import ast
import asyncio
import selectors
import contextlib
import functools
import gc
import hashlib
import inspect
import io
import json
import os
import pathlib
import pickle
import platform
import random
import shutil
import subprocess
import sys
import threading
import time
import traceback
import zipfile
import tarfile
from array import array
from collections import deque
from concurrent.futures import ProcessPoolExecutor
from datetime import datetime
from typing import Tuple, Callable, Dict
from queue import Queue, Empty
from concurrent.futures import ThreadPoolExecutor
from urllib.parse import urlparse
import filelock
import fire
import numpy as np
import pandas as pd
import psutil
import requests
import uuid
import re
from packaging import version
import tabulate
from fire import inspectutils
from joblib import Parallel
from tqdm.auto import tqdm
from enums import split_google, invalid_json_str, docs_joiner_default, git_hash_unset, is_json_model, \
openai_supports_functiontools, openai_supports_parallel_functiontools, does_support_functiontools
from utils_procs import reulimit
reulimit()
def H2O_Fire(component=None):
config_prefix = "H2OGPT_"
args = sys.argv[1:]
query_args = [arg.split("=")[0].split(" ")[0].lstrip("-") for arg in args]
fn_spec = inspectutils.GetFullArgSpec(component)
for key, value in os.environ.items():
if not (
(key.startswith(config_prefix) or key.startswith(config_prefix.lower()))
and len(key) > len(config_prefix)
):
continue # ignore as non H2OGPT argument
new_key = key[len(config_prefix):].lower()
if new_key in query_args:
continue # ignore as already passed as script argument
if new_key not in fn_spec.args:
continue # ignore as not a valid H2OGPT argument
args.append(f"--{new_key}={value}")
fire.Fire(component=component, command=args)
def set_seed(seed: int):
"""
Sets the seed of the entire notebook so results are the same every time we run.
This is for REPRODUCIBILITY.
"""
import torch
np.random.seed(seed)
random_state = np.random.RandomState(seed)
random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False
os.environ['PYTHONHASHSEED'] = str(seed)
return random_state
def flatten_list(lis):
"""Given a list, possibly nested to any level, return it flattened."""
new_lis = []
for item in lis:
if type(item) == type([]):
new_lis.extend(flatten_list(item))
else:
new_lis.append(item)
return new_lis
def clear_torch_cache(allow_skip=False):
if allow_skip and os.getenv('CLEAR_CLEAR_TORCH', '2') == '1' or os.getenv('CLEAR_CLEAR_TORCH', '2') == '0':
return
try:
import torch
if torch.cuda.is_available():
torch.cuda.empty_cache()
torch.cuda.ipc_collect()
gc.collect()
except RuntimeError as e:
print("clear_torch_cache error: %s" % ''.join(traceback.format_tb(e.__traceback__)), flush=True)
def ping():
try:
print('Ping: %s' % str(datetime.now()), flush=True)
except AttributeError:
# some programs wrap print and will fail with flush passed
pass
def ping_gpu():
try:
print('Ping_GPU: %s %s' % (str(datetime.now()), system_info()), flush=True)
except AttributeError:
# some programs wrap print and will fail with flush passed
pass
try:
ping_gpu_memory()
except Exception as e:
print('Ping_GPU memory failure: %s' % str(e), flush=True)
def ping_gpu_memory():
from models.gpu_mem_track import MemTracker
gpu_tracker = MemTracker() # define a GPU tracker
from torch.cuda import memory_summary
gpu_tracker.track()
def get_torch_allocated():
import torch
return torch.cuda.memory_allocated()
def get_device(n_gpus=None):
import torch
if torch.cuda.is_available() and n_gpus != 0:
device = "cuda"
elif torch.backends.mps.is_built():
device = "mps"
else:
device = "cpu"
return device
def system_info():
import psutil
system = {}
# https://stackoverflow.com/questions/48951136/plot-multiple-graphs-in-one-plot-using-tensorboard
# https://arshren.medium.com/monitoring-your-devices-in-python-5191d672f749
try:
temps = psutil.sensors_temperatures(fahrenheit=False)
if 'coretemp' in temps:
coretemp = temps['coretemp']
temp_dict = {k.label: k.current for k in coretemp}
for k, v in temp_dict.items():
system['CPU_C/%s' % k] = v
except AttributeError:
pass
# https://github.com/gpuopenanalytics/pynvml/blob/master/help_query_gpu.txt
try:
from pynvml.smi import nvidia_smi
nvsmi = nvidia_smi.getInstance()
gpu_power_dict = {'W_gpu%d' % i: x['power_readings']['power_draw'] for i, x in
enumerate(nvsmi.DeviceQuery('power.draw')['gpu'])}
for k, v in gpu_power_dict.items():
system['GPU_W/%s' % k] = v
gpu_temp_dict = {'C_gpu%d' % i: x['temperature']['gpu_temp'] for i, x in
enumerate(nvsmi.DeviceQuery('temperature.gpu')['gpu'])}
for k, v in gpu_temp_dict.items():
system['GPU_C/%s' % k] = v
gpu_memory_free_dict = {'MiB_gpu%d' % i: x['fb_memory_usage']['free'] for i, x in
enumerate(nvsmi.DeviceQuery('memory.free')['gpu'])}
gpu_memory_total_dict = {'MiB_gpu%d' % i: x['fb_memory_usage']['total'] for i, x in
enumerate(nvsmi.DeviceQuery('memory.total')['gpu'])}
gpu_memory_frac_dict = {k: gpu_memory_free_dict[k] / gpu_memory_total_dict[k] for k in gpu_memory_total_dict}
for k, v in gpu_memory_frac_dict.items():
system[f'GPU_M/%s' % k] = v
except (KeyError, ModuleNotFoundError):
pass
system['hash'] = get_githash()
debug_mem = False
if debug_mem:
try:
# pip install guppy3
from guppy import hpy
h = hpy()
print(h.heap())
print(h.heap().byvia)
print(h.heap().byid)
except:
pass
return system
def system_info_print():
try:
df = pd.DataFrame.from_dict(system_info(), orient='index')
# avoid slamming GPUs
time.sleep(1)
return df.to_markdown()
except Exception as e:
return "Error: %s" % str(e)
def zip_data(root_dirs=None, zip_file=None, base_dir='./', fail_any_exception=False):
try:
return _zip_data(zip_file=zip_file, base_dir=base_dir, root_dirs=root_dirs)
except Exception as e:
traceback.print_exc()
print('Exception in zipping: %s' % str(e))
if not fail_any_exception:
raise
def _zip_data(root_dirs=None, zip_file=None, base_dir='./'):
if isinstance(root_dirs, str):
root_dirs = [root_dirs]
if zip_file is None:
datetime_str = str(datetime.now()).replace(" ", "_").replace(":", "_")
host_name = os.getenv('HF_HOSTNAME', 'emptyhost')
zip_file = "data_%s_%s.zip" % (datetime_str, host_name)
assert root_dirs is not None
base_path = os.path.dirname(zip_file)
if not os.path.isdir(base_path) and os.path.dirname(zip_file):
base_path = makedirs(base_path, exist_ok=True, tmp_ok=True, use_base=True)
zip_file = os.path.join(base_path, os.path.basename(zip_file))
with zipfile.ZipFile(zip_file, "w") as expt_zip:
for root_dir in root_dirs:
if root_dir is None:
continue
for root, d, files in os.walk(root_dir):
for file in files:
file_to_archive = os.path.join(root, file)
assert os.path.exists(file_to_archive)
path_to_archive = os.path.relpath(file_to_archive, base_dir)
expt_zip.write(filename=file_to_archive, arcname=path_to_archive)
return zip_file, zip_file
def tar_data(root_dirs=None, tar_file=None, base_dir='./', fail_any_exception=False):
try:
return _tar_data(tar_file=tar_file, base_dir=base_dir, root_dirs=root_dirs)
except Exception as e:
traceback.print_exc()
print('Exception in tar archiving: %s' % str(e))
if not fail_any_exception:
raise
def _tar_data(root_dirs=None, tar_file=None, base_dir='./'):
if isinstance(root_dirs, str):
root_dirs = [root_dirs]
if tar_file is None:
datetime_str = str(datetime.now()).replace(" ", "_").replace(":", "_")
host_name = os.getenv('HF_HOSTNAME', 'emptyhost')
tar_file = "data_%s_%s.tar.gz" % (datetime_str, host_name)
assert root_dirs is not None
base_path = os.path.dirname(tar_file)
if not os.path.isdir(base_path) and os.path.dirname(tar_file):
base_path = makedirs(base_path, exist_ok=True, tmp_ok=True, use_base=True)
tar_file = os.path.join(base_path, os.path.basename(tar_file))
with tarfile.open(tar_file, "w:gz") as expt_tar:
for root_dir in root_dirs:
if root_dir is None:
continue
for root, d, files in os.walk(root_dir):
for file in files:
file_to_archive = os.path.join(root, file)
assert os.path.exists(file_to_archive)
path_to_archive = os.path.relpath(file_to_archive, base_dir)
expt_tar.add(name=file_to_archive, arcname=path_to_archive)
return tar_file, tar_file
def save_generate_output(prompt=None, output=None, base_model=None, save_dir=None, where_from='unknown where from',
extra_dict={}, error='', sources=[], which_api='', valid_key=None,
h2ogpt_key='', return_dict=False, **kwargs_extra):
if not save_dir:
return
try:
return _save_generate_output(prompt=prompt, output=output, base_model=base_model, save_dir=save_dir,
where_from=where_from, extra_dict=extra_dict, error=error, sources=sources,
which_api=which_api, valid_key=valid_key, h2ogpt_key=h2ogpt_key,
return_dict=return_dict, **kwargs_extra)
except Exception as e:
traceback.print_exc()
print('Exception in saving: %s' % str(e))
def _save_generate_tokens(response_no_refs, extra_dict):
# tokenize at end if need to, so doesn't block generation in multi-generator case
if extra_dict.get('ntokens') is None:
extra_dict['ntokens'] = FakeTokenizer().num_tokens_from_string(str(response_no_refs))
# only do below if didn't already compute ntokens, else assume also computed rate
if extra_dict.get('ntokens') is not None and extra_dict.get('t_generate') is not None:
extra_dict['tokens_persecond'] = extra_dict['ntokens'] / extra_dict['t_generate']
return extra_dict
def _save_generate_output(prompt=None, output=None, base_model=None, save_dir=None, where_from='unknown where from',
extra_dict={}, error='', sources=[], which_api='',
valid_key=None, h2ogpt_key='',
return_dict=False, **kwargs_extra):
"""
Save conversation to .json, row by row.
json_file_path is path to final JSON file. If not in ., then will attempt to make directories.
