Mirrors/text-generation-webui
1
0
Fork 0
mirror of https://github.com/oobabooga/text-generation-webui.git synced 2024-11-26 01:30:20 +01:00
Code Issues Packages Projects Releases Wiki Activity

refactor quant models loader and add support of OPT

Browse Source
This commit is contained in:
Ayanami Rei 2023-03-13 19:59:57 +03:00
parent 2c4699a7e9
commit 345b6dee8c
1 changed files with 9 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

26
modules/quantized_LLaMA.py → modules/quant_loader.py
View File

@ -7,28 +7,20 @@ import torch
import modules.shared as shared import modules.shared as shared
sys.path.insert(0, str(Path("repositories/GPTQ-for-LLaMa"))) sys.path.insert(0, str(Path("repositories/GPTQ-for-LLaMa")))
from llama import load_quant
# 4-bit LLaMA # 4-bit LLaMA
def load_quantized_LLaMA(model_name): def load_quant(model_name, model_type):
if shared.args.load_in_4bit: if model_type == 'llama':
bits = 4 from llama import load_quant
elif model_type == 'opt':
from opt import load_quant
else: else:
bits = shared.args.gptq_bits print("Unknown pre-quantized model type specified. Only 'llama' and 'opt' are supported")
exit()
path_to_model = Path(f'models/{model_name}') path_to_model = Path(f'models/{model_name}')
pt_model = '' pt_model = f'{model_name}-{shared.args.gptq_bits}bit.pt'
if path_to_model.name.lower().startswith('llama-7b'):
pt_model = f'llama-7b-{bits}bit.pt'
elif path_to_model.name.lower().startswith('llama-13b'):
pt_model = f'llama-13b-{bits}bit.pt'
elif path_to_model.name.lower().startswith('llama-30b'):
pt_model = f'llama-30b-{bits}bit.pt'
elif path_to_model.name.lower().startswith('llama-65b'):
pt_model = f'llama-65b-{bits}bit.pt'
else:
pt_model = f'{model_name}-{bits}bit.pt'
# Try to find the .pt both in models/ and in the subfolder # Try to find the .pt both in models/ and in the subfolder
pt_path = None pt_path = None
@ -40,7 +32,7 @@ def load_quantized_LLaMA(model_name):
print(f"Could not find {pt_model}, exiting...") print(f"Could not find {pt_model}, exiting...")
exit() exit()
model = load_quant(path_to_model, str(pt_path), bits) model = load_quant(path_to_model, str(pt_path), shared.args.gptq_bits)
# Multiple GPUs or GPU+CPU # Multiple GPUs or GPU+CPU
if shared.args.gpu_memory: if shared.args.gpu_memory: