Spaces:
Running
on
CPU Upgrade
Running
on
CPU Upgrade
File size: 2,356 Bytes
c2b7f1b d2805fc ed6c708 c2b7f1b ed6c708 b9d391c c2b7f1b ed6c708 c2b7f1b d2805fc da60a9e d2805fc 97bea1c d2805fc c2b7f1b ed6c708 c2b7f1b 8695aa8 c2b7f1b 8695aa8 c2b7f1b |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 |
"""
Central configuration for models and providers
This file centralizes all configurations related to models and providers used in the application.
"""
# Definition of preferred providers, used in get_available_model_provider.py
# PREFERRED_PROVIDERS = ["sambanova", "novita"]
# increased number of providers
PREFERRED_PROVIDERS = ["fireworks-ai", "cerebras", "sambanova", "together", "nebius", "novita", "hyperbolic", "cohere", "hf-inference"]
# Default models to evaluate for evaluation
DEFAULT_EVALUATION_MODELS = [
"Qwen/Qwen3-30B-A3B",
"Qwen/Qwen3-235B-A22B",
"meta-llama/Llama-4-Maverick-17B-128E-Instruct",
"meta-llama/Llama-4-Scout-17B-16E-Instruct",
"deepseek-ai/DeepSeek-V3-0324",
"google/gemma-3-27b-it",
"mistralai/Mistral-Small-3.1-24B-Instruct-2503"
]
# DEFAULT_EVALUATION_MODELS = [
# "Qwen/QwQ-32B",
# "Qwen/Qwen2.5-72B-Instruct",
# "Qwen/Qwen2.5-32B-Instruct",
# "meta-llama/Llama-3.1-8B-Instruct",
# "meta-llama/Llama-3.3-70B-Instruct",
# "deepseek-ai/DeepSeek-R1-Distill-Llama-70B",
# "mistralai/Mistral-Small-24B-Instruct-2501",
# ]
# Modèles alternatifs à utiliser si le modèle par défaut n'est pas disponible
ALTERNATIVE_BENCHMARK_MODELS = [
"deepseek-ai/DeepSeek-R1-Distill-Llama-70B",
"meta-llama/Llama-3.3-70B-Instruct",
"meta-llama/Llama-3.1-8B-Instruct",
"Qwen/Qwen2.5-72B-Instruct",
"mistralai/Mistral-Small-24B-Instruct-2501",
# Modèles open-source qui peuvent fonctionner sans authentification
"HuggingFaceH4/zephyr-7b-beta",
"TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"microsoft/phi-2",
]
# Required model for create_bench_config_file.py (only one default model)
DEFAULT_BENCHMARK_MODEL = "Qwen/Qwen3-30B-A3B"
# DEFAULT_BENCHMARK_MODEL = "Qwen/Qwen2.5-32B-Instruct"
# Models by roles for benchmark configuration
# All roles use the default model except chunking
BENCHMARK_MODEL_ROLES = {
"ingestion": [DEFAULT_BENCHMARK_MODEL],
"summarization": [DEFAULT_BENCHMARK_MODEL],
"chunking": ["intfloat/multilingual-e5-large-instruct"],
"single_shot_question_generation": [DEFAULT_BENCHMARK_MODEL],
"multi_hop_question_generation": [DEFAULT_BENCHMARK_MODEL],
}
# Default evaluation timeout (in seconds)
DEFAULT_EVALUATION_TIMEOUT = 60.0
# Default benchmark timeout (in seconds)
DEFAULT_BENCHMARK_TIMEOUT = 300.0 |