Appends if file exists
"""
prompt = '<not set>' if prompt is None else prompt
output = '<not set>' if output is None else output
extra_dict = _save_generate_tokens(output, extra_dict)
dict_to_save = dict(prompt=prompt, text=output, time=time.ctime(),
base_model=base_model,
where_from=where_from,
error=error,
sources=sources,
which_api=which_api,
valid_key=valid_key,
h2ogpt_key=h2ogpt_key,
)
dict_to_save.update(extra_dict)
dict_to_save.update(kwargs_extra)
if return_dict:
return dict_to_save
if os.path.exists(save_dir) and not os.path.isdir(save_dir):
raise RuntimeError("save_dir already exists and is not a directory!")
makedirs(save_dir, exist_ok=True) # already should be made, can't change at this point
import json
with filelock.FileLock("%s.lock" % os.path.basename(save_dir)):
# lock logging in case have concurrency
with open(os.path.join(save_dir, "history.json"), "a") as f:
# just add [ at start, and ] at end, and have proper JSON dataset
f.write(
" " + json.dumps(
dict_to_save
) + ",\n"
)
def s3up(filename):
try:
return _s3up(filename)
except Exception as e:
traceback.print_exc()
print('Exception for file %s in s3up: %s' % (filename, str(e)))
return "Failed to upload %s: Error: %s" % (filename, str(e))
def _s3up(filename):
import boto3
aws_access_key_id = os.getenv('AWS_SERVER_PUBLIC_KEY')
aws_secret_access_key = os.getenv('AWS_SERVER_SECRET_KEY')
bucket = os.getenv('AWS_BUCKET')
assert aws_access_key_id, "Set AWS key"
assert aws_secret_access_key, "Set AWS secret"
assert bucket, "Set AWS Bucket"
s3 = boto3.client('s3',
aws_access_key_id=os.getenv('AWS_SERVER_PUBLIC_KEY'),
aws_secret_access_key=os.getenv('AWS_SERVER_SECRET_KEY'),
)
ret = s3.upload_file(
Filename=filename,
Bucket=os.getenv('AWS_BUCKET'),
Key=filename,
)
if ret in [None, '']:
return "Successfully uploaded %s" % filename
def get_githash():
githash = git_hash_unset
try:
githash = subprocess.run(['git', 'rev-parse', 'HEAD'], stdout=subprocess.PIPE).stdout.decode('utf-8')[0:-1]
if githash in ['', None]:
githash = git_hash_unset
except Exception as e:
print("git failed to run: %s" % str(e))
if githash == git_hash_unset:
try:
from version import __version__
githash = __version__
except:
pass
if os.getenv('HARD_ASSERTS'):
assert is_full_git_hash(githash)
return githash
def copy_code(run_id):
"""
copy code to track changes
:param run_id:
:return:
"""
rnd_num = str(random.randint(0, 2 ** 31))
run_id = 'run_' + str(run_id)
os.makedirs(run_id, exist_ok=True)
me_full = os.path.join(pathlib.Path(__file__).parent.resolve(), __file__)
me_file = os.path.basename(__file__)
new_me = os.path.join(run_id, me_file + '_' + get_githash())
if os.path.isfile(new_me):
new_me = os.path.join(run_id, me_file + '_' + get_githash() + '_' + rnd_num)
shutil.copy(me_full, new_me)
else:
shutil.copy(me_full, new_me)
class NullContext(threading.local):
"""No-op context manager, executes block without doing any additional processing.
Used as a stand-in if a particular block of code is only sometimes
used with a normal context manager:
"""
def __init__(self, *args, **kwargs):
pass
def __enter__(self):
return self
def __exit__(self, exc_type, exc_value, exc_traceback):
self.finally_act()
def finally_act(self):
pass
class AsyncNullContext(threading.local):
"""No-op async context manager, executes block without doing any additional processing.
Used as a stand-in if a particular block of code is only sometimes
used with a normal async context manager:
"""
def __init__(self, *args, **kwargs):
pass
async def __aenter__(self):
return self
async def __aexit__(self, exc_type, exc_value, exc_traceback):
await self.finally_act()
async def finally_act(self):
pass
def wrapped_partial(func, *args, **kwargs):
"""
Give partial properties of normal function, like __name__ attribute etc.
:param func:
:param args:
:param kwargs:
:return:
"""
partial_func = functools.partial(func, *args, **kwargs)
functools.update_wrapper(partial_func, func)
return partial_func
class ThreadException(Exception):
pass
class EThread(threading.Thread):
# Function that raises the custom exception
def __init__(self, group=None, target=None, name=None,
args=(), kwargs=None, *, daemon=None, streamer=None, bucket=None,
async_output=False):
self.bucket = bucket
self.streamer = streamer
self.exc = None
self._return = None
self.async_output = async_output
super().__init__(group=group, target=target, name=name, args=args, kwargs=kwargs, daemon=daemon)
def run(self):
# Variable that stores the exception, if raised by someFunction
try:
if self._target is not None:
if self.async_output:
self._return = asyncio.run(self._target(*self._args, **self._kwargs))
else:
self._return = self._target(*self._args, **self._kwargs)
except BaseException as e:
print("thread exception: %s" % str(traceback.format_exc()))
self.bucket.put(sys.exc_info())
self.exc = e
if self.streamer:
print("make stop: %s" % str(traceback.format_exc()), flush=True)
self.streamer.do_stop = True
finally:
# Avoid a refcycle if the thread is running a function with
# an argument that has a member that points to the thread.
del self._target, self._args, self._kwargs
def join(self, timeout=None):
threading.Thread.join(self)
# Since join() returns in caller thread
# we re-raise the caught exception
# if any was caught
if self.exc:
raise self.exc
return self._return
def import_matplotlib():
import matplotlib
matplotlib.use('agg')
# KEEP THESE HERE! START
import matplotlib.pyplot as plt
import pandas as pd
# to avoid dlopen deadlock in fork
import pandas.core.computation.expressions as pd_expressions
import pandas.core.algorithms as pd_algorithms
import pandas.core.common as pd_com
import numpy as np
# KEEP THESE HERE! END
def get_sha(value):
return hashlib.md5(str(value).encode('utf-8')).hexdigest()
def sanitize_filename(name, file_length_limit=250):
"""
Sanitize file *base* names.
:param name: name to sanitize
:param file_length_limit: bit smaller than 256 for safety
:return:
"""
bad_chars = ['[', ']', ',', '/', '\\', '\\w', '\\s', '-', '+', '\"', '\'', '>', '<', ' ', '=', ')', '(', ':', '^']
for char in bad_chars:
name = name.replace(char, "_")
length = len(name)
sha_length = 32
real_length_limit = file_length_limit - (sha_length + 2)
assert real_length_limit > 0, "Bad file limit length: %s %s" % (file_length_limit, real_length_limit)
if length > file_length_limit:
sha = get_sha(name)
half_real_length_limit = max(1, int(real_length_limit / 2))
name = name[0:half_real_length_limit] + "_" + sha + "_" + name[length - half_real_length_limit:length]
return name
def shutil_rmtree(*args, **kwargs):
path = args[0]
assert not os.path.samefile(path,
'/'), "Should not be trying to remove entire root directory: %s" % str(path)
assert not os.path.samefile(path,
'./'), "Should not be trying to remove entire local directory: %s" % str(path)
return shutil.rmtree(*args, **kwargs)
def remove(path: str):
try:
if path is not None and os.path.exists(path):
if os.path.isdir(path):
shutil_rmtree(path, ignore_errors=True)
else:
with contextlib.suppress(FileNotFoundError):
os.remove(path)
except:
pass
def makedirs(path, exist_ok=True, tmp_ok=False, use_base=False):
"""
Avoid some inefficiency in os.makedirs()
:param path:
:param exist_ok:
:param tmp_ok: use /tmp if can't write locally
:param use_base:
:return:
"""
if path is None:
return path
# if base path set, make relative to that, unless user_path absolute path
if use_base:
if os.path.normpath(path) == os.path.normpath(os.path.abspath(path)):
pass
else:
if os.getenv('H2OGPT_BASE_PATH') is not None:
base_dir = os.path.normpath(os.getenv('H2OGPT_BASE_PATH'))
path = os.path.normpath(path)
if not path.startswith(base_dir):
path = os.path.join(os.getenv('H2OGPT_BASE_PATH', ''), path)
path = os.path.normpath(path)
if os.path.isdir(path) and os.path.exists(path):
assert exist_ok, "Path already exists"
return path
try:
os.makedirs(path, exist_ok=exist_ok)
return path
except FileExistsError:
# e.g. soft link
return path
except PermissionError:
if tmp_ok:
path0 = path
path = os.path.join('/tmp/', path)
print("Permission denied to %s, using %s instead" % (path0, path), flush=True)
os.makedirs(path, exist_ok=exist_ok)
return path
else:
raise
def atomic_move_simple(src, dst):
try:
shutil.move(src, dst)
except (shutil.Error, FileExistsError):
pass
remove(src)
def atomic_copy(src="", dst=None, content=None):
my_uuid = uuid.uuid4()
src_tmp = None
if content is not None:
src_tmp = os.path.join('./', str(my_uuid))
with open(src_tmp, 'wt') as f:
f.write(content)
elif src != "":
src_tmp = src + str(my_uuid)
shutil.copy(src, src_tmp)
if src_tmp is not None:
makedirs(os.path.dirname(dst), exist_ok=True)
shutil.move(src_tmp, dst)
remove(src_tmp)
def move_tree(src, dst, include_root=True):
makedirs(dst, exist_ok=True)
if include_root:
shutil.move(src, dst)
else:
for (path, dirs, files) in os.walk(src):
new_path = path.replace(src, dst)
makedirs(new_path, exist_ok=True)
for file in files:
filename = os.path.join(path, file)
new_filename = os.path.join(new_path, file)
# print("%s -> %s" % (filename, new_filename))
try:
# only move if file doesn't already exist
# this ensures use earliest installation if used for pip install race avoidance
if not os.path.isfile(new_filename):
shutil.move(filename, new_filename)
except FileExistsError:
pass
for (path, dirs, files) in os.walk(src):
shutil.rmtree(path, ignore_errors=True)
def copy_tree(src, dst, follow_symlink=False):
makedirs(dst, exist_ok=True)
for (path, dirs, files) in os.walk(src, followlinks=follow_symlink):
new_path = path.replace(src, dst)
makedirs(new_path, exist_ok=True)
for file in files:
filename = os.path.join(path, file)
new_filename = os.path.join(new_path, file)
# print("%s -> %s" % (filename, new_filename))
try:
atomic_copy(filename, new_filename)
except FileNotFoundError:
pass
def download_simple(url, dest=None, overwrite=False, verbose=False):
if dest is None:
dest = os.path.basename(url)
base_path = os.path.dirname(dest)
if base_path: # else local path
base_path = makedirs(base_path, exist_ok=True, tmp_ok=True, use_base=True)
dest = os.path.join(base_path, os.path.basename(dest))
if os.path.isfile(dest):
if not overwrite:
if verbose:
print("Already have %s from url %s, delete file if invalid" % (dest, str(url)), flush=True)
return dest
else:
remove(dest)
if verbose:
print("BEGIN get url %s" % str(url), flush=True)
if url.startswith("file://"):
from requests_file import FileAdapter
s = requests.Session()
s.mount('file://', FileAdapter())
url_data = s.get(url, stream=True)
else:
url_data = requests.get(url, stream=True)
if verbose:
print("GOT url %s" % str(url), flush=True)
if url_data.status_code != requests.codes.ok:
msg = "Cannot get url %s, code: %s, reason: %s" % (
str(url),
str(url_data.status_code),
str(url_data.reason),
)
raise requests.exceptions.RequestException(msg)
url_data.raw.decode_content = True
uuid_tmp = str(uuid.uuid4())[:6]
dest_tmp = dest + "_dl_" + uuid_tmp + ".tmp"
# Sizes in bytes.
total_size = int(url_data.headers.get("content-length", 0))
block_size = 1024
with tqdm(total=total_size, unit="B", unit_scale=True) as progress_bar:
with open(dest_tmp, "wb") as file:
for data in url_data.iter_content(block_size):
progress_bar.update(len(data))
file.write(data)
if total_size != 0 and progress_bar.n != total_size:
raise RuntimeError("Could not download file")
atomic_move_simple(dest_tmp, dest)
if verbose:
print("DONE url %s" % str(url), flush=True)
return dest
def download(url, dest=None, dest_path=None):
if dest_path is not None:
dest = os.path.join(dest_path, os.path.basename(url))
if os.path.isfile(dest):
print("already downloaded %s -> %s" % (url, dest))
return dest
elif dest is not None:
if os.path.exists(dest):
print("already downloaded %s -> %s" % (url, dest))
return dest
else:
uuid_tmp = "dl2_" + str(uuid.uuid4())[:6]
dest = uuid_tmp + os.path.basename(url)
print("downloading %s to %s" % (url, dest))
if url.startswith("file://"):
from requests_file import FileAdapter
s = requests.Session()
s.mount('file://', FileAdapter())
url_data = s.get(url, stream=True)
else:
url_data = requests.get(url, stream=True)
if url_data.status_code != requests.codes.ok:
msg = "Cannot get url %s, code: %s, reason: %s" % (
str(url), str(url_data.status_code), str(url_data.reason))
raise requests.exceptions.RequestException(msg)
url_data.raw.decode_content = True
dirname = os.path.dirname(dest)
if dirname != "" and not os.path.isdir(dirname):
base_path = os.path.dirname(dest)
base_path = makedirs(base_path, exist_ok=True, tmp_ok=True, use_base=True)
dest = os.path.join(base_path, os.path.basename(dest))
uuid_tmp = "dl3_" + str(uuid.uuid4())[:6]
dest_tmp = dest + "_" + uuid_tmp + ".tmp"
with open(dest_tmp, 'wb') as f:
shutil.copyfileobj(url_data.raw, f)
try:
shutil.move(dest_tmp, dest)
except FileExistsError:
pass
remove(dest_tmp)
return dest
def get_doc(x):
return x.page_content
def get_source(x):
return x.metadata.get('source', "UNKNOWN SOURCE")
def markdown_to_html(content):
import markdown
# Create a Markdown object
markdowner = markdown.Markdown()
# Convert the Markdown block to HTML
try:
html = markdowner.reset().convert(content)
except Exception as e:
# FIXME:
print("Invalid conversion of markdown to html: %s\n\n%s" % (content, str(e)))
html = content
return html
def is_markdown(string):
"""Returns True if the string is markdown, False otherwise."""
# Check for the presence of double square brackets
if re.search(r'\[\[.+?\]\]', string):
return True
# Check for the presence of angle brackets
if re.search(r'<.+?>', string):
return False
# If neither of the above patterns are found, assume the string is markdown
return True
def get_accordion_named(content, title, font_size=8):
# content = content.replace('\n', '<br>')
if is_markdown(content):
content = markdown_to_html(content)
return f"""<details><summary><font size="{font_size}">{title}</font></summary><font size="{font_size}">{content}</font></details>"""
def hyde_titles(level):
if level == 0:
title = "HYDE 0: LLM"
elif level == 1:
title = "HYDE 1: Prompt+LLM embedding"
elif level == 2:
title = "HYDE 2: Prompt+LLM+HYDE 1 embedding"
elif level == 3:
title = "HYDE 3: Prompt+LLM+HYDE 1&2 embedding"
else:
title = "HYDE 4: Prompt+LLM+HYDE 1&2&3 embedding"
return title
def get_accordion(x, font_size=2, head_acc=50):
title = x.page_content[:head_acc].replace("\n", ' ').replace("<br>", ' ').replace("<p>", ' ').replace("\r", ' ')
content = x.page_content
return f"""<details><summary><font size="{font_size}">{title}</font></summary><font size="{font_size}">{content}</font></details>"""
def get_url(x, from_str=False, short_name=False, font_size=2):
if not from_str:
source = x.metadata['source']
else:
source = x
if short_name:
source_name = get_short_name(source)
else:
source_name = source
if source.startswith('http://') or source.startswith('https://'):
return """<font size="%s"><a href="%s" target="_blank" rel="noopener noreferrer">%s</a></font>""" % (
font_size, source, source_name)
elif '<a href=' not in source:
return """<font size="%s"><a href="file:///%s" target="_blank" rel="noopener noreferrer">%s</a></font>""" % (
font_size, source, source_name)
else:
# already filled
return source
def get_short_name(name, maxl=50):
if name is None:
return ''
length = len(name)
if length > maxl:
allow_length = maxl - 3
half_allowed = max(1, int(allow_length / 2))
name = name[0:half_allowed] + "..." + name[length - half_allowed:length]
return name
def cuda_vis_check(total_gpus):
"""Helper function to count GPUs by environment variable
Stolen from Jon's h2o4gpu utils
"""
cudavis = os.getenv("CUDA_VISIBLE_DEVICES")
which_gpus = []
if cudavis is not None:
# prune away white-space, non-numerics,
# except commas for simple checking
cudavis = "".join(cudavis.split())
import re
cudavis = re.sub("[^0-9,]", "", cudavis)
lencudavis = len(cudavis)
if lencudavis == 0:
total_gpus = 0
else:
total_gpus = min(
total_gpus,
os.getenv("CUDA_VISIBLE_DEVICES").count(",") + 1)
which_gpus = os.getenv("CUDA_VISIBLE_DEVICES").split(",")
which_gpus = [int(x) for x in which_gpus]
else:
which_gpus = list(range(0, total_gpus))
return total_gpus, which_gpus
def get_ngpus_vis(raise_if_exception=True):
ngpus_vis1 = None
shell = False
if shell:
cmd = "nvidia-smi -L 2> /dev/null"
else:
cmd = ["nvidia-smi", "-L"]
try:
timeout = 5 * 3
o = subprocess.check_output(cmd, shell=shell, timeout=timeout)
lines = o.decode("utf-8").splitlines()
ngpus_vis1 = 0
for line in lines:
if 'Failed to initialize NVML' not in line:
ngpus_vis1 += 1
except (FileNotFoundError, subprocess.CalledProcessError, OSError):
# GPU systems might not have nvidia-smi, so can't fail
pass
except subprocess.TimeoutExpired as e:
print('Failed get_ngpus_vis: %s' % str(e))
if raise_if_exception:
raise
if ngpus_vis1 is None:
import torch
if get_device() == 'cuda':
ngpus_vis1 = torch.cuda.device_count() if torch.cuda.is_available() else 0
else:
ngpus_vis1 = 0
ngpus_vis1, which_gpus = cuda_vis_check(ngpus_vis1)
return ngpus_vis1
def get_mem_gpus(raise_if_exception=True, ngpus=None):
totalmem_gpus1 = 0
usedmem_gpus1 = 0
freemem_gpus1 = 0
if ngpus == 0:
return totalmem_gpus1, usedmem_gpus1, freemem_gpus1
try:
cmd = "nvidia-smi -q 2> /dev/null | grep -A 3 'FB Memory Usage'"
o = subprocess.check_output(cmd, shell=True, timeout=15)
lines = o.decode("utf-8").splitlines()
for line in lines:
if 'Total' in line:
totalmem_gpus1 += int(line.split()[2]) * 1024 ** 2
if 'Used' in line:
usedmem_gpus1 += int(line.split()[2]) * 1024 ** 2
if 'Free' in line:
freemem_gpus1 += int(line.split()[2]) * 1024 ** 2
except (FileNotFoundError, subprocess.CalledProcessError, OSError):
# GPU systems might not have nvidia-smi, so can't fail
pass
except subprocess.TimeoutExpired as e:
print('Failed get_mem_gpus: %s' % str(e))
if raise_if_exception:
raise
return totalmem_gpus1, usedmem_gpus1, freemem_gpus1
n_gpus_global = get_ngpus_vis()
class ForkContext(threading.local):
"""
Set context for forking
Ensures state is returned once done
"""
def __init__(self, args=None, kwargs=None, forkdata_capable=True):
"""
:param args:
:param kwargs:
:param forkdata_capable: whether fork is forkdata capable and will use copy-on-write forking of args/kwargs
"""
self.forkdata_capable = forkdata_capable
if self.forkdata_capable:
self.has_args = args is not None
self.has_kwargs = kwargs is not None
forkdatacontext.args = args
forkdatacontext.kwargs = kwargs
else:
self.has_args = False
self.has_kwargs = False
def __enter__(self):
try:
# flush all outputs so doesn't happen during fork -- don't print/log inside ForkContext contexts!
sys.stdout.flush()
sys.stderr.flush()
except BaseException as e:
# exit not called if exception, and don't want to leave forkdatacontext filled in that case
print("ForkContext failure on enter: %s" % str(e))
self.finally_act()
raise
return self
def __exit__(self, exc_type, exc_value, exc_traceback):
self.finally_act()
def finally_act(self):
"""
Done when exception hit or exit is reached in context
first reset forkdatacontext as crucial to have reset even if later 2 calls fail
:return: None
"""
if self.forkdata_capable and (self.has_args or self.has_kwargs):
forkdatacontext._reset()
class _ForkDataContext(threading.local):
def __init__(
self,
args=None,
kwargs=None,
):
"""
Global context for fork to carry data to subprocess instead of relying upon copy/pickle/serialization
:param args: args
:param kwargs: kwargs
"""
assert isinstance(args, (tuple, type(None)))
assert isinstance(kwargs, (dict, type(None)))
self.__args = args
self.__kwargs = kwargs
@property
def args(self) -> Tuple:
"""returns args"""
return self.__args
@args.setter
def args(self, args):
if self.__args is not None:
raise AttributeError(
"args cannot be overwritten: %s %s" % (str(self.__args), str(self.__kwargs))
)
self.__args = args
@property
def kwargs(self) -> Dict:
"""returns kwargs"""
return self.__kwargs
@kwargs.setter
def kwargs(self, kwargs):
if self.__kwargs is not None:
raise AttributeError(
"kwargs cannot be overwritten: %s %s" % (str(self.__args), str(self.__kwargs))
)
self.__kwargs = kwargs
def _reset(self):
"""Reset fork arg-kwarg context to default values"""
self.__args = None
self.__kwargs = None
def get_args_kwargs(self, func, args, kwargs) -> Tuple[Callable, Tuple, Dict]:
if self.__args:
args = self.__args[1:]
if not func:
assert len(self.__args) > 0, "if have no func, must have in args"
func = self.__args[0] # should always be there
if self.__kwargs:
kwargs = self.__kwargs
try:
return func, args, kwargs
finally:
forkdatacontext._reset()
@staticmethod
def get_args_kwargs_for_traced_func(func, args, kwargs):
"""
Return args/kwargs out of forkdatacontext when using copy-on-write way of passing args/kwargs
:param func: actual function ran by _traced_func, which itself is directly what mppool treats as function
:param args:
:param kwargs:
:return: func, args, kwargs from forkdatacontext if used, else originals
"""
# first 3 lines are debug
func_was_None = func is None
args_was_None_or_empty = args is None or len(args) == 0
kwargs_was_None_or_empty = kwargs is None or len(kwargs) == 0
forkdatacontext_args_was_None = forkdatacontext.args is None
forkdatacontext_kwargs_was_None = forkdatacontext.kwargs is None
func, args, kwargs = forkdatacontext.get_args_kwargs(func, args, kwargs)
using_forkdatacontext = func_was_None and func is not None # pulled func out of forkdatacontext.__args[0]
assert forkdatacontext.args is None, "forkdatacontext.args should be None after get_args_kwargs"
assert forkdatacontext.kwargs is None, "forkdatacontext.kwargs should be None after get_args_kwargs"
proc_type = kwargs.get('proc_type', 'SUBPROCESS')
if using_forkdatacontext:
assert proc_type == "SUBPROCESS" or proc_type == "SUBPROCESS"
if proc_type == "NORMAL":
assert forkdatacontext_args_was_None, "if no fork, expect forkdatacontext.args None entering _traced_func"
assert forkdatacontext_kwargs_was_None, "if no fork, expect forkdatacontext.kwargs None entering _traced_func"
assert func is not None, "function should not be None, indicates original args[0] was None or args was None"
return func, args, kwargs
def using_conda():
"""
Whether using conda and want to use conda
:return:
"""
import os, sys
return os.path.exists(os.path.join(sys.prefix, 'conda-meta')) and os.environ.get('AVOID_FULL_CONDA') is None
def get_python_paths():
"""
Various python paths, same as make/get_python_paths.sh
:return:
"""
import os, sys
exec_file = sys.executable
bpath = os.path.dirname(sys.executable)
rootpath = os.path.dirname(os.path.dirname(sys.executable))
libpath = os.path.join(rootpath, "lib")
includepath = os.path.join(rootpath, "include")
from sysconfig import get_paths
info = get_paths()
spackagespath = info['purelib']
pincludepath = info['platinclude']
plibpath = info['platstdlib']
from distutils.sysconfig import get_config_var
plibfile = '%s/%s' % (get_config_var('LIBDIR'), get_config_var('INSTSONAME'))
return dict(exec_file=exec_file, bpath=bpath, rootpath=rootpath, libpath=libpath, includepath=includepath,
spackagespath=spackagespath, pincludepath=pincludepath, plibpath=plibpath, plibfile=plibfile)
forkdatacontext = _ForkDataContext()
def _traced_func(func, *args, **kwargs):
try:
func, args, kwargs = forkdatacontext.get_args_kwargs_for_traced_func(func, args, kwargs)
return func(*args, **kwargs)
except BaseException as e:
print(e)
ex = traceback.format_exc()
raise RuntimeError(str(ex))
def call_subprocess_onetask(func, args=None, kwargs=None):
if platform.system() in ['Darwin', 'Windows']:
return func(*args, **kwargs)
if isinstance(args, list):
args = tuple(args)
if args is None:
args = ()
if kwargs is None:
kwargs = {}
args = list(args)
args = [func] + args
args = tuple(args)
with ForkContext(args=args, kwargs=kwargs):
args = (None,)
kwargs = {}
with ProcessPoolExecutor(max_workers=1) as executor:
future = executor.submit(_traced_func, *args, **kwargs)
return future.result()
class ProgressParallel(Parallel):
def __init__(self, use_tqdm=True, total=None, *args, **kwargs):
self._use_tqdm = use_tqdm
self._total = total
super().__init__(*args, **kwargs)
def __call__(self, *args, **kwargs):
with tqdm(disable=not self._use_tqdm, total=self._total) as self._pbar:
return Parallel.__call__(self, *args, **kwargs)
def print_progress(self):
if self._total is None:
self._pbar.total = self.n_dispatched_tasks
self._pbar.n = self.n_completed_tasks
self._pbar.refresh()
def get_kwargs(func, exclude_names=None, **kwargs):
func_names = list(inspect.signature(func).parameters)
missing_kwargs = [x for x in func_names if x not in kwargs]
if exclude_names:
for k in exclude_names:
if k in missing_kwargs:
missing_kwargs.remove(k)
if k in func_names:
func_names.remove(k)
assert not missing_kwargs, "Missing %s" % missing_kwargs
kwargs = {k: v for k, v in kwargs.items() if k in func_names}
return kwargs
from importlib.metadata import distribution, PackageNotFoundError
have_faiss = False
try:
assert distribution('faiss') is not None
have_faiss = True
except (PackageNotFoundError, AssertionError):
pass
try:
assert distribution('faiss_gpu') is not None
have_faiss = True
except (PackageNotFoundError, AssertionError):
pass
try:
assert distribution('faiss_cpu') is not None
have_faiss = True
except (PackageNotFoundError, AssertionError):
pass
have_serpapi = False
try:
assert distribution('google-search-results') is not None
have_serpapi = True
except (PackageNotFoundError, AssertionError):
pass
have_autogen = False
try:
assert distribution('pyautogen') is not None
have_autogen = True
except (PackageNotFoundError, AssertionError):
pass
def hash_file(file):
try:
