Publish Tiny Hinglish Turn Detector development preview

release inventory sha256: fa73d6c3defbd256e431aa99c8b4fae4ea6e74cbb2fb736f57eb43fbbe65f41a

MODEL_CARD.md

@@ -303,7 +303,7 @@ failure aggregates, silence perturbation report, exact environment files, CPU
 benchmarks, and guarded release tooling. Per-example validation predictions and
 the text-bearing split manifest remain in the private local workspace; the
 public package contains aggregate reports and hashes, not those sensitive rows.
-The quality gate completed with Ruff 0.15.16 passing and 127/127 automated tests
+The quality gate completed with Ruff 0.15.16 passing and 130/130 automated tests
 passing via Python `unittest` discovery. The configured Hugging Face owner is
 `suvradeepp`.
 

README.md

@@ -303,7 +303,7 @@ failure aggregates, silence perturbation report, exact environment files, CPU
 benchmarks, and guarded release tooling. Per-example validation predictions and
 the text-bearing split manifest remain in the private local workspace; the
 public package contains aggregate reports and hashes, not those sensitive rows.
-The quality gate completed with Ruff 0.15.16 passing and 127/127 automated tests
+The quality gate completed with Ruff 0.15.16 passing and 130/130 automated tests
 passing via Python `unittest` discovery. The configured Hugging Face owner is
 `suvradeepp`.
 

REPORT.md

@@ -525,8 +525,8 @@ calendar days or the missing external data collection occurred.
 | 13 | Integrate controller and demo | replay, three-state controller, Gradio, Hinglish recording prompts | Preview model and synthetic controller integration complete; real sequence replay pending |
 | 14 | Freeze test, package, publish | sealed-test guard, atomic release builder, exact-mirror publisher | Development packaging path ready; official-test/final gates not satisfied |
 
-The repository quality gate passed Ruff 0.15.16 and all 127 automated tests
-(127 passed via Python `unittest` discovery). That validates the implemented behavior under test; it does not fill
+The repository quality gate passed Ruff 0.15.16 and all 130 automated tests
+(130 passed via Python `unittest` discovery). That validates the implemented behavior under test; it does not fill
 the missing empirical datasets or rights review.
 
 ## 9. Decision log and next experiments

SOURCE_README.md

@@ -26,7 +26,7 @@ engineering evidence, not a production-readiness claim.
 | Domain-shift stress | Measured, confounded | Source-held-out AP 0.5500/0.5368 for TinyTCN/baseline; both near chance in ranking, with wide nine-source bootstrap intervals |
 | Runtime/export | Measured | 611 KiB FP32 ONNX; exact PyTorch/ONNX parity; ONNX p95 0.956 ms neural-only and 1.629 ms waveform-to-probability |
 | Controller integration | Synthetic only | Metadata-bound 8-checkpoint/3-turn fixture emits 3 response edges with 0 duplicates; not model or product-quality evidence |
-| Code quality gate | Passed | Ruff 0.15.16; 127/127 automated tests passed via `unittest` discovery |
+| Code quality gate | Passed | Ruff 0.15.16; 130/130 automated tests passed via `unittest` discovery |
 | Accuracy/generalization | **Not established** | No full-corpus run, collected Hinglish recordings, real conversation replay, or official-test evaluation |
 | Hugging Face publication | Not claimed here | Local guarded release/publisher path exists; verify any remote publication separately |
 
@@ -321,6 +321,13 @@ uv run python scripts/publish_hf.py --username suvradeepp
 uv run python scripts/publish_hf.py --username suvradeepp --execute
 ```
 
+For a personal account without Hugging Face PRO, the publisher requests the
+free-account-compatible ZeroGPU tier (`--space-hardware zero-a10g`) only when it
+must create a new Space. Hugging Face currently gates new personal Gradio CPU
+Spaces behind PRO even when CPU Basic has no hourly charge. ZeroGPU eligibility
+still requires a verified account in good standing and an available hosted
+Space slot. Paid accounts may override this with `--space-hardware cpu-basic`.
+
 The release builder refuses smoke checkpoints or a release without measured
 test metrics unless `--allow-development-artifact` is passed explicitly. That
 override must remain visibly development-only. For a reviewable partial-data

export_manifest.json

@@ -171,9 +171,9 @@
       "sha256": "89167b269b2103ad0bd94918f2ec4f7b605cf001d391e570683ceef39066b6c2"
     },
     {
-      "bytes": 42576,
+      "bytes": 43530,
       "path": "scripts/publish_hf.py",
-      "sha256": "97b861d22e6fa0f6b6f998ca1026151f9720cfe3607670b3c5735120a5312760"
+      "sha256": "36311a2457bb0e93cbd6fd0ed444e57a2ec2c44ab9bd17ed06132e9640a9ddc3"
     },
     {
       "bytes": 3752,
@@ -371,7 +371,7 @@
       "sha256": "da0eac829fe43615f7bb149998c2f2762f148790fb0c14d4ba6db4d6b25298f4"
     }
   ],
-  "source_inventory_sha256": "786e0982f03db47ca1ce264cea005f6d379fce76f72870bfe5c95b76502877eb",
+  "source_inventory_sha256": "259931c509dc4abd54a3c5c72ca12948449bd19d3aed2b2c2ea0fac88a6a705b",
   "task": "audio-turn-end-detection",
   "threshold": 0.741000771522522,
   "training_data": {

release_provenance.json

@@ -16,7 +16,7 @@
       },
       "MODEL_CARD.md": {
         "bytes": 14805,
-        "sha256": "bf5ca1a9f856ebf39de50094b784936cb0751d103bf585dd9523f100837d336e"
+        "sha256": "6add0853cdc0f09ff1abb594149c59e9d53f29ab4dba5de52bed7278cfb13c1b"
       },
       "Makefile": {
         "bytes": 777,
@@ -28,15 +28,15 @@
       },
       "README.md": {
         "bytes": 14805,
-        "sha256": "bf5ca1a9f856ebf39de50094b784936cb0751d103bf585dd9523f100837d336e"
+        "sha256": "6add0853cdc0f09ff1abb594149c59e9d53f29ab4dba5de52bed7278cfb13c1b"
       },
       "REPORT.md": {
         "bytes": 32272,
-        "sha256": "fd5ec03ec01af960c043d5828ab5d1204eed2ced1c6f74223f1222b2f6721d82"
+        "sha256": "9205011a4c547b8cf537d5b3538d25f1b8a9545c9bb020694f6d94deb26683f7"
       },
       "SOURCE_README.md": {
-        "bytes": 20065,
-        "sha256": "ca0f08b2c18fc5b3c16b10c36922add96ef24048dea67ddfb120cf5bc51f5c49"
+        "bytes": 20535,
+        "sha256": "64539b78efae390a42ca19a718b8e7354327826fff81dc6c1a9c7841b11132e0"
       },
       "app.py": {
         "bytes": 10381,
@@ -108,7 +108,7 @@
       },
       "export_manifest.json": {
         "bytes": 12430,
-        "sha256": "48a7c3c711c23ad66e9df35c88153544a2107be41b0c9c6613f2b7fbdb0d235d"
+        "sha256": "1f70d8ec79c9c2430c9d3a16942f51254202279c984048e02a781a33313f242b"
       },
       "model.onnx": {
         "bytes": 625431,
@@ -283,8 +283,8 @@
         "sha256": "89167b269b2103ad0bd94918f2ec4f7b605cf001d391e570683ceef39066b6c2"
       },
       "scripts/publish_hf.py": {
-        "bytes": 42576,
-        "sha256": "97b861d22e6fa0f6b6f998ca1026151f9720cfe3607670b3c5735120a5312760"
+        "bytes": 43530,
+        "sha256": "36311a2457bb0e93cbd6fd0ed444e57a2ec2c44ab9bd17ed06132e9640a9ddc3"
       },
       "scripts/replay_stream.py": {
         "bytes": 3752,
@@ -527,8 +527,8 @@
         "sha256": "12a9a0edda07265de293076d12073a707b10e6fb9ebbf2e7c70ba9a219289bdc"
       },
       "tests/test_publish_hf.py": {
-        "bytes": 22529,
-        "sha256": "91f825c0ad9c5a6a25079ff4d956c27564b254f4db46ddb494770425d20ebd2f"
+        "bytes": 25161,
+        "sha256": "5af81ad7995b6bfd3816b7d97ee1226027659dc5b8a7cb4fe4376fa850dcd14e"
       },
       "tests/test_release_guardrails.py": {
         "bytes": 17399,
@@ -566,7 +566,7 @@
       },
       "MODEL_CARD.md": {
         "bytes": 14805,
-        "sha256": "bf5ca1a9f856ebf39de50094b784936cb0751d103bf585dd9523f100837d336e"
+        "sha256": "6add0853cdc0f09ff1abb594149c59e9d53f29ab4dba5de52bed7278cfb13c1b"
       },
       "NOTICE": {
         "bytes": 563,
@@ -626,5 +626,5 @@
   "metrics_file": "development_metrics.json",
   "metrics_scope": "development",
   "model_file": "model.onnx",
-  "source_and_artifact_inventory_sha256": "5d8ac3bec5bdc21d874024019012fefebe1c6e53cd002bc5a2aa8dd4d9e83ef3"
+  "source_and_artifact_inventory_sha256": "4c327046a4f44513d65a2a920a67566c646bf9c82e1f770a7f1ee6614006a47e"
 }

scripts/publish_hf.py

@@ -306,6 +306,12 @@ def _failure_kind(exc: BaseException) -> str:
     status = _http_status(exc)
     if status in {401, 403}:
         return f"authentication/authorization rejection (HTTP {status})"
+    if status == 402:
+        return (
+            "Hugging Face plan/hardware entitlement required (HTTP 402); "
+            "verify free ZeroGPU eligibility and an available slot for "
+            "--space-hardware zero-a10g, otherwise upgrade the account plan"
+        )
     if status is not None:
         return f"HTTP {status}"
     chain = _exception_chain(exc)
@@ -495,6 +501,7 @@ def _ensure_repo(
     repo_id: str,
     repo_type: str,
     private: bool,
+    space_hardware: str | None,
     policy: RetryPolicy,
 ) -> Any:
     info = _repo_info_or_none(api, repo_id=repo_id, repo_type=repo_type, policy=policy)
@@ -507,6 +514,8 @@ def _ensure_repo(
         }
         if repo_type == "space":
             create_kwargs["space_sdk"] = "gradio"
+            if space_hardware is not None:
+                create_kwargs["space_hardware"] = space_hardware
         _call_with_retry(
             f"create {repo_type} repository",
             lambda: api.create_repo(**create_kwargs),
@@ -604,12 +613,14 @@ def _publish_component(
     release_kind: str,
     inventory_digest: str,
     policy: RetryPolicy,
+    space_hardware: str | None = None,
 ) -> ComponentResult:
     _ensure_repo(
         api,
         repo_id=repo_id,
         repo_type=repo_type,
         private=private,
+        space_hardware=space_hardware if repo_type == "space" else None,
         policy=policy,
     )
     info, initial = _current_verification(
@@ -840,6 +851,10 @@ def _execute_publish(
             "model": f"{args.username}/{args.model_name}",
             "space": f"{args.username}/{args.space_name}",
         },
+        "space_creation": {
+            "sdk": "gradio",
+            "hardware_if_created": args.space_hardware,
+        },
         "status": "starting",
         "updated_at": _utc_now(),
     }
@@ -914,6 +929,7 @@ def _execute_publish(
             release_kind=release_kind,
             inventory_digest=manifest["release_inventory_sha256"],
             policy=policy,
+            space_hardware=args.space_hardware,
         )
         receipt["space"] = space.as_dict()
         receipt["status"] = "space_content_verified"
@@ -1018,6 +1034,14 @@ def parse_args() -> argparse.Namespace:
     parser.add_argument("--username", default="suvradeepp")
     parser.add_argument("--model-name", default="tiny-hinglish-turn-detector")
     parser.add_argument("--space-name", default="tiny-hinglish-turn-detector")
+    parser.add_argument(
+        "--space-hardware",
+        default="zero-a10g",
+        help=(
+            "Hardware requested only when creating a new Space (default: zero-a10g, "
+            "the free-account-compatible ZeroGPU tier)"
+        ),
+    )
     parser.add_argument("--release-dir", default="release")
     parser.add_argument("--receipt", default="reports/hf_publish_receipt.json")
     parser.add_argument("--private", action="store_true")

tests/test_publish_hf.py

@@ -10,6 +10,7 @@ from io import StringIO
 from pathlib import Path
 from types import SimpleNamespace
 from typing import Any
+from unittest.mock import patch
 
 from scripts.publish_hf import (
     PublishFailure,
@@ -24,6 +25,7 @@ from scripts.publish_hf import (
     _publish_component,
     _validate_inventory,
     _wait_for_space,
+    parse_args,
 )
 
 
@@ -37,6 +39,7 @@ class FakeHubApi:
     def __init__(self) -> None:
         self.repos: dict[tuple[str, str], dict[str, Any]] = {}
         self.history: dict[tuple[str, str, str], dict[str, bytes]] = {}
+        self.create_calls: list[dict[str, Any]] = []
         self.upload_calls: list[dict[str, Any]] = []
         self.runtime_stages = ["BUILDING", "RUNNING"]
         self.upload_mode = "normal"
@@ -94,6 +97,7 @@ class FakeHubApi:
         )
 
     def create_repo(self, **kwargs: Any) -> None:
+        self.create_calls.append(dict(kwargs))
         key = (kwargs["repo_type"], kwargs["repo_id"])
         if key not in self.repos:
             self.add_repo(
@@ -183,6 +187,7 @@ class PublishHfTest(unittest.TestCase):
         *,
         repo_type: str = "model",
         replace_existing: bool = False,
+        space_hardware: str | None = None,
     ):
         return _publish_component(
             api,
@@ -197,8 +202,63 @@ class PublishHfTest(unittest.TestCase):
             release_kind="development preview",
             inventory_digest="a" * 64,
             policy=self.policy,
+            space_hardware=space_hardware,
         )
 
+    def test_space_hardware_is_sent_only_when_creating_a_new_space(self) -> None:
+        api = FakeHubApi()
+        with tempfile.TemporaryDirectory() as directory:
+            folder = Path(directory)
+            (folder / "payload.bin").write_bytes(b"payload")
+            self._component(api, folder)
+            self._component(
+                api,
+                folder,
+                repo_type="space",
+                space_hardware="zero-a10g",
+            )
+            self._component(
+                api,
+                folder,
+                repo_type="space",
+                space_hardware="cpu-basic",
+            )
+
+        model_create, space_create = api.create_calls
+        self.assertNotIn("space_hardware", model_create)
+        self.assertEqual(space_create["space_sdk"], "gradio")
+        self.assertEqual(space_create["space_hardware"], "zero-a10g")
+
+    def test_space_hardware_cli_defaults_to_zerogpu(self) -> None:
+        with patch("sys.argv", ["publish_hf.py"]):
+            self.assertEqual(parse_args().space_hardware, "zero-a10g")
+
+    def test_space_create_402_is_actionable_sanitized_and_not_retried(self) -> None:
+        secret = "hf_do_not_print_me"
+
+        class PlanGatedApi(FakeHubApi):
+            def create_repo(self, **kwargs: Any) -> None:
+                self.create_calls.append(dict(kwargs))
+                raise FakeHttpError(402, f"billing failure containing {secret}")
+
+        api = PlanGatedApi()
+        with tempfile.TemporaryDirectory() as directory:
+            folder = Path(directory)
+            (folder / "payload.bin").write_bytes(b"payload")
+            with self.assertRaises(PublishFailure) as caught:
+                self._component(
+                    api,
+                    folder,
+                    repo_type="space",
+                    space_hardware="zero-a10g",
+                )
+
+        message = str(caught.exception)
+        self.assertEqual(len(api.create_calls), 1)
+        self.assertIn("HTTP 402", message)
+        self.assertIn("--space-hardware zero-a10g", message)
+        self.assertNotIn(secret, message)
+
     def test_unmanaged_existing_repo_requires_explicit_replacement(self) -> None:
         api = FakeHubApi()
         api.add_repo("suvradeepp/test-model", "model", {"README.md": b"unrelated"})
@@ -387,6 +447,7 @@ class PublishHfTest(unittest.TestCase):
                 username="suvradeepp",
                 model_name="test-model",
                 space_name="test-space",
+                space_hardware="zero-a10g",
                 receipt=str(receipt_path),
                 private=False,
                 replace_existing=False,
@@ -474,6 +535,7 @@ class PublishHfTest(unittest.TestCase):
                 username="suvradeepp",
                 model_name="test-model",
                 space_name="test-space",
+                space_hardware="zero-a10g",
                 receipt=str(receipt),
                 private=False,
                 replace_existing=False,
@@ -491,6 +553,10 @@ class PublishHfTest(unittest.TestCase):
                 policy=self.policy,
             )
             self.assertEqual(result["status"], "content_verified_runtime_running")
+            self.assertEqual(
+                result["space_creation"],
+                {"sdk": "gradio", "hardware_if_created": "zero-a10g"},
+            )
             persisted = receipt.read_text(encoding="utf-8")
             self.assertNotIn("secret-token", persisted)
             self.assertEqual(json.loads(persisted)["space"]["runtime_stage"], "RUNNING")
@@ -543,6 +609,7 @@ class PublishHfTest(unittest.TestCase):
                 username="suvradeepp",
                 model_name="test-model",
                 space_name="test-space",
+                space_hardware="zero-a10g",
                 receipt=str(receipt_path),
                 private=False,
                 replace_existing=False,