vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-11 01:24:36 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

downloading fallbacks for aria2, added minimum size (+1 squashed commits)

Browse source 
Squashed commits:

[86b49095] downloading fallbacks for aria2, added minimum size
This commit is contained in:
Concedo 2025-02-14 23:46:49 +08:00
parent 53486b6713
commit 979088320d
1 changed files with 54 additions and 51 deletions
Download patch file Download diff file Expand all files Collapse all files

97
koboldcpp.py
View file

@ -3102,7 +3102,7 @@ def show_gui():
args.model_param = askopenfilename(title="Select ggml model .bin or .gguf file or .kcpps config") args.model_param = askopenfilename(title="Select ggml model .bin or .gguf file or .kcpps config")
root.withdraw() root.withdraw()
root.quit() root.quit()
if args.model_param and args.model_param!="" and (args.model_param.lower().endswith('.kcpps') or args.model_param.lower().endswith('.kcppt')): if args.model_param and args.model_param!="" and (args.model_param.lower().endswith('.kcpps') or args.model_param.lower().endswith('.kcppt') or args.model_param.lower().endswith('.kcpps?download=true') or args.model_param.lower().endswith('.kcppt?download=true')):
dlfile = download_model_from_url(args.model_param,[".kcpps",".kcppt"]) # maybe download from url dlfile = download_model_from_url(args.model_param,[".kcpps",".kcppt"]) # maybe download from url
if dlfile: if dlfile:
args.model_param = dlfile args.model_param = dlfile
@ -4641,8 +4641,12 @@ def convert_outdated_args(args):
dict["usecpu"] = True dict["usecpu"] = True
if "failsafe" in dict and dict["failsafe"]: #failsafe implies noavx2 if "failsafe" in dict and dict["failsafe"]: #failsafe implies noavx2
dict["noavx2"] = True dict["noavx2"] = True
if ("model_param" not in dict or not dict["model_param"]) and ("model" in dict and dict["model"]): if ("model_param" not in dict or not dict["model_param"]) and ("model" in dict):
dict["model_param"] = dict["model"] model_value = dict["model"] #may be null, empty/non-empty string, empty/non empty array
if isinstance(model_value, str) and model_value: # Non-empty string
dict["model_param"] = model_value
elif isinstance(model_value, list) and model_value: # Non-empty list
dict["model_param"] = model_value[0] # Take the first file in the list
return args return args
def setuptunnel(global_memory, has_sd): def setuptunnel(global_memory, has_sd):
@ -4703,16 +4707,16 @@ def setuptunnel(global_memory, has_sd):
tunnelproc.wait() tunnelproc.wait()
if os.name == 'nt': if os.name == 'nt':
downloader_internal("https://github.com/cloudflare/cloudflared/releases/latest/download/cloudflared-windows-amd64.exe", "cloudflared.exe", True) downloader_internal("https://github.com/cloudflare/cloudflared/releases/latest/download/cloudflared-windows-amd64.exe", "cloudflared.exe", True, 500000)
elif sys.platform=="darwin": elif sys.platform=="darwin":
downloader_internal("https://github.com/cloudflare/cloudflared/releases/latest/download/cloudflared-darwin-amd64.tgz", "cloudflared-darwin-amd64.tgz", True) downloader_internal("https://github.com/cloudflare/cloudflared/releases/latest/download/cloudflared-darwin-amd64.tgz", "cloudflared-darwin-amd64.tgz", True, 500000)
subprocess.run("tar -xzf cloudflared-darwin-amd64.tgz", shell=True) subprocess.run("tar -xzf cloudflared-darwin-amd64.tgz", shell=True)
subprocess.run("chmod +x 'cloudflared'", shell=True) subprocess.run("chmod +x 'cloudflared'", shell=True)
elif sys.platform == "linux" and platform.machine().lower() == "aarch64": elif sys.platform == "linux" and platform.machine().lower() == "aarch64":
downloader_internal("https://github.com/cloudflare/cloudflared/releases/latest/download/cloudflared-linux-arm64", "cloudflared-linux-arm64", True) downloader_internal("https://github.com/cloudflare/cloudflared/releases/latest/download/cloudflared-linux-arm64", "cloudflared-linux-arm64", True, 500000)
subprocess.run("chmod +x 'cloudflared-linux-arm64'", shell=True) subprocess.run("chmod +x 'cloudflared-linux-arm64'", shell=True)
else: else:
downloader_internal("https://github.com/cloudflare/cloudflared/releases/latest/download/cloudflared-linux-amd64", "cloudflared-linux-amd64", True) downloader_internal("https://github.com/cloudflare/cloudflared/releases/latest/download/cloudflared-linux-amd64", "cloudflared-linux-amd64", True, 500000)
subprocess.run("chmod +x 'cloudflared-linux-amd64'", shell=True) subprocess.run("chmod +x 'cloudflared-linux-amd64'", shell=True)
print("Attempting to start tunnel thread...", flush=True) print("Attempting to start tunnel thread...", flush=True)
tunnel_thread = threading.Thread(target=run_tunnel) tunnel_thread = threading.Thread(target=run_tunnel)
@ -4856,39 +4860,35 @@ def sanitize_string(input_string):
sanitized_string = re.sub( r'[^\w\d\.\-_]', '', input_string) sanitized_string = re.sub( r'[^\w\d\.\-_]', '', input_string)
return sanitized_string return sanitized_string
def downloader_internal(input_url, output_filename, capture_output): def downloader_internal(input_url, output_filename, capture_output, min_file_size=64): #64 bytes required by default
import shutil import shutil
import subprocess import subprocess
import os import os
input_url = input_url.split(',') if "https://huggingface.co/" in input_url and "/blob/main/" in input_url:
for i in input_url: input_url = input_url.replace("/blob/main/", "/resolve/main/")
if "https://huggingface.co/" in i and "/blob/main/" in i:
i = i.replace("/blob/main/", "/resolve/main/")
if output_filename == "auto": if output_filename == "auto":
current_filename = os.path.basename(i).split('?')[0] output_filename = os.path.basename(input_url).split('?')[0].split('#')[0]
else: if os.path.exists(output_filename) and os.path.getsize(output_filename) > min_file_size:
current_filename = output_filename print(f"{output_filename} already exists, using existing file.")
if os.path.exists(current_filename) and os.path.getsize(current_filename) > 1000: #1KB trigger return output_filename
print(f"{current_filename} already exists, using existing file.") print(f"Downloading {input_url}", flush=True)
continue dl_success = False
print(f"Downloading {current_filename}", flush=True) if shutil.which("aria2c") is not None:
if shutil.which("aria2c") is None: rc = subprocess.run(f"aria2c -x 16 -s 16 --summary-interval=30 --console-log-level=error --log-level=error --download-result=default --allow-overwrite=true --file-allocation=none -o {output_filename} {input_url}", shell=True, capture_output=capture_output, text=True, check=True, encoding='utf-8')
if shutil.which("curl") is None: dl_success = (rc.returncode==0 and os.path.exists(output_filename) and os.path.getsize(output_filename) > min_file_size)
if shutil.which("wget") is None: if not dl_success and shutil.which("curl") is not None:
rc = subprocess.run(f"curl -fLo {output_filename} {input_url}", shell=True, capture_output=capture_output, text=True, check=True, encoding='utf-8')
dl_success = (rc.returncode==0 and os.path.exists(output_filename) and os.path.getsize(output_filename) > min_file_size)
if not dl_success and shutil.which("wget") is None:
rc = subprocess.run(f"wget -O {output_filename} {input_url}", shell=True, capture_output=capture_output, text=True, check=True, encoding='utf-8')
dl_success = (rc.returncode==0 and os.path.exists(output_filename) and os.path.getsize(output_filename) > min_file_size)
if not dl_success:
print("Could not find suitable download software, please install aria2 or curl.") print("Could not find suitable download software, please install aria2 or curl.")
return None return None
subprocess.run(f"wget -O {current_filename} {i}", shell=True, capture_output=capture_output, text=True, check=True, encoding='utf-8')
continue
subprocess.run(f"curl -fLo {current_filename} {i}", shell=True, capture_output=capture_output, text=True, check=True, encoding='utf-8')
continue
subprocess.run(f"aria2c -x 16 -s 16 --summary-interval=10 --download-result=default --allow-overwrite=true --file-allocation=none -o {current_filename} {i}", shell=True, capture_output=capture_output, text=True, check=True, encoding='utf-8')
continue
if output_filename == "auto":
output_filename = os.path.basename(input_url[0]).split('?')[0]
return output_filename return output_filename
def download_model_from_url(url,permitted_types=[".gguf",".safetensors", ".ggml", ".bin"]): def download_model_from_url(url, permitted_types=[".gguf",".safetensors", ".ggml", ".bin"], min_file_size=64):
if url and url!="": if url and url!="":
if url.endswith("?download=true"): if url.endswith("?download=true"):
url = url.replace("?download=true","") url = url.replace("?download=true","")
@ -4898,7 +4898,7 @@ def download_model_from_url(url,permitted_types=[".gguf",".safetensors", ".ggml"
end_ext_ok = True end_ext_ok = True
break break
if ((url.startswith("http://") or url.startswith("https://")) and end_ext_ok): if ((url.startswith("http://") or url.startswith("https://")) and end_ext_ok):
dlfile = downloader_internal(url, "auto", False) dlfile = downloader_internal(url, "auto", False, min_file_size)
return dlfile return dlfile
return None return None
@ -4941,7 +4941,7 @@ def main(launch_args):
args = convert_outdated_args(args) args = convert_outdated_args(args)
temp_hide_print = ((args.model_param or args.model) and args.prompt and not args.benchmark and not (args.debugmode >= 1)) temp_hide_print = (args.model_param and args.prompt and not args.benchmark and not (args.debugmode >= 1))
if not temp_hide_print: if not temp_hide_print:
print(f"***\nWelcome to KoboldCpp - Version {KcppVersion}") print(f"***\nWelcome to KoboldCpp - Version {KcppVersion}")
@ -4977,7 +4977,7 @@ def main(launch_args):
args = convert_outdated_args(args) args = convert_outdated_args(args)
#positional handling for kcpps files (drag and drop) #positional handling for kcpps files (drag and drop)
if args.model_param and args.model_param!="" and (args.model_param.lower().endswith('.kcpps') or args.model_param.lower().endswith('.kcppt')): if args.model_param and args.model_param!="" and (args.model_param.lower().endswith('.kcpps') or args.model_param.lower().endswith('.kcppt') or args.model_param.lower().endswith('.kcpps?download=true') or args.model_param.lower().endswith('.kcppt?download=true')):
dlfile = download_model_from_url(args.model_param,[".kcpps",".kcppt"]) # maybe download from url dlfile = download_model_from_url(args.model_param,[".kcpps",".kcppt"]) # maybe download from url
if dlfile: if dlfile:
args.model_param = dlfile args.model_param = dlfile
@ -5073,7 +5073,7 @@ def kcpp_main_process(launch_args, g_memory=None, gui_launcher=False):
using_gui_launcher = gui_launcher using_gui_launcher = gui_launcher
start_time = time.time() start_time = time.time()
if (args.model_param or args.model) and args.prompt and not args.benchmark and not (args.debugmode >= 1): if args.model_param and args.prompt and not args.benchmark and not (args.debugmode >= 1):
suppress_stdout() suppress_stdout()
if args.model_param and (args.benchmark or args.prompt): if args.model_param and (args.benchmark or args.prompt):
@ -5153,47 +5153,50 @@ def kcpp_main_process(launch_args, g_memory=None, gui_launcher=False):
# handle model downloads if needed # handle model downloads if needed
if args.model_param and args.model_param!="": if args.model_param and args.model_param!="":
dlfile = download_model_from_url(args.model_param,[".gguf",".bin", ".ggml"]) dlfile = download_model_from_url(args.model_param,[".gguf",".bin", ".ggml"],min_file_size=500000)
if dlfile: if dlfile:
args.model_param = dlfile args.model_param = dlfile
if args.model and isinstance(args.model, list) and len(args.model)>1: #handle multi file downloading
for extramodel in args.model[1:]:
download_model_from_url(extramodel,[".gguf",".bin", ".ggml"],min_file_size=500000)
if args.sdmodel and args.sdmodel!="": if args.sdmodel and args.sdmodel!="":
dlfile = download_model_from_url(args.sdmodel,[".gguf",".safetensors"]) dlfile = download_model_from_url(args.sdmodel,[".gguf",".safetensors"],min_file_size=500000)
if dlfile: if dlfile:
args.sdmodel = dlfile args.sdmodel = dlfile
if args.sdt5xxl and args.sdt5xxl!="": if args.sdt5xxl and args.sdt5xxl!="":
dlfile = download_model_from_url(args.sdt5xxl,[".gguf",".safetensors"]) dlfile = download_model_from_url(args.sdt5xxl,[".gguf",".safetensors"],min_file_size=500000)
if dlfile: if dlfile:
args.sdt5xxl = dlfile args.sdt5xxl = dlfile
if args.sdclipl and args.sdclipl!="": if args.sdclipl and args.sdclipl!="":
dlfile = download_model_from_url(args.sdclipl,[".gguf",".safetensors"]) dlfile = download_model_from_url(args.sdclipl,[".gguf",".safetensors"],min_file_size=500000)
if dlfile: if dlfile:
args.sdclipl = dlfile args.sdclipl = dlfile
if args.sdclipg and args.sdclipg!="": if args.sdclipg and args.sdclipg!="":
dlfile = download_model_from_url(args.sdclipg,[".gguf",".safetensors"]) dlfile = download_model_from_url(args.sdclipg,[".gguf",".safetensors"],min_file_size=500000)
if dlfile: if dlfile:
args.sdclipg = dlfile args.sdclipg = dlfile
if args.sdvae and args.sdvae!="": if args.sdvae and args.sdvae!="":
dlfile = download_model_from_url(args.sdvae,[".gguf",".safetensors"]) dlfile = download_model_from_url(args.sdvae,[".gguf",".safetensors"],min_file_size=500000)
if dlfile: if dlfile:
args.sdvae = dlfile args.sdvae = dlfile
if args.mmproj and args.mmproj!="": if args.mmproj and args.mmproj!="":
dlfile = download_model_from_url(args.mmproj,[".gguf"]) dlfile = download_model_from_url(args.mmproj,[".gguf"],min_file_size=500000)
if dlfile: if dlfile:
args.mmproj = dlfile args.mmproj = dlfile
if args.whispermodel and args.whispermodel!="": if args.whispermodel and args.whispermodel!="":
dlfile = download_model_from_url(args.whispermodel,[".gguf",".bin"]) dlfile = download_model_from_url(args.whispermodel,[".gguf",".bin"],min_file_size=500000)
if dlfile: if dlfile:
args.whispermodel = dlfile args.whispermodel = dlfile
if args.draftmodel and args.draftmodel!="": if args.draftmodel and args.draftmodel!="":
dlfile = download_model_from_url(args.draftmodel,[".gguf"]) dlfile = download_model_from_url(args.draftmodel,[".gguf"],min_file_size=500000)
if dlfile: if dlfile:
args.draftmodel = dlfile args.draftmodel = dlfile
if args.ttsmodel and args.ttsmodel!="": if args.ttsmodel and args.ttsmodel!="":
dlfile = download_model_from_url(args.ttsmodel,[".gguf"]) dlfile = download_model_from_url(args.ttsmodel,[".gguf"],min_file_size=500000)
if dlfile: if dlfile:
args.ttsmodel = dlfile args.ttsmodel = dlfile
if args.ttswavtokenizer and args.ttswavtokenizer!="": if args.ttswavtokenizer and args.ttswavtokenizer!="":
dlfile = download_model_from_url(args.ttswavtokenizer,[".gguf"]) dlfile = download_model_from_url(args.ttswavtokenizer,[".gguf"],min_file_size=500000)
if dlfile: if dlfile:
args.ttswavtokenizer = dlfile args.ttswavtokenizer = dlfile
@ -5687,7 +5690,7 @@ if __name__ == '__main__':
parser = argparse.ArgumentParser(description=f'KoboldCpp Server - Version {KcppVersion}') parser = argparse.ArgumentParser(description=f'KoboldCpp Server - Version {KcppVersion}')
modelgroup = parser.add_mutually_exclusive_group() #we want to be backwards compatible with the unnamed positional args modelgroup = parser.add_mutually_exclusive_group() #we want to be backwards compatible with the unnamed positional args
modelgroup.add_argument("--model", metavar=('[filename]'), help="Model file to load", type=str, default="") modelgroup.add_argument("--model", metavar=('[filenames]'), help="Model file to load. Accepts multiple values if they are URLs.", type=str, nargs='+', default=[])
modelgroup.add_argument("model_param", help="Model file to load (positional)", nargs="?") modelgroup.add_argument("model_param", help="Model file to load (positional)", nargs="?")
portgroup = parser.add_mutually_exclusive_group() #we want to be backwards compatible with the unnamed positional args portgroup = parser.add_mutually_exclusive_group() #we want to be backwards compatible with the unnamed positional args
portgroup.add_argument("--port", metavar=('[portnumber]'), help=f"Port to listen on. (Defaults to {defaultport})", default=defaultport, type=int, action='store') portgroup.add_argument("--port", metavar=('[portnumber]'), help=f"Port to listen on. (Defaults to {defaultport})", default=defaultport, type=int, action='store')