chatgpt send request and fix logic for max tokens

models/openai_model.py

@@ -1018,7 +1018,7 @@ class Model:
         # Validate that  all the parameters are in a good state before we send the request
 
         if not max_tokens_override:
-            if model:
+            if model and model not in Models.GPT4_MODELS and model not in Models.CHATGPT_MODELS:
                 max_tokens_override = Models.get_max_tokens(model) - tokens
 
         print(f"The prompt about to be sent is {prompt}")

tests/test_requests.py

@@ -16,6 +16,16 @@ async def test_send_req():
     model = Model(usage_service)
     prompt = 'how many hours are in a day?'
     tokens = len(GPT2TokenizerFast.from_pretrained("gpt2")(prompt)["input_ids"])
-    # tokens = 60
     res = await model.send_request(prompt, tokens)
     assert '24' in res['choices'][0]['text']
+
+
+# ChatGPT version
+@pytest.mark.asyncio
+async def test_send_req_gpt():
+
+    usage_service = UsageService(Path("../tests"))
+    model = Model(usage_service)
+    prompt = 'how many hours are in a day?'
+    res = await model.send_request(prompt, None, is_chatgpt_request=True, model="gpt-3.5-turbo")
+    assert '24' in res['choices'][0]['message']['content']