orca-sdk 0.1.5__py3-none-any.whl → 0.1.6__py3-none-any.whl

orca_sdk/_shared/metrics.py

@@ -20,7 +20,9 @@ from numpy.typing import NDArray
 def softmax(logits: np.ndarray, axis: int = -1) -> np.ndarray:
     shifted = logits - np.max(logits, axis=axis, keepdims=True)
     exps = np.exp(shifted)
-    return exps / np.sum(exps, axis=axis, keepdims=True)
+    sums = np.sum(exps, axis=axis, keepdims=True)
+    # Guard against division by zero (can happen if all logits are -inf or NaN)
+    return exps / np.where(sums > 0, sums, 1.0)
 
 
 # We don't want to depend on transformers just for the eval_pred type in orca_sdk
@@ -300,7 +302,9 @@ def convert_logits_to_probabilities(logits: NDArray[np.float32]) -> NDArray[np.f
             probabilities = cast(NDArray[np.float32], softmax(logits))
         elif not np.allclose(logits.sum(-1, keepdims=True), 1.0):
             # Rows don't sum to 1: normalize to probabilities
-            probabilities = cast(NDArray[np.float32], logits / logits.sum(-1, keepdims=True))
+            row_sums = logits.sum(-1, keepdims=True)
+            # Guard against division by zero (can happen if all values in a row are 0 or NaN)
+            probabilities = cast(NDArray[np.float32], logits / np.where(row_sums > 0, row_sums, 1.0))
         else:
             # Already normalized probabilities
             probabilities = logits
@@ -349,7 +353,7 @@ def calculate_classification_metrics(
     num_classes_references = len(set(references))
     num_classes_predictions = len(set(predictions))
     num_none_predictions = np.isnan(probabilities).all(axis=-1).sum()
-    coverage = 1 - num_none_predictions / len(probabilities)
+    coverage = 1 - (num_none_predictions / len(probabilities) if len(probabilities) > 0 else 0)
 
     if average is None:
         average = "binary" if num_classes_references == 2 and num_none_predictions == 0 else "weighted"
@@ -503,7 +507,7 @@ def calculate_regression_metrics(
     # Filter out NaN values from predictions (expected_scores are already validated to be non-NaN)
     valid_mask = ~np.isnan(predictions)
     num_none_predictions = (~valid_mask).sum()
-    coverage = 1 - num_none_predictions / len(predictions)
+    coverage = 1 - (num_none_predictions / len(predictions) if len(predictions) > 0 else 0)
     if num_none_predictions > 0:
         references = references[valid_mask]
         predictions = predictions[valid_mask]

orca_sdk/async_client.py

@@ -555,16 +555,7 @@ class PredictiveModelUpdate(TypedDict):
 
 
 PretrainedEmbeddingModelName = Literal[
-    "CLIP_BASE",
-    "GTE_BASE",
-    "CDE_SMALL",
-    "DISTILBERT",
-    "GTE_SMALL",
-    "MXBAI_LARGE",
-    "E5_LARGE",
-    "QWEN2_1_5B",
-    "BGE_BASE",
-    "GIST_LARGE",
+    "CLIP_BASE", "GTE_BASE", "CDE_SMALL", "DISTILBERT", "GTE_SMALL", "MXBAI_LARGE", "E5_LARGE", "BGE_BASE", "GIST_LARGE"
 ]
 
 
@@ -1175,7 +1166,14 @@ class BootstrapClassificationModelRequest(TypedDict):
     num_examples_per_label: NotRequired[int]
 
 
-class BootstrapClassificationModelResult(TypedDict):
+class BootstrapLabeledMemoryDataInput(TypedDict):
+    model_description: str
+    label_names: list[str]
+    initial_examples: NotRequired[list[LabeledExample]]
+    num_examples_per_label: NotRequired[int]
+
+
+class BootstrapLabeledMemoryDataResult(TypedDict):
     model_description: str
     label_names: list[str]
     model_name: str
@@ -1680,7 +1678,7 @@ class BootstrapClassificationModelMeta(TypedDict):
     datasource_meta: DatasourceMetadata
     memoryset_meta: MemorysetMetadata
     model_meta: ClassificationModelMetadata
-    agent_output: BootstrapClassificationModelResult
+    agent_output: BootstrapLabeledMemoryDataResult
 
 
 class BootstrapClassificationModelResponse(TypedDict):
@@ -2570,7 +2568,7 @@ class OrcaAsyncClient(AsyncClient):
         timeout: TimeoutTypes | UseClientDefault = USE_CLIENT_DEFAULT,
         extensions: RequestExtensions | None = None,
     ) -> BootstrapClassificationModelResponse:
-        """Get the status of a bootstrap classification model job"""
+        """Get the status of a bootstrap labeled memory data job"""
         pass
 
     async def GET(
@@ -3292,6 +3290,32 @@ class OrcaAsyncClient(AsyncClient):
         """Get row count from a specific datasource with optional filtering."""
         pass
 
+    @overload
+    async def POST(
+        self,
+        path: Literal["/datasource/bootstrap_memory_data"],
+        *,
+        params: None = None,
+        json: BootstrapLabeledMemoryDataInput,
+        data: None = None,
+        files: None = None,
+        content: None = None,
+        parse_as: Literal["json"] = "json",
+        headers: HeaderTypes | None = None,
+        cookies: CookieTypes | None = None,
+        auth: AuthTypes | UseClientDefault = USE_CLIENT_DEFAULT,
+        follow_redirects: bool | UseClientDefault = USE_CLIENT_DEFAULT,
+        timeout: TimeoutTypes | UseClientDefault = USE_CLIENT_DEFAULT,
+        extensions: RequestExtensions | None = None,
+    ) -> BootstrapLabeledMemoryDataResult:
+        """
+        Bootstrap memory data using an AI agent.
+
+        This endpoint uses the bootstrap labeled memory data agent to generate
+        high-quality, diverse training examples for a classification model.
+        """
+        pass
+
     @overload
     async def POST(
         self,
@@ -3540,7 +3564,7 @@ class OrcaAsyncClient(AsyncClient):
         """
         Bootstrap a classification model by creating a memoryset with generated memories and a classification model.
 
-        This endpoint uses the bootstrap_classification_model agent to generate:
+        This endpoint uses the bootstrap_labeled_memory_data agent to generate:
         1. Memoryset configuration with appropriate settings
         2. Model configuration with optimal parameters
         3. High-quality training memories for each label

orca_sdk/classification_model.py

@@ -12,12 +12,10 @@ from ._utils.common import UNSET, CreateMode, DropMode
 from .async_client import OrcaAsyncClient
 from .client import (
     BootstrapClassificationModelMeta,
-    BootstrapClassificationModelResult,
-    ClassificationEvaluationRequest,
+    BootstrapLabeledMemoryDataResult,
     ClassificationModelMetadata,
     ClassificationPredictionRequest,
     OrcaClient,
-    PostClassificationModelByModelNameOrIdEvaluationParams,
     PredictiveModelUpdate,
     RACHeadType,
 )
@@ -43,7 +41,7 @@ class BootstrappedClassificationModel:
     datasource: Datasource | None
     memoryset: LabeledMemoryset | None
     classification_model: ClassificationModel | None
-    agent_output: BootstrapClassificationModelResult | None
+    agent_output: BootstrapLabeledMemoryDataResult | None
 
     def __init__(self, metadata: BootstrapClassificationModelMeta):
         self.datasource = Datasource.open(metadata["datasource_meta"]["id"])

orca_sdk/client.py

@@ -553,16 +553,7 @@ class PredictiveModelUpdate(TypedDict):
 
 
 PretrainedEmbeddingModelName = Literal[
-    "CLIP_BASE",
-    "GTE_BASE",
-    "CDE_SMALL",
-    "DISTILBERT",
-    "GTE_SMALL",
-    "MXBAI_LARGE",
-    "E5_LARGE",
-    "QWEN2_1_5B",
-    "BGE_BASE",
-    "GIST_LARGE",
+    "CLIP_BASE", "GTE_BASE", "CDE_SMALL", "DISTILBERT", "GTE_SMALL", "MXBAI_LARGE", "E5_LARGE", "BGE_BASE", "GIST_LARGE"
 ]
 
 
@@ -1173,7 +1164,14 @@ class BootstrapClassificationModelRequest(TypedDict):
     num_examples_per_label: NotRequired[int]
 
 
-class BootstrapClassificationModelResult(TypedDict):
+class BootstrapLabeledMemoryDataInput(TypedDict):
+    model_description: str
+    label_names: list[str]
+    initial_examples: NotRequired[list[LabeledExample]]
+    num_examples_per_label: NotRequired[int]
+
+
+class BootstrapLabeledMemoryDataResult(TypedDict):
     model_description: str
     label_names: list[str]
     model_name: str
@@ -1678,7 +1676,7 @@ class BootstrapClassificationModelMeta(TypedDict):
     datasource_meta: DatasourceMetadata
     memoryset_meta: MemorysetMetadata
     model_meta: ClassificationModelMetadata
-    agent_output: BootstrapClassificationModelResult
+    agent_output: BootstrapLabeledMemoryDataResult
 
 
 class BootstrapClassificationModelResponse(TypedDict):
@@ -2568,7 +2566,7 @@ class OrcaClient(Client):
         timeout: TimeoutTypes | UseClientDefault = USE_CLIENT_DEFAULT,
         extensions: RequestExtensions | None = None,
     ) -> BootstrapClassificationModelResponse:
-        """Get the status of a bootstrap classification model job"""
+        """Get the status of a bootstrap labeled memory data job"""
         pass
 
     def GET(
@@ -3290,6 +3288,32 @@ class OrcaClient(Client):
         """Get row count from a specific datasource with optional filtering."""
         pass
 
+    @overload
+    def POST(
+        self,
+        path: Literal["/datasource/bootstrap_memory_data"],
+        *,
+        params: None = None,
+        json: BootstrapLabeledMemoryDataInput,
+        data: None = None,
+        files: None = None,
+        content: None = None,
+        parse_as: Literal["json"] = "json",
+        headers: HeaderTypes | None = None,
+        cookies: CookieTypes | None = None,
+        auth: AuthTypes | UseClientDefault = USE_CLIENT_DEFAULT,
+        follow_redirects: bool | UseClientDefault = USE_CLIENT_DEFAULT,
+        timeout: TimeoutTypes | UseClientDefault = USE_CLIENT_DEFAULT,
+        extensions: RequestExtensions | None = None,
+    ) -> BootstrapLabeledMemoryDataResult:
+        """
+        Bootstrap memory data using an AI agent.
+
+        This endpoint uses the bootstrap labeled memory data agent to generate
+        high-quality, diverse training examples for a classification model.
+        """
+        pass
+
     @overload
     def POST(
         self,
@@ -3538,7 +3562,7 @@ class OrcaClient(Client):
         """
         Bootstrap a classification model by creating a memoryset with generated memories and a classification model.
 
-        This endpoint uses the bootstrap_classification_model agent to generate:
+        This endpoint uses the bootstrap_labeled_memory_data agent to generate:
         1. Memoryset configuration with appropriate settings
         2. Model configuration with optimal parameters
         3. High-quality training memories for each label

orca_sdk/embedding_model.py

@@ -340,7 +340,6 @@ class PretrainedEmbeddingModel(EmbeddingModelBase):
     - **`E5_LARGE`**: E5-Large instruction-tuned embedding model from Hugging Face ([intfloat/multilingual-e5-large-instruct](https://huggingface.co/intfloat/multilingual-e5-large-instruct))
     - **`GIST_LARGE`**: GIST-Large embedding model from Hugging Face ([avsolatorio/GIST-large-Embedding-v0](https://huggingface.co/avsolatorio/GIST-large-Embedding-v0))
     - **`MXBAI_LARGE`**: Mixbreas's Large embedding model from Hugging Face ([mixedbread-ai/mxbai-embed-large-v1](https://huggingface.co/mixedbread-ai/mxbai-embed-large-v1))
-    - **`QWEN2_1_5B`**: Alibaba's Qwen2-1.5B instruction-tuned embedding model from Hugging Face ([Alibaba-NLP/gte-Qwen2-1.5B-instruct](https://huggingface.co/Alibaba-NLP/gte-Qwen2-1.5B-instruct))
     - **`BGE_BASE`**: BAAI's BGE-Base instruction-tuned embedding model from Hugging Face ([BAAI/bge-base-en-v1.5](https://huggingface.co/BAAI/bge-base-en-v1.5))
 
     **Instruction Support:**
@@ -373,7 +372,6 @@ class PretrainedEmbeddingModel(EmbeddingModelBase):
     E5_LARGE = _ModelDescriptor("E5_LARGE")
     GIST_LARGE = _ModelDescriptor("GIST_LARGE")
     MXBAI_LARGE = _ModelDescriptor("MXBAI_LARGE")
-    QWEN2_1_5B = _ModelDescriptor("QWEN2_1_5B")
     BGE_BASE = _ModelDescriptor("BGE_BASE")
 
     name: PretrainedEmbeddingModelName

{orca_sdk-0.1.5.dist-info → orca_sdk-0.1.6.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: orca_sdk
-Version: 0.1.5
+Version: 0.1.6
 Summary: SDK for interacting with Orca Services
 License-Expression: Apache-2.0
 Author: Orca DB Inc.

{orca_sdk-0.1.5.dist-info → orca_sdk-0.1.6.dist-info}/RECORD

@@ -1,6 +1,6 @@
 orca_sdk/__init__.py,sha256=xyjNwkLQXaX8A-UYgGwYDjv2btOXArT_yiMTfmW7KA8,1003
 orca_sdk/_shared/__init__.py,sha256=3Kt0Hu3QLI5FEp9nqGTxqAm3hAoBJKcagfaGQZ-lbJQ,223
-orca_sdk/_shared/metrics.py,sha256=m-d2-AsHI12REWev1WeniOcQRhF5cXxNjUgC4skM2o4,19412
+orca_sdk/_shared/metrics.py,sha256=a_FdsPGDjR3CMOEBaEhEBqMfWUg7sqz9Jeh26XzAeg0,19756
 orca_sdk/_shared/metrics_test.py,sha256=n7eEAT8e6RqbI94ftEDljTBzOuh-YkFpXfF3DOoZA10,12905
 orca_sdk/_utils/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 orca_sdk/_utils/analysis_ui.py,sha256=nT-M_YcNRCVPQzvuqYNFKnNHhYkADYBvq1GlIUePrWw,9232
@@ -17,16 +17,16 @@ orca_sdk/_utils/prediction_result_ui.py,sha256=Ur_FY7dz3oWNmtPiP3Wl3yRlEMgK8q9Uf
 orca_sdk/_utils/tqdm_file_reader.py,sha256=Lw7Cg1UgNuRUoN6jjqZb-IlV00H-kbRcrZLdudr1GxE,324
 orca_sdk/_utils/value_parser.py,sha256=c3qMABCCDQcIjn9N1orYYnlRwDW9JWdGwW_2TDZPLdI,1286
 orca_sdk/_utils/value_parser_test.py,sha256=OybsiC-Obi32RRi9NIuwrVBRAnlyPMV1xVAaevSrb7M,1079
-orca_sdk/async_client.py,sha256=V16wWwdJFvAzmKd5zHsFo3ny_-7B34UrONl80bZzKKs,131628
-orca_sdk/classification_model.py,sha256=90r-PfJ3ZW7ZJ7jrZPTbhuXRds46f7Ooe8FTp-iUJgg,46350
+orca_sdk/async_client.py,sha256=y2D3fPQZmbwmtYWAk5acJ45atSZen9MNfjP2tKjpP6Q,132737
+orca_sdk/classification_model.py,sha256=4AcQvAm0EN7w0qx0WpgEs7VUoIIPTqIVE86wtkaIAYs,46249
 orca_sdk/classification_model_test.py,sha256=vBn7KBb9-ACuJEdzW50n54Fn6Mh9iEYbn1197lE8-yI,36997
-orca_sdk/client.py,sha256=J3Od1sWO7YK2M5afcRNeJcjzNEgZ4zt6e7vLJdk6Nbs,130695
+orca_sdk/client.py,sha256=oQd8Lm0agetLyAdVRP8IZqe6S5mjxhFSnbVHqhT7dmI,131798
 orca_sdk/conftest.py,sha256=0O1VY-SPKNAvi9fBLdY1RMnYVgZvMjP92y99bNAqqiw,12461
 orca_sdk/credentials.py,sha256=80_1r8n5jruEvN_E629SaRrRhKvF_NhWUEZyZzPXkqQ,6620
 orca_sdk/credentials_test.py,sha256=TLbXJMz3IlThvtSrHeLM7jRsKnrncA_ahOTpHg15Ei4,4089
 orca_sdk/datasource.py,sha256=6QaccghiyFEUSFcqnwjIJzpgIh9Id0snJk2EqViqPsU,22356
 orca_sdk/datasource_test.py,sha256=sCk3IcQJbDut5oN4Wf7PXhTxyMwalxMuCXJekSxy9wk,16665
-orca_sdk/embedding_model.py,sha256=bZhbNJBimWc9Ryklza3q9HS0MRWsiH5Lhn6p7pff0RI,28165
+orca_sdk/embedding_model.py,sha256=4xxfo26b5X_YJtU8KyqoMmJQ6VgfHEcYftVSz-RfDng,27920
 orca_sdk/embedding_model_test.py,sha256=-NItbNb3tTVj5jAvSi3WjV3FP448q08lmT5iObg9vwA,8133
 orca_sdk/job.py,sha256=wHwVt-s7i-v8udhLGybB-90Kp4dwOLrY806bE4Tam5Q,13092
 orca_sdk/job_test.py,sha256=nRSWxd_1UIfrj9oMVvrXjt6OBkBpddYAjb2y6P-DTUg,4327
@@ -36,6 +36,6 @@ orca_sdk/regression_model.py,sha256=vXdY2Fbfc0MyECUR3fa_IR-nETPrDN7VFAdjvsgHPrs,
 orca_sdk/regression_model_test.py,sha256=DfWLkqxB835jjwM-sj1uxQ6Yz_ZBMnt8EHjdfnHsRnU,25103
 orca_sdk/telemetry.py,sha256=ZyCMiyyo_SchjadWZH55TlLrC4Ucq5S316NbW26LL4Y,27834
 orca_sdk/telemetry_test.py,sha256=eT66C5lFdNg-pQdo2I__BP7Tn5fTc9aTkVo9ZhWwhU0,5519
-orca_sdk-0.1.5.dist-info/METADATA,sha256=NsXoCiKQ-frqwZeydk_OzvK-QqD1_SnGdRuERXM1ILc,3659
-orca_sdk-0.1.5.dist-info/WHEEL,sha256=zp0Cn7JsFoX2ATtOhtaFYIiE2rmFAD4OcMhtUki8W3U,88
-orca_sdk-0.1.5.dist-info/RECORD,,
+orca_sdk-0.1.6.dist-info/METADATA,sha256=85QDZDP9Uxda4oZ3BMPP_kI5T4GPy1mFMYtWh1-nI54,3659
+orca_sdk-0.1.6.dist-info/WHEEL,sha256=zp0Cn7JsFoX2ATtOhtaFYIiE2rmFAD4OcMhtUki8W3U,88
+orca_sdk-0.1.6.dist-info/RECORD,,