2023-09-29 17:54:46 +03:00
|
|
|
from __future__ import annotations
|
2023-09-18 00:23:54 +03:00
|
|
|
from dataclasses import dataclass
|
2023-09-29 17:54:46 +03:00
|
|
|
from .typing import Union
|
|
|
|
from .Provider import BaseProvider, RetryProvider
|
|
|
|
from .Provider import (
|
2023-09-18 00:37:14 +03:00
|
|
|
ChatgptLogin,
|
2023-09-29 17:54:46 +03:00
|
|
|
ChatgptAi,
|
|
|
|
ChatBase,
|
|
|
|
Vercel,
|
|
|
|
DeepAi,
|
|
|
|
Aivvm,
|
|
|
|
Bard,
|
2023-09-21 21:10:59 +03:00
|
|
|
H2o,
|
|
|
|
GptGo,
|
|
|
|
Bing,
|
|
|
|
PerplexityAi,
|
|
|
|
Wewordle,
|
|
|
|
Yqcloud,
|
|
|
|
AItianhu,
|
2023-09-26 02:02:02 +03:00
|
|
|
AItianhuSpace,
|
2023-09-21 21:10:59 +03:00
|
|
|
Aichat,
|
2023-09-25 16:52:19 +03:00
|
|
|
Myshell,
|
2023-09-26 11:03:37 +03:00
|
|
|
Aibn,
|
|
|
|
ChatgptDuo,
|
2023-09-18 00:37:14 +03:00
|
|
|
)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
2023-09-29 17:54:46 +03:00
|
|
|
|
2023-09-20 07:12:34 +03:00
|
|
|
@dataclass(unsafe_hash=True)
|
2023-09-18 00:23:54 +03:00
|
|
|
class Model:
|
|
|
|
name: str
|
|
|
|
base_provider: str
|
2023-09-23 16:35:17 +03:00
|
|
|
best_provider: Union[type[BaseProvider], RetryProvider] = None
|
2023-09-18 00:23:54 +03:00
|
|
|
|
2023-09-29 17:54:46 +03:00
|
|
|
|
2023-09-18 00:23:54 +03:00
|
|
|
# Config for HuggingChat, OpenAssistant
|
|
|
|
# Works for Liaobots, H2o, OpenaiChat, Yqcloud, You
|
|
|
|
default = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name="",
|
|
|
|
base_provider="",
|
|
|
|
best_provider=RetryProvider([
|
|
|
|
Bing, # Not fully GPT 3 or 4
|
|
|
|
PerplexityAi, # Adds references to sources
|
|
|
|
Wewordle, # Responds with markdown
|
|
|
|
Yqcloud, # Answers short questions in chinese
|
|
|
|
ChatBase, # Don't want to answer creatively
|
|
|
|
ChatgptDuo, # Include search results
|
2023-09-26 11:03:37 +03:00
|
|
|
DeepAi, ChatgptLogin, ChatgptAi, Aivvm, GptGo, AItianhu, AItianhuSpace, Aichat, Myshell, Aibn,
|
2023-09-21 21:10:59 +03:00
|
|
|
])
|
|
|
|
)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
# GPT-3.5 / GPT-4
|
|
|
|
gpt_35_turbo = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='gpt-3.5-turbo',
|
|
|
|
base_provider='openai',
|
|
|
|
best_provider=RetryProvider([
|
2023-09-26 11:03:37 +03:00
|
|
|
DeepAi, ChatgptLogin, ChatgptAi, Aivvm, GptGo, AItianhu, Aichat, AItianhuSpace, Myshell, Aibn,
|
2023-09-21 21:10:59 +03:00
|
|
|
])
|
2023-09-18 00:23:54 +03:00
|
|
|
)
|
|
|
|
|
|
|
|
gpt_4 = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='gpt-4',
|
|
|
|
base_provider='openai',
|
|
|
|
best_provider=RetryProvider([
|
2023-09-29 17:21:18 +03:00
|
|
|
Myshell, AItianhuSpace, Aivvm
|
2023-09-23 02:44:09 +03:00
|
|
|
])
|
|
|
|
)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
# Bard
|
|
|
|
palm = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='palm',
|
|
|
|
base_provider='google',
|
|
|
|
best_provider=Bard)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
# H2o
|
|
|
|
falcon_7b = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='h2oai/h2ogpt-gm-oasst1-en-2048-falcon-7b-v3',
|
|
|
|
base_provider='huggingface',
|
|
|
|
best_provider=H2o)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
falcon_40b = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='h2oai/h2ogpt-gm-oasst1-en-2048-falcon-40b-v1',
|
|
|
|
base_provider='huggingface',
|
|
|
|
best_provider=H2o)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
llama_13b = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='h2oai/h2ogpt-gm-oasst1-en-2048-open-llama-13b',
|
|
|
|
base_provider='huggingface',
|
|
|
|
