Update gpt-3.5-turbo-instruct to act as Completion model instead of Instruct (as OpenAI API doesn't honor <|endofprompt|> special token)

Author: Harsha-Nori

guidance/models/_openai.py

@@ -68,8 +68,8 @@ def __init__(self, model, tokenizer=None, echo=True, caching=True, api_key=None,
                 found_subclass = OpenAIChat
 
             # instruct
-            elif "instruct" in model:
-                found_subclass = OpenAIInstruct
+            # elif "instruct" in model: # All current OpenAI instruct models behave as Completion models. 
+            #     found_subclass = OpenAIInstruct
 
             # regular completion
             else:

tests/models/test_openai.py

@@ -14,8 +14,8 @@ def test_openai_class_detection():
         "ft:gpt-4": guidance.models.OpenAIChat,
         "ft:gpt-4-vision-preview": guidance.models.OpenAIChat,
         "ft:gpt-3.5-turbo:my-org:custom_suffix:id": guidance.models.OpenAIChat,
-        "gpt-3.5-turbo-instruct": guidance.models.OpenAIInstruct,
-        "ft:gpt-3.5-turbo-instruct": guidance.models.OpenAIInstruct,
+        "gpt-3.5-turbo-instruct": guidance.models.OpenAICompletion,
+        "ft:gpt-3.5-turbo-instruct": guidance.models.OpenAICompletion,
         "text-curie-001": guidance.models.OpenAICompletion,
         "ft:text-curie-001": guidance.models.OpenAICompletion,
         "text-davinci-003": guidance.models.OpenAICompletion,
@@ -40,19 +40,6 @@ def test_openai_basic():
     lm += f"""{gen(max_tokens=1, suffix=nl)}aaaaaa"""
     assert str(lm)[-5:] == "aaaaa"
 
-def test_openai_gpt35_instruct():
-    from guidance import gen, instruction
-
-    # this relies on the environment variable OPENAI_API_KEY being set
-    try:
-        lm = guidance.models.OpenAI('gpt-3.5-turbo-instruct')
-    except:
-        pytest.skip("Skipping OpenAI test because we can't load the model!")
-
-    with instruction():
-        lm += "What is a popular flavor?"
-    lm += gen('flavor', max_tokens=2, stop=".")
-    assert len(lm['flavor']) > 0
 
 def test_openai_select():
     try: