diff --git a/fastchat/conversation.py b/fastchat/conversation.py index 539bef2916..1768933d3b 100644 --- a/fastchat/conversation.py +++ b/fastchat/conversation.py @@ -1403,56 +1403,6 @@ def get_conv_template(name: str) -> Conversation: ) ) -register_conv_template( - Conversation( - name="meta-llama-3.2", - system_message=( - """*Objective:** -You are an advanced language model optimized for generating well-structured, user-friendly, and contextually appropriate responses. Your primary focus is on clear, precise formatting that enhances the readability and usefulness of information. At the same time, you must engage users in a way that feels natural, helpful, and context-aware. - -**Guidelines:** - -1. User-Centric Formatting: - -Use bullet points, tables, headers, and lists to break down complex information. -For step-by-step instructions, numbered lists are preferred. -If presenting options, bold or italicize key terms to highlight choices. -Maintain a consistent format throughout your responses to ensure clarity. - -2. Content Quality: - -Prioritize accuracy and relevance in your responses. -Provide concise answers that directly address the user's queries. If the query requires elaboration, break it into sections or use examples. -Be prepared to adjust your tone according to the user's needs—whether they require a formal, technical response or a more casual, conversational one. - -3. Math and Coding Expertise: - -When addressing math problems, ensure that calculations are correct and explain steps clearly. Use LaTeX formatting for mathematical expressions to enhance readability. Reason step-by-step. Check your work. -For coding tasks, provide clean, well-commented code snippets in the appropriate programming language. Explain the logic behind the code and suggest improvements or alternatives when necessary. -Be prepared to handle edge cases and offer error-checking advice in coding responses. For math, double-check your work to avoid inaccuracies. - -4. Engagement and Adaptability: - -Anticipate follow-up questions and provide information that guides the conversation forward. -Offer suggestions or clarifications proactively if the user's request is vague or open-ended. -Be polite, professional, and empathetic in your responses. Adapt to the user's emotional tone, providing encouragement or reassurance when appropriate. - -5. Quality: - -Remember that high-quality and clear responses are your goal. Always aim to provide responses that exemplify excellence in clarity, usefulness, and user engagement. -Avoid over-complicating responses with unnecessary details. Keep your answers focused and to the point. - -6. Continuous Improvement: - -Learn from user feedback and adjust your strategies to improve the quality of your formatting and content over time. -Keep refining your balance between aesthetic formatting and substantive content to enhance user experience.""" - ), - roles=("user", "assistant"), - sep_style=SeparatorStyle.DEFAULT, - sep=None, - ) -) - # MetaMath default template # reference: https://github.com/meta-math/MetaMath/blob/7b338b5e4692b4c75a2653ec9d65982a61762f6c/eval_math.py#L58 register_conv_template( @@ -2321,19 +2271,6 @@ def get_conv_template(name: str) -> Conversation: ) ) -register_conv_template( - Conversation( - name="column-r", - system_message=( - "You are a smart and helpful AI assistant created by Column AI. " - "Please think step by step, provide detailed and professional response." - ), - roles=("user", "assistant"), - sep_style=SeparatorStyle.DEFAULT, - sep=None, - ) -) - register_conv_template( Conversation( name="grok-2", @@ -2360,21 +2297,6 @@ def get_conv_template(name: str) -> Conversation: ) ) -register_conv_template( - Conversation( - name="toto", - system_message=( - "You are Toto, a brilliant and helpful Large Language Model made by Toto AI.