mirror of
https://github.com/FoundationAgents/MetaGPT.git
synced 2026-06-08 15:05:17 +02:00
Fixed max tokens; Corrected typo
This commit is contained in:
JGalego
parent
58a5720f01
commit
ddb5409118
3 changed files with 51 additions and 51 deletions
|
|
@ -1,7 +1,7 @@
|
|||
from metagpt.logs import logger
|
||||
|
||||
# max_tokens for each model
|
||||
NOT_SUUPORT_STREAM_MODELS = {
|
||||
NOT_SUPPORT_STREAM_MODELS = {
|
||||
# Jurassic-2 Mid-v1 and Ultra-v1
|
||||
# + Legacy date: 2024-04-30 (us-west-2/Oregon)
|
||||
# + EOL date: 2024-08-31 (us-west-2/Oregon)
|
||||
|
|
@ -11,72 +11,72 @@ NOT_SUUPORT_STREAM_MODELS = {
|
|||
|
||||
SUPPORT_STREAM_MODELS = {
|
||||
# Jamba-Instruct
|
||||
"ai21.jamba-instruct-v1:0": 256000,
|
||||
"ai21.jamba-instruct-v1:0": 4096,
|
||||
# Titan Text G1 - Express
|
||||
"amazon.titan-text-express-v1": 8192,
|
||||
"amazon.titan-text-express-v1:0:8k": 8192,
|
||||
# Titan Text G1 - Lite
|
||||
"amazon.titan-text-lite-v1:0:4k": 4096,
|
||||
"amazon.titan-text-lite-v1": 4096,
|
||||
# Titan Text Premier
|
||||
"amazon.titan-text-premier-v1:0": 3072,
|
||||
"amazon.titan-text-premier-v1:0:32k": 3072,
|
||||
# Titan Text Large
|
||||
"amazon.titan-tg1-large": 8000,
|
||||
# Titan Text G1 - Express
|
||||
"amazon.titan-text-express-v1": 8000,
|
||||
"amazon.titan-text-express-v1:0:8k": 8000,
|
||||
# Titan Text G1 - Lite
|
||||
"amazon.titan-text-lite-v1:0:4k": 4000,
|
||||
"amazon.titan-text-lite-v1": 4000,
|
||||
# Titan Text Premier
|
||||
"amazon.titan-text-premier-v1:0": 32000,
|
||||
"amazon.titan-text-premier-v1:0:32k": 32000,
|
||||
# Claude Instant v1
|
||||
"anthropic.claude-instant-v1": 100000,
|
||||
"anthropic.claude-instant-v1:2:100k": 100000,
|
||||
"anthropic.claude-instant-v1": 4096,
|
||||
"anthropic.claude-instant-v1:2:100k": 4096,
|
||||
# Claude v2
|
||||
"anthropic.claude-v2": 100000,
|
||||
"anthropic.claude-v2:0:18k": 18000,
|
||||
"anthropic.claude-v2:0:100k": 100000,
|
||||
"anthropic.claude-v2": 4096,
|
||||
"anthropic.claude-v2:0:18k": 4096,
|
||||
"anthropic.claude-v2:0:100k": 4096,
|
||||
# Claude v2.1
|
||||
"anthropic.claude-v2:1": 200000,
|
||||
"anthropic.claude-v2:1:18k": 18000,
|
||||
"anthropic.claude-v2:1:200k": 200000,
|
||||
"anthropic.claude-v2:1": 4096,
|
||||
"anthropic.claude-v2:1:18k": 4096,
|
||||
"anthropic.claude-v2:1:200k": 4096,
|
||||
# Claude 3 Sonnet
|
||||
"anthropic.claude-3-sonnet-20240229-v1:0": 200000,
|
||||
"anthropic.claude-3-sonnet-20240229-v1:0:28k": 28000,
|
||||
"anthropic.claude-3-sonnet-20240229-v1:0:200k": 200000,
|
||||
"anthropic.claude-3-sonnet-20240229-v1:0": 4096,
|
||||
"anthropic.claude-3-sonnet-20240229-v1:0:28k": 4096,
|
||||
"anthropic.claude-3-sonnet-20240229-v1:0:200k": 4096,
|
||||
# Claude 3 Haiku
|
||||
"anthropic.claude-3-haiku-20240307-v1:0": 200000,
|
||||
"anthropic.claude-3-haiku-20240307-v1:0:48k": 48000,
|
||||
"anthropic.claude-3-haiku-20240307-v1:0:200k": 200000,
|
||||
"anthropic.claude-3-haiku-20240307-v1:0": 4096,
|
||||
"anthropic.claude-3-haiku-20240307-v1:0:48k": 4096,
|
||||
"anthropic.claude-3-haiku-20240307-v1:0:200k": 4096,
|
||||
# Claude 3 Opus
|
||||
"anthropic.claude-3-opus-20240229-v1:0": 200000,
|
||||
"anthropic.claude-3-opus-20240229-v1:0": 4096,
|
||||
# Claude 3.5 Sonnet
|
||||
"anthropic.claude-3-5-sonnet-20240620-v1:0": 200000,
|
||||
"anthropic.claude-3-5-sonnet-20240620-v1:0": 8192,
|
||||
# Command Text
|
||||
"cohere.command-text-v14": 4000,
|
||||
"cohere.command-text-v14:7:4k": 4000,
|
||||
"cohere.command-text-v14": 4096,
|
||||
"cohere.command-text-v14:7:4k": 4096,
|
||||
# Command Light Text
|
||||
"cohere.command-light-text-v14": 4000,
|
||||
"cohere.command-light-text-v14:7:4k": 4000,
|
||||
"cohere.command-light-text-v14": 4096,
|
||||
"cohere.command-light-text-v14:7:4k": 4096,
|
||||
# Command R
|
||||
"cohere.command-r-v1:0": 128000,
|
||||
"cohere.command-r-v1:0": 4096,
|
||||
# Command R+
|
||||
"cohere.command-r-plus-v1:0": 128000,
|
||||
"cohere.command-r-plus-v1:0": 4096,
|
||||
# Llama 2 (--> Llama 3/3.1) !!!
|
||||
# + Legacy: 2024-05-12
|
||||
# + EOL: 2024-10-30
|
||||
"meta.llama2-13b-chat-v1": 2000,
|
||||
"meta.llama2-13b-chat-v1:0:4k": 4000,
|
||||
"meta.llama2-70b-v1": 4000,
|
||||
"meta.llama2-70b-v1:0:4k": 4000,
|
||||
"meta.llama2-70b-chat-v1": 2000,
|
||||
"meta.llama2-70b-chat-v1:0:4k": 2000,
|
||||
"meta.llama2-13b-chat-v1": 2048,
|
||||
"meta.llama2-13b-chat-v1:0:4k": 2048,
|
||||
"meta.llama2-70b-v1": 2048,
|
||||
"meta.llama2-70b-v1:0:4k": 2048,
|
||||
"meta.llama2-70b-chat-v1": 2048,
|
||||
"meta.llama2-70b-chat-v1:0:4k": 2048,
|
||||
# Llama 3 8B Instruct
|
||||
"meta.llama3-8b-instruct-v1:0": 2000,
|
||||
"meta.llama3-8b-instruct-v1:0": 2048,
|
||||
# Llama 3 70B Instruct
|
||||
"meta.llama3-70b-instruct-v1:0": 2000,
|
||||
"meta.llama3-70b-instruct-v1:0": 2048,
|
||||
# Mistral 7B Instruct
|
||||
"mistral.mistral-7b-instruct-v0:2": 32000,
|
||||
"mistral.mistral-7b-instruct-v0:2": 8192,
|
||||
# Mixtral 8x7B Instruct
|
||||
"mistral.mixtral-8x7b-instruct-v0:1": 32000,
|
||||
"mistral.mixtral-8x7b-instruct-v0:1": 8192,
|
||||
# Mistral Large (24.02)
|
||||
"mistral.mistral-large-2402-v1:0": 32000,
|
||||
"mistral.mistral-large-2402-v1:0": 8192,
|
||||
# Mistral Large 2 (24.07)
|
||||
"mistral.mistral-large-2407-v1:0": 128000
|
||||
"mistral.mistral-large-2407-v1:0": 8192
|
||||
}
|
||||
|
||||
# TODO:use a more general function for constructing chat templates.
|
||||
|
|
@ -136,7 +136,7 @@ def messages_to_prompt_claude2(messages: list[dict]) -> str:
