KoboldAI-Client/utils.py

379 lines
17 KiB
Python
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

from threading import Timer
import re
import shutil
import json
import subprocess
import tempfile
import requests
import requests.adapters
import time
from transformers import __version__ as transformers_version
from transformers import PreTrainedModel
import packaging.version
from tqdm.auto import tqdm
import os
import itertools
from typing import List, Optional
HAS_ACCELERATE = packaging.version.parse(transformers_version) >= packaging.version.parse("4.20.0.dev0")
try:
import accelerate
except ImportError:
HAS_ACCELERATE = False
vars = None
num_shards: Optional[int] = None
current_shard = 0
from_pretrained_model_name = ""
from_pretrained_index_filename: Optional[str] = None
from_pretrained_kwargs = {}
bar = None
layers_module_names: Optional[List[str]] = None
module_names: Optional[List[str]] = None
named_buffers: Optional[List[tuple]] = None
default_sampler_order = [0, 1, 2, 3, 4, 5]
#==================================================================#
# Decorator to prevent a function's actions from being run until
# at least x seconds have passed without the function being called
#==================================================================#
def debounce(wait):
def decorator(fun):
def debounced(*args, **kwargs):
def call_it():
fun(*args, **kwargs)
try:
debounced.t.cancel()
except AttributeError:
pass
debounced.t = Timer(wait, call_it)
debounced.t.start()
return debounced
return decorator
#==================================================================#
# Replace fancy quotes and apostrope's with standard ones
#==================================================================#
def fixquotes(txt):
txt = txt.replace("", '"')
txt = txt.replace("", '"')
txt = txt.replace("", "'")
txt = txt.replace("`", "'")
return txt
#==================================================================#
#
#==================================================================#
def trimincompletesentence(txt):
# Cache length of text
ln = len(txt)
# Find last instance of punctuation (Borrowed from Clover-Edition by cloveranon)
lastpunc = max(txt.rfind("."), txt.rfind("!"), txt.rfind("?"))
# Is this the end of a quote?
if(lastpunc < ln-1):
if(txt[lastpunc+1] == '"'):
lastpunc = lastpunc + 1
if(lastpunc >= 0):
txt = txt[:lastpunc+1]
return txt
#==================================================================#
#
#==================================================================#
def replaceblanklines(txt):
txt = txt.replace("\n\n", "\n")
return txt
#==================================================================#
#
#==================================================================#
def removespecialchars(txt, vars=None):
if vars is None or vars.actionmode == 0:
txt = re.sub(r"[#/@%<>{}+=~|\^]", "", txt)
else:
txt = re.sub(r"[#/@%{}+=~|\^]", "", txt)
return txt
#==================================================================#
# If the next action follows a sentence closure, add a space
#==================================================================#
def addsentencespacing(txt, vars):
# Don't add sentence spacing if submission is empty or starts with whitespace
if(len(txt) == 0 or len(txt) != len(txt.lstrip())):
return txt
# Get last character of last action
if(len(vars.actions) > 0):
if(len(vars.actions[vars.actions.get_last_key()]) > 0):
action = vars.actions[vars.actions.get_last_key()]
lastchar = action[-1] if len(action) else ""
else:
# Last action is blank, this should never happen, but
# since it did let's bail out.
return txt
else:
action = vars.prompt
lastchar = action[-1] if len(action) else ""
if(lastchar != " "):
txt = " " + txt
return txt
def singlelineprocessing(txt, vars):
txt = vars.regex_sl.sub('', txt)
if(len(vars.actions) > 0):
if(len(vars.actions[vars.actions.get_last_key()]) > 0):
action = vars.actions[vars.actions.get_last_key()]
lastchar = action[-1] if len(action) else ""
else:
# Last action is blank, this should never happen, but
# since it did let's bail out.
return txt
else:
action = vars.prompt
lastchar = action[-1] if len(action) else ""
if(lastchar != "\n"):
txt = txt + "\n"
return txt
#==================================================================#
# Cleans string for use in file name
#==================================================================#
def cleanfilename(filename):
filteredcharacters = ('/','\\')
filename = "".join(c for c in filename if c not in filteredcharacters).rstrip()
return filename
#==================================================================#
# Newline substitution for fairseq models
#==================================================================#
def encodenewlines(txt):
if(vars.newlinemode == "s"):
return txt.replace('\n', "</s>")
return txt
def decodenewlines(txt):
if(vars.newlinemode == "s"):
return txt.replace("</s>", '\n')
if(vars.newlinemode == "ns"):
return txt.replace("</s>", '')
return txt
#==================================================================#
# Returns number of layers given an HF model config
#==================================================================#
def num_layers(config):
return config.num_layers if hasattr(config, "num_layers") else config.n_layer if hasattr(config, "n_layer") else config.num_hidden_layers if hasattr(config, 'num_hidden_layers') else None
#==================================================================#
# Downloads huggingface checkpoints using aria2c if possible
#==================================================================#
from flask_socketio import emit
class Send_to_socketio(object):
def write(self, bar):
print("should be emitting: ", bar, end="")
time.sleep(0.01)
try:
emit('from_server', {'cmd': 'model_load_status', 'data': bar.replace(" ", "&nbsp;")}, broadcast=True)
except:
pass
def aria2_hook(pretrained_model_name_or_path: str, force_download=False, cache_dir=None, proxies=None, resume_download=False, local_files_only=False, use_auth_token=None, user_agent=None, revision=None, mirror=None, **kwargs):
import transformers
import transformers.modeling_utils
from huggingface_hub import HfFolder
if shutil.which("aria2c") is None: # Don't do anything if aria2 is not installed
return
if local_files_only: # If local_files_only is true, we obviously don't need to download anything
return
if os.path.isdir(pretrained_model_name_or_path) or os.path.isfile(pretrained_model_name_or_path) or os.path.isfile(pretrained_model_name_or_path + ".index") or transformers.modeling_utils.is_remote_url(pretrained_model_name_or_path):
return
if proxies:
print("WARNING: KoboldAI does not support using aria2 to download models from huggingface.co through a proxy. Disabling aria2 download mode.")
return
if use_auth_token:
if isinstance(use_auth_token, str):
token = use_auth_token
else:
token = HfFolder.get_token()
if token is None:
raise EnvironmentError("You specified use_auth_token=True, but a huggingface token was not found.")
_cache_dir = str(cache_dir) if cache_dir is not None else transformers.TRANSFORMERS_CACHE
sharded = False
headers = {"user-agent": transformers.file_utils.http_user_agent(user_agent)}
if use_auth_token:
headers["authorization"] = f"Bearer {use_auth_token}"
def is_cached(url):
try:
transformers.file_utils.get_from_cache(url, cache_dir=cache_dir, local_files_only=True)
except (FileNotFoundError, transformers.file_utils.EntryNotFoundError):
return False
return True
while True: # Try to get the huggingface.co URL of the model's pytorch_model.bin or pytorch_model.bin.index.json file
try:
filename = transformers.modeling_utils.WEIGHTS_INDEX_NAME if sharded else transformers.modeling_utils.WEIGHTS_NAME
except AttributeError:
return
url = transformers.file_utils.hf_bucket_url(pretrained_model_name_or_path, filename, revision=revision, mirror=mirror)
if is_cached(url) or requests.head(url, allow_redirects=True, proxies=proxies, headers=headers):
break
if sharded:
return
else:
sharded = True
if not sharded: # If the model has a pytorch_model.bin file, that's the only file to download
filenames = [transformers.modeling_utils.WEIGHTS_NAME]
else: # Otherwise download the pytorch_model.bin.index.json and then let aria2 download all the pytorch_model-#####-of-#####.bin files mentioned inside it
map_filename = transformers.file_utils.cached_path(url, cache_dir=cache_dir, force_download=force_download, proxies=proxies, resume_download=resume_download, use_auth_token=use_auth_token, user_agent=user_agent)
with open(map_filename) as f:
map_data = json.load(f)
filenames = set(map_data["weight_map"].values())
urls = [transformers.file_utils.hf_bucket_url(pretrained_model_name_or_path, n, revision=revision, mirror=mirror) for n in filenames]
if not force_download:
urls = [u for u in urls if not is_cached(u)]
if not urls:
return
etags = [h.get("X-Linked-Etag") or h.get("ETag") for u in urls for h in [requests.head(u, headers=headers, allow_redirects=False, proxies=proxies, timeout=10).headers]]
headers = [requests.head(u, headers=headers, allow_redirects=True, proxies=proxies, timeout=10).headers for u in urls]
filenames = [transformers.file_utils.url_to_filename(u, t) for u, t in zip(urls, etags)]
for n in filenames:
path = os.path.join(_cache_dir, "kai-tempfile." + n + ".aria2")
if os.path.exists(path):
os.remove(path)
path = os.path.join(_cache_dir, "kai-tempfile." + n)
if os.path.exists(path):
os.remove(path)
if force_download:
path = os.path.join(_cache_dir, n + ".json")
if os.path.exists(path):
os.remove(path)
path = os.path.join(_cache_dir, n)
if os.path.exists(path):
os.remove(path)
total_length = sum(int(h["Content-Length"]) for h in headers)
lengths = {}
aria2_config = "\n".join(f"{u}\n out=kai-tempfile.{n}" for u, n in zip(urls, filenames)).encode()
s = requests.Session()
s.mount("http://", requests.adapters.HTTPAdapter(max_retries=requests.adapters.Retry(total=120, backoff_factor=1)))
bar = None
done = False
secret = os.urandom(17).hex()
try:
with tempfile.NamedTemporaryFile("w+b", delete=False) as f:
f.write(aria2_config)
f.flush()
p = subprocess.Popen(["aria2c", "-x", "10", "-s", "10", "-j", "10", "--enable-rpc=true", f"--rpc-secret={secret}", "--rpc-listen-port", str(vars.aria2_port), "--disable-ipv6", "--file-allocation=trunc", "--allow-overwrite", "--auto-file-renaming=false", "-d", _cache_dir, "-i", f.name, "-U", transformers.file_utils.http_user_agent(user_agent)] + (["-c"] if not force_download else []) + ([f"--header='Authorization: Bearer {token}'"] if use_auth_token else []), stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
while p.poll() is None:
r = s.post(f"http://localhost:{vars.aria2_port}/jsonrpc", json={"jsonrpc": "2.0", "id": "kai", "method": "aria2.tellActive", "params": [f"token:{secret}"]}).json()["result"]
if not r:
s.close()
if bar is not None:
bar.n = bar.total
bar.close()
p.terminate()
done = True
break
if bar is None:
bar = tqdm(total=total_length, desc=f"[aria2] Downloading model", unit="B", unit_scale=True, unit_divisor=1000, file=Send_to_socketio())
visited = set()
for x in r:
filename = x["files"][0]["path"]
lengths[filename] = (int(x["completedLength"]), int(x["totalLength"]))
visited.add(filename)
for k, v in lengths.items():
if k not in visited:
lengths[k] = (v[1], v[1])
bar.n = sum(v[0] for v in lengths.values())
bar.update()
time.sleep(0.1)
path = f.name
except Exception as e:
p.terminate()
raise e
finally:
try:
os.remove(path)
except OSError:
pass
code = p.wait()
if not done and code:
raise OSError(f"aria2 exited with exit code {code}")
for u, t, n in zip(urls, etags, filenames):
os.rename(os.path.join(_cache_dir, "kai-tempfile." + n), os.path.join(_cache_dir, n))
with open(os.path.join(_cache_dir, n + ".json"), "w") as f:
json.dump({"url": u, "etag": t}, f)
#==================================================================#
# Given the path to a pytorch_model.bin.index.json, returns how many
# shards there are in the model
#==================================================================#
def get_num_shards(filename):
with open(filename) as f:
map_data = json.load(f)
return len(set(map_data["weight_map"].values()))
#==================================================================#
# Given the name/path of a sharded model and the path to a
# pytorch_model.bin.index.json, returns a list of weight names in the
# sharded model. Requires lazy loader to be enabled to work properl
#==================================================================#
def get_sharded_checkpoint_num_tensors(pretrained_model_name_or_path, filename, cache_dir=None, force_download=False, proxies=None, resume_download=False, local_files_only=False, use_auth_token=None, user_agent=None, revision=None, mirror=None, **kwargs):
import transformers.modeling_utils
import torch
shard_paths, _ = transformers.modeling_utils.get_checkpoint_shard_files(pretrained_model_name_or_path, filename, cache_dir=cache_dir, force_download=force_download, proxies=proxies, resume_download=resume_download, local_files_only=local_files_only, use_auth_token=use_auth_token, user_agent=user_agent, revision=revision, mirror=mirror)
return list(itertools.chain(*(torch.load(p, map_location="cpu").keys() for p in shard_paths)))
#==================================================================#
# Given a PreTrainedModel, returns the list of module names that correspond
# to the model's hidden layers.
#==================================================================#
def get_layers_module_names(model: PreTrainedModel) -> List[str]:
names: List[str] = []
def recurse(module, head=""):
for c in module.named_children():
name = head + c[0]
if c[0].isnumeric() and any(c[1].__class__.__name__.endswith(suffix) for suffix in ("Block", "Layer")):
names.append(name)
else:
recurse(c[1], head=name + ".")
recurse(model)
return names
#==================================================================#
# Given a PreTrainedModel, returns the module name that corresponds
# to the model's input embeddings.
#==================================================================#
def get_input_embeddings_module_name(model: PreTrainedModel) -> str:
embeddings = model.get_input_embeddings()
def recurse(module, head=""):
for c in module.named_children():
name = head + c[0]
if c[1] is embeddings:
return name
else:
return recurse(c[1], head=name + ".")
return recurse(model)
#==================================================================#
# Given a PreTrainedModel and a list of module names, returns a list
# of module names such that the union of the set of modules given as input
# and the set of modules returned as output contains all modules in the model.
#==================================================================#
def get_missing_module_names(model: PreTrainedModel, names: List[str]) -> List[str]:
missing_names: List[str] = []
def recurse(module, head=""):
for c in module.named_children():
name = head + c[0]
if any(name.startswith(n) for n in names):
continue
if next(c[1].named_children(), None) is None:
missing_names.append(name)
else:
recurse(c[1], head=name + ".")
recurse(model)
return missing_names