\n" - "When presented with complicated questions that require multiple reasoning steps, " - "you work them step by step. You follow safety guidelines and avoid toxic, " - "violent or illegal content. When needed, you ask the user for clarifications or " - "extra information in order to perform the task as well as you can." - ), - roles=("user", "assistant"), - sep_style=SeparatorStyle.DEFAULT, - sep=None, - ) -) if __name__ == "__main__": from fastchat.conversation import get_conv_template diff --git a/fastchat/model/model_adapter.py b/fastchat/model/model_adapter.py index d2cb28521e..9625df6dbf 100644 --- a/fastchat/model/model_adapter.py +++ b/fastchat/model/model_adapter.py @@ -84,12 +84,9 @@ "gpt2-chatbot", "im-also-a-good-gpt2-chatbot", "im-a-good-gpt2-chatbot", - "upcoming-gpt-mini", "gpt-4o-mini-2024-07-18", "gpt-4o-2024-05-13", "gpt-4o-2024-08-06", - "anonymous-chatbot", - "anonymous-chatbot-0903", "chatgpt-4o-latest-20240903", "chatgpt-4o-latest", "o1-preview", @@ -1622,14 +1619,6 @@ def get_default_conv_template(self, model_path: str) -> Conversation: return get_conv_template("meta-llama-3.1") -class ColumnAdapter(BaseModelAdapter): - def match(self, model_path: str): - return "sus-column-r" in model_path.lower() - - def get_default_conv_template(self, model_path: str) -> Conversation: - return get_conv_template("column-r") - - class GrokAdapter(BaseModelAdapter): def match(self, model_path: str): return "grok" in model_path.lower() @@ -1640,14 +1629,6 @@ def get_default_conv_template(self, model_path: str) -> Conversation: return get_conv_template("grok-2") -class TotoAdapter(BaseModelAdapter): - def match(self, model_path: str): - return "toto" in model_path.lower() - - def get_default_conv_template(self, model_path: str) -> Conversation: - return get_conv_template("toto") - - class CuteGPTAdapter(BaseModelAdapter): """The model adapter for CuteGPT""" @@ -2619,8 +2600,6 @@ def get_default_conv_template(self, model_path: str) -> Conversation: register_model_adapter(SmaugChatAdapter) register_model_adapter(Llama3Adapter) register_model_adapter(Llama31Adapter) -register_model_adapter(ColumnAdapter) -register_model_adapter(TotoAdapter) register_model_adapter(GrokAdapter) register_model_adapter(NoSystemAdapter) diff --git a/fastchat/model/model_registry.py b/fastchat/model/model_registry.py index f59adba1bf..2eed9649e1 100644 --- a/fastchat/model/model_registry.py +++ b/fastchat/model/model_registry.py @@ -47,8 +47,6 @@ def get_model_info(name: str) -> ModelInfo: "gpt-4o-mini-2024-07-18", "gpt-4o-2024-08-06", "gpt-4o-2024-05-13", - "anonymous-chatbot", - "anonymous-chatbot-0903", ], "GPT-4o", "https://openai.com/index/hello-gpt-4o/", @@ -128,8 +126,6 @@ def get_model_info(name: str) -> ModelInfo: "gpt2-chatbot", "im-also-a-good-gpt2-chatbot", "im-a-good-gpt2-chatbot", - "upcoming-gpt-mini", - "sus-column-r", ], "GPT-4-Turbo", "https://platform.openai.com/docs/models/gpt-4-and-gpt-4-turbo", @@ -192,7 +188,7 @@ def get_model_info(name: str) -> ModelInfo: ) register_model_info( - ["athene-70b", "athene-70b-0725", "im-a-little-birdie"], + ["athene-70b", "athene-70b-0725"], "Athene-70B", "https://nexusflow.ai/blogs/athene", "A large language model by NexusFlow", @@ -288,8 +284,6 @@ def get_model_info(name: str) -> ModelInfo: "reka-flash-preview-20240611", "reka-flash", "reka-flash-online", - "pizza-model-small", - "pizza-model-large", ], "Reka Flash", "https://www.reka.ai/news/reka-flash-efficient-and-capable-multimodal-language-models", @@ -336,8 +330,6 @@ def get_model_info(name: str) -> ModelInfo: "mistral-next", "mistral-7b-instruct-v0.2", "mistral-7b-instruct", - "experimental-mf-router", - "experimental-causal-router", "pixtral-12b-2409", ], "Mixtral of experts", @@ -488,8 +480,6 @@ def get_model_info(name: str) -> ModelInfo: "vicuna-13b-v1.5", "vicuna-7b", "vicuna-7b-v1.5", - "column-r", - "column-u", ], "Vicuna", "https://lmsys.org/blog/2023-03-30-vicuna/", diff --git a/fastchat/serve/call_monitor.py b/fastchat/serve/call_monitor.py index d55414bdb9..bc456f107b 100644 --- a/fastchat/serve/call_monitor.py +++ b/fastchat/serve/call_monitor.py @@ -8,26 +8,7 @@ import asyncio REFRESH_INTERVAL_SEC = 300 -LOG_DIR_LIST = [ - "/home/gcpuser/fastchat_logs/server0", - "/home/gcpuser/fastchat_logs/server1", - "/home/gcpuser/fastchat_logs/server2", - "/home/gcpuser/fastchat_logs/server3", - "/home/gcpuser/fastchat_logs/server4", - "/home/gcpuser/fastchat_logs/server5", - "/home/gcpuser/fastchat_logs/server6", - "/home/gcpuser/fastchat_logs/server7", - "/home/gcpuser/fastchat_logs/server8", - "/home/gcpuser/fastchat_logs/server9", - "/home/gcpuser/fastchat_logs/server10", - "/home/gcpuser/fastchat_logs/server11", - "/home/gcpuser/fastchat_logs/server12", - "/home/gcpuser/fastchat_logs/server13", - "/home/gcpuser/fastchat_logs/server14", - "/home/gcpuser/fastchat_logs/server15", - "/home/gcpuser/fastchat_logs/server16", - "/home/gcpuser/fastchat_logs/server17", -] +LOG_DIR_LIST = [] # LOG_DIR = "/home/vicuna/tmp/test_env" @@ -38,58 +19,8 @@ def __init__(self, log_dir_list: list): self.log_dir_list = log_dir_list self.model_call = {} self.user_call = {} - self.model_call_limit_global = { - "gpt-4-1106-preview": 100, - "gpt-4-0125-preview": 100, - "gpt-4-turbo-browsing": 200, - "gpt-4-turbo-2024-04-09": 100, - "mistral-large-2402": 200, - "llama-3-70b-instruct": 2000, - "claude-3-opus-20240229": 1000, - "claude-3-sonnet-20240229": 1000, - "dbrx-instruct": 1000, - "command-r-plus": 1000, - "gemini-1.5-pro-api-0409-preview": 1000, - "gemini-1.5-pro-api-preview": 1000, - "gpt2-chatbot": 1000, - "im-a-good-gpt2-chatbot": 3000, - "im-also-a-good-gpt2-chatbot": 3000, - "gpt-4o-2024-05-13": 1000, - "gpt-4o-2024-08-06": 1000, - "yi-large": 300, - "deepseek-coder-v2": 500, - "deepseek-v2-api-0628": 300, - "claude-3-5-sonnet-20240620": 2000, - "experimental-mf-router": 100, - "experimental-causal-router": 100, - "llama-3.1-405b-instruct": 2000, - "gemini-1.5-pro-exp-0801": 800, - "sus-column-r": 3000, - "chatgpt-4o-latest-20240903": 1500, - } - self.model_call_day_limit_per_user = { - "gpt-4-1106-preview": 5, - "gpt-4-0125-preview": 5, - "gpt-4-turbo-2024-04-09": 5, - "gpt-4-turbo-browsing": 8, - "mistral-large-2402": 8, - "claude-3-opus-20240229": 15, - "claude-3-sonnet-20240229": 32, - "command-r-plus": 32, - "gemini-1.5-pro-api-0409-preview": 16, - "gemini-1.5-pro-api-preview": 16, - "gpt2-chatbot": 8, - "im-a-good-gpt2-chatbot": 16, - "im-also-a-good-gpt2-chatbot": 16, - "gpt-4o-2024-05-13": 8, - "gpt-4o-2024-08-06": 8, - "deepseek-coder-v2": 16, - "claude-3-5-sonnet-20240620": 16, - "llama-3.1-405b-instruct": 16, - "gemini-1.5-pro-exp-0801": 8, - "sus-column-r": 16, - "chatgpt-4o-latest-20240903": 8, - } + self.model_call_limit_global = {} + self.model_call_day_limit_per_user = {} async def update_stats(self, num_file=1) -> None: while True: