mmarch/text-generation-webui
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

Add gpu_split param to ExLlama

Browse source 
Adapted from code created by Ph0rk0z. Thank you Ph0rk0z.
This commit is contained in:
oobabooga 2023-06-16 20:49:36 -03:00
parent cb9be5db1c
commit 5f392122fd
6 changed files with 20 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

2
modules/ui.py
View file

@ -30,7 +30,7 @@ theme = gr.themes.Default(
def list_model_elements():
elements = ['loader', 'cpu_memory', 'auto_devices', 'disk', 'cpu', 'bf16', 'load_in_8bit', 'trust_remote_code', 'load_in_4bit', 'compute_dtype', 'quant_type', 'use_double_quant', 'wbits', 'groupsize', 'model_type', 'pre_layer', 'triton', 'desc_act', 'no_inject_fused_attention', 'no_inject_fused_mlp', 'threads', 'n_batch', 'no_mmap', 'mlock', 'n_gpu_layers', 'n_ctx', 'llama_cpp_seed']
elements = ['loader', 'cpu_memory', 'auto_devices', 'disk', 'cpu', 'bf16', 'load_in_8bit', 'trust_remote_code', 'load_in_4bit', 'compute_dtype', 'quant_type', 'use_double_quant', 'wbits', 'groupsize', 'model_type', 'pre_layer', 'triton', 'desc_act', 'no_inject_fused_attention', 'no_inject_fused_mlp', 'threads', 'n_batch', 'no_mmap', 'mlock', 'n_gpu_layers', 'n_ctx', 'llama_cpp_seed', 'gpu_split']
for i in range(torch.cuda.device_count()):
elements.append(f'gpu_memory_{i}')