Mirrors/text-generation-webui
1
0
Fork 0
mirror of https://github.com/oobabooga/text-generation-webui.git synced 2024-11-22 16:17:57 +01:00
Code Issues Packages Projects Releases Wiki Activity

determine model type from model name

Browse Source
This commit is contained in:
Ayanami Rei 2023-03-13 22:11:32 +03:00
parent b6c5c57f2e
commit a6a6522b6a
1 changed files with 21 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
modules/quant_loader.py
View File

@ -9,8 +9,17 @@ 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")))
# 4-bit LLaMA def load_quantized(model_name):
def load_quantized(model_name, model_type): if not shared.args.gptq_model_type:
# Try to determine model type from model name
model_type = model_name.split('-')[0].lower()
if model_type not in ('llama', 'opt'):
print("Can't determine model type from model name. Please specify it manually using --gptq-model-type "
"argument")
exit()
else:
model_type = shared.args.gptq_model_type.lower()
if model_type == 'llama': if model_type == 'llama':
from llama import load_quant from llama import load_quant
elif model_type == 'opt': elif model_type == 'opt':
@ -20,6 +29,15 @@ def load_quantized(model_name, model_type):
exit() exit()
path_to_model = Path(f'models/{model_name}') path_to_model = Path(f'models/{model_name}')
if path_to_model.name.lower().startswith('llama-7b'):
pt_model = f'llama-7b-{shared.args.gptq_bits}bit.pt'
elif path_to_model.name.lower().startswith('llama-13b'):
pt_model = f'llama-13b-{shared.args.gptq_bits}bit.pt'
elif path_to_model.name.lower().startswith('llama-30b'):
pt_model = f'llama-30b-{shared.args.gptq_bits}bit.pt'
elif path_to_model.name.lower().startswith('llama-65b'):
pt_model = f'llama-65b-{shared.args.gptq_bits}bit.pt'
else:
pt_model = f'{model_name}-{shared.args.gptq_bits}bit.pt' pt_model = f'{model_name}-{shared.args.gptq_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