2023-03-07 10:52:22 -07:00
|
|
|
import json
|
|
|
|
import requests
|
2024-05-15 07:40:07 -06:00
|
|
|
import warnings
|
2023-03-07 10:52:22 -07:00
|
|
|
|
|
|
|
from aiohttp import ClientSession, ClientTimeout
|
|
|
|
from pydantic import ValidationError
|
2024-02-28 03:10:27 -07:00
|
|
|
from typing import Dict, Optional, List, AsyncIterator, Iterator, Union
|
2023-03-07 10:52:22 -07:00
|
|
|
|
2024-05-15 07:40:07 -06:00
|
|
|
from text_generation import DEPRECATION_WARNING
|
2023-03-07 10:52:22 -07:00
|
|
|
from text_generation.types import (
|
|
|
|
StreamResponse,
|
|
|
|
Response,
|
|
|
|
Request,
|
|
|
|
Parameters,
|
2024-02-15 02:28:10 -07:00
|
|
|
Grammar,
|
2024-05-23 07:37:09 -06:00
|
|
|
CompletionRequest,
|
|
|
|
Completion,
|
|
|
|
CompletionComplete,
|
2024-02-28 03:10:27 -07:00
|
|
|
ChatRequest,
|
|
|
|
ChatCompletionChunk,
|
|
|
|
ChatComplete,
|
|
|
|
Message,
|
|
|
|
Tool,
|
2023-03-07 10:52:22 -07:00
|
|
|
)
|
|
|
|
from text_generation.errors import parse_error
|
|
|
|
|
2024-05-15 07:40:07 -06:00
|
|
|
# emit deprecation warnings
|
|
|
|
warnings.simplefilter("always", DeprecationWarning)
|
|
|
|
|
2023-03-07 10:52:22 -07:00
|
|
|
|
|
|
|
class Client:
|
|
|
|
"""Client to make calls to a text-generation-inference instance
|
|
|
|
|
|
|
|
Example:
|
|
|
|
|
|
|
|
```python
|
|
|
|
>>> from text_generation import Client
|
|
|
|
|
|
|
|
>>> client = Client("https://api-inference.huggingface.co/models/bigscience/bloomz")
|
|
|
|
>>> client.generate("Why is the sky blue?").generated_text
|
|
|
|
' Rayleigh scattering'
|
|
|
|
|
|
|
|
>>> result = ""
|
|
|
|
>>> for response in client.generate_stream("Why is the sky blue?"):
|
|
|
|
>>> if not response.token.special:
|
|
|
|
>>> result += response.token.text
|
|
|
|
>>> result
|
|
|
|
' Rayleigh scattering'
|
|
|
|
```
|
|
|
|
"""
|
|
|
|
|
|
|
|
def __init__(
|
2023-03-23 11:01:01 -06:00
|
|
|
self,
|
|
|
|
base_url: str,
|
|
|
|
headers: Optional[Dict[str, str]] = None,
|
|
|
|
cookies: Optional[Dict[str, str]] = None,
|
|
|
|
timeout: int = 10,
|
2023-03-07 10:52:22 -07:00
|
|
|
):
|
|
|
|
"""
|
|
|
|
Args:
|
|
|
|
base_url (`str`):
|
|
|
|
text-generation-inference instance base url
|
|
|
|
headers (`Optional[Dict[str, str]]`):
|
|
|
|
Additional headers
|
2023-03-23 11:01:01 -06:00
|
|
|
cookies (`Optional[Dict[str, str]]`):
|
|
|
|
Cookies to include in the requests
|
2023-03-07 10:52:22 -07:00
|
|
|
timeout (`int`):
|
|
|
|
Timeout in seconds
|
|
|
|
"""
|
2024-05-15 07:40:07 -06:00
|
|
|
warnings.warn(DEPRECATION_WARNING, DeprecationWarning)
|
2023-03-07 10:52:22 -07:00
|
|
|
self.base_url = base_url
|
|
|
|
self.headers = headers
|
2023-03-23 11:01:01 -06:00
|
|
|
self.cookies = cookies
|
2023-03-07 10:52:22 -07:00
|
|
|
self.timeout = timeout
|
|
|
|
|
2024-05-23 07:37:09 -06:00
|
|
|
def completion(
|
|
|
|
self,
|
|
|
|
prompt: str,
|
|
|
|
frequency_penalty: Optional[float] = None,
|
|
|
|
max_tokens: Optional[int] = None,
|
|
|
|
repetition_penalty: Optional[float] = None,
|
|
|
|
seed: Optional[int] = None,
|
|
|
|
stream: bool = False,
|
|
|
|
temperature: Optional[float] = None,
|
|
|
|
top_p: Optional[float] = None,
|
|
|
|
stop: Optional[List[str]] = None,
|
|
|
|
):
|
|
|
|
"""
|
|
|
|
Given a prompt, generate a response synchronously
|
|
|
|
|
|
|
|
Args:
|
|
|
|
prompt (`str`):
|
|
|
|
Prompt
|
|
|
|
frequency_penalty (`float`):
|
|
|
|
The parameter for frequency penalty. 0.0 means no penalty
|
|
|
|
Penalize new tokens based on their existing frequency in the text so far,
|
|
|
|
decreasing the model's likelihood to repeat the same line verbatim.
|
|
|
|
max_tokens (`int`):
|
|
|
|
Maximum number of generated tokens
|
|
|
|
repetition_penalty (`float`):
|
|
|
|
The parameter for frequency penalty. 0.0 means no penalty. See [this
|
|
|
|
paper](https://arxiv.org/pdf/1909.05858.pdf) for more details.
|
|
|
|
seed (`int`):
|
|
|
|
Random sampling seed
|
|
|
|
stream (`bool`):
|
|
|
|
Stream the response
|
|
|
|
temperature (`float`):
|
|
|
|
The value used to module the logits distribution.
|
|
|
|
top_p (`float`):
|
|
|
|
If set to < 1, only the smallest set of most probable tokens with probabilities that add up to `top_p` or
|
|
|
|
higher are kept for generation
|
|
|
|
stop (`List[str]`):
|
|
|
|
Stop generating tokens if a member of `stop` is generated
|
|
|
|
"""
|
|
|
|
request = CompletionRequest(
|
|
|
|
model="tgi",
|
|
|
|
prompt=prompt,
|
|
|
|
frequency_penalty=frequency_penalty,
|
|
|
|
max_tokens=max_tokens,
|
|
|
|
repetition_penalty=repetition_penalty,
|
|
|
|
seed=seed,
|
|
|
|
stream=stream,
|
|
|
|
temperature=temperature,
|
|
|
|
top_p=top_p,
|
|
|
|
stop=stop,
|
|
|
|
)
|
|
|
|
if not stream:
|
|
|
|
resp = requests.post(
|
|
|
|
f"{self.base_url}/v1/completions",
|
|
|
|
json=request.dict(),
|
|
|
|
headers=self.headers,
|
|
|
|
cookies=self.cookies,
|
|
|
|
timeout=self.timeout,
|
|
|
|
)
|
|
|
|
payload = resp.json()
|
|
|
|
if resp.status_code != 200:
|
|
|
|
raise parse_error(resp.status_code, payload)
|
|
|
|
return Completion(**payload)
|
|
|
|
else:
|
|
|
|
return self._completion_stream_response(request)
|
|
|
|
|
|
|
|
def _completion_stream_response(self, request):
|
|
|
|
resp = requests.post(
|
|
|
|
f"{self.base_url}/v1/completions",
|
|
|
|
json=request.dict(),
|
|
|
|
headers=self.headers,
|
|
|
|
cookies=self.cookies,
|
|
|
|
timeout=self.timeout,
|
|
|
|
stream=True,
|
|
|
|
)
|
|
|
|
# iterate and print stream
|
|
|
|
for byte_payload in resp.iter_lines():
|
|
|
|
if byte_payload == b"\n":
|
|
|
|
continue
|
|
|
|
payload = byte_payload.decode("utf-8")
|
|
|
|
if payload.startswith("data:"):
|
|
|
|
json_payload = json.loads(payload.lstrip("data:").rstrip("\n"))
|
|
|
|
try:
|
|
|
|
response = CompletionComplete(**json_payload)
|
|
|
|
yield response
|
|
|
|
except ValidationError:
|
|
|
|
raise parse_error(resp.status, json_payload)
|
|
|
|
|
2024-02-28 03:10:27 -07:00
|
|
|
def chat(
|
|
|
|
self,
|
|
|
|
messages: List[Message],
|
2024-03-22 10:10:56 -06:00
|
|
|
repetition_penalty: Optional[float] = None,
|
2024-02-28 03:10:27 -07:00
|
|
|
frequency_penalty: Optional[float] = None,
|
|
|
|
logit_bias: Optional[List[float]] = None,
|
|
|
|
logprobs: Optional[bool] = None,
|
|
|
|
top_logprobs: Optional[int] = None,
|
|
|
|
max_tokens: Optional[int] = None,
|
|
|
|
n: Optional[int] = None,
|
|
|
|
presence_penalty: Optional[float] = None,
|
|
|
|
stream: bool = False,
|
|
|
|
seed: Optional[int] = None,
|
|
|
|
temperature: Optional[float] = None,
|
|
|
|
top_p: Optional[float] = None,
|
|
|
|
tools: Optional[List[Tool]] = None,
|
2024-04-30 03:07:17 -06:00
|
|
|
tool_prompt: Optional[str] = None,
|
2024-02-28 03:10:27 -07:00
|
|
|
tool_choice: Optional[str] = None,
|
2024-05-23 07:37:09 -06:00
|
|
|
stop: Optional[List[str]] = None,
|
2024-02-28 03:10:27 -07:00
|
|
|
):
|
|
|
|
"""
|
|
|
|
Given a list of messages, generate a response asynchronously
|
|
|
|
|
|
|
|
Args:
|
|
|
|
messages (`List[Message]`):
|
|
|
|
List of messages
|
2024-03-22 10:10:56 -06:00
|
|
|
repetition_penalty (`float`):
|
|
|
|
The parameter for repetition penalty. 0.0 means no penalty. See [this
|
2024-02-28 03:10:27 -07:00
|
|
|
paper](https://arxiv.org/pdf/1909.05858.pdf) for more details.
|
2024-03-22 10:10:56 -06:00
|
|
|
frequency_penalty (`float`):
|
|
|
|
The parameter for frequency penalty. 0.0 means no penalty
|
|
|
|
Penalize new tokens based on their existing frequency in the text so far,
|
|
|
|
decreasing the model's likelihood to repeat the same line verbatim.
|
2024-02-28 03:10:27 -07:00
|
|
|
logit_bias (`List[float]`):
|
|
|
|
Adjust the likelihood of specified tokens
|
|
|
|
logprobs (`bool`):
|
|
|
|
Include log probabilities in the response
|
|
|
|
top_logprobs (`int`):
|
|
|
|
Include the `n` most likely tokens at each step
|
|
|
|
max_tokens (`int`):
|
|
|
|
Maximum number of generated tokens
|
|
|
|
n (`int`):
|
|
|
|
Generate `n` completions
|
|
|
|
presence_penalty (`float`):
|
|
|
|
The parameter for presence penalty. 0.0 means no penalty. See [this
|
|
|
|
paper](https://arxiv.org/pdf/1909.05858.pdf) for more details.
|
|
|
|
stream (`bool`):
|
|
|
|
Stream the response
|
|
|
|
seed (`int`):
|
|
|
|
Random sampling seed
|
|
|
|
temperature (`float`):
|
|
|
|
The value used to module the logits distribution.
|
|
|
|
top_p (`float`):
|
|
|
|
If set to < 1, only the smallest set of most probable tokens with probabilities that add up to `top_p` or
|
|
|
|
higher are kept for generation
|
|
|
|
tools (`List[Tool]`):
|
|
|
|
List of tools to use
|
2024-04-30 03:07:17 -06:00
|
|
|
tool_prompt (`str`):
|
|
|
|
A prompt to be appended before the tools
|
2024-02-28 03:10:27 -07:00
|
|
|
tool_choice (`str`):
|
|
|
|
The tool to use
|
2024-05-23 07:37:09 -06:00
|
|
|
stop (`List[str]`):
|
|
|
|
Stop generating tokens if a member of `stop` is generated
|
2024-02-28 03:10:27 -07:00
|
|
|
|
|
|
|
"""
|
|
|
|
request = ChatRequest(
|
|
|
|
model="tgi",
|
|
|
|
messages=messages,
|
2024-03-22 10:10:56 -06:00
|
|
|
repetition_penalty=repetition_penalty,
|
2024-02-28 03:10:27 -07:00
|
|
|
frequency_penalty=frequency_penalty,
|
|
|
|
logit_bias=logit_bias,
|
|
|
|
logprobs=logprobs,
|
|
|
|
top_logprobs=top_logprobs,
|
|
|
|
max_tokens=max_tokens,
|
|
|
|
n=n,
|
|
|
|
presence_penalty=presence_penalty,
|
|
|
|
stream=stream,
|
|
|
|
seed=seed,
|
|
|
|
temperature=temperature,
|
|
|
|
top_p=top_p,
|
|
|
|
tools=tools,
|
2024-04-30 03:07:17 -06:00
|
|
|
tool_prompt=tool_prompt,
|
2024-02-28 03:10:27 -07:00
|
|
|
tool_choice=tool_choice,
|
2024-05-23 07:37:09 -06:00
|
|
|
stop=stop,
|
2024-02-28 03:10:27 -07:00
|
|
|
)
|
|
|
|
if not stream:
|
|
|
|
resp = requests.post(
|
|
|
|
f"{self.base_url}/v1/chat/completions",
|
|
|
|
json=request.dict(),
|
|
|
|
headers=self.headers,
|
|
|
|
cookies=self.cookies,
|
|
|
|
timeout=self.timeout,
|
|
|
|
)
|
|
|
|
payload = resp.json()
|
|
|
|
if resp.status_code != 200:
|
|
|
|
raise parse_error(resp.status_code, payload)
|
|
|
|
return ChatComplete(**payload)
|
|
|
|
else:
|
|
|
|
return self._chat_stream_response(request)
|
|
|
|
|
|
|
|
def _chat_stream_response(self, request):
|
|
|
|
resp = requests.post(
|
|
|
|
f"{self.base_url}/v1/chat/completions",
|
|
|
|
json=request.dict(),
|
|
|
|
headers=self.headers,
|
|
|
|
cookies=self.cookies,
|
|
|
|
timeout=self.timeout,
|
|
|
|
stream=True,
|
|
|
|
)
|
|
|
|
# iterate and print stream
|
|
|
|
for byte_payload in resp.iter_lines():
|
|
|
|
if byte_payload == b"\n":
|
|
|
|
continue
|
|
|
|
payload = byte_payload.decode("utf-8")
|
|
|
|
if payload.startswith("data:"):
|
|
|
|
json_payload = json.loads(payload.lstrip("data:").rstrip("\n"))
|
|
|
|
try:
|
|
|
|
response = ChatCompletionChunk(**json_payload)
|
|
|
|
yield response
|
|
|
|
except ValidationError:
|
|
|
|
raise parse_error(resp.status, json_payload)
|
|
|
|
|
2023-03-07 10:52:22 -07:00
|
|
|
def generate(
|
|
|
|
self,
|
|
|
|
prompt: str,
|
|
|
|
do_sample: bool = False,
|
2024-02-01 07:36:10 -07:00
|
|
|
max_new_tokens: int = 20,
|
2023-03-09 08:05:33 -07:00
|
|
|
best_of: Optional[int] = None,
|
2023-03-07 10:52:22 -07:00
|
|
|
repetition_penalty: Optional[float] = None,
|
2024-03-22 10:10:56 -06:00
|
|
|
frequency_penalty: Optional[float] = None,
|
2023-03-07 10:52:22 -07:00
|
|
|
return_full_text: bool = False,
|
|
|
|
seed: Optional[int] = None,
|
|
|
|
stop_sequences: Optional[List[str]] = None,
|
|
|
|
temperature: Optional[float] = None,
|
|
|
|
top_k: Optional[int] = None,
|
|
|
|
top_p: Optional[float] = None,
|
2023-03-09 08:05:33 -07:00
|
|
|
truncate: Optional[int] = None,
|
|
|
|
typical_p: Optional[float] = None,
|
2023-03-08 03:06:59 -07:00
|
|
|
watermark: bool = False,
|
2023-06-02 09:12:30 -06:00
|
|
|
decoder_input_details: bool = False,
|
2023-08-28 03:43:47 -06:00
|
|
|
top_n_tokens: Optional[int] = None,
|
2024-02-15 02:28:10 -07:00
|
|
|
grammar: Optional[Grammar] = None,
|
2023-03-07 10:52:22 -07:00
|
|
|
) -> Response:
|
|
|
|
"""
|
|
|
|
Given a prompt, generate the following text
|
|
|
|
|
|
|
|
Args:
|
|
|
|
prompt (`str`):
|
|
|
|
Input text
|
|
|
|
do_sample (`bool`):
|
|
|
|
Activate logits sampling
|
|
|
|
max_new_tokens (`int`):
|
|
|
|
Maximum number of generated tokens
|
2023-03-09 08:05:33 -07:00
|
|
|
best_of (`int`):
|
|
|
|
Generate best_of sequences and return the one if the highest token logprobs
|
2023-03-07 10:52:22 -07:00
|
|
|
repetition_penalty (`float`):
|
|
|
|
The parameter for repetition penalty. 1.0 means no penalty. See [this
|
|
|
|
paper](https://arxiv.org/pdf/1909.05858.pdf) for more details.
|
2024-03-22 10:10:56 -06:00
|
|
|
frequency_penalty (`float`):
|
|
|
|
The parameter for frequency penalty. 1.0 means no penalty
|
|
|
|
Penalize new tokens based on their existing frequency in the text so far,
|
|
|
|
decreasing the model's likelihood to repeat the same line verbatim.
|
2023-03-07 10:52:22 -07:00
|
|
|
return_full_text (`bool`):
|
|
|
|
Whether to prepend the prompt to the generated text
|
|
|
|
seed (`int`):
|
|
|
|
Random sampling seed
|
|
|
|
stop_sequences (`List[str]`):
|
|
|
|
Stop generating tokens if a member of `stop_sequences` is generated
|
|
|
|
temperature (`float`):
|
|
|
|
The value used to module the logits distribution.
|
|
|
|
top_k (`int`):
|
|
|
|
The number of highest probability vocabulary tokens to keep for top-k-filtering.
|
|
|
|
top_p (`float`):
|
|
|
|
If set to < 1, only the smallest set of most probable tokens with probabilities that add up to `top_p` or
|
|
|
|
higher are kept for generation.
|
2023-03-09 08:05:33 -07:00
|
|
|
truncate (`int`):
|
|
|
|
Truncate inputs tokens to the given size
|
|
|
|
typical_p (`float`):
|
|
|
|
Typical Decoding mass
|
|
|
|
See [Typical Decoding for Natural Language Generation](https://arxiv.org/abs/2202.00666) for more information
|
2023-03-08 03:06:59 -07:00
|
|
|
watermark (`bool`):
|
2023-03-07 10:52:22 -07:00
|
|
|
Watermarking with [A Watermark for Large Language Models](https://arxiv.org/abs/2301.10226)
|
2023-06-02 09:12:30 -06:00
|
|
|
decoder_input_details (`bool`):
|
|
|
|
Return the decoder input token logprobs and ids
|
2023-08-28 03:43:47 -06:00
|
|
|
top_n_tokens (`int`):
|
|
|
|
Return the `n` most likely tokens at each step
|
2024-02-29 08:44:20 -07:00
|
|
|
grammar (`Grammar`):
|
|
|
|
Whether to use a grammar for the generation and the grammar to use. Grammars will constrain the generation
|
|
|
|
of the text to match a regular expression or JSON schema.
|
2023-03-07 10:52:22 -07:00
|
|
|
|
|
|
|
Returns:
|
|
|
|
Response: generated response
|
|
|
|
"""
|
|
|
|
# Validate parameters
|
|
|
|
parameters = Parameters(
|
2023-03-09 08:05:33 -07:00
|
|
|
best_of=best_of,
|
2023-03-07 10:52:22 -07:00
|
|
|
details=True,
|
|
|
|
do_sample=do_sample,
|
|
|
|
max_new_tokens=max_new_tokens,
|
|
|
|
repetition_penalty=repetition_penalty,
|
2024-03-22 10:10:56 -06:00
|
|
|
frequency_penalty=frequency_penalty,
|
2023-03-07 10:52:22 -07:00
|
|
|
return_full_text=return_full_text,
|
|
|
|
seed=seed,
|
|
|
|
stop=stop_sequences if stop_sequences is not None else [],
|
|
|
|
temperature=temperature,
|
|
|
|
top_k=top_k,
|
|
|
|
top_p=top_p,
|
2023-03-09 08:05:33 -07:00
|
|
|
truncate=truncate,
|
|
|
|
typical_p=typical_p,
|
2023-03-08 03:06:59 -07:00
|
|
|
watermark=watermark,
|
2023-06-02 09:12:30 -06:00
|
|
|
decoder_input_details=decoder_input_details,
|
2023-09-27 04:22:09 -06:00
|
|
|
top_n_tokens=top_n_tokens,
|
2024-02-15 02:28:10 -07:00
|
|
|
grammar=grammar,
|
2023-03-07 10:52:22 -07:00
|
|
|
)
|
|
|
|
request = Request(inputs=prompt, stream=False, parameters=parameters)
|
|
|
|
|
|
|
|
resp = requests.post(
|
|
|
|
self.base_url,
|
|
|
|
json=request.dict(),
|
|
|
|
headers=self.headers,
|
2023-03-23 11:01:01 -06:00
|
|
|
cookies=self.cookies,
|
2023-03-07 10:52:22 -07:00
|
|
|
timeout=self.timeout,
|
|
|
|
)
|
|
|
|
payload = resp.json()
|
|
|
|
if resp.status_code != 200:
|
|
|
|
raise parse_error(resp.status_code, payload)
|
|
|
|
return Response(**payload[0])
|
|
|
|
|
|
|
|
def generate_stream(
|
|
|
|
self,
|
|
|
|
prompt: str,
|
|
|
|
do_sample: bool = False,
|
2024-02-01 07:36:10 -07:00
|
|
|
max_new_tokens: int = 20,
|
2023-03-07 10:52:22 -07:00
|
|
|
repetition_penalty: Optional[float] = None,
|
2024-03-22 10:10:56 -06:00
|
|
|
frequency_penalty: Optional[float] = None,
|
2023-03-07 10:52:22 -07:00
|
|
|
return_full_text: bool = False,
|
|
|
|
seed: Optional[int] = None,
|
|
|
|
stop_sequences: Optional[List[str]] = None,
|
|
|
|
temperature: Optional[float] = None,
|
|
|
|
top_k: Optional[int] = None,
|
|
|
|
top_p: Optional[float] = None,
|
2023-03-09 08:05:33 -07:00
|
|
|
truncate: Optional[int] = None,
|
|
|
|
typical_p: Optional[float] = None,
|
2023-03-08 03:06:59 -07:00
|
|
|
watermark: bool = False,
|
2023-08-28 03:43:47 -06:00
|
|
|
top_n_tokens: Optional[int] = None,
|
2024-02-15 02:28:10 -07:00
|
|
|
grammar: Optional[Grammar] = None,
|
2023-03-07 10:52:22 -07:00
|
|
|
) -> Iterator[StreamResponse]:
|
|
|
|
"""
|
|
|
|
Given a prompt, generate the following stream of tokens
|
|
|
|
|
|
|
|
Args:
|
|
|
|
prompt (`str`):
|
|
|
|
Input text
|
|
|
|
do_sample (`bool`):
|
|
|
|
Activate logits sampling
|
|
|
|
max_new_tokens (`int`):
|
|
|
|
Maximum number of generated tokens
|
|
|
|
repetition_penalty (`float`):
|
|
|
|
The parameter for repetition penalty. 1.0 means no penalty. See [this
|
|
|
|
paper](https://arxiv.org/pdf/1909.05858.pdf) for more details.
|
2024-03-22 10:10:56 -06:00
|
|
|
frequency_penalty (`float`):
|
|
|
|
The parameter for frequency penalty. 1.0 means no penalty
|
|
|
|
Penalize new tokens based on their existing frequency in the text so far,
|
|
|
|
decreasing the model's likelihood to repeat the same line verbatim.
|
2023-03-07 10:52:22 -07:00
|
|
|
return_full_text (`bool`):
|
|
|
|
Whether to prepend the prompt to the generated text
|
|
|
|
seed (`int`):
|
|
|
|
Random sampling seed
|
|
|
|
stop_sequences (`List[str]`):
|
|
|
|
Stop generating tokens if a member of `stop_sequences` is generated
|
|
|
|
temperature (`float`):
|
|
|
|
The value used to module the logits distribution.
|
|
|
|
top_k (`int`):
|
|
|
|
The number of highest probability vocabulary tokens to keep for top-k-filtering.
|
|
|
|
top_p (`float`):
|
|
|
|
If set to < 1, only the smallest set of most probable tokens with probabilities that add up to `top_p` or
|
|
|
|
higher are kept for generation.
|
2023-03-09 08:05:33 -07:00
|
|
|
truncate (`int`):
|
|
|
|
Truncate inputs tokens to the given size
|
|
|
|
typical_p (`float`):
|
|
|
|
Typical Decoding mass
|
|
|
|
See [Typical Decoding for Natural Language Generation](https://arxiv.org/abs/2202.00666) for more information
|
2023-03-08 03:06:59 -07:00
|
|
|
watermark (`bool`):
|
2023-03-07 10:52:22 -07:00
|
|
|
Watermarking with [A Watermark for Large Language Models](https://arxiv.org/abs/2301.10226)
|
2023-08-28 03:43:47 -06:00
|
|
|
top_n_tokens (`int`):
|
|
|
|
Return the `n` most likely tokens at each step
|
2024-02-29 08:44:20 -07:00
|
|
|
grammar (`Grammar`):
|
|
|
|
Whether to use a grammar for the generation and the grammar to use. Grammars will constrain the generation
|
|
|
|
of the text to match a regular expression or JSON schema.
|
2023-03-07 10:52:22 -07:00
|
|
|
|
|
|
|
Returns:
|
|
|
|
Iterator[StreamResponse]: stream of generated tokens
|
|
|
|
"""
|
|
|
|
# Validate parameters
|
|
|
|
parameters = Parameters(
|
2023-03-24 11:21:41 -06:00
|
|
|
best_of=None,
|
2023-03-07 10:52:22 -07:00
|
|
|
details=True,
|
2023-06-02 09:12:30 -06:00
|
|
|
decoder_input_details=False,
|
2023-03-07 10:52:22 -07:00
|
|
|
do_sample=do_sample,
|
|
|
|
max_new_tokens=max_new_tokens,
|
|
|
|
repetition_penalty=repetition_penalty,
|
2024-03-22 10:10:56 -06:00
|
|
|
frequency_penalty=frequency_penalty,
|
2023-03-07 10:52:22 -07:00
|
|
|
return_full_text=return_full_text,
|
|
|
|
seed=seed,
|
|
|
|
stop=stop_sequences if stop_sequences is not None else [],
|
|
|
|
temperature=temperature,
|
|
|
|
top_k=top_k,
|
|
|
|
top_p=top_p,
|
2023-03-09 08:05:33 -07:00
|
|
|
truncate=truncate,
|
|
|
|
typical_p=typical_p,
|
2023-03-08 03:06:59 -07:00
|
|
|
watermark=watermark,
|
2023-08-28 03:43:47 -06:00
|
|
|
top_n_tokens=top_n_tokens,
|
2024-02-15 02:28:10 -07:00
|
|
|
grammar=grammar,
|
2023-03-07 10:52:22 -07:00
|
|
|
)
|
|
|
|
request = Request(inputs=prompt, stream=True, parameters=parameters)
|
|
|
|
|
|
|
|
resp = requests.post(
|
|
|
|
self.base_url,
|
|
|
|
json=request.dict(),
|
|
|
|
headers=self.headers,
|
2023-03-23 11:01:01 -06:00
|
|
|
cookies=self.cookies,
|
2023-03-07 10:52:22 -07:00
|
|
|
timeout=self.timeout,
|
2023-03-08 08:48:16 -07:00
|
|
|
stream=True,
|
2023-03-07 10:52:22 -07:00
|
|
|
)
|
|
|
|
|
|
|
|
if resp.status_code != 200:
|
|
|
|
raise parse_error(resp.status_code, resp.json())
|
|
|
|
|
|
|
|
# Parse ServerSentEvents
|
|
|
|
for byte_payload in resp.iter_lines():
|
|
|
|
# Skip line
|
|
|
|
if byte_payload == b"\n":
|
|
|
|
continue
|
|
|
|
|
|
|
|
payload = byte_payload.decode("utf-8")
|
|
|
|
|
|
|
|
# Event data
|
|
|
|
if payload.startswith("data:"):
|
|
|
|
# Decode payload
|
|
|
|
json_payload = json.loads(payload.lstrip("data:").rstrip("/n"))
|
|
|
|
# Parse payload
|
|
|
|
try:
|
|
|
|
response = StreamResponse(**json_payload)
|
|
|
|
except ValidationError:
|
|
|
|
# If we failed to parse the payload, then it is an error payload
|
|
|
|
raise parse_error(resp.status_code, json_payload)
|
|
|
|
yield response
|
|
|
|
|
|
|
|
|
|
|
|
class AsyncClient:
|
|
|
|
"""Asynchronous Client to make calls to a text-generation-inference instance
|
|
|
|
|
|
|
|
Example:
|
|
|
|
|
|
|
|
```python
|
|
|
|
>>> from text_generation import AsyncClient
|
|
|
|
|
|
|
|
>>> client = AsyncClient("https://api-inference.huggingface.co/models/bigscience/bloomz")
|
|
|
|
>>> response = await client.generate("Why is the sky blue?")
|
|
|
|
>>> response.generated_text
|
|
|
|
' Rayleigh scattering'
|
|
|
|
|
|
|
|
>>> result = ""
|
|
|
|
>>> async for response in client.generate_stream("Why is the sky blue?"):
|
|
|
|
>>> if not response.token.special:
|
|
|
|
>>> result += response.token.text
|
|
|
|
>>> result
|
|
|
|
' Rayleigh scattering'
|
|
|
|
```
|
|
|
|
"""
|
|
|
|
|
|
|
|
def __init__(
|
2023-03-23 11:01:01 -06:00
|
|
|
self,
|
|
|
|
base_url: str,
|
|
|
|
headers: Optional[Dict[str, str]] = None,
|
|
|
|
cookies: Optional[Dict[str, str]] = None,
|
|
|
|
timeout: int = 10,
|
2023-03-07 10:52:22 -07:00
|
|
|
):
|
|
|
|
"""
|
|
|
|
Args:
|
|
|
|
base_url (`str`):
|
|
|
|
text-generation-inference instance base url
|
|
|
|
headers (`Optional[Dict[str, str]]`):
|
|
|
|
Additional headers
|
2023-03-23 11:01:01 -06:00
|
|
|
cookies (`Optional[Dict[str, str]]`):
|
|
|
|
Cookies to include in the requests
|
2023-03-07 10:52:22 -07:00
|
|
|
timeout (`int`):
|
|
|
|
Timeout in seconds
|
|
|
|
"""
|
2024-05-15 07:40:07 -06:00
|
|
|
warnings.warn(DEPRECATION_WARNING, DeprecationWarning)
|
2023-03-07 10:52:22 -07:00
|
|
|
self.base_url = base_url
|
|
|
|
self.headers = headers
|
2023-03-23 11:01:01 -06:00
|
|
|
self.cookies = cookies
|
2024-02-29 07:41:49 -07:00
|
|
|
self.timeout = ClientTimeout(timeout)
|
2023-03-07 10:52:22 -07:00
|
|
|
|
2024-05-23 07:37:09 -06:00
|
|
|
async def completion(
|
|
|
|
self,
|
|
|
|
prompt: str,
|
|
|
|
frequency_penalty: Optional[float] = None,
|
|
|
|
max_tokens: Optional[int] = None,
|
|
|
|
repetition_penalty: Optional[float] = None,
|
|
|
|
seed: Optional[int] = None,
|
|
|
|
stream: bool = False,
|
|
|
|
temperature: Optional[float] = None,
|
|
|
|
top_p: Optional[float] = None,
|
|
|
|
stop: Optional[List[str]] = None,
|
|
|
|
) -> Union[Completion, AsyncIterator[CompletionComplete]]:
|
|
|
|
"""
|
|
|
|
Given a prompt, generate a response asynchronously
|
|
|
|
|
|
|
|
Args:
|
|
|
|
prompt (`str`):
|
|
|
|
Prompt
|
|
|
|
frequency_penalty (`float`):
|
|
|
|
The parameter for frequency penalty. 0.0 means no penalty
|
|
|
|
Penalize new tokens based on their existing frequency in the text so far,
|
|
|
|
decreasing the model's likelihood to repeat the same line verbatim.
|
|
|
|
max_tokens (`int`):
|
|
|
|
Maximum number of generated tokens
|
|
|
|
repetition_penalty (`float`):
|
|
|
|
The parameter for frequency penalty. 0.0 means no penalty. See [this
|
|
|
|
paper](https://arxiv.org/pdf/1909.05858.pdf) for more details.
|
|
|
|
seed (`int`):
|
|
|
|
Random sampling seed
|
|
|
|
stream (`bool`):
|
|
|
|
Stream the response
|
|
|
|
temperature (`float`):
|
|
|
|
The value used to module the logits distribution.
|
|
|
|
top_p (`float`):
|
|
|
|
If set to < 1, only the smallest set of most probable tokens with probabilities that add up to `top_p` or
|
|
|
|
higher are kept for generation
|
|
|
|
stop (`List[str]`):
|
|
|
|
Stop generating tokens if a member of `stop` is generated
|
|
|
|
"""
|
|
|
|
request = CompletionRequest(
|
|
|
|
model="tgi",
|
|
|
|
prompt=prompt,
|
|
|
|
frequency_penalty=frequency_penalty,
|
|
|
|
max_tokens=max_tokens,
|
|
|
|
repetition_penalty=repetition_penalty,
|
|
|
|
seed=seed,
|
|
|
|
stream=stream,
|
|
|
|
temperature=temperature,
|
|
|
|
top_p=top_p,
|
|
|
|
stop=stop,
|
|
|
|
)
|
|
|
|
if not stream:
|
|
|
|
return await self._completion_single_response(request)
|
|
|
|
else:
|
|
|
|
return self._completion_stream_response(request)
|
|
|
|
|
|
|
|
async def _completion_single_response(self, request):
|
|
|
|
async with ClientSession(
|
|
|
|
headers=self.headers, cookies=self.cookies, timeout=self.timeout
|
|
|
|
) as session:
|
|
|
|
async with session.post(
|
|
|
|
f"{self.base_url}/v1/completions", json=request.dict()
|
|
|
|
) as resp:
|
|
|
|
payload = await resp.json()
|
|
|
|
if resp.status != 200:
|
|
|
|
raise parse_error(resp.status, payload)
|
|
|
|
return Completion(**payload)
|
|
|
|
|
|
|
|
async def _completion_stream_response(self, request):
|
|
|
|
async with ClientSession(
|
|
|
|
headers=self.headers, cookies=self.cookies, timeout=self.timeout
|
|
|
|
) as session:
|
|
|
|
async with session.post(
|
|
|
|
f"{self.base_url}/v1/completions", json=request.dict()
|
|
|
|
) as resp:
|
|
|
|
async for byte_payload in resp.content:
|
|
|
|
if byte_payload == b"\n":
|
|
|
|
continue
|
|
|
|
payload = byte_payload.decode("utf-8")
|
|
|
|
if payload.startswith("data:"):
|
|
|
|
json_payload = json.loads(payload.lstrip("data:").rstrip("\n"))
|
|
|
|
try:
|
|
|
|
response = CompletionComplete(**json_payload)
|
|
|
|
yield response
|
|
|
|
except ValidationError:
|
|
|
|
raise parse_error(resp.status, json_payload)
|
|
|
|
|
2024-02-28 03:10:27 -07:00
|
|
|
async def chat(
|
|
|
|
self,
|
|
|
|
messages: List[Message],
|
2024-03-22 10:10:56 -06:00
|
|
|
repetition_penalty: Optional[float] = None,
|
2024-02-28 03:10:27 -07:00
|
|
|
frequency_penalty: Optional[float] = None,
|
|
|
|
logit_bias: Optional[List[float]] = None,
|
|
|
|
logprobs: Optional[bool] = None,
|
|
|
|
top_logprobs: Optional[int] = None,
|
|
|
|
max_tokens: Optional[int] = None,
|
|
|
|
n: Optional[int] = None,
|
|
|
|
presence_penalty: Optional[float] = None,
|
|
|
|
stream: bool = False,
|
|
|
|
seed: Optional[int] = None,
|
|
|
|
temperature: Optional[float] = None,
|
|
|
|
top_p: Optional[float] = None,
|
|
|
|
tools: Optional[List[Tool]] = None,
|
2024-04-30 03:07:17 -06:00
|
|
|
tool_prompt: Optional[str] = None,
|
2024-02-28 03:10:27 -07:00
|
|
|
tool_choice: Optional[str] = None,
|
2024-05-23 07:37:09 -06:00
|
|
|
stop: Optional[List[str]] = None,
|
2024-02-28 03:10:27 -07:00
|
|
|
) -> Union[ChatComplete, AsyncIterator[ChatCompletionChunk]]:
|
|
|
|
"""
|
|
|
|
Given a list of messages, generate a response asynchronously
|
|
|
|
|
|
|
|
Args:
|
|
|
|
messages (`List[Message]`):
|
|
|
|
List of messages
|
2024-03-22 10:10:56 -06:00
|
|
|
repetition_penalty (`float`):
|
2024-02-28 03:10:27 -07:00
|
|
|
The parameter for frequency penalty. 0.0 means no penalty. See [this
|
|
|
|
paper](https://arxiv.org/pdf/1909.05858.pdf) for more details.
|
2024-03-22 10:10:56 -06:00
|
|
|
frequency_penalty (`float`):
|
|
|
|
The parameter for frequency penalty. 0.0 means no penalty
|
|
|
|
Penalize new tokens based on their existing frequency in the text so far,
|
|
|
|
decreasing the model's likelihood to repeat the same line verbatim.
|
2024-02-28 03:10:27 -07:00
|
|
|
logit_bias (`List[float]`):
|
|
|
|
Adjust the likelihood of specified tokens
|
|
|
|
logprobs (`bool`):
|
|
|
|
Include log probabilities in the response
|
|
|
|
top_logprobs (`int`):
|
|
|
|
Include the `n` most likely tokens at each step
|
|
|
|
max_tokens (`int`):
|
|
|
|
Maximum number of generated tokens
|
|
|
|
n (`int`):
|
|
|
|
Generate `n` completions
|
|
|
|
presence_penalty (`float`):
|
|
|
|
The parameter for presence penalty. 0.0 means no penalty. See [this
|
|
|
|
paper](https://arxiv.org/pdf/1909.05858.pdf) for more details.
|
|
|
|
stream (`bool`):
|
|
|
|
Stream the response
|
|
|
|
seed (`int`):
|
|
|
|
Random sampling seed
|
|
|
|
temperature (`float`):
|
|
|
|
The value used to module the logits distribution.
|
|
|
|
top_p (`float`):
|
|
|
|
If set to < 1, only the smallest set of most probable tokens with probabilities that add up to `top_p` or
|
|
|
|
higher are kept for generation
|
|
|
|
tools (`List[Tool]`):
|
|
|
|
List of tools to use
|
2024-04-30 03:07:17 -06:00
|
|
|
tool_prompt (`str`):
|
|
|
|
A prompt to be appended before the tools
|
2024-02-28 03:10:27 -07:00
|
|
|
tool_choice (`str`):
|
|
|
|
The tool to use
|
2024-05-23 07:37:09 -06:00
|
|
|
stop (`List[str]`):
|
|
|
|
Stop generating tokens if a member of `stop` is generated
|
2024-02-28 03:10:27 -07:00
|
|
|
|
|
|
|
"""
|
|
|
|
request = ChatRequest(
|
|
|
|
model="tgi",
|
|
|
|
messages=messages,
|
2024-03-22 10:10:56 -06:00
|
|
|
repetition_penalty=repetition_penalty,
|
2024-02-28 03:10:27 -07:00
|
|
|
frequency_penalty=frequency_penalty,
|
|
|
|
logit_bias=logit_bias,
|
|
|
|
logprobs=logprobs,
|
|
|
|
top_logprobs=top_logprobs,
|
|
|
|
max_tokens=max_tokens,
|
|
|
|
n=n,
|
|
|
|
presence_penalty=presence_penalty,
|
|
|
|
stream=stream,
|
|
|
|
seed=seed,
|
|
|
|
temperature=temperature,
|
|
|
|
top_p=top_p,
|
|
|
|
tools=tools,
|
2024-04-30 03:07:17 -06:00
|
|
|
tool_prompt=tool_prompt,
|
2024-02-28 03:10:27 -07:00
|
|
|
tool_choice=tool_choice,
|
2024-05-23 07:37:09 -06:00
|
|
|
stop=stop,
|
2024-02-28 03:10:27 -07:00
|
|
|
)
|
|
|
|
if not stream:
|
|
|
|
return await self._chat_single_response(request)
|
|
|
|
else:
|
|
|
|
return self._chat_stream_response(request)
|
|
|
|
|
|
|
|
async def _chat_single_response(self, request):
|
|
|
|
async with ClientSession(
|
|
|
|
headers=self.headers, cookies=self.cookies, timeout=self.timeout
|
|
|
|
) as session:
|
|
|
|
async with session.post(
|
|
|
|
f"{self.base_url}/v1/chat/completions", json=request.dict()
|
|
|
|
) as resp:
|
|
|
|
payload = await resp.json()
|
|
|
|
if resp.status != 200:
|
|
|
|
raise parse_error(resp.status, payload)
|
|
|
|
return ChatComplete(**payload)
|
|
|
|
|
|
|
|
async def _chat_stream_response(self, request):
|
|
|
|
async with ClientSession(
|
|
|
|
headers=self.headers, cookies=self.cookies, timeout=self.timeout
|
|
|
|
) as session:
|
|
|
|
async with session.post(
|
|
|
|
f"{self.base_url}/v1/chat/completions", json=request.dict()
|
|
|
|
) as resp:
|
|
|
|
async for byte_payload in resp.content:
|
|
|
|
if byte_payload == b"\n":
|
|
|
|
continue
|
|
|
|
payload = byte_payload.decode("utf-8")
|
|
|
|
if payload.startswith("data:"):
|
|
|
|
json_payload = json.loads(payload.lstrip("data:").rstrip("\n"))
|
|
|
|
try:
|
|
|
|
response = ChatCompletionChunk(**json_payload)
|
|
|
|
yield response
|
|
|
|
except ValidationError:
|
|
|
|
raise parse_error(resp.status, json_payload)
|
|
|
|
|
2023-03-07 10:52:22 -07:00
|
|
|
async def generate(
|
|
|
|
self,
|
|
|
|
prompt: str,
|
|
|
|
do_sample: bool = False,
|
2024-02-01 07:36:10 -07:00
|
|
|
max_new_tokens: int = 20,
|
2023-03-09 08:05:33 -07:00
|
|
|
best_of: Optional[int] = None,
|
2023-03-07 10:52:22 -07:00
|
|
|
repetition_penalty: Optional[float] = None,
|
2024-03-22 10:10:56 -06:00
|
|
|
frequency_penalty: Optional[float] = None,
|
2023-03-07 10:52:22 -07:00
|
|
|
return_full_text: bool = False,
|
|
|
|
seed: Optional[int] = None,
|
|
|
|
stop_sequences: Optional[List[str]] = None,
|
|
|
|
temperature: Optional[float] = None,
|
|
|
|
top_k: Optional[int] = None,
|
|
|
|
top_p: Optional[float] = None,
|
2023-03-09 08:05:33 -07:00
|
|
|
truncate: Optional[int] = None,
|
|
|
|
typical_p: Optional[float] = None,
|
2023-03-08 03:06:59 -07:00
|
|
|
watermark: bool = False,
|
2023-06-02 09:12:30 -06:00
|
|
|
decoder_input_details: bool = False,
|
2023-08-28 03:43:47 -06:00
|
|
|
top_n_tokens: Optional[int] = None,
|
2024-02-15 02:28:10 -07:00
|
|
|
grammar: Optional[Grammar] = None,
|
2023-03-07 10:52:22 -07:00
|
|
|
) -> Response:
|
|
|
|
"""
|
|
|
|
Given a prompt, generate the following text asynchronously
|
|
|
|
|
|
|
|
Args:
|
|
|
|
prompt (`str`):
|
|
|
|
Input text
|
|
|
|
do_sample (`bool`):
|
|
|
|
Activate logits sampling
|
|
|
|
max_new_tokens (`int`):
|
|
|
|
Maximum number of generated tokens
|
2023-03-09 08:05:33 -07:00
|
|
|
best_of (`int`):
|
|
|
|
Generate best_of sequences and return the one if the highest token logprobs
|
2023-03-07 10:52:22 -07:00
|
|
|
repetition_penalty (`float`):
|
|
|
|
The parameter for repetition penalty. 1.0 means no penalty. See [this
|
|
|
|
paper](https://arxiv.org/pdf/1909.05858.pdf) for more details.
|
2024-03-22 10:10:56 -06:00
|
|
|
frequency_penalty (`float`):
|
|
|
|
The parameter for frequency penalty. 1.0 means no penalty
|
|
|
|
Penalize new tokens based on their existing frequency in the text so far,
|
|
|
|
decreasing the model's likelihood to repeat the same line verbatim.
|
2023-03-07 10:52:22 -07:00
|
|
|
return_full_text (`bool`):
|
|
|
|
Whether to prepend the prompt to the generated text
|
|
|
|
seed (`int`):
|
|
|
|
Random sampling seed
|
|
|
|
stop_sequences (`List[str]`):
|
|
|
|
Stop generating tokens if a member of `stop_sequences` is generated
|
|
|
|
temperature (`float`):
|
|
|
|
The value used to module the logits distribution.
|
|
|
|
top_k (`int`):
|
|
|
|
The number of highest probability vocabulary tokens to keep for top-k-filtering.
|
|
|
|
top_p (`float`):
|
|
|
|
If set to < 1, only the smallest set of most probable tokens with probabilities that add up to `top_p` or
|
|
|
|
higher are kept for generation.
|
2023-03-09 08:05:33 -07:00
|
|
|
truncate (`int`):
|
|
|
|
Truncate inputs tokens to the given size
|
|
|
|
typical_p (`float`):
|
|
|
|
Typical Decoding mass
|
|
|
|
See [Typical Decoding for Natural Language Generation](https://arxiv.org/abs/2202.00666) for more information
|
2023-03-08 03:06:59 -07:00
|
|
|
watermark (`bool`):
|
2023-03-07 10:52:22 -07:00
|
|
|
Watermarking with [A Watermark for Large Language Models](https://arxiv.org/abs/2301.10226)
|
2023-06-02 09:12:30 -06:00
|
|
|
decoder_input_details (`bool`):
|
|
|
|
Return the decoder input token logprobs and ids
|
2023-08-28 03:43:47 -06:00
|
|
|
top_n_tokens (`int`):
|
|
|
|
Return the `n` most likely tokens at each step
|
2024-02-29 08:44:20 -07:00
|
|
|
grammar (`Grammar`):
|
|
|
|
Whether to use a grammar for the generation and the grammar to use. Grammars will constrain the generation
|
|
|
|
of the text to match a regular expression or JSON schema.
|
2023-03-07 10:52:22 -07:00
|
|
|
|
|
|
|
Returns:
|
|
|
|
Response: generated response
|
|
|
|
"""
|
2024-02-15 02:28:10 -07:00
|
|
|
|
2023-03-07 10:52:22 -07:00
|
|
|
# Validate parameters
|
|
|
|
parameters = Parameters(
|
2023-03-09 08:05:33 -07:00
|
|
|
best_of=best_of,
|
2023-03-07 10:52:22 -07:00
|
|
|
details=True,
|
2023-06-02 09:12:30 -06:00
|
|
|
decoder_input_details=decoder_input_details,
|
2023-03-07 10:52:22 -07:00
|
|
|
do_sample=do_sample,
|
|
|
|
max_new_tokens=max_new_tokens,
|
|
|
|
repetition_penalty=repetition_penalty,
|
2024-03-22 10:10:56 -06:00
|
|
|
frequency_penalty=frequency_penalty,
|
2023-03-07 10:52:22 -07:00
|
|
|
return_full_text=return_full_text,
|
|
|
|
seed=seed,
|
|
|
|
stop=stop_sequences if stop_sequences is not None else [],
|
|
|
|
temperature=temperature,
|
|
|
|
top_k=top_k,
|
|
|
|
top_p=top_p,
|
2023-03-09 08:05:33 -07:00
|
|
|
truncate=truncate,
|
|
|
|
typical_p=typical_p,
|
2023-03-08 03:06:59 -07:00
|
|
|
watermark=watermark,
|
2023-08-28 03:43:47 -06:00
|
|
|
top_n_tokens=top_n_tokens,
|
2024-02-15 02:28:10 -07:00
|
|
|
grammar=grammar,
|
2023-03-07 10:52:22 -07:00
|
|
|
)
|
|
|
|
request = Request(inputs=prompt, stream=False, parameters=parameters)
|
|
|
|
|
2023-03-23 11:01:01 -06:00
|
|
|
async with ClientSession(
|
|
|
|
headers=self.headers, cookies=self.cookies, timeout=self.timeout
|
|
|
|
) as session:
|
2023-03-07 10:52:22 -07:00
|
|
|
async with session.post(self.base_url, json=request.dict()) as resp:
|
|
|
|
payload = await resp.json()
|
|
|
|
|
|
|
|
if resp.status != 200:
|
|
|
|
raise parse_error(resp.status, payload)
|
|
|
|
return Response(**payload[0])
|
|
|
|
|
|
|
|
async def generate_stream(
|
|
|
|
self,
|
|
|
|
prompt: str,
|
|
|
|
do_sample: bool = False,
|
2024-02-01 07:36:10 -07:00
|
|
|
max_new_tokens: int = 20,
|
2023-03-07 10:52:22 -07:00
|
|
|
repetition_penalty: Optional[float] = None,
|
2024-03-22 10:10:56 -06:00
|
|
|
frequency_penalty: Optional[float] = None,
|
2023-03-07 10:52:22 -07:00
|
|
|
return_full_text: bool = False,
|
|
|
|
seed: Optional[int] = None,
|
|
|
|
stop_sequences: Optional[List[str]] = None,
|
|
|
|
temperature: Optional[float] = None,
|
|
|
|
top_k: Optional[int] = None,
|
|
|
|
top_p: Optional[float] = None,
|
2023-03-09 08:05:33 -07:00
|
|
|
truncate: Optional[int] = None,
|
|
|
|
typical_p: Optional[float] = None,
|
2023-03-08 03:06:59 -07:00
|
|
|
watermark: bool = False,
|
2023-08-28 03:43:47 -06:00
|
|
|
top_n_tokens: Optional[int] = None,
|
2024-02-15 02:28:10 -07:00
|
|
|
grammar: Optional[Grammar] = None,
|
2023-03-07 10:52:22 -07:00
|
|
|
) -> AsyncIterator[StreamResponse]:
|
|
|
|
"""
|
|
|
|
Given a prompt, generate the following stream of tokens asynchronously
|
|
|
|
|
|
|
|
Args:
|
|
|
|
prompt (`str`):
|
|
|
|
Input text
|
|
|
|
do_sample (`bool`):
|
|
|
|
Activate logits sampling
|
|
|
|
max_new_tokens (`int`):
|
|
|
|
Maximum number of generated tokens
|
|
|
|
repetition_penalty (`float`):
|
|
|
|
The parameter for repetition penalty. 1.0 means no penalty. See [this
|
|
|
|
paper](https://arxiv.org/pdf/1909.05858.pdf) for more details.
|
2024-03-22 10:10:56 -06:00
|
|
|
frequency_penalty (`float`):
|
|
|
|
The parameter for frequency penalty. 1.0 means no penalty
|
|
|
|
Penalize new tokens based on their existing frequency in the text so far,
|
|
|
|
decreasing the model's likelihood to repeat the same line verbatim.
|
2023-03-07 10:52:22 -07:00
|
|
|
return_full_text (`bool`):
|
|
|
|
Whether to prepend the prompt to the generated text
|
|
|
|
seed (`int`):
|
|
|
|
Random sampling seed
|
|
|
|
stop_sequences (`List[str]`):
|
|
|
|
Stop generating tokens if a member of `stop_sequences` is generated
|
|
|
|
temperature (`float`):
|
|
|
|
The value used to module the logits distribution.
|
|
|
|
top_k (`int`):
|
|
|
|
The number of highest probability vocabulary tokens to keep for top-k-filtering.
|
|
|
|
top_p (`float`):
|
|
|
|
If set to < 1, only the smallest set of most probable tokens with probabilities that add up to `top_p` or
|
|
|
|
higher are kept for generation.
|
2023-03-09 08:05:33 -07:00
|
|
|
truncate (`int`):
|
|
|
|
Truncate inputs tokens to the given size
|
|
|
|
typical_p (`float`):
|
|
|
|
Typical Decoding mass
|
|
|
|
See [Typical Decoding for Natural Language Generation](https://arxiv.org/abs/2202.00666) for more information
|
2023-03-08 03:06:59 -07:00
|
|
|
watermark (`bool`):
|
2023-03-07 10:52:22 -07:00
|
|
|
Watermarking with [A Watermark for Large Language Models](https://arxiv.org/abs/2301.10226)
|
2023-08-28 03:43:47 -06:00
|
|
|
top_n_tokens (`int`):
|
|
|
|
Return the `n` most likely tokens at each step
|
2024-02-29 08:44:20 -07:00
|
|
|
grammar (`Grammar`):
|
|
|
|
Whether to use a grammar for the generation and the grammar to use. Grammars will constrain the generation
|
|
|
|
of the text to match a regular expression or JSON schema.
|
2023-03-07 10:52:22 -07:00
|
|
|
|
|
|
|
Returns:
|
|
|
|
AsyncIterator[StreamResponse]: stream of generated tokens
|
|
|
|
"""
|
|
|
|
# Validate parameters
|
|
|
|
parameters = Parameters(
|
2023-03-24 11:21:41 -06:00
|
|
|
best_of=None,
|
2023-03-07 10:52:22 -07:00
|
|
|
details=True,
|
2023-06-02 09:12:30 -06:00
|
|
|
decoder_input_details=False,
|
2023-03-07 10:52:22 -07:00
|
|
|
do_sample=do_sample,
|
|
|
|
max_new_tokens=max_new_tokens,
|
|
|
|
repetition_penalty=repetition_penalty,
|
2024-03-22 10:10:56 -06:00
|
|
|
frequency_penalty=frequency_penalty,
|
2023-03-07 10:52:22 -07:00
|
|
|
return_full_text=return_full_text,
|
|
|
|
seed=seed,
|
|
|
|
stop=stop_sequences if stop_sequences is not None else [],
|
|
|
|
temperature=temperature,
|
|
|
|
top_k=top_k,
|
|
|
|
top_p=top_p,
|
2023-03-09 08:05:33 -07:00
|
|
|
truncate=truncate,
|
|
|
|
typical_p=typical_p,
|
2023-03-08 03:06:59 -07:00
|
|
|
watermark=watermark,
|
2023-08-28 03:43:47 -06:00
|
|
|
top_n_tokens=top_n_tokens,
|
2024-02-15 02:28:10 -07:00
|
|
|
grammar=grammar,
|
2023-03-07 10:52:22 -07:00
|
|
|
)
|
|
|
|
request = Request(inputs=prompt, stream=True, parameters=parameters)
|
|
|
|
|
2023-03-23 11:01:01 -06:00
|
|
|
async with ClientSession(
|
|
|
|
headers=self.headers, cookies=self.cookies, timeout=self.timeout
|
|
|
|
) as session:
|
2023-03-07 10:52:22 -07:00
|
|
|
async with session.post(self.base_url, json=request.dict()) as resp:
|
|
|
|
if resp.status != 200:
|
|
|
|
raise parse_error(resp.status, await resp.json())
|
|
|
|
|
|
|
|
# Parse ServerSentEvents
|
|
|
|
async for byte_payload in resp.content:
|
|
|
|
# Skip line
|
|
|
|
if byte_payload == b"\n":
|
|
|
|
continue
|
|
|
|
|
|
|
|
payload = byte_payload.decode("utf-8")
|
|
|
|
|
|
|
|
# Event data
|
|
|
|
if payload.startswith("data:"):
|
|
|
|
# Decode payload
|
|
|
|
json_payload = json.loads(payload.lstrip("data:").rstrip("/n"))
|
|
|
|
# Parse payload
|
|
|
|
try:
|
|
|
|
response = StreamResponse(**json_payload)
|
|
|
|
except ValidationError:
|
|
|
|
# If we failed to parse the payload, then it is an error payload
|
|
|
|
raise parse_error(resp.status, json_payload)
|
|
|
|
yield response
|