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

AutoGPTQ: Add UI and command line support for disabling fused attention and fused MLP (#2648)

Browse source
This commit is contained in:
Tom Jobbins 2023-06-16 03:59:54 +01:00 • committed by GitHub
parent 909d8c6ae3
commit 646b0c889f
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
5 changed files with 11 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

2
modules/AutoGPTQ_loader.py
View file

@ -43,6 +43,8 @@ def load_quantized(model_name):
'model_basename': pt_path.stem,
'device': "cuda:0" if not shared.args.cpu else "cpu",
'use_triton': shared.args.triton,
'inject_fused_attention': not shared.args.no_inject_fused_attention,
'inject_fused_mlp': not shared.args.no_inject_fused_mlp,
'use_safetensors': use_safetensors,
'trust_remote_code': shared.args.trust_remote_code,
'max_memory': get_max_memory_dict(),

2
modules/shared.py
View file

@ -142,6 +142,8 @@ parser.add_argument('--fused_mlp', action='store_true', help='(triton) Enable fu
parser.add_argument('--gptq-for-llama', action='store_true', help='Use GPTQ-for-LLaMa to load the GPTQ model instead of AutoGPTQ.')
parser.add_argument('--autogptq', action='store_true', help='DEPRECATED')
parser.add_argument('--triton', action='store_true', help='Use triton.')
parser.add_argument('--no_inject_fused_attention', action='store_true', help='Do not use fused attention (lowers VRAM requirements).')
parser.add_argument('--no_inject_fused_mlp', action='store_true', help='Triton mode only: Do not use fused MLP (lowers VRAM requirements).')
parser.add_argument('--desc_act', action='store_true', help='For models that don\'t have a quantize_config.json, this parameter is used to define whether to set desc_act or not in BaseQuantizeConfig.')
# FlexGen

2
modules/ui.py
View file

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