import hashlib
# BUF_SIZE is totally arbitrary, change for your app!
BUF_SIZE = 65536 # lets read stuff in 64kb chunks!
md5 = hashlib.md5()
# sha1 = hashlib.sha1()
if not os.path.isfile(file):
md5.update(file.encode(encoding='UTF-8'))
else:
with open(file, 'rb') as f:
while True:
data = f.read(BUF_SIZE)
if not data:
break
md5.update(data)
# sha1.update(data)
except BaseException as e:
print("Cannot hash %s due to %s" % (file, str(e)))
traceback.print_exc()
return ''
return md5.hexdigest()
def start_faulthandler():
# If hit server or any subprocess with signal SIGUSR1, it'll print out all threads stack trace, but wont't quit or coredump
# If more than one fork tries to write at same time, then looks corrupted.
import faulthandler
# SIGUSR1 in h2oai/__init__.py as well
faulthandler.enable()
if hasattr(faulthandler, 'register'):
# windows/mac
import signal
faulthandler.register(signal.SIGUSR1)
def get_hf_server(inference_server):
inf_split = inference_server.split(" ")
if len(inf_split) == 3:
assert len(inf_split) == 1 or len(inf_split) == 3
inference_server = inf_split[0]
headers = {"authorization": "%s %s" % (inf_split[1], inf_split[2])}
user = None
password = None
else:
ip_port_vllm = ':'.join(inference_server.split(':')[0:])
if ip_port_vllm.startswith('https://'):
http_prefix = 'https://'
ip_port_vllm = ip_port_vllm[len(http_prefix):]
elif ip_port_vllm.startswith('http://'):
http_prefix = 'http://'
ip_port_vllm = ip_port_vllm[len(http_prefix):]
else:
http_prefix = 'http://'
inf_split = ip_port_vllm.split(":")
if len(inf_split) <= 2:
# i.e. just DNS or IP and no port or IP + port
user = None
password = None
elif len(inf_split) == 3:
# i.e. just DNS or IP, no port + user + pass = 3
user = inf_split[len(inf_split) - 2]
password = inf_split[len(inf_split) - 1]
ip_port_vllm = ':'.join(inf_split[:len(inf_split) - 2])
elif len(inf_split) == 4:
# i.e. DNS/IP + port + user + pass = 4
port = inf_split[len(inf_split) - 3]
user = inf_split[len(inf_split) - 2]
password = inf_split[len(inf_split) - 1]
if port not in [None, 'None']:
ip_port_vllm = ':'.join([inf_split[0], port])
else:
ip_port_vllm = inf_split[0]
else:
raise ValueError("Malformed inference_server=%s" % inference_server)
headers = None
# remove None if port was None
if 'None' in ip_port_vllm.split(':'):
ip_port_vllm = ':'.join([x for x in ip_port_vllm.split(':') if x != 'None'])
inference_server = http_prefix + ip_port_vllm
return inference_server, headers, user, password
class FakeTokenizer:
"""
1) For keeping track of model_max_length
2) For when model doesn't directly expose tokenizer but need to count tokens
"""
def __init__(self, model_max_length=2048,
encoding_name="cl100k_base",
is_openai=False,
is_anthropic=False,
is_google=False,
is_hf=False,
tokenizer=None,
is_llama_cpp=False,
is_super_fake=False,
is_mistral=False,
):
if model_max_length is None:
assert not (
is_openai or is_anthropic or is_google), "Should have set model_max_length for OpenAI or Anthropic or Google"
model_max_length = 2048
self.is_openai = is_openai
self.is_anthropic = is_anthropic
self.is_google = is_google
self.is_hf = is_hf
self.is_llama_cpp = is_llama_cpp
self.is_super_fake = is_super_fake
self.is_mistral = is_mistral
self.tokenizer = tokenizer
self.model_max_length = model_max_length
if not self.is_openai and not self.is_anthropic and not self.is_llama_cpp:
# don't push limit, since if using fake tokenizer, only estimate, and seen underestimates by order 250
self.model_max_length -= 250
self.encoding_name = encoding_name
if self.is_super_fake:
self.encoding = None
# The first time this runs, it will require an internet connection to download. Later runs won't need an internet connection.
elif not (self.is_anthropic or self.is_google or self.is_mistral):
import tiktoken
self.encoding = tiktoken.get_encoding(self.encoding_name)
else:
self.encoding = None
def encode(self, x, *args, return_tensors="pt", **kwargs):
if not x:
return dict(input_ids=[])
if self.is_super_fake:
input_ids = self.heuristic_encode(x)
# avoid torch tensor
return dict(input_ids=input_ids)
elif self.is_llama_cpp: # and len(x) < 4 * 4 * self.model_max_length: # don't use llama.cpp if too much
input_ids = self.tokenizer.tokenize(b" " + x.encode("utf-8"))
elif self.is_anthropic:
from anthropic import Anthropic
client = Anthropic()
tokenizer = client.get_tokenizer()
input_ids = tokenizer.encode(x).ids
elif self.is_google:
input_ids = [0] * self.tokenizer(x).total_tokens # fake tokens
elif self.is_hf:
input_ids = self.tokenizer.encode(x)
elif self.is_mistral:
from mistral_common.protocol.instruct.request import ChatCompletionRequest
input_ids = self.tokenizer.encode_chat_completion(
ChatCompletionRequest(messages=[dict(role='user', content=x)])).tokens
else:
input_ids = self.encoding.encode(x, disallowed_special=())
if return_tensors == 'pt' and isinstance(input_ids, list):
import torch
input_ids = torch.tensor(input_ids)
return dict(input_ids=input_ids)
def decode(self, x, *args, **kwargs):
if self.is_super_fake:
return ['aaaa'] * len(x) # fake
elif self.is_llama_cpp: # and len(x) < 4 * self.model_max_length: # don't use llama.cpp if too much
return self.tokenizer.detokenize(x)
elif self.is_anthropic:
from anthropic import Anthropic
client = Anthropic()
tokenizer = client.get_tokenizer()
return tokenizer.decode(x)
elif self.is_google:
return ['a'] * len(x) # fake
elif self.is_mistral:
return ['a'] * len(x) # fake
elif self.is_hf:
return self.tokenizer.decode(x)
# input is input_ids[0] form
return self.encoding.decode(x)
def num_tokens_from_string(self, prompt: str) -> int:
"""Returns the number of tokens in a text string."""
if self.is_super_fake:
return len(self.heuristic_encode(prompt))
elif self.is_anthropic:
from anthropic import Anthropic
client = Anthropic()
return client.count_tokens(prompt)
elif self.is_google:
return self.tokenizer(prompt)
elif self.is_mistral:
return len(self.encode(prompt))
elif self.is_hf:
return len(self.tokenizer.encode(prompt))
num_tokens = len(self.encode(prompt)['input_ids'])
return num_tokens
def heuristic_encode(self, text: str) -> list:
"""
A heuristic-based approach to estimate token counts.
