Cargo.lock

@@ -138,7 +138,7 @@ dependencies = [
 
 [[package]]
 name = "letify-agent"
-version = "2.4.5"
+version = "2.4.6"
 dependencies = [
  "letify-wire",
  "libloading",
@@ -146,7 +146,7 @@ dependencies = [
 
 [[package]]
 name = "letify-driver"
-version = "2.4.5"
+version = "2.4.6"
 dependencies = [
  "letify-wire",
 ]
@@ -164,7 +164,7 @@ dependencies = [
 
 [[package]]
 name = "letify-wire"
-version = "2.4.5"
+version = "2.4.6"
 
 [[package]]
 name = "libc"

{letify-2.4.5 → letify-2.4.6}/letify-core/Cargo.toml

@@ -8,7 +8,7 @@ resolver = "3"
 members = ["letify-wire", "letify-driver", "letify-agent", "letify-quic"]
 
 [workspace.package]
-version = "2.4.5"
+version = "2.4.6"
 edition = "2024"
 license = "MIT"
 repository = "https://github.com/DarkPyonix/letify"

{letify-2.4.5 → letify-2.4.6}/letify-ext/package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "letify-status",
-  "version": "2.4.5",
+  "version": "2.4.6",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "letify-status",
-      "version": "2.4.5",
+      "version": "2.4.6",
       "license": "SEE LICENSE IN LICENSE",
       "devDependencies": {
         "@types/node": "^18.19.0",

{letify-2.4.5 → letify-2.4.6}/letify-ext/package.json

@@ -2,7 +2,7 @@
   "name": "letify-status",
   "displayName": "letify Status",
   "description": "letify quota and GPU activity in the status bar, read from the letify CLI.",
-  "version": "2.4.5",
+  "version": "2.4.6",
   "publisher": "darkpyonix",
   "license": "SEE LICENSE IN LICENSE",
   "repository": {

{letify-2.4.5 → letify-2.4.6}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "letify"
-version = "2.4.5"
+version = "2.4.6"
 description = "Declarations that become infrastructure."
 readme = "README.md"
 requires-python = ">=3.11"

{letify-2.4.5 → letify-2.4.6}/tests/test_kaggle.py

@@ -1496,3 +1496,48 @@ def test_a_session_with_the_cards_and_a_driver_passes() -> None:
         session.check_devices()
     finally:
         Runtime.eval = original
+
+
+def test_a_session_whose_cuinit_fails_is_refused() -> None:
+    """Spec "Kaggle": cuInit is the call CUDA initialization makes.
+
+    A live session listed both T4s and had /dev/nvidiactl, and the first .cuda() still
+    raised "Found no NVIDIA driver on your system" after 30 GB of shards had been sent.
+    """
+    from letify.errors import InsufficientDevices
+    from letify.runtime.session import Runtime
+
+    provider = kaggle_provider()
+    session = object.__new__(Runtime)
+    session.name = "letify-t4-fc7e27"
+    session.provider = provider
+    session.instance = type("I", (), {"gpu": "T4", "accelerator": "T4"})()
+    answer = {"cards": ["Tesla T4", "Tesla T4"], "driver": True, "cuinit": 100}
+    monkey = lambda self, source, timeout=None: answer  # noqa: E731
+    Runtime.eval, original = monkey, Runtime.eval
+    try:
+        with pytest.raises(InsufficientDevices) as failure:
+            session.check_devices()
+    finally:
+        Runtime.eval = original
+    message = str(failure.value)
+    assert "cuInit" in message
+    assert "100" in message
+
+
+def test_a_session_whose_cuinit_succeeds_passes() -> None:
+    """Spec: zero from cuInit is what CUDA starting looks like."""
+    from letify.runtime.session import Runtime
+
+    provider = kaggle_provider()
+    session = object.__new__(Runtime)
+    session.name = "letify-t4-ok"
+    session.provider = provider
+    session.instance = type("I", (), {"gpu": "T4", "accelerator": "T4"})()
+    answer = {"cards": ["Tesla T4", "Tesla T4"], "driver": True, "cuinit": 0}
+    monkey = lambda self, source, timeout=None: answer  # noqa: E731
+    Runtime.eval, original = monkey, Runtime.eval
+    try:
+        session.check_devices()
+    finally:
+        Runtime.eval = original