mirror of
https://github.com/ggerganov/llama.cpp.git
synced 2025-02-06 00:20:34 +01:00
0cec062a63
* add glm edge chat model * use config partial_rotary_factor as rope ratio * support for glm edge model * vision model support * remove debug info * fix format * llava.cpp trailing whitespace * remove unused AutoTokenizer * Update src/llama.cpp for not contain <|end|> or </s> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * add edge template * fix chat template * fix confict * fix confict * fix ci err * fix format err * fix template err * 9b hf chat support * format * format clip.cpp * fix format * Apply suggestions from code review * Apply suggestions from code review * Update examples/llava/clip.cpp * fix format * minor : style --------- Co-authored-by: liyuhang <yuhang.li@zhipuai.cn> Co-authored-by: piDack <pcdack@hotmail.co> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> Co-authored-by: liyuhang <yuhang.li@aminer.cn> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
34 lines
1.3 KiB
Python
34 lines
1.3 KiB
Python
import argparse
|
|
import os
|
|
import torch
|
|
from transformers import AutoModel
|
|
|
|
ap = argparse.ArgumentParser()
|
|
ap.add_argument("-m", "--model", help="Path to GLM model")
|
|
args = ap.parse_args()
|
|
|
|
# find the model part that includes the the multimodal projector weights
|
|
model = AutoModel.from_pretrained(args.model, trust_remote_code=True, local_files_only=True)
|
|
checkpoint = model.state_dict()
|
|
|
|
# get a list of mm tensor names
|
|
mm_tensors = [k for k, v in checkpoint.items() if k.startswith("vision.adapter.")]
|
|
|
|
# store these tensors in a new dictionary and torch.save them
|
|
projector = {name: checkpoint[name].float() for name in mm_tensors}
|
|
torch.save(projector, f"{args.model}/glm.projector")
|
|
|
|
clip_tensors = [k for k, v in checkpoint.items() if k.startswith("vision.vit.model.vision_model.")]
|
|
if len(clip_tensors) > 0:
|
|
clip = {name.replace("vision.vit.model.", ""): checkpoint[name].float() for name in clip_tensors}
|
|
torch.save(clip, f"{args.model}/glm.clip")
|
|
|
|
# added tokens should be removed to be able to convert Mistral models
|
|
if os.path.exists(f"{args.model}/added_tokens.json"):
|
|
with open(f"{args.model}/added_tokens.json", "w") as f:
|
|
f.write("{}\n")
|
|
|
|
print("Done!")
|
|
print(f"Now you can convert {args.model} to a regular LLaMA GGUF file.")
|
|
print(f"Also, use {args.model}glm.projector to prepare a glm-encoder.gguf file.")
|