"""
total_tokens = len(text) // 4 if len(text) >= 4 else 1
return [0] * total_tokens
def __call__(self, x, *args, **kwargs):
return self.encode(x, *args, **kwargs)
def get_local_ip():
import socket
s = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
try:
# doesn't even have to be reachable
s.connect(('10.255.255.255', 1))
IP = s.getsockname()[0]
except Exception:
IP = '127.0.0.1'
finally:
s.close()
return IP
try:
assert distribution('langchain') is not None
have_langchain = True
except (PackageNotFoundError, AssertionError):
have_langchain = False
import distutils.spawn
have_tesseract = distutils.spawn.find_executable("tesseract")
have_libreoffice = distutils.spawn.find_executable("libreoffice")
try:
from weasyprint import HTML
import doctr
have_doctr = True
except:
have_doctr = False
try:
assert distribution('arxiv') is not None
assert distribution('pymupdf') is not None
have_arxiv = True
except (PackageNotFoundError, AssertionError):
have_arxiv = False
try:
assert distribution('pymupdf') is not None
have_pymupdf = True
except (PackageNotFoundError, AssertionError):
have_pymupdf = False
have_pymupdf4llm = False
try:
assert distribution('pymupdf4llm') is not None
have_pymupdf4llm = False # too slow, avoid for now
except (PackageNotFoundError, AssertionError):
pass
try:
assert distribution('selenium') is not None
have_selenium = True
except (PackageNotFoundError, AssertionError):
have_selenium = False
try:
assert distribution('pillow') is not None
have_pillow = True
except (PackageNotFoundError, AssertionError):
have_pillow = False
try:
assert distribution('playwright') is not None
have_playwright = True
except (PackageNotFoundError, AssertionError):
have_playwright = False
try:
assert distribution('jq') is not None
have_jq = True
except (PackageNotFoundError, AssertionError):
have_jq = False
try:
assert distribution('optimum') is not None
have_optimum = True
except (PackageNotFoundError, AssertionError):
have_optimum = False
try:
assert distribution('librosa') is not None
have_librosa = True
except (PackageNotFoundError, AssertionError):
have_librosa = False
try:
assert distribution('wavio') is not None
have_wavio = True
except (PackageNotFoundError, AssertionError):
have_wavio = False
try:
assert distribution('soundfile') is not None
have_soundfile = True
except (PackageNotFoundError, AssertionError):
have_soundfile = False
try:
assert distribution('deepspeed') is not None
have_deepspeed = True
except (PackageNotFoundError, AssertionError):
have_deepspeed = False
try:
assert distribution('emoji') is not None
have_emoji = True
except (PackageNotFoundError, AssertionError):
have_emoji = False
try:
assert distribution('langid') is not None
have_langid = True
except (PackageNotFoundError, AssertionError):
have_langid = False
try:
assert distribution('TTS') is not None
have_TTS = True
except (PackageNotFoundError, AssertionError):
have_TTS = False
try:
assert distribution('faster_whisper') is not None
have_use_faster = True
except (PackageNotFoundError, AssertionError):
have_use_faster = False
try:
assert distribution('flash_attn') is not None
have_flash_attention = True
have_flash_attention_2 = distribution('flash_attn').version.startswith('2.')
except (PackageNotFoundError, AssertionError):
have_flash_attention = False
have_flash_attention_2 = False
try:
assert distribution('gradio') is not None
have_gradio = True
is_gradio_version4 = distribution('gradio').version.startswith('4.')
except (PackageNotFoundError, AssertionError):
have_gradio = False
is_gradio_version4 = False
try:
assert distribution('gradio_pdf') is not None
have_gradio_pdf = is_gradio_version4
except (PackageNotFoundError, AssertionError):
have_gradio_pdf = False
try:
assert distribution('pyrubberband') is not None
have_pyrubberband = True
except (PackageNotFoundError, AssertionError):
have_pyrubberband = False
try:
assert distribution('fiftyone') is not None
have_fiftyone = True
except (PackageNotFoundError, AssertionError):
have_fiftyone = False
try:
assert distribution('diffusers') is not None
have_diffusers = True
except (PackageNotFoundError, AssertionError):
have_diffusers = False
try:
assert distribution('opencv-python-headless') is not None
have_cv2 = True
except (PackageNotFoundError, AssertionError):
try:
assert distribution('opencv-python') is not None
have_cv2 = True
except (PackageNotFoundError, AssertionError):
have_cv2 = False
only_unstructured_urls = os.environ.get("ONLY_UNSTRUCTURED_URLS", "0") == "1"
only_selenium = os.environ.get("ONLY_SELENIUM", "0") == "1"
only_playwright = os.environ.get("ONLY_PLAYWRIGHT", "0") == "1"
def set_openai(inference_server, model_name=None):
if inference_server.startswith('sglang'):
inference_server_split = inference_server.split(':')
inference_server_split[1] = None
inference_server = ':'.join([x for x in inference_server_split if x is not None])
if inference_server.startswith('vllm') or inference_server.startswith('sglang'):
api_key = "EMPTY"
inf_type = inference_server.split(':')[0].strip()
ip_port = ':'.join(inference_server.split(':')[1:])
if ip_port.startswith('https://'):
http_prefix = 'https://'
ip_port = ip_port[len(http_prefix):]
auto_v1 = False
elif ip_port.startswith('http://'):
http_prefix = 'http://'
ip_port = ip_port[len(http_prefix):]
auto_v1 = False
else:
http_prefix = 'http://'
auto_v1 = True
if inference_server.startswith('sglang') and '/v1' not in inference_server:
auto_v1 = True
address = ':'.join(ip_port.split(':')[0:1]).strip()
api_base = http_prefix + address
if len(ip_port.split(':')) >= 2:
port = ip_port.split(':')[1].strip()
if port not in [None, 'None']:
api_base += ':' + port
if len(ip_port.split(':')) >= 3:
# if not there, use EMPTY as default
url_path = ip_port.split(':')[2].strip()
if url_path not in [None, 'None']:
api_base += url_path # assume includes prefix of / and /v1
if auto_v1 and not api_base.endswith('/v1'):
api_base += '/v1'
if len(ip_port.split(':')) >= 4:
# if not there, use EMPTY as default
api_key = ip_port.split(':')[3].strip()
from openai import OpenAI, AsyncOpenAI
client_args = dict(base_url=api_base, api_key=api_key)
client = OpenAI(**client_args)
async_client = AsyncOpenAI(**client_args)
return client, async_client, inf_type, None, api_base, None, api_key
else:
api_key = os.getenv("OPENAI_API_KEY")
base_url = None
deployment_type = None
api_version = None
inf_type = inference_server.split(':')[0].strip()
if len(inference_server.split(':')) >= 2:
deployment_type = inference_server.split(':')[1].strip()
if len(inference_server.split(':')) >= 3:
base_url = inference_server.split(':')[2].strip()
base_url = 'https://' + base_url
if len(inference_server.split(':')) >= 4:
api_version = inference_server.split(':')[3].strip()
if inference_server.startswith('openai_azure'):
if api_version in ['None', None]:
# for function tools support
# https://github.com/Azure/azure-rest-api-specs/tree/main/specification/cognitiveservices/data-plane/AzureOpenAI/inference/preview/2023-12-01-preview
# https://learn.microsoft.com/en-us/azure/ai-services/openai/api-version-deprecation
# https://learn.microsoft.com/en-us/azure/ai-services/openai/how-to/function-calling
api_version = "2024-07-01-preview"
if os.getenv('OPENAI_AZURE_KEY') is not None:
# use this instead if exists
api_key = os.getenv("OPENAI_AZURE_KEY")
elif api_version in ['None', None]:
api_version = None
if len(inference_server.split(':')) >= 5:
api_key0 = inference_server.split(':')[4].strip()
if api_key0 not in ['None', None]:
api_key = api_key0
if deployment_type == 'None':
deployment_type = None
if base_url == 'None':
base_url = None
if base_url == 'None':
base_url = None
# cannot use non-chat model, uses old openai. stuff if go through to H2OOpenAI with chat model
if model_name:
chat_model = (model_name.startswith("gpt-3.5-turbo") or model_name.startswith(
"gpt-4")) and "-instruct" not in model_name
if chat_model and inf_type == 'openai_azure':
inf_type = 'openai_azure_chat'
if chat_model and inf_type == 'openai':
inf_type = 'openai_chat'
from openai import OpenAI, AzureOpenAI, AsyncOpenAI, AsyncAzureOpenAI
if inf_type in ['openai_azure', 'openai_azure_chat']:
client_args = dict(azure_deployment=deployment_type, azure_endpoint=base_url, api_version=api_version,
api_key=api_key)
client = AzureOpenAI(**client_args)
async_client = AsyncAzureOpenAI(**client_args)
else:
client_args = dict(base_url=base_url, api_key=api_key)
client = OpenAI(**client_args)
async_client = AsyncOpenAI(**client_args)
return client, async_client, inf_type, deployment_type, base_url, api_version, api_key
def get_model_name(model_name, openai_client):
if os.getenv('DISABLE_OPENAI_AUTO_MODEL_NAME', '0') == '1':
return model_name
# override, required for lmdeploy
# https://github.com/InternLM/lmdeploy/issues/1674
# https://github.com/InternLM/lmdeploy/blob/e6468e7afda6b29d4c065f296a4e893b52bd33d5/lmdeploy/serve/proxy/proxy.py#L320
# https://lmdeploy.readthedocs.io/en/latest/serving/api_server.html#restful-api
try:
model_names = openai_client.models.list().data
if len(model_names) == 1:
model_name = openai_client.models.list().data[0].id
else:
print("Too few or too many models in list so do not know which to chose: given: %s list: %s" % (
model_name, model_names))
except Exception as e:
print(f"Failed to get model name from OpenAI client, using default {model_name}: {str(e)}")
return model_name
def get_list_or_str(x):
if isinstance(x, list):
return x
elif isinstance(x, str):
try:
x1 = ast.literal_eval(x)
assert isinstance(x1, list)
return x1
except:
return x
else:
return x
def deepcopy_by_pickle_object(object):
"""
Faster deepcopy, can only work on things that are picklable. Naive Deepcopy is more general.
Same method as for class Individual
:param object:
:return:
"""
gc.disable()
new_object = pickle.loads(pickle.dumps(object, -1))
gc.enable()
return new_object
def url_alive(url):
if not isinstance(url, str):
return False
try:
response = requests.head(url)
except Exception as e:
return False
else:
if response.status_code in [200, 301, 302, 307]:
return True
else:
return False
def return_good_url(url):
# ignore status code, just see if exists or not
for prefix in ['', 'https://', 'http://', 'https://www.', 'http://www.']:
try:
url_test = prefix + url
response = requests.head(url_test, timeout=10)
except requests.exceptions.Timeout as e:
response = None
url_test = None
except Exception as e:
response = None
url_test = None
if response is not None:
# and response.status_code < 400:
# don't do status check, if got status, then is real URL regardless of goodness, not text
return url_test
return None
def is_probably_url(url):
if not isinstance(url, str):
return False
# url_alive too slow
return any(url.startswith(prefix) for prefix in ['www.', 'http://', 'https://', 'https://www.', 'http://www.'])
def dict_to_html(x, small=True, api=False):
x = {k: v if not in_gradio_root(v) and not is_probably_url(v) else get_url(v, from_str=True, short_name=True) for
k, v in x.items()}
df = pd.DataFrame(x.items(), columns=['Key', 'Value'])
df.index = df.index + 1
df.index.name = 'index'
if api:
return tabulate.tabulate(df, headers='keys')
else:
res = tabulate.tabulate(df, headers='keys', tablefmt='unsafehtml')
if small:
return "<small>" + res + "</small>"
else:
return res
def split_into_sentences(text):
# Split text by specified punctuation followed by space or end of text
sentences = re.split(r'(?<=[.!?]) +', text)
return sentences
def text_to_html(x, api=False):
if api:
return x
return """
<style>
pre {
overflow-x: auto;
white-space: pre-wrap;
white-space: -moz-pre-wrap;
white-space: -pre-wrap;
white-space: -o-pre-wrap;
word-wrap: break-word;
}
</style>
<pre>
%s
</pre>
""" % '<br>'.join(split_into_sentences(x))
def lg_to_gr(
**kwargs,
):
# translate:
import torch
n_gpus = torch.cuda.device_count() if torch.cuda.is_available() else 0
n_gpus, _ = cuda_vis_check(n_gpus)
image_audio_loaders_options = ['Caption']
if n_gpus != 0:
image_audio_loaders_options.extend(['CaptionLarge', 'Pix2Struct'])
if have_tesseract:
image_audio_loaders_options.append('OCR')
if have_doctr:
image_audio_loaders_options.append('DocTR')
if have_librosa:
image_audio_loaders_options.append('ASR')
if n_gpus != 0:
image_audio_loaders_options.append('ASRLarge')
if kwargs['enable_llava'] and kwargs['llava_model']:
image_audio_loaders_options.append('LLaVa')
image_audio_loaders_options0 = []
if have_tesseract and kwargs['enable_ocr']:
image_audio_loaders_options0.append('OCR')
if have_doctr and kwargs['enable_doctr']:
image_audio_loaders_options0.append('DocTR')
if kwargs['enable_captions']:
if kwargs['max_quality'] and n_gpus > 0:
# BLIP2 only on GPU
image_audio_loaders_options0.append('CaptionLarge')
else:
image_audio_loaders_options0.append('Caption')
if have_librosa and kwargs['enable_transcriptions']:
if kwargs['max_quality'] and n_gpus > 0:
image_audio_loaders_options0.append('ASRLarge')
else:
image_audio_loaders_options0.append('ASR')
if kwargs['enable_llava'] and kwargs['llava_model'] and 'vllm' not in kwargs['llava_model']:
# Caption like llava model is only gradio based, legacy method
# and n_gpus > 0 # don't require local GPUs
# LLaVa better and faster if present
# and kwargs['max_quality']
image_audio_loaders_options0.append('LLaVa')
if 'Caption' in image_audio_loaders_options0:
image_audio_loaders_options0.remove('Caption')
if 'CaptionLarge' in image_audio_loaders_options0:
image_audio_loaders_options0.remove('CaptionLarge')
pdf_loaders_options = ['Unstructured', 'PyPDF', 'TryHTML']
if have_pymupdf:
pdf_loaders_options = ['PyMuPDF'] + pdf_loaders_options
if have_tesseract:
pdf_loaders_options.append('OCR')
if have_doctr:
pdf_loaders_options.append('DocTR')
pdf_loaders_options0 = []
if have_pymupdf and kwargs['use_pymupdf'] in [True, 'auto', 'on']:
pdf_loaders_options0.append('PyMuPDF')
if kwargs['enable_pdf_ocr'] in [True, 'on']:
pdf_loaders_options0.append('OCR')
if have_doctr and kwargs['enable_pdf_doctr'] in [True, 'on']:
pdf_loaders_options0.append('DocTR')
# in case my pymupdf, use pypdf as backup default
if kwargs['use_pypdf'] in [True, 'on'] and have_pymupdf or kwargs['use_pypdf'] in [True, 'auto',
'on'] and not have_pymupdf:
pdf_loaders_options0.append('PyPDF')
if kwargs['use_unstructured_pdf'] in [True, 'on']:
pdf_loaders_options0.append('Unstructured')
if kwargs['try_pdf_as_html'] in [True, 'on']:
pdf_loaders_options0.append('TryHTML')
url_loaders_options = []
if only_unstructured_urls:
url_loaders_options.append('Unstructured')
elif have_selenium and only_selenium:
url_loaders_options.append('Selenium')
elif have_playwright and only_playwright:
url_loaders_options.append('PlayWright')
else:
url_loaders_options.append('Unstructured')
if have_selenium:
url_loaders_options.append('Selenium')
if have_playwright:
url_loaders_options.append('PlayWright')
url_loaders_options.append('ScrapeWithPlayWright')
url_loaders_options.append('ScrapeWithHttp')
url_loaders_options0 = [url_loaders_options[0]]
assert set(image_audio_loaders_options0).issubset(image_audio_loaders_options), "%s %s" % (
image_audio_loaders_options0, image_audio_loaders_options)
assert set(pdf_loaders_options0).issubset(pdf_loaders_options), "%s %s" % (
pdf_loaders_options0, pdf_loaders_options)
assert set(url_loaders_options0).issubset(url_loaders_options), "%s %s" % (
url_loaders_options0, url_loaders_options)
return image_audio_loaders_options0, image_audio_loaders_options, \
pdf_loaders_options0, pdf_loaders_options, \
url_loaders_options0, url_loaders_options
def enqueue_output(file, queue):
# for line in iter(file.readline, ''):
for line in iter(file.readline, b'' if isinstance(file, io.BufferedReader) else ''):
queue.put(line)
file.close()
def read_popen_pipes(p):
with ThreadPoolExecutor(2) as pool:
q_stdout, q_stderr = Queue(), Queue()
pool.submit(enqueue_output, p.stdout, q_stdout)
pool.submit(enqueue_output, p.stderr, q_stderr)
while True:
if p.poll() is not None and q_stdout.empty() and q_stderr.empty():
break
out_line = err_line = ''
try:
out_line = q_stdout.get_nowait()
except Empty:
pass
try:
err_line = q_stderr.get_nowait()
except Empty:
pass
yield out_line, err_line
def start_process(cmd):
start_cmd = sys.executable + " -i -q -u"
print_cmd = 'print("{}")'
cmd = [start_cmd] + [cmd]
process = subprocess.Popen(cmd, stdout=subprocess.PIPE)
for c in iter(lambda: process.stdout.read(1), b''):
sys.stdout.write(c)
def execute_cmd_stream(cmd=None, script_content=None, cwd=None, env=None, timeout=None, capture_output=True,
text=True, print_tags=False, print_literal=True, print_func=print,
guard_func=None, sleep=0.05,
max_stream_length=4096, max_memory_usage=16*1024**3):
if script_content is None and cmd is None:
raise ValueError("Either script_content or cmd must be provided")
if script_content is not None:
script_path = 'temp_script.py'
with open(script_path, 'w') as f:
f.write(script_content)
cmd = [sys.executable, script_path]
else:
script_path = None
assert cmd, "cmd must be provided if script_content is None"
length = 0
try:
# Prepare Popen arguments
popen_kwargs = {
'cwd': cwd,
'env': env,
'bufsize': 1, # Line-buffered
'stdout': subprocess.PIPE,
'stderr': subprocess.PIPE,
'universal_newlines': text,
}
with subprocess.Popen(cmd, **popen_kwargs) as p:
# Start psutil process to monitor memory usage
psutil_process = psutil.Process(p.pid)
sel = selectors.DefaultSelector()
sel.register(p.stdout, selectors.EVENT_READ)
sel.register(p.stderr, selectors.EVENT_READ)
stdout_data = []
stderr_data = []
start_time = time.time()
while True:
if timeout and time.time() - start_time > timeout:
p.terminate()
raise subprocess.TimeoutExpired(cmd, timeout)
# Monitor memory usage for the main process and all its children
if max_memory_usage:
measure_t0 = time.time()
try:
# Get memory usage of the main process and its children
mem_info = psutil_process.memory_info().rss
children = psutil_process.children(recursive=True)
for child in children:
mem_info += child.memory_info().rss
except psutil.NoSuchProcess:
mem_info = 0
# Check if the total memory usage exceeds the limit
if mem_info > max_memory_usage:
try:
p.terminate()
except Exception as e:
print(f"Error terminating process: {e}")
try:
p.kill()
except Exception as e:
print(f"Error killing process: {e}")
error = f"Process and its children used memory {mem_info} that exceeded memory limit of {max_memory_usage} bytes detected in {time.time() - measure_t0}."
stderr_data.append(error)
print(f"OOM on cmd:\n\n{cmd}\n\n", flush=True, file=sys.stderr)
events = sel.select(timeout=1)
if not events and p.poll() is not None:
break # No more events and the process has exited
for key, _ in events:
data = key.fileobj.readline()
if not data: # EOF
sel.unregister(key.fileobj)
continue
if guard_func:
data = guard_func(data)
if key.fileobj is p.stdout:
stdout_data.append(data)
if length + len(data) <= max_stream_length:
if print_tags:
if data.strip():
print_func(f"STDOUT: {data.strip()}")
elif print_literal:
print_func(data, end='')
else:
print_func(data)
length += len(data)
elif key.fileobj is p.stderr:
stderr_data.append(data)
if length + len(data) <= max_stream_length:
if print_tags:
if data.strip():
print_func(f"STDERR: {data.strip()}")
elif print_literal:
print_func(data, end='')
else:
print_func(data)
length += len(data)
if p.poll() is not None and not sel.get_map():
break # Process has exited and no more data to read
# sleep shouldn't be too long or else will get chunky streaming and not detect memory usage rapidly enough
# sleep shouldn't be too short or else will constantly be doing psutil stuff
time.sleep(sleep)
p.wait(timeout=timeout)
# Prepare return object similar to subprocess.CompletedProcess
return subprocess.CompletedProcess(
args=cmd,
returncode=p.returncode,
stdout=''.join(stdout_data) if capture_output else None,
stderr=''.join(stderr_data) if capture_output else None
)
finally:
if script_path and os.path.exists(script_path):
os.remove(script_path)
def str_to_list(x, allow_none=False):
if isinstance(x, str):
if len(x.strip()) > 0:
if x.strip().startswith('['):
try:
x = ast.literal_eval(x.strip())
except Exception:
print("bad x: %s" % x, flush=True)
raise
else:
raise ValueError("Invalid str_to_list for %s" % x)
else:
x = []
elif x is None and not allow_none:
x = []
if allow_none:
assert isinstance(x, (type(None), list))
else:
assert isinstance(x, list)
return x
def str_to_dict(x):
if isinstance(x, str):
if len(x.strip()) > 0:
if x.strip().startswith('{'):
x = ast.literal_eval(x.strip())
else:
raise ValueError("Invalid str_to_dict for %s" % x)
else:
x = {}
elif x is None:
x = {}
assert isinstance(x, dict)
return x
def get_token_count(x, tokenizer, token_count_fun=None, add_special_tokens=True):
# NOTE: Somewhat duplicates H2OTextGenerationPipeline.get_token_count()
# handle ambiguity in if get dict or list
other_kwargs = dict(add_special_tokens=add_special_tokens) if hasattr(tokenizer, 'add_special_tokens') else {}
if tokenizer is not None:
if hasattr(tokenizer, 'encode'):
tokens = tokenizer.encode(x, **other_kwargs)
else:
tokens = tokenizer(x, **other_kwargs)
if isinstance(tokens, dict) and 'input_ids' in tokens:
tokens = tokens['input_ids']
if isinstance(tokens, list):
n_tokens = len(tokens)
elif len(tokens.shape) == 2:
n_tokens = tokens.shape[1]
elif len(tokens.shape) == 1:
n_tokens = tokens.shape[0]
else:
raise RuntimeError("Cannot handle tokens: %s" % tokens)
elif token_count_fun is not None:
assert callable(token_count_fun)
other_kwargs = dict(add_special_tokens=add_special_tokens) if hasattr(token_count_fun,
'add_special_tokens') else {}
n_tokens = token_count_fun(x, **other_kwargs)
else:
tokenizer = FakeTokenizer()
n_tokens = tokenizer.num_tokens_from_string(x)
return n_tokens
def reverse_ucurve_list(lst):
if not lst:
return []
if len(lst) == 1:
return lst
if len(lst) == 2:
return [lst[1], lst[0]]
front_list = []
end_list = []
for i, item in enumerate(lst):
if i % 2 == 0:
end_list.append(item)
else:
front_list.append(item)
return front_list + end_list[::-1]
def undo_reverse_ucurve_list(lst):
if not lst:
return []
if len(lst) == 1:
return lst
if len(lst) == 2:
return [lst[1], lst[0]]
# Split the list into two halves: the first half and the second half (reversed)
mid = len(lst) // 2
first_half = lst[:mid]
second_half = lst[mid:][::-1]
# Merge the two halves by taking elements alternatively from the second half and then the first half
result = []
for i in range(mid):
result.append(second_half[i])
result.append(first_half[i])
# If the length of the list is odd, append the last element of the second half
if len(lst) % 2 != 0:
result.append(second_half[-1])
return result
def get_size(start_path='.'):
total_size = 0
for dirpath, dirnames, filenames in os.walk(start_path):
for f in filenames:
fp = os.path.join(dirpath, f)
# skip if it is symbolic link
if not os.path.islink(fp):
total_size += os.path.getsize(fp)
return total_size
def get_test_name_core():
tn = os.environ['PYTEST_CURRENT_TEST'].split(':')[-1]
tn = "_".join(tn.split(' ')[:-1]) # skip (call) at end
return sanitize_filename(tn)
class FullSet(set):
def __contains__(self, item):
return True
import os
def create_relative_symlink(target, link_name):
"""
Creates a relative symlink to a target from a link location, ensuring parent directories exist.
The target can be either a file or a directory.
Parameters:
- target: The path to the target file or directory. This can be an absolute or a relative path.
- link_name: The path where the symlink will be created. This should include the name of the symlink itself.
Raises:
- ValueError: If the target does not exist.
"""
# Ensure the target exists
if not os.path.exists(target):
raise ValueError("Target does not exist: " + target)
# Calculate the absolute paths
target_abs = os.path.abspath(target)
link_dir = os.path.dirname(os.path.abspath(link_name))
# Ensure the parent directory of the link exists
os.makedirs(link_dir, exist_ok=True)
# Calculate the relative path for the symlink
relative_path = os.path.relpath(target_abs, link_dir)
# Remove the link if it already exists
if os.path.exists(link_name) or os.path.islink(link_name):
os.remove(link_name)
# Create the symlink
os.symlink(relative_path, link_name)
print(f"Symlink created: {link_name} -> {relative_path}")
def get_gradio_tmp():
gradio_tmp = '/tmp/gradio'
makedirs(gradio_tmp, exist_ok=True) # won't hurt if soft link if exists
gradio_tmp = os.path.realpath(gradio_tmp)
return gradio_tmp
def in_gradio_root(file):
ret = False
ret |= isinstance(file, str) and os.path.isfile(file) and os.path.abspath(file).startswith('/tmp/gradio')
ret |= isinstance(file, str) and os.path.isfile(file) and os.path.abspath(file).startswith(get_gradio_tmp())
return ret
def get_is_gradio_h2oai():
try:
import gradio as gr
return gr.__h2oai__
except:
return False
def split_list(input_list, split_size):
for i in range(0, len(input_list), split_size):
yield input_list[i:i + split_size]
def get_lock_file(name):
lock_type = name
base_path = os.path.join('locks', '%s_locks' % name)
base_path = makedirs(base_path, exist_ok=True, tmp_ok=True, use_base=True)
lock_file = os.path.join(base_path, "%s.lock" % lock_type)
makedirs(os.path.dirname(lock_file)) # ensure made
return lock_file
def merge_dict(dict1, dict2):
ret = dict1.copy()
ret.update(dict2)
return ret
def is_uuid4(string):
# Regular expression to match the UUID v4 format
pattern = re.compile(r'^[a-f0-9]{8}-[a-f0-9]{4}-4[a-f0-9]{3}-[89ab][a-f0-9]{3}-[a-f0-9]{12}$', re.IGNORECASE)
return bool(pattern.match(string))
def is_full_git_hash(s):
# This regex checks for exactly 40 hexadecimal characters.
return bool(re.fullmatch(r'[0-9a-f]{40}', s))
def get_show_username(username1):
if split_google in username1:
show_username = split_google.join(username1.split(split_google)[0:1])
else:
show_username = username1
return show_username
# for extracting code blocks
pattern = re.compile(r"```(.*?)(\n[\s\S]*?)?```", re.DOTALL)
def get_code_blocks(response):
return pattern.findall(response)
def get_json(response, fixup=True, json_schema_type=None):
is_list = isinstance(response, list)
if not is_list:
response = [response]
response_new = [_get_json(x, fixup=fixup, json_schema_type=json_schema_type) for x in response]
if not is_list:
response_new = response_new[0]
return response_new
def extract_values(data):
if isinstance(data, dict):
if 'type' in data and 'value' in data:
return data['value']
elif 'items' in data:
return [extract_values(item) for item in data['items']]
elif 'properties' in data:
return {key: extract_values(value) for key, value in data['properties'].items()}
elif 'enum' in data:
return data['enum'] # return the enum values
elif 'const' in data:
return data['const'] # return the const value
elif 'oneOf' in data:
return [extract_values(item) for item in data['oneOf']]
elif 'anyOf' in data:
return [extract_values(item) for item in data['anyOf']]
elif 'allOf' in data:
return [extract_values(item) for item in data['allOf']]
else:
return {key: extract_values(value) for key, value in data.items()}
elif isinstance(data, list):
return [extract_values(item) for item in data]
else:
return data
# Function to check if JSON contains schema information
def contains_schema(data):
if isinstance(data, dict):
if 'type' in data and 'value' in data:
return True
for key, value in data.items():
if contains_schema(value):
return True
elif isinstance(data, list):
for item in data:
if contains_schema(item):
return True
return False
# Main function to handle both schema and regular JSON
def handle_json(data):
if contains_schema(data):
return extract_values(data)
else:
return data
def repair_json_by_type(response, json_schema_type=None):
# WIP for later
if json_schema_type in ['object', None]:
from json_repair import repair_json
response_str = response
response = repair_json(response)
if response in ['""', """''""", '', None]:
return {}
try:
# assumes already dict
response = handle_json(json.loads(response))
if isinstance(response, list) and len(response) >= 1 and not response_str.startswith('['):
response = response[-1] # take last if list, if was not pure list response
return json.dumps(response)
except Exception as e:
print("Did not extract_values: %s" % str(e))
return response
else:
from json_repair import repair_json
return repair_json(response)
def _get_json(response, fixup=True, json_schema_type=None):
if fixup:
# first rely upon json_repair package, handles code block extraction as well automatically
try:
response0 = repair_json_by_type(response, json_schema_type=json_schema_type)
if response0:
return response0
except Exception as e:
# FIXME: best effort, don't understand if package will hae issues
print("repair_json exception1: %s: %s" % (str(e), response))
# if json_repair fails, try to extract code block content
# sIf content is found (not an empty string), return None (or possibly an empty string as per updated logic)
response0 = extract_code_block_content(response)
if response0:
if fixup:
try:
response0 = repair_json_by_type(response0, json_schema_type=json_schema_type)
except Exception as e:
# FIXME: best effort, don't understand if package will hae issues
print("repair_json exception2: %s: %s" % (str(e), response))
return response0
# Next, check if the response looks like JSON, return it if so
if looks_like_json(response):
response = response.strip()
if response.endswith('```'):
response = response[:-3].strip()
if fixup:
try:
response = repair_json_by_type(response, json_schema_type=json_schema_type)
except Exception as e:
# FIXME: best effort, don't understand if package will hae issues
print("repair_json exception3: %s: %s" % (str(e), response))
return response
# If it doesn't look like JSON, return an empty string as a default case
return invalid_json_str
# Adjusted pattern to match code block content accurately
pattern_extract_codeblock = re.compile(r"```(?:[a-zA-Z]*)\s*(.*?)(```|$)", re.DOTALL)
def preprocess_code_blocks(stream_content):
# Remove consecutive starting code block delimiters, but keep the inner content
stream_content = re.sub(r"```[a-zA-Z]*\n```[a-zA-Z]*", "```", stream_content)
# Remove consecutive ending code block delimiters
stream_content = re.sub(r"```\n```", "```", stream_content)
return stream_content
def extract_code_block_content(stream_content):
# Postprocess to handle nested or consecutive code block delimiters
stream_content = preprocess_code_blocks(stream_content)
match = pattern_extract_codeblock.search(stream_content)
if match:
return match.group(1).strip()
else:
return ''
def has_starting_code_block(text):
pattern_partial_codeblock = re.compile(r"(^|\n|\r|<br\s*/?>)\s*```")
return bool(pattern_partial_codeblock.search(text))
def looks_like_json(text):
# Strip leading whitespace and check the first non-whitespace character
stripped_text = text.lstrip()
# Check if the text starts with '{', '[', or potentially a JSON string
if stripped_text.startswith(('{', '[', '"')):
return True
# Optionally, check for simple numeric values or null, true, false which are valid JSON
if re.match(r'(-?\d+(\.\d+)?([eE][+-]?\d+)?|null|true|false)\s*($|[,\]}])', stripped_text):
return True
return False
def is_json_vllm(model, base_model, inference_server, verbose=False):
if inference_server and not inference_server.startswith('vllm') or not inference_server:
return False
if isinstance(model, dict) and 'client' in model:
openai_client = model['client']
else:
openai_client, _, _, _, _, _, _ = set_openai(inference_server, model_name=base_model)
vllm_version = get_vllm_version(openai_client, inference_server, verbose=verbose)
json_vllm_version = "0.4.0" # The version to compare against
# Parse the version strings into comparable objects
parsed_vllm_version = version.parse(vllm_version)
parsed_json_vllm_version = version.parse(json_vllm_version)
# Compare the versions
if parsed_vllm_version >= parsed_json_vllm_version:
return True
else:
return False
def get_vllm_version(openai_client, inference_server, verbose=False):
vllm_version = '0.3.0'
if inference_server.startswith('vllm'):
# https://github.com/vllm-project/vllm/blob/main/vllm/entrypoints/openai/api_server.py
parsed_url = str(openai_client.base_url).replace("/v1", "/version")
try:
response = requests.get(parsed_url, timeout=int(os.getenv('REQUEST_TIMEOUT', '30')))
if response.status_code == 200:
# Parsing the JSON response content to a dictionary
data = response.json()
# Accessing the version from the response
vllm_version = data.get('version', vllm_version)
if verbose:
print(f"vLLM Server version: {vllm_version}")
else:
if verbose:
print(f"Failed to retrieve version, status code: {response.status_code}")
except (requests.exceptions.Timeout, requests.exceptions.JSONDecodeError, requests.exceptions.ConnectionError):
# if times out, assume older version, with no JSON. Or might not be real vllm
vllm_version = '0.3.0'
print(f"vLLM Server version timeout, assuming: {vllm_version}")
return vllm_version
def get_docs_tokens(tokenizer, text_context_list=[], max_input_tokens=None, docs_joiner=docs_joiner_default):
"""
max_input_tokens: Over all LLM calls, upper limit of total token count,
or single LLM call if want to know what docs fit into single call
"""
if text_context_list is None or len(text_context_list) == 0:
return 0, None, 0
assert max_input_tokens is not None, "Must set max_input_tokens"
tokens = [get_token_count(x + docs_joiner, tokenizer) for x in text_context_list]
tokens_cumsum = np.cumsum(tokens)
where_res = np.where(tokens_cumsum <= max_input_tokens)[0]
# if below condition fails, then keep top_k_docs=-1 and trigger special handling next
if where_res.shape[0] > 0:
top_k_docs = 1 + where_res[-1]
one_doc_size = None
num_doc_tokens = tokens_cumsum[top_k_docs - 1] # by index
else:
# if here, means 0 and just do best with 1 doc
top_k_docs = 1
text_context_list = text_context_list[:top_k_docs]
# critical protection
from h2oai_pipeline import H2OTextGenerationPipeline
doc_content = text_context_list[0]
doc_content, new_tokens0 = H2OTextGenerationPipeline.limit_prompt(doc_content,
tokenizer,
max_prompt_length=max_input_tokens)
text_context_list[0] = doc_content
one_doc_size = len(doc_content)
num_doc_tokens = get_token_count(doc_content + docs_joiner, tokenizer)
print(
"Unexpected large chunks and can't add to context, will add 1 anyways. Tokens %s -> %s for max_input_tokens=%s" % (
tokens[0], new_tokens0, max_input_tokens), flush=True)
return top_k_docs, one_doc_size, num_doc_tokens
def get_limited_text(hard_limit_tokens, text, tokenizer, verbose=False):
if tokenizer is None:
return text[:4 * hard_limit_tokens]
low = 0
high = len(text)
best_guess = text # Initialize best_guess to ensure it's defined
ntokens0 = len(tokenizer.tokenize(best_guess))
ntokens = None
max_steps = 5
steps = 0
while low <= high:
mid = low + (high - low) // 2 # Calculate midpoint for current search interval
# Estimate a trial cut of the text based on mid
trial_text_length = max(int(mid * 4), 1) # Using mid * 4 as an estimation, ensuring at least 1 character
trial_text = text[-trial_text_length:] # Take text from the end, based on trial_text_length
# Tokenize the trial text and count tokens
ntokens = len(tokenizer.tokenize(trial_text))
if ntokens > hard_limit_tokens:
# If the trial exceeds the token limit, reduce 'high' to exclude the current trial length
high = mid - 1
else:
# If the trial does not exceed the token limit, update 'best_guess' and increase 'low'
best_guess = trial_text # Update best_guess with the current trial_text
low = mid + 1 # Attempt to include more text in the next trial
if steps >= max_steps:
break
steps += 1
# 'best_guess' now contains the text that best fits the criteria
if verbose:
print("steps: %s ntokens0: %s/%s text0: %s ntokens: %s/%s text: %s" % (
steps, ntokens0, hard_limit_tokens, len(text), ntokens, hard_limit_tokens, len(best_guess)))
return best_guess
def deduplicate_names(names):
# Dictionary to hold the counts of each name
name_counts = {}
# List to store the final results
deduplicated_names = []
for name in names:
# Check if the name already exists in the dictionary
if name in name_counts:
# Increment the count for this name
name_counts[name] += 1
# Append the new name with the count as a suffix
deduplicated_names.append(f"{name}_{name_counts[name]}")
else:
# Add the name to the dictionary with a count of 0
name_counts[name] = 0
# Append the name as it is the first occurrence
deduplicated_names.append(name)
return deduplicated_names
def download_image(image_url, save_dir):
"""
Download an image from a URL and save it to a specified directory.
Parameters:
image_url (str): The URL of the image to download.
save_dir (str): The directory path where the image will be saved.
Returns:
str or None: The file path where the image was saved, or None if an error occurred.
"""
try:
response = requests.get(image_url)
response.raise_for_status() # Check if the request was successful
# Extract the file name from the URL
parsed_url = urlparse(image_url)
file_name = os.path.basename(parsed_url.path)
# Create the full save path
save_path = os.path.join(save_dir, file_name)
makedirs(save_dir, exist_ok=True)
# Save the image
with open(save_path, 'wb') as file:
file.write(response.content)
return save_path
except requests.exceptions.RequestException as e:
print(f"Error downloading the image: {e}")
return None
# Check if the input is a URL
url_pattern = re.compile(
r'^(?:http|ftp)s?://' # http:// or https://
r'(?:(?:[A-Z0-9](?:[A-Z0-9-]{0,61}[A-Z0-9])?\.)+(?:[A-Z]{2,6}\.?|[A-Z0-9-]{2,}\.?)|' # domain...
r'localhost|' # localhost...
r'\d{1,3}\.\d{1,3}\.\d{1,3}\.\d{1,3}|' # ...or ipv4
r'\[?[A-F0-9]*:[A-F0-9:]+\]?)' # ...or ipv6
r'(?::\d+)?' # optional port
r'(?:/?|[/?]\S+)$', re.IGNORECASE)
def check_input_type(input_string):
"""
Check if the input string is a file path, URL, or a base64 encoded image.
Parameters:
input_string (str): The input string to check.
Returns:
str: 'file', 'url', 'base64', or 'unknown' based on the input type.
"""
if not isinstance(input_string, str):
return 'unknown'
# Check if the input string looks like a base64 encoded image
if input_string.startswith("data:image/") or input_string.startswith("b'data:image/"):
return 'base64'
if re.match(url_pattern, input_string):
return 'url'
is_youtube = any(
input_string.replace('http://', '').replace('https://', '').replace('www.', '').startswith(prefix) for prefix in
url_prefixes_youtube)
if is_youtube:
return 'youtube'
# Check if the input is a file path
if os.path.isfile(input_string):
return 'file'
return 'unknown'
def get_youtube_urls():
# https://www.netify.ai/resources/applications/youtube
base = ['googlevideo.com',
'video.google.com',
'video.l.google.com',
'wide-youtube.l.google.com',
'youtu.be',
'youtube.ae',
'youtube.al',
'youtube.am',
'youtube.at',
'youtube.az',
'youtube.ba',
'youtube.be',
'youtube.bg',
'youtube.bh',
'youtube.bo',
'youtube.by',
'youtube.ca',
'youtube.cat',
'youtube.ch',
'youtube.cl',
'youtube.co',
'youtube.co.ae',
'youtube.co.at',
'youtube.co.cr',
'youtube.co.hu',
'youtube.co.id',
'youtube.co.il',
'youtube.co.in',
'youtube.co.jp',
'youtube.co.ke',
'youtube.co.kr',
'youtube.com',
'youtube.co.ma',
'youtube.com.ar',
'youtube.com.au',
'youtube.com.az',
'youtube.com.bd',
'youtube.com.bh',
'youtube.com.bo',
'youtube.com.br',
'youtube.com.by',
'youtube.com.co',
'youtube.com.do',
'youtube.com.ec',
'youtube.com.ee',
'youtube.com.eg',
'youtube.com.es',
'youtube.com.gh',
'youtube.com.gr',
'youtube.com.gt',
'youtube.com.hk',
'youtube.com.hn',
'youtube.com.hr',
'youtube.com.jm',
'youtube.com.jo',
'youtube.com.kw',
'youtube.com.lb',
'youtube.com.lv',
'youtube.com.ly',
'youtube.com.mk',
'youtube.com.mt',
'youtube.com.mx',
'youtube.com.my',
'youtube.com.ng',
'youtube.com.ni',
'youtube.com.om',
'youtube.com.pa',
'youtube.com.pe',
'youtube.com.ph',
'youtube.com.pk',
'youtube.com.pt',
'youtube.com.py',
'youtube.com.qa',
'youtube.com.ro',
'youtube.com.sa',
'youtube.com.sg',
'youtube.com.sv',
'youtube.com.tn',
'youtube.com.tr',
'youtube.com.tw',
'youtube.com.ua',
'youtube.com.uy',
'youtube.com.ve',
'youtube.co.nz',
'youtube.co.th',
'youtube.co.tz',
'youtube.co.ug',
'youtube.co.uk',
'youtube.co.ve',
'youtube.co.za',
'youtube.co.zw',
'youtube.cr',
'youtube.cz',
'youtube.de',
'youtube.dk',
'youtubeeducation.com',
'youtube.ee',
'youtubeembeddedplayer.googleapis.com',
'youtube.es',
'youtube.fi',
'youtube.fr',
'youtube.ge',
'youtube.googleapis.com',
'youtube.gr',
'youtube.gt',
'youtube.hk',
'youtube.hr',
'youtube.hu',
'youtube.ie',
'youtubei.googleapis.com',
'youtube.in',
'youtube.iq',
'youtube.is',
'youtube.it',
'youtube.jo',
'youtube.jp',
'youtubekids.com',
'youtube.kr',
'youtube.kz',
'youtube.la',
'youtube.lk',
'youtube.lt',
'youtube.lu',
'youtube.lv',
'youtube.ly',
'youtube.ma',
'youtube.md',
'youtube.me',
'youtube.mk',
'youtube.mn',
'youtube.mx',
'youtube.my',
'youtube.ng',
'youtube.ni',
'youtube.nl',
'youtube.no',
'youtube-nocookie.com',
'youtube.pa',
'youtube.pe',
'youtube.ph',
'youtube.pk',
'youtube.pl',
'youtube.pr',
'youtube.pt',
'youtube.qa',
'youtube.ro',
'youtube.rs',
'youtube.ru',
'youtube.sa',
'youtube.se',
'youtube.sg',
'youtube.si',
'youtube.sk',
'youtube.sn',
'youtube.soy',
'youtube.sv',
'youtube.tn',
'youtube.tv',
'youtube.ua',
'youtube.ug',
'youtube-ui.l.google.com',
'youtube.uy',
'youtube.vn',
'yt3.ggpht.com',
'yt.be',
'ytimg.com',
'ytimg.l.google.com',
'ytkids.app.goo.gl',
'yt-video-upload.l.google.com']
url_prefixes_youtube1 = []
for x in base:
url_prefixes_youtube1.extend([
# '%s/watch?v=' % x,
'%s' % x,
# '%s/shorts/' % x,
])
return set(url_prefixes_youtube1)
url_prefixes_youtube = get_youtube_urls()
def get_llama_lower_hf(llama_lower):
if 'huggingface.co' in llama_lower and '/resolve/' in llama_lower and len(llama_lower.split('huggingface.co')) == 2:
llama_lower_hf = llama_lower.split('huggingface.co')[1].split('resolve/')[0]
else:
llama_lower_hf = None
return llama_lower_hf
def get_depth_normal(lst):
if isinstance(lst, list) and lst:
return 1 + max(get_depth_normal(item) for item in lst)
else:
return 0
def get_gradio_depth(lst):
def get_depth(lst):
if isinstance(lst, (tuple, list)) and lst:
depths = [get_depth(item) for item in lst]
return 1 + max(depths)
else:
return 0
def has_single_element_sublist(lst, depth):
if depth == 1:
return isinstance(lst, (tuple, list)) and len(lst) == 1
if isinstance(lst, (tuple, list)):
return any(has_single_element_sublist(item, depth - 1) for item in lst)
return False
depth = get_depth(lst)
if has_single_element_sublist(lst, depth):
depth -= 1
return depth
def is_empty(obj):
if obj is None:
return True
if isinstance(obj, (str, list, tuple, dict, set)):
return len(obj) == 0
if isinstance(obj, bool):
return False
if isinstance(obj, (int, float)):
# Numbers can't be "empty" in the traditional sense, so go by value for them
return False if 0 else True
if isinstance(obj, complex):
return obj == 0
if isinstance(obj, bytes):
return len(obj) == 0
if isinstance(obj, bytearray):
return len(obj) == 0
if isinstance(obj, memoryview):
return len(obj) == 0
if isinstance(obj, range):
return len(obj) == 0
if isinstance(obj, frozenset):
return len(obj) == 0
if isinstance(obj, deque):
return len(obj) == 0
if isinstance(obj, array):
return len(obj) == 0
if isinstance(obj, (map, filter, zip)):
# These are iterators and need to be converted to a list to check if they are empty
return len(list(obj)) == 0
if hasattr(obj, '__len__'):
return len(obj) == 0
return False
from typing import Any, Dict, List, Union
from typing_extensions import TypedDict
def create_typed_dict(schema: Dict[str, Any], name: str = "Schema") -> type:
properties = schema.get("properties", {})
required = set(schema.get("required", []))
fields: Dict[str, Union[type, Any]] = {}
total = len(required) == len(properties)
for prop, details in properties.items():
prop_type = details.get("type")
if prop_type == "string":
field_type = str
elif prop_type == "integer":
field_type = int
elif prop_type == "number":
field_type = float
elif prop_type == "boolean":
field_type = bool
elif prop_type == "array":
items = details.get("items", {})
if items.get("type") == "string":
field_type = List[str]
elif items.get("type") == "object":
field_type = List[create_typed_dict(items, f"{name}Item")]
else:
field_type = List[Any]
elif prop_type == "object":
field_type = create_typed_dict(details, f"{name}{prop.capitalize()}")
else:
field_type = Any
if prop in required:
fields[prop] = field_type
else:
fields[prop] = Union[field_type, None]
return TypedDict(name, fields, total=total)
def get_supports_schema(inference_server, base_model, response_format='json_object', guided_json={}, json_vllm=False,
just_test=False):
if just_test:
supports_schema = True
else:
supports_schema = not is_empty(guided_json) and \
response_format == 'json_object'
supports_schema &= is_json_model(base_model, inference_server, json_vllm=json_vllm)
supports_schema &= json_vllm or \
not is_empty(inference_server) and \
any(inference_server.startswith(x) for x in ['openai_chat', 'openai_azure_chat']) and \
not is_empty(
base_model) and base_model in openai_supports_functiontools + openai_supports_parallel_functiontools or \
not is_empty(inference_server) and \
inference_server.startswith('anthropic') or \
not is_empty(inference_server) and \
inference_server.startswith('google') and base_model == 'gemini-1.5-pro-latest' or \
not is_empty(inference_server) and \
inference_server.startswith('mistralai') and \
does_support_functiontools(inference_server, base_model)
return supports_schema
def dedup_list(x):
x = [x.text if hasattr(x, 'text') else x for x in x]
return list(dict.fromkeys(x))