minor (7eec38ff) · Commits · 张泽凯 / VAGEN

scripts/exps/prompting/frozenlake/model_config.yaml

+22 −23

Original line number	Diff line number	Diff line
		models:
		# qwen_0.5b:
		# provider: vllm
		# model_name: Qwen/Qwen2.5-0.5B-Instruct
		# max_tokens: 1024
		# temperature: 0.7
		# tensor_parallel_size: 1
		# gpu_memory_utilization: 0.9
		models:
		# qwen_0.5b:
		# provider: vllm
		@@ -23,15 +15,15 @@ models:
		# tensor_parallel_size: 2
		# gpu_memory_utilization: 0.9

		gpt4o:
		provider: openai
		model_name: gpt-4o
		max_tokens: 1024
		temperature: 0.7
		presence_penalty: 0.0
		frequency_penalty: 0.0
		max_retries: 3
		timeout: 60
		# gpt4o:
		# provider: openai
		# model_name: gpt-4o
		# max_tokens: 1024
		# temperature: 0.7
		# presence_penalty: 0.0
		# frequency_penalty: 0.0
		# max_retries: 3
		# timeout: 60

		# gpt4_vision:
		# provider: openai
		@@ -43,11 +35,11 @@ models:
		# max_retries: 3
		# timeout: 60

		claude_3_sonnet:
		provider: claude
		model_name: claude-3-7-sonnet-20250219
		max_tokens: 1024
		temperature: 0.7
		# claude_3_sonnet:
		# provider: claude
		# model_name: claude-3-7-sonnet-20250219
		# max_tokens: 1024
		# temperature: 0.7

		# # Batch processing configuration example
		# claude_3_haiku_batch:
		@@ -58,3 +50,10 @@ models:
		# use_batch_api: true # Use batch API for cost optimization
		# batch_poll_interval: 10 # Poll every 10 seconds
		# batch_max_wait_time: 7200 # Wait up to 2 hours

		Gemini_2.5_flask:
		provider: gemini
		model_name: gemini
		max_tokens: 1024
		temperature: 0.7

		No newline at end of file

deleted100644 → 0

−6.07 KiB

Loading image diff...

+4 −1

Original line number	Diff line number	Diff line
		@@ -8,7 +8,10 @@ export OPENAI_API_KEY=<sk-your_openai_api_key>
		pip install anthropic
		export ANTHROPIC_API_KEY=<your_claude_api_key>
		```

		# Use Gemini API
		```
		export GOOGLE_API_KEY=<your_gemini_api_key>
		```
		## Test your inference pipeline
		```
		python vagen/server/server.py

+5 −0

Original line number	Diff line number	Diff line
		from .vllm import VLLMModelInterface, VLLMModelConfig
		from .openai import OpenAIModelInterface, OpenAIModelConfig
		from .claude import ClaudeModelInterface, ClaudeModelConfig
		from .gemini import GeminiModelInterface, GeminiModelConfig

		REGISTERED_MODEL = {
		"vllm": {
		@@ -14,5 +15,9 @@ REGISTERED_MODEL = {
		"claude": {
		"model_cls": ClaudeModelInterface,
		"config_cls": ClaudeModelConfig
		},
		"gemini": {
		"model_cls": GeminiModelInterface,
		"config_cls": GeminiModelConfig
		}
		}
		No newline at end of file

+18 −5

Original line number	Diff line number	Diff line
		@@ -38,11 +38,24 @@ class ModelFactory:
		available_providers = list(REGISTERED_MODEL.keys())
		raise ValueError(f"Unknown provider '{provider}'. Available providers: {available_providers}")

		try:
		# Get model and config classes from registry
		model_cls = REGISTERED_MODEL[provider]["model_cls"]
		config_cls = REGISTERED_MODEL[provider]["config_cls"]

		# Get provider info to validate model_name
		supported_models = []
		if hasattr(config_cls, 'get_provider_info'):
		provider_info = config_cls.get_provider_info()
		supported_models = provider_info.get("supported_models", [])

		# Validate model name before initialization
		if model_name and supported_models and model_name not in supported_models:
		raise ValueError(
		f"Invalid model '{model_name}' for provider '{provider}'. "
		f"Supported models: {supported_models}"
		)

		try:
		# Create config instance
		model_config = config_cls(**config)

		@@ -51,7 +64,7 @@ class ModelFactory:

		except Exception as e:
		logger.error(f"Failed to initialize model interface: {str(e)}")
		raise
		raise ValueError(f"Failed to initialize model interface: {str(e)}")

		@staticmethod
		def create_from_config_instance(config: BaseModelConfig) -> BaseModelInterface: