code-loader 1.0.166.dev2__tar.gz → 1.0.168__tar.gz

{code_loader-1.0.166.dev2 → code_loader-1.0.168}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: code-loader
-Version: 1.0.166.dev2
+Version: 1.0.168
 Summary: 
 Home-page: https://github.com/tensorleap/code-loader
 License: MIT

{code_loader-1.0.166.dev2 → code_loader-1.0.168}/code_loader/inner_leap_binder/leapbinder_decorators.py

@@ -406,12 +406,27 @@ def tensorleap_load_model(prediction_types: Optional[List[PredictionTypeHandler]
                     coerced = {}
                     meta = {i.name: i for i in self.model.get_inputs()}
 
-                    for name, arr in float_arrays_inputs.items():
-                        if name not in meta:
-                            # Keep as-is unless extra inputs are disallowed
-                            coerced[name] = arr
-                            continue
+                    unexpected = [n for n in float_arrays_inputs.keys() if n not in meta]
+                    if unexpected:
+                        raise ValueError(
+                            f"Input name(s) {sorted(unexpected)} do not match any ONNX model input. "
+                            f"Expected input names: {[i.name for i in self.model.get_inputs()]}"
+                        )
+
+                    missing = [n for n in meta if n not in float_arrays_inputs]
+                    if missing:
+                        raise KeyError(f"Missing required input(s): {sorted(missing)}")
 
+                    onnx_input_order = [i.name for i in self.model.get_inputs()]
+                    provided_order = list(float_arrays_inputs.keys())
+                    if provided_order != onnx_input_order:
+                        raise ValueError(
+                            f"Input order mismatch. Provided: {provided_order}, "
+                            f"expected ONNX order: {onnx_input_order}. "
+                            f"Reorder your input functions to match the ONNX model's input order."
+                        )
+
+                    for name, arr in float_arrays_inputs.items():
                         info = meta[name]
                         onnx_type = info.type
                         want_dtype = ONNX_TYPE_TO_NP.get(onnx_type)
@@ -419,17 +434,11 @@ def tensorleap_load_model(prediction_types: Optional[List[PredictionTypeHandler]
                         if want_dtype is None:
                             raise TypeError(f"Unsupported ONNX input type: {onnx_type}")
 
-                        # Cast dtype if needed
                         if arr.dtype != want_dtype:
                             arr = arr.astype(want_dtype, copy=False)
 
                         coerced[name] = arr
 
-                    # Verify required inputs are present
-                    missing = [n for n in meta if n not in coerced]
-                    if missing:
-                        raise KeyError(f"Missing required input(s): {sorted(missing)}")
-
                     return coerced
 
                 # onnx runtime interface
@@ -1067,10 +1076,16 @@ def tensorleap_preprocess():
                 else f"{user_function.__name__}() validation failed: expected to return a single list[{PreprocessResponse.__name__}] object, "
                      f"but returned {len(result)} objects instead."
             )
+            assert len(result) >= 2, \
+                (f'{user_function.__name__}() validation failed: '
+                 f'Expected at least 2 PreprocessResponse objects (training and validation), but got {len(result)}.')
             for i, response in enumerate(result):
                 assert isinstance(response, PreprocessResponse), \
                     (f'{user_function.__name__}() validation failed: '
                      f'Element #{i} in the return list should be a PreprocessResponse. Got {type(response)}.')
+                assert len(response) > 0, \
+                    (f'{user_function.__name__}() validation failed: '
+                     f'PreprocessResponse #{i} has length 0. Each PreprocessResponse must have at least one sample.')
             assert len(set(result)) == len(result), \
                 (f'{user_function.__name__}() validation failed: '
                  f'The return list should not contain duplicate PreprocessResponse objects.')
@@ -1389,11 +1404,7 @@ def tensorleap_input_encoder(name: str, channel_dim=_UNSET, model_input_index=No
 
         def mapping_inner(*args, **kwargs):
             class TempMapping:
-                def __getitem__(self, key):
-                    return self
-
-                def __iter__(self):
-                    return iter([])
+                pass
 
             ret = TempMapping()
             ret.node_mapping = mapping_inner.node_mapping
@@ -1643,8 +1654,8 @@ def tensorleap_status_table():
     code_mapping_failure = [0]
 
     table = [
-        {"name": "tensorleap_preprocess", "Added to integration": UNKNOWN},
         {"name": "tensorleap_integration_test", "Added to integration": UNKNOWN},
+        {"name": "tensorleap_preprocess", "Added to integration": UNKNOWN},
         {"name": "tensorleap_input_encoder", "Added to integration": UNKNOWN},
         {"name": "tensorleap_gt_encoder", "Added to integration": UNKNOWN},
         {"name": "tensorleap_load_model", "Added to integration": UNKNOWN},
@@ -1728,12 +1739,8 @@ def tensorleap_status_table():
         mandatory_ready_mess = "\nAll mandatory parts have been successfully set. If no errors accured, you can now push the project to the Tensorleap system or continue to the next optional reccomeded interface,adding: "
         code_mapping_failure_mes = "Tensorleap_integration_test code flow failed, check raised exception."
 
-        # Sort table: CHECK first, then CROSS, then UNKNOWN
-        status_order = {CHECK: 0, CROSS: 1, UNKNOWN: 2}
-        sorted_table = sorted(table, key=lambda row: status_order.get(row["Added to integration"], 3))
-
-        name_width = max(len(row["name"]) for row in sorted_table)
-        status_width = max(len(row["Added to integration"]) for row in sorted_table)
+        name_width = max(len(row["name"]) for row in table)
+        status_width = max(len(row["Added to integration"]) for row in table)
 
         header = f"{'Decorator Name'.ljust(name_width)} | {'Added to integration'.ljust(status_width)}"
         sep = "-" * len(header)
@@ -1741,7 +1748,7 @@ def tensorleap_status_table():
         print("\n" + header)
         print(sep)
 
-        for row in sorted_table:
+        for row in table:
             print(f"{row['name'].ljust(name_width)} | {row['Added to integration'].ljust(status_width)}")
 
         # Determine next_step based on original table order (recommended integration order)

{code_loader-1.0.166.dev2 → code_loader-1.0.168}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "code-loader"
-version = "1.0.166.dev2"
+version = "1.0.168"
 description = ""
 authors = ["dorhar <doron.harnoy@tensorleap.ai>"]
 license = "MIT"
@@ -28,6 +28,8 @@ grappa = "^1.0.1"
 pytest-cov = "^3.0.0"
 pytest-watch = "^4.2.0"
 mypy = "1.3.0"
+onnxruntime = {version = ">=1.0.0", python = ">=3.10"}
+onnx = {version = ">=1.0.0", python = ">=3.10"}
 
 [build-system]
 requires = ["poetry-core>=1.0.0"]