Deprecate openai/chat-gpt in favor of openai/gpt-3.5-turbo-0613 (#1809)

stanford-crfm · Aug 24, 2023 · 4810bd5 · 4810bd5
1 parent be6f08d
commit 4810bd5
Show file tree

Hide file tree

Showing 9 changed files with 0 additions and 376 deletions.
diff --git a/requirements-freeze.txt b/requirements-freeze.txt
@@ -127,7 +127,6 @@ regex==2022.9.13
 requests==2.28.1
 responses==0.18.0
 retrying==1.3.4
-revChatGPT==0.1.1
 rouge-score==0.1.2
 rsa==4.9
 s3transfer==0.6.0

diff --git a/setup.cfg b/setup.cfg
@@ -62,7 +62,6 @@ install_requires=
     anthropic~=0.2.5
     icetk~=0.0.4  # for ice_tokenizer_client
     openai~=0.27.8
-    revChatGPT~=0.1.1
     sentencepiece~=0.1.97  # For palmyra_client and yalm_tokenizer
     tiktoken~=0.3.3  # for openai_client
     tokenizers~=0.13.3  # for aleph_alpha_client

diff --git a/src/helm/benchmark/presentation/run_specs_chat_gpt.conf b/src/helm/benchmark/presentation/run_specs_chat_gpt.conf
diff --git a/src/helm/benchmark/run_expander.py b/src/helm/benchmark/run_expander.py
@@ -369,7 +369,6 @@ def values_dict(self):
             "full_functionality_text": get_model_names_with_tag(FULL_FUNCTIONALITY_TEXT_MODEL_TAG),
             "ai21/j1-jumbo": ["ai21/j1-jumbo"],
             "openai/curie": ["openai/curie"],
-            "chat_run": ["openai/chat-gpt", "openai/text-davinci-003"],  # Compare ChatGPT to text-davinci-003
             "all": get_all_models(),
             "text_code": get_all_text_models() + get_all_code_models(),
             "text": get_all_text_models(),

diff --git a/src/helm/benchmark/static/schema.yaml b/src/helm/benchmark/static/schema.yaml
@@ -715,13 +715,6 @@ models:
     creator_organization: OpenAI
     access: limited
     release_date: 2023-06-13
-  - name: openai/chat-gpt
-    display_name: ChatGPT
-    description: Sibling model to InstructGPT which interacts in a conversational way. See [OpenAI's announcement](https://openai.com/blog/chatgpt/). The size of the model is unknown.
-    creator_organization: OpenAI
-    access: limited
-    release_date: 2022-11-30
-    todo: true
 
   # Together
   - name: together/Together-gpt-JT-6B-v1

diff --git a/src/helm/proxy/clients/auto_client.py b/src/helm/proxy/clients/auto_client.py
@@ -90,26 +90,12 @@ def _get_client(self, model: str) -> Client:
             elif organization == "neurips":
                 client = HTTPModelClient(cache_config=cache_config)
             elif organization == "openai":
-                from helm.proxy.clients.chat_gpt_client import ChatGPTClient
                 from helm.proxy.clients.openai_client import OpenAIClient
 
-                # TODO: add ChatGPT to the OpenAIClient when it's supported.
-                #       We're using a separate client for now since we're using an unofficial Python library.
-                # See https://github.com/acheong08/ChatGPT/wiki/Setup on how to get a valid session token.
-                chat_gpt_client: ChatGPTClient = ChatGPTClient(
-                    session_token=self.credentials.get("chatGPTSessionToken", ""),
-                    lock_file_path=os.path.join(self.cache_path, "ChatGPT.lock"),
-                    # TODO: use `cache_config` above. Since this feature is still experimental,
-                    #       save queries and responses in a separate collection.
-                    cache_config=self._build_cache_config("ChatGPT"),
-                    tokenizer_client=self._get_tokenizer_client("huggingface"),
-                )
-
                 org_id = self.credentials.get("openaiOrgId", None)
                 api_key = self.credentials.get("openaiApiKey", None)
                 client = OpenAIClient(
                     cache_config=cache_config,
-                    chat_gpt_client=chat_gpt_client,
                     api_key=api_key,
                     org_id=org_id,
                 )