best_provider=H2o)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
# Vercel
|
|
|
|
claude_instant_v1 = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='claude-instant-v1',
|
|
|
|
base_provider='anthropic',
|
|
|
|
best_provider=Vercel)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
claude_v1 = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='claude-v1',
|
|
|
|
base_provider='anthropic',
|
|
|
|
best_provider=Vercel)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
claude_v2 = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='claude-v2',
|
|
|
|
base_provider='anthropic',
|
|
|
|
best_provider=Vercel)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
command_light_nightly = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='command-light-nightly',
|
|
|
|
base_provider='cohere',
|
|
|
|
best_provider=Vercel)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
command_nightly = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='command-nightly',
|
|
|
|
base_provider='cohere',
|
|
|
|
best_provider=Vercel)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
gpt_neox_20b = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='EleutherAI/gpt-neox-20b',
|
|
|
|
base_provider='huggingface',
|
|
|
|
best_provider=Vercel)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
oasst_sft_1_pythia_12b = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='OpenAssistant/oasst-sft-1-pythia-12b',
|
|
|
|
base_provider='huggingface',
|
|
|
|
best_provider=Vercel)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
oasst_sft_4_pythia_12b_epoch_35 = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='OpenAssistant/oasst-sft-4-pythia-12b-epoch-3.5',
|
|
|
|
base_provider='huggingface',
|
|
|
|
best_provider=Vercel)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
santacoder = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='bigcode/santacoder',
|
|
|
|
base_provider='huggingface',
|
|
|
|
best_provider=Vercel)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
bloom = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='bigscience/bloom',
|
|
|
|
base_provider='huggingface',
|
|
|
|
best_provider=Vercel)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
flan_t5_xxl = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='google/flan-t5-xxl',
|
|
|
|
base_provider='huggingface',
|
|
|
|
best_provider=Vercel)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
code_davinci_002 = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='code-davinci-002',
|
|
|
|
base_provider='openai',
|
|
|
|
best_provider=Vercel)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
gpt_35_turbo_16k = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='gpt-3.5-turbo-16k',
|
|
|
|
base_provider='openai',
|
|
|
|
best_provider=Aivvm)
|
2023-09-20 06:00:19 +03:00
|
|
|
|
2023-09-18 00:23:54 +03:00
|
|
|
gpt_35_turbo_16k_0613 = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='gpt-3.5-turbo-16k-0613',
|
|
|
|
base_provider='openai')
|
2023-09-22 22:36:44 +03:00
|
|
|
|
|
|
|
gpt_35_turbo_0613 = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='gpt-3.5-turbo-0613',
|
|
|
|
base_provider='openai',
|
|
|
|
best_provider=RetryProvider([
|
2023-09-25 16:52:19 +03:00
|
|
|
Aivvm, ChatgptLogin
|
|
|
|
])
|
|
|
|
)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
gpt_4_0613 = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='gpt-4-0613',
|
|
|
|
base_provider='openai',
|
|
|
|
best_provider=Aivvm)
|
2023-09-29 17:21:18 +03:00
|
|
|
|
|
|
|
gpt_4_32k = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='gpt-4-32k',
|
|
|
|
base_provider='openai',
|
|
|
|
best_provider=Aivvm)
|
2023-09-29 17:21:18 +03:00
|
|
|
|
|
|
|
gpt_4_32k_0613 = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='gpt-4-32k-0613',
|
|
|
|
base_provider='openai',
|
|
|
|
best_provider=Aivvm)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
text_ada_001 = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='text-ada-001',
|
|
|
|
base_provider='openai',
|
|
|
|
best_provider=Vercel)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
text_babbage_001 = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='text-babbage-001',
|
|
|
|
base_provider='openai',
|
|
|
|
best_provider=Vercel)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
text_curie_001 = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='text-curie-001',
|
|
|
|
base_provider='openai',
|
|
|
|
best_provider=Vercel)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
text_davinci_002 = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='text-davinci-002',
|
|
|
|
base_provider='openai',
|
|
|
|
best_provider=Vercel)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
text_davinci_003 = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='text-davinci-003',
|
|
|
|
base_provider='openai',
|
|
|
|
best_provider=Vercel)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
llama13b_v2_chat = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='replicate:a16z-infra/llama13b-v2-chat',
|
|
|
|
base_provider='replicate',
|
|
|
|
best_provider=Vercel)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
llama7b_v2_chat = Model(
|
2023-09-29 17:54:46 +03:00
|
|
|
name='replicate:a16z-infra/llama7b-v2-chat',
|
|
|
|
base_provider='replicate',
|
|
|
|
best_provider=Vercel)
|
2023-09-18 00:23:54 +03:00
|
|
|
|
|
|
|
|
|
|
|
class ModelUtils:
|
|
|
|
convert: dict[str, Model] = {
|
2023-09-23 03:31:16 +03:00
|
|
|
# gpt-3.5 / gpt-4
|
2023-09-29 17:54:46 +03:00
|
|
|
'gpt-3.5-turbo': gpt_35_turbo,
|
|
|
|
'gpt-3.5-turbo-16k': gpt_35_turbo_16k,
|
|
|
|
'gpt-3.5-turbo-16k-0613': gpt_35_turbo_16k_0613,
|
|
|
|
'gpt-4': gpt_4,
|
|
|
|
'gpt-4-0613': gpt_4_0613,
|
|
|
|
'gpt-4-32k': gpt_4_32k,
|
|
|
|
'gpt-4-32k-0613': gpt_4_32k_0613,
|
|
|
|
|
2023-09-18 00:23:54 +03:00
|
|
|
# Bard
|
2023-09-29 17:54:46 +03:00
|
|
|
'palm2': palm,
|
|
|
|
'palm': palm,
|
|
|
|
'google': palm,
|
|
|
|
'google-bard': palm,
|
|
|
|
'google-palm': palm,
|
|
|
|
'bard': palm,
|
|
|
|
|
2023-09-18 00:23:54 +03:00
|
|
|
# H2o
|
2023-09-29 17:54:46 +03:00
|
|
|
'falcon-40b': falcon_40b,
|
|
|
|
'falcon-7b': falcon_7b,
|
|
|
|
'llama-13b': llama_13b,
|
|
|
|
|
2023-09-18 00:23:54 +03:00
|
|
|
# Vercel
|
2023-09-29 17:54:46 +03:00
|
|
|
'claude-instant-v1': claude_instant_v1,
|
|
|
|
'claude-v1': claude_v1,
|
|
|
|
'claude-v2': claude_v2,
|
|
|
|
'command-nightly': command_nightly,
|
|
|
|
'gpt-neox-20b': gpt_neox_20b,
|
|
|
|
'santacoder': santacoder,
|
|
|
|
'bloom': bloom,
|
|
|
|
'flan-t5-xxl': flan_t5_xxl,
|
|
|
|
'code-davinci-002': code_davinci_002,
|
|
|
|
'text-ada-001': text_ada_001,
|
|
|
|
'text-babbage-001': text_babbage_001,
|
|
|
|
'text-curie-001': text_curie_001,
|
|
|
|
'text-davinci-002': text_davinci_002,
|
|
|
|
'text-davinci-003': text_davinci_003,
|
|
|
|
'llama13b-v2-chat': llama13b_v2_chat,
|
|
|
|
'llama7b-v2-chat': llama7b_v2_chat,
|
|
|
|
|
|
|
|
'oasst-sft-1-pythia-12b': oasst_sft_1_pythia_12b,
|
|
|
|
'oasst-sft-4-pythia-12b-epoch-3.5': oasst_sft_4_pythia_12b_epoch_35,
|
|
|
|
'command-light-nightly': command_light_nightly,
|
2023-09-22 22:36:44 +03:00
|
|
|
}
|