summaryrefslogtreecommitdiffstats
path: root/g4f/Provider/ReplicateHome.py
diff options
context:
space:
mode:
Diffstat (limited to 'g4f/Provider/ReplicateHome.py')
-rw-r--r--g4f/Provider/ReplicateHome.py227
1 files changed, 117 insertions, 110 deletions
diff --git a/g4f/Provider/ReplicateHome.py b/g4f/Provider/ReplicateHome.py
index 48336831..7f443a7d 100644
--- a/g4f/Provider/ReplicateHome.py
+++ b/g4f/Provider/ReplicateHome.py
@@ -1,136 +1,143 @@
from __future__ import annotations
-from typing import Generator, Optional, Dict, Any, Union, List
-import random
+
+import json
import asyncio
-import base64
+from aiohttp import ClientSession, ContentTypeError
-from .base_provider import AsyncGeneratorProvider, ProviderModelMixin
from ..typing import AsyncResult, Messages
-from ..requests import StreamSession, raise_for_status
-from ..errors import ResponseError
+from .base_provider import AsyncGeneratorProvider, ProviderModelMixin
+from .helper import format_prompt
from ..image import ImageResponse
class ReplicateHome(AsyncGeneratorProvider, ProviderModelMixin):
url = "https://replicate.com"
- parent = "Replicate"
+ api_endpoint = "https://homepage.replicate.com/api/prediction"
working = True
- default_model = 'stability-ai/sdxl'
- models = [
- # image
- 'stability-ai/sdxl',
- 'ai-forever/kandinsky-2.2',
-
- # text
- 'meta/llama-2-70b-chat',
- 'mistralai/mistral-7b-instruct-v0.2'
+ supports_stream = True
+ supports_system_message = True
+ supports_message_history = True
+
+ default_model = 'meta/meta-llama-3-70b-instruct'
+
+ text_models = [
+ 'meta/meta-llama-3-70b-instruct',
+ 'mistralai/mixtral-8x7b-instruct-v0.1',
+ 'google-deepmind/gemma-2b-it',
+ 'yorickvp/llava-13b',
]
- versions = {
- # image
- 'stability-ai/sdxl': [
- "39ed52f2a78e934b3ba6e2a89f5b1c712de7dfea535525255b1aa35c5565e08b",
- "2b017d9b67edd2ee1401238df49d75da53c523f36e363881e057f5dc3ed3c5b2",
- "7762fd07cf82c948538e41f63f77d685e02b063e37e496e96eefd46c929f9bdc"
- ],
- 'ai-forever/kandinsky-2.2': [
- "ad9d7879fbffa2874e1d909d1d37d9bc682889cc65b31f7bb00d2362619f194a"
- ],
+ image_models = [
+ 'black-forest-labs/flux-schnell',
+ 'stability-ai/stable-diffusion-3',
+ 'bytedance/sdxl-lightning-4step',
+ 'playgroundai/playground-v2.5-1024px-aesthetic',
+ ]
-
- # Text
- 'meta/llama-2-70b-chat': [
- "dp-542693885b1777c98ef8c5a98f2005e7"
- ],
- 'mistralai/mistral-7b-instruct-v0.2': [
- "dp-89e00f489d498885048e94f9809fbc76"
- ]
+ models = text_models + image_models
+
+ model_aliases = {
+ "flux-schnell": "black-forest-labs/flux-schnell",
+ "sd-3": "stability-ai/stable-diffusion-3",
+ "sdxl": "bytedance/sdxl-lightning-4step",
+ "playground-v2.5": "playgroundai/playground-v2.5-1024px-aesthetic",
+ "llama-3-70b": "meta/meta-llama-3-70b-instruct",
+ "mixtral-8x7b": "mistralai/mixtral-8x7b-instruct-v0.1",
+ "gemma-2b": "google-deepmind/gemma-2b-it",
+ "llava-13b": "yorickvp/llava-13b",
}
- image_models = {"stability-ai/sdxl", "ai-forever/kandinsky-2.2"}
- text_models = {"meta/llama-2-70b-chat", "mistralai/mistral-7b-instruct-v0.2"}
+ model_versions = {
+ "meta/meta-llama-3-70b-instruct": "fbfb20b472b2f3bdd101412a9f70a0ed4fc0ced78a77ff00970ee7a2383c575d",
+ "mistralai/mixtral-8x7b-instruct-v0.1": "5d78bcd7a992c4b793465bcdcf551dc2ab9668d12bb7aa714557a21c1e77041c",
+ "google-deepmind/gemma-2b-it": "dff94eaf770e1fc211e425a50b51baa8e4cac6c39ef074681f9e39d778773626",
+ "yorickvp/llava-13b": "80537f9eead1a5bfa72d5ac6ea6414379be41d4d4f6679fd776e9535d1eb58bb",
+ 'black-forest-labs/flux-schnell': "f2ab8a5bfe79f02f0789a146cf5e73d2a4ff2684a98c2b303d1e1ff3814271db",
+ 'stability-ai/stable-diffusion-3': "527d2a6296facb8e47ba1eaf17f142c240c19a30894f437feee9b91cc29d8e4f",
+ 'bytedance/sdxl-lightning-4step': "5f24084160c9089501c1b3545d9be3c27883ae2239b6f412990e82d4a6210f8f",
+ 'playgroundai/playground-v2.5-1024px-aesthetic': "a45f82a1382bed5c7aeb861dac7c7d191b0fdf74d8d57c4a0e6ed7d4d0bf7d24",
+ }
@classmethod
- async def create_async_generator(
- cls,
- model: str,
- messages: Messages,
- **kwargs: Any
- ) -> Generator[Union[str, ImageResponse], None, None]:
- yield await cls.create_async(messages[-1]["content"], model, **kwargs)
+ def get_model(cls, model: str) -> str:
+ if model in cls.models:
+ return model
+ elif model in cls.model_aliases:
+ return cls.model_aliases[model]
+ else:
+ return cls.default_model
@classmethod
- async def create_async(
+ async def create_async_generator(
cls,
- prompt: str,
model: str,
- api_key: Optional[str] = None,
- proxy: Optional[str] = None,
- timeout: int = 180,
- version: Optional[str] = None,
- extra_data: Dict[str, Any] = {},
- **kwargs: Any
- ) -> Union[str, ImageResponse]:
+ messages: Messages,
+ proxy: str = None,
+ **kwargs
+ ) -> AsyncResult:
+ model = cls.get_model(model)
+
headers = {
- 'Accept-Encoding': 'gzip, deflate, br',
- 'Accept-Language': 'en-US',
- 'Connection': 'keep-alive',
- 'Origin': cls.url,
- 'Referer': f'{cls.url}/',
- 'Sec-Fetch-Dest': 'empty',
- 'Sec-Fetch-Mode': 'cors',
- 'Sec-Fetch-Site': 'same-site',
- 'User-Agent': 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/119.0.0.0 Safari/537.36',
- 'sec-ch-ua': '"Google Chrome";v="119", "Chromium";v="119", "Not?A_Brand";v="24"',
- 'sec-ch-ua-mobile': '?0',
- 'sec-ch-ua-platform': '"macOS"',
+ "accept": "*/*",
+ "accept-language": "en-US,en;q=0.9",
+ "cache-control": "no-cache",
+ "content-type": "application/json",
+ "origin": "https://replicate.com",
+ "pragma": "no-cache",
+ "priority": "u=1, i",
+ "referer": "https://replicate.com/",
+ "sec-ch-ua": '"Not;A=Brand";v="24", "Chromium";v="128"',
+ "sec-ch-ua-mobile": "?0",
+ "sec-ch-ua-platform": '"Linux"',
+ "sec-fetch-dest": "empty",
+ "sec-fetch-mode": "cors",
+ "sec-fetch-site": "same-site",
+ "user-agent": "Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/128.0.0.0 Safari/537.36"
}
-
- if version is None:
- version = random.choice(cls.versions.get(model, []))
- if api_key is not None:
- headers["Authorization"] = f"Bearer {api_key}"
-
- async with StreamSession(
- proxies={"all": proxy},
- headers=headers,
- timeout=timeout
- ) as session:
+
+ async with ClientSession(headers=headers) as session:
+ if model in cls.image_models:
+ prompt = messages[-1]['content'] if messages else ""
+ else:
+ prompt = format_prompt(messages)
+
data = {
- "input": {
- "prompt": prompt,
- **extra_data
- },
- "version": version
+ "model": model,
+ "version": cls.model_versions[model],
+ "input": {"prompt": prompt},
}
- if api_key is None:
- data["model"] = cls.get_model(model)
- url = "https://homepage.replicate.com/api/prediction"
- else:
- url = "https://api.replicate.com/v1/predictions"
- async with session.post(url, json=data) as response:
- await raise_for_status(response)
+
+ async with session.post(cls.api_endpoint, json=data, proxy=proxy) as response:
+ response.raise_for_status()
result = await response.json()
- if "id" not in result:
- raise ResponseError(f"Invalid response: {result}")
+ prediction_id = result['id']
+
+ poll_url = f"https://homepage.replicate.com/api/poll?id={prediction_id}"
+ max_attempts = 30
+ delay = 5
+ for _ in range(max_attempts):
+ async with session.get(poll_url, proxy=proxy) as response:
+ response.raise_for_status()
+ try:
+ result = await response.json()
+ except ContentTypeError:
+ text = await response.text()
+ try:
+ result = json.loads(text)
+ except json.JSONDecodeError:
+ raise ValueError(f"Unexpected response format: {text}")
- while True:
- if api_key is None:
- url = f"https://homepage.replicate.com/api/poll?id={result['id']}"
- else:
- url = f"https://api.replicate.com/v1/predictions/{result['id']}"
- async with session.get(url) as response:
- await raise_for_status(response)
- result = await response.json()
- if "status" not in result:
- raise ResponseError(f"Invalid response: {result}")
- if result["status"] == "succeeded":
- output = result['output']
- if model in cls.text_models:
- return ''.join(output) if isinstance(output, list) else output
- elif model in cls.image_models:
- images: List[Any] = output
- images = images[0] if len(images) == 1 else images
- return ImageResponse(images, prompt)
- elif result["status"] == "failed":
- raise ResponseError(f"Prediction failed: {result}")
- await asyncio.sleep(0.5)
+ if result['status'] == 'succeeded':
+ if model in cls.image_models:
+ image_url = result['output'][0]
+ yield ImageResponse(image_url, "Generated image")
+ return
+ else:
+ for chunk in result['output']:
+ yield chunk
+ break
+ elif result['status'] == 'failed':
+ raise Exception(f"Prediction failed: {result.get('error')}")
+ await asyncio.sleep(delay)
+
+ if result['status'] != 'succeeded':
+ raise Exception("Prediction timed out")
generated by cgit v1.2.3 (git 2.25.1) at 2024-09-17 17:17:00 +0200