|
|||
|
||||
def get_max_tokens(model_id: str) -> int:
|
||||
try:
|
||||
max_tokens = (NOT_SUUPORT_STREAM_MODELS | SUPPORT_STREAM_MODELS)[model_id]
|
||||
max_tokens = (NOT_SUPPORT_STREAM_MODELS | SUPPORT_STREAM_MODELS)[model_id]
|
||||
except KeyError:
|
||||
logger.warning(f"Couldn't find model:{model_id} , max tokens has been set to 2048")
|
||||
max_tokens = 2048
|
||||
|
|
|
|||
|
|
@ -12,7 +12,7 @@ from metagpt.const import USE_CONFIG_TIMEOUT
|
|||
from metagpt.logs import log_llm_stream, logger
|
||||
from metagpt.provider.base_llm import BaseLLM
|
||||
from metagpt.provider.bedrock.bedrock_provider import get_provider
|
||||
from metagpt.provider.bedrock.utils import NOT_SUUPORT_STREAM_MODELS, get_max_tokens
|
||||
from metagpt.provider.bedrock.utils import NOT_SUPPORT_STREAM_MODELS, get_max_tokens
|
||||
from metagpt.provider.llm_provider_registry import register_provider
|
||||
from metagpt.utils.cost_manager import CostManager
|
||||
from metagpt.utils.token_counter import BEDROCK_TOKEN_COSTS
|
||||
|
|
@ -25,7 +25,7 @@ class BedrockLLM(BaseLLM):
|
|||
self.__client = self.__init_client("bedrock-runtime")
|
||||
self.__provider = get_provider(self.config.model)
|
||||
self.cost_manager = CostManager(token_costs=BEDROCK_TOKEN_COSTS)
|
||||
if self.config.model in NOT_SUUPORT_STREAM_MODELS:
|
||||
if self.config.model in NOT_SUPPORT_STREAM_MODELS:
|
||||
logger.warning(f"model {self.config.model} doesn't support streaming output!")
|
||||
|
||||
def __init_client(self, service_name: Literal["bedrock-runtime", "bedrock"]):
|
||||
|
|
@ -113,7 +113,7 @@ class BedrockLLM(BaseLLM):
|
|||
return await self.acompletion(messages)
|
||||
|
||||
async def _achat_completion_stream(self, messages: list[dict], timeout=USE_CONFIG_TIMEOUT) -> str:
|
||||
if self.config.model in NOT_SUUPORT_STREAM_MODELS:
|
||||
if self.config.model in NOT_SUPPORT_STREAM_MODELS:
|
||||
rsp = await self.acompletion(messages)
|
||||
full_text = self.get_choice_text(rsp)
|
||||
log_llm_stream(full_text)
|
||||
|
|
|
|||
|
|
@ -3,7 +3,7 @@ import json
|
|||
import pytest
|
||||
|
||||
from metagpt.provider.bedrock.utils import (
|
||||
NOT_SUUPORT_STREAM_MODELS,
|
||||
NOT_SUPPORT_STREAM_MODELS,
|
||||
SUPPORT_STREAM_MODELS,
|
||||
)
|
||||
from metagpt.provider.bedrock_api import BedrockLLM
|
||||
|
|
@ -14,7 +14,7 @@ from tests.metagpt.provider.req_resp_const import (
|
|||
)
|
||||
|
||||
# all available model from bedrock
|
||||
models = SUPPORT_STREAM_MODELS | NOT_SUUPORT_STREAM_MODELS
|
||||
models = SUPPORT_STREAM_MODELS | NOT_SUPPORT_STREAM_MODELS
|
||||
messages = [{"role": "user", "content": "Hi!"}]
|
||||
usage = {
|
||||
"prompt_tokens": 1000000,
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue