fix: v2 API need special handling

- in V2 client API the embeddings are not returned as a list but as EmbedByTypeResponseEmbeddings
- v2 client api also mandates embedding_types arg for embed()

src/cohere/base_client.py

@@ -2251,9 +2251,17 @@ def classify(
         )
         """
         # TODO @coderham: Cache the LR model to speedup the classification process and save on API calls
-        example_embeds = self.embed(model=model, texts=[example.text for example in examples], input_type="classification")
-        example_embeddings = np.zeros((len(examples), len(example_embeds.embeddings[0])))
-        for i, embed in enumerate(example_embeds.embeddings):
+        example_embeds = self.embed(model=model, embedding_types=["float"], input_type="classification",
+                                    texts=[example.text for example in examples])
+
+        # in V2 client API the embeddings are not returned as a list but as EmbedByTypeResponseEmbeddings
+        if type(example_embeds.embeddings) is not list:
+            example_embedding_float = example_embeds.embeddings.float_
+        else:
+            example_embedding_float = example_embeds.embeddings
+
+        example_embeddings = np.zeros((len(examples), len(example_embedding_float[0])))
+        for i, embed in enumerate(example_embedding_float):
             example_embeddings[i] = np.array(embed)
         example_labels = np.array([example.label for example in examples])
 
@@ -2267,9 +2275,17 @@ def classify(
             print(f"ConvergenceWarning: {convergence_warning}")
 
         # Make predictions on the inputs
-        input_embeds = self.embed(model=model, texts=inputs, input_type="classification")
-        input_embeddings = np.zeros((len(inputs), len(input_embeds.embeddings[0])))
-        for i, embed in enumerate(input_embeds.embeddings):
+        input_embeds = self.embed(model=model, embedding_types=["float"], input_type="classification",
+                                  texts=inputs)
+
+        # in V2 client API the embeddings are not returned as a list but as EmbedByTypeResponseEmbeddings
+        if type(input_embeds.embeddings) is not list:
+            input_embedding_float = input_embeds.embeddings.float_
+        else:
+            input_embedding_float = input_embeds.embeddings
+
+        input_embeddings = np.zeros((len(inputs), len(input_embedding_float[0])))
+        for i, embed in enumerate(input_embedding_float):
             input_embeddings[i] = np.array(embed)
 
         class_probs = clf.predict_proba(input_embeddings)