Mirrors/text-generation-webui
1
0
Fork 0
mirror of https://github.com/oobabooga/text-generation-webui.git synced 2025-01-23 01:59:21 +01:00
Code Issues Packages Projects Releases Wiki Activity

Detect ggml*.bin files automatically

Browse Source
This commit is contained in:
oobabooga 2023-03-31 17:18:21 -03:00
parent 0aee7341d8
commit 3a47a602a3
1 changed files with 4 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
modules/models.py
View File

@ -42,7 +42,7 @@ def load_model(model_name):
t0 = time.time() t0 = time.time()
shared.is_RWKV = 'rwkv-' in model_name.lower() shared.is_RWKV = 'rwkv-' in model_name.lower()
shared.is_llamacpp = model_name.lower().startswith(('llamacpp', 'alpaca-cpp')) shared.is_llamacpp = len(list(Path(f'models/{model_name}').glob('ggml*.bin'))) > 0
# Default settings # Default settings
if not any([shared.args.cpu, shared.args.load_in_8bit, shared.args.wbits, shared.args.auto_devices, shared.args.disk, shared.args.gpu_memory is not None, shared.args.cpu_memory is not None, shared.args.deepspeed, shared.args.flexgen, shared.is_RWKV, shared.is_llamacpp]): if not any([shared.args.cpu, shared.args.load_in_8bit, shared.args.wbits, shared.args.auto_devices, shared.args.disk, shared.args.gpu_memory is not None, shared.args.cpu_memory is not None, shared.args.deepspeed, shared.args.flexgen, shared.is_RWKV, shared.is_llamacpp]):
@ -105,12 +105,10 @@ def load_model(model_name):
elif shared.is_llamacpp: elif shared.is_llamacpp:
from modules.llamacpp_model import LlamaCppModel from modules.llamacpp_model import LlamaCppModel
if model_name.lower().startswith('alpaca-cpp'): model_file = list(Path(f'models/{model_name}').glob('ggml*.bin'))[0]
model_file = f'models/{model_name}/ggml-alpaca-7b-q4.bin' print(f"llama.cpp weights detected: {model_file}\n")
else:
model_file = f'models/{model_name}/ggml-model-q4_0.bin'
model, tokenizer = LlamaCppModel.from_pretrained(Path(model_file)) model, tokenizer = LlamaCppModel.from_pretrained(model_file)
return model, tokenizer return model, tokenizer
# Custom # Custom