langfun 0.1.2.dev202502060804__tar.gz → 0.1.2.dev202502070804__tar.gz

{langfun-0.1.2.dev202502060804 → langfun-0.1.2.dev202502070804}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: langfun
-Version: 0.1.2.dev202502060804
+Version: 0.1.2.dev202502070804
 Summary: Langfun: Language as Functions.
 Home-page: https://github.com/google/langfun
 Author: Langfun Authors

{langfun-0.1.2.dev202502060804 → langfun-0.1.2.dev202502070804}/langfun/core/llms/gemini.py

@@ -90,7 +90,7 @@ SUPPORTED_MODELS_AND_SETTINGS = {
         experimental=True,
         in_service=True,
         supported_modalities=ALL_MODALITIES,
-        rpm_free=2000,
+        rpm_free=10_000,
         tpm_free=40_000_000,
         rpm_paid=0,
         tpm_paid=0,
@@ -106,32 +106,32 @@ SUPPORTED_MODELS_AND_SETTINGS = {
         experimental=True,
         in_service=True,
         supported_modalities=ALL_MODALITIES,
-        rpm_free=2000,
+        rpm_free=10_000,
         tpm_free=40_000_000,
         rpm_paid=0,
         tpm_paid=0,
-        cost_per_1m_input_tokens_up_to_128k=0,
-        cost_per_1m_output_tokens_up_to_128k=0,
-        cost_per_1m_cached_tokens_up_to_128k=0,
-        cost_per_1m_input_tokens_longer_than_128k=0,
-        cost_per_1m_output_tokens_longer_than_128k=0,
-        cost_per_1m_cached_tokens_longer_than_128k=0,
+        cost_per_1m_input_tokens_up_to_128k=0.1,
+        cost_per_1m_output_tokens_up_to_128k=0.4,
+        cost_per_1m_cached_tokens_up_to_128k=0.025,
+        cost_per_1m_input_tokens_longer_than_128k=0.1,
+        cost_per_1m_output_tokens_longer_than_128k=0.4,
+        cost_per_1m_cached_tokens_longer_than_128k=0.025,
     ),
     'gemini-2.0-pro-exp-02-05': pg.Dict(
         latest_update='2025-02-05',
         experimental=True,
         in_service=True,
         supported_modalities=ALL_MODALITIES,
-        rpm_free=1000,
-        tpm_free=40_000_000,
+        rpm_free=10,
+        tpm_free=4_000_000,
         rpm_paid=0,
         tpm_paid=0,
-        cost_per_1m_input_tokens_up_to_128k=0,
-        cost_per_1m_output_tokens_up_to_128k=0,
-        cost_per_1m_cached_tokens_up_to_128k=0,
-        cost_per_1m_input_tokens_longer_than_128k=0,
-        cost_per_1m_output_tokens_longer_than_128k=0,
-        cost_per_1m_cached_tokens_longer_than_128k=0,
+        cost_per_1m_input_tokens_up_to_128k=0.1,
+        cost_per_1m_output_tokens_up_to_128k=0.4,
+        cost_per_1m_cached_tokens_up_to_128k=0.025,
+        cost_per_1m_input_tokens_longer_than_128k=0.1,
+        cost_per_1m_output_tokens_longer_than_128k=0.4,
+        cost_per_1m_cached_tokens_longer_than_128k=0.025,
     ),
     'gemini-2.0-flash-thinking-exp-01-21': pg.Dict(
         latest_update='2024-01-21',

{langfun-0.1.2.dev202502060804 → langfun-0.1.2.dev202502070804}/langfun/core/llms/rest.py

@@ -63,15 +63,14 @@ class REST(lf.LanguageModel):
   def _initialize(self) -> None:
     """Initializes the API. Subclasses can override."""
 
-  @functools.cached_property
-  def _session(self) -> requests.Session:
+  def session(self) -> requests.Session:
     assert self._api_initialized
-    s = self._create_session()
+    s = self._session()
     # Placeholder for Google-internal session adapter.
     s.headers.update(self.headers or {})
     return s
 
-  def _create_session(self) -> requests.Session:
+  def _session(self) -> requests.Session:
     """Creates a new session."""
     return requests.Session()
 
@@ -88,12 +87,14 @@ class REST(lf.LanguageModel):
 
   def _sample_single(self, prompt: lf.Message) -> lf.LMSamplingResult:
     try:
-      response = self._session.post(
-          self.api_endpoint,
-          json=self.request(prompt, self.sampling_options),
-          timeout=self.timeout,
-      )
-      return self._parse_response(response)
+      with self.session() as session:
+        return self._parse_response(
+            session.post(
+                self.api_endpoint,
+                json=self.request(prompt, self.sampling_options),
+                timeout=self.timeout,
+            )
+        )
     except (requests.exceptions.ReadTimeout,
             requests.exceptions.ConnectTimeout) as e:
       raise lf.TemporaryLMError(str(e)) from e

{langfun-0.1.2.dev202502060804 → langfun-0.1.2.dev202502070804}/langfun/core/llms/vertexai.py

@@ -118,7 +118,7 @@ class VertexAI(rest.REST):
     """Returns a string to identify the model."""
     return f'VertexAI({self.model})'
 
-  def _create_session(self):
+  def _session(self):
     assert self._credentials is not None
     assert auth_requests is not None
     return auth_requests.AuthorizedSession(self._credentials)

{langfun-0.1.2.dev202502060804 → langfun-0.1.2.dev202502070804}/langfun/core/llms/vertexai_test.py

@@ -45,7 +45,7 @@ class VertexAITest(unittest.TestCase):
     self.assertTrue(model.model_id.startswith('VertexAI('))
     self.assertIn('us-central1', model.api_endpoint)
     self.assertTrue(model._api_initialized)
-    self.assertIsNotNone(model._session)
+    self.assertIsNotNone(model.session())
     del os.environ['VERTEXAI_PROJECT']
     del os.environ['VERTEXAI_LOCATION']
 

{langfun-0.1.2.dev202502060804 → langfun-0.1.2.dev202502070804}/langfun.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: langfun
-Version: 0.1.2.dev202502060804
+Version: 0.1.2.dev202502070804
 Summary: Langfun: Language as Functions.
 Home-page: https://github.com/google/langfun
 Author: Langfun Authors