vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

auto scale

Browse source
This commit is contained in:
ngxson 2024-07-15 11:41:18 +02:00
parent 703573f608
commit 42415a4874
4 changed files with 36 additions and 16 deletions
Download patch file Download diff file Expand all files Collapse all files

2
convert_lora_to_gguf.py
View file

@ -366,9 +366,11 @@ if __name__ == '__main__':
lparams: dict[str, Any] = json.load(f)
alpha = lparams["lora_alpha"]
rank = lparams["r"]
model_instance.gguf_writer.add_string("training.type", "finetune_lora")
model_instance.gguf_writer.add_float32("training.lora.alpha", float(alpha))
model_instance.gguf_writer.add_float32("training.lora.scale", float(alpha) / float(rank))
model_instance.gguf_writer.add_quantization_version(gguf.GGML_QUANT_VERSION)
logger.info("Exporting model...")