Skip to content

Conversation

@prusnak
Copy link
Contributor

@prusnak prusnak commented Mar 13, 2023

Alternative to #17 suggested in #17 (comment)

@ggerganov ggerganov merged commit d1f2247 into master Mar 13, 2023
@ggerganov ggerganov deleted the quantize-sh branch March 13, 2023 16:15
@Jettford
Copy link

I wrote up a basic python implementation of the same script for Windows users, would it be worth making a pull request to replace the batch script?

import os
import re
import sys

def print_usage():
    print("Usage: llama-quantize.py 7B|13B|30B|65B [--remove-f16]")
    exit(0)

if not len(sys.argv) > 1:
    print_usage()
    
regex_test = re.compile("^[0-9]{1,2}B$")

if not regex_test.match(sys.argv[1]):
    print_usage()
    
model_directory = f"./models/{sys.argv[1]}/"

if not os.path.exists(model_directory):
    print("Failed to find model directory")
    
    print_usage()
    
for file in os.listdir(model_directory):
    if not "ggml-model-f16.bin" in file:
        continue
        
    file = os.path.join(model_directory, file)
    
    new_name = file.replace("f16", "q4_0")
    
    binary_name = "./quantize"
    
    if sys.platform == "win32":
        binary_name += ".exe"
        
        binary_name = binary_name[2:]
        
    os.system(f"{binary_name} {file} {new_name} 2")
    
    if len(sys.argv) > 2:
        if sys.argv[2] == "--remove-f16":
            os.remove(file)

@ghost

This comment was marked as outdated.

@tmzncty
Copy link

tmzncty commented Mar 16, 2023

我觉得,直接编译出quantize.exe,然后在CMD中运行就好了。
‘‘‘quantize.exe ggml-model-f16.bin ggml-model-q4.bin 2’’’
image
image
image
image

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants