AI/text-generation-webui
1
0
Fork 0
mirror of https://github.com/oobabooga/text-generation-webui.git synced 2024-10-01 01:26:03 -04:00
Code Issues Packages Projects Releases Wiki Activity

Change GPTQ triton default settings

Browse Source
This commit is contained in:
oobabooga 2023-04-22 12:27:30 -03:00
parent e03b873460
commit 7438f4f6ba
3 changed files with 10 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
README.md
View File

@ -230,9 +230,9 @@ Optionally, you can use the following command-line flags:
| `--groupsize GROUPSIZE` | Group size. |
| `--pre_layer PRE_LAYER` | The number of layers to allocate to the GPU. Setting this parameter enables CPU offloading for 4-bit models. |
| `--monkey-patch` | Apply the monkey patch for using LoRAs with quantized models.
| `--no-quant_attn` | (triton) Disable quant attention. If you encounter incoherent results try disabling this.
| `--no-warmup_autotune` | (triton) Disable warmup autotune.
| `--no-fused_mlp` | (triton) Disable fused mlp. If you encounter "Unexpected mma -> mma layout conversion" try disabling this.
| `--quant_attn` | (triton) Enable quant attention.
| `--warmup_autotune` | (triton) Enable warmup autotune.
| `--fused_mlp` | (triton) Enable fused mlp.
#### FlexGen

8
modules/GPTQ_loader.py
View File

@ -79,14 +79,14 @@ def _load_quant(model, checkpoint, wbits, groupsize=-1, faster_kernel=False, exc
model.load_state_dict(torch.load(checkpoint), strict=False)
if is_triton:
if not shared.args.no_quant_attn:
if shared.args.quant_attn:
quant.make_quant_attn(model)
if eval and not shared.args.no_fused_mlp:
if eval and shared.args.fused_mlp:
quant.make_fused_mlp(model)
if not shared.args.no_warmup_autotune:
if shared.args.warmup_autotune:
quant.autotune_warmup_linear(model, transpose=not eval)
if eval and not shared.args.no_fused_mlp:
if eval and shared.args.fused_mlp:
quant.autotune_warmup_fused(model)
model.seqlen = 2048

6
modules/shared.py
View File

@ -124,9 +124,9 @@ parser.add_argument('--model_type', type=str, help='Model type of pre-quantized
parser.add_argument('--groupsize', type=int, default=-1, help='Group size.')
parser.add_argument('--pre_layer', type=int, default=0, help='The number of layers to allocate to the GPU. Setting this parameter enables CPU offloading for 4-bit models.')
parser.add_argument('--monkey-patch', action='store_true', help='Apply the monkey patch for using LoRAs with quantized models.')
parser.add_argument('--no-quant_attn', action='store_true', help='(triton) Disable quant attention. If you encounter incoherent results try disabling this.')
parser.add_argument('--no-warmup_autotune', action='store_true', help='(triton) Disable warmup autotune.')
parser.add_argument('--no-fused_mlp', action='store_true', help='(triton) Disable fused mlp. If you encounter "Unexpected mma -> mma layout conversion" try disabling this.')
parser.add_argument('--quant_attn', action='store_true', help='(triton) Enable quant attention.')
parser.add_argument('--warmup_autotune', action='store_true', help='(triton) Enable warmup autotune.')
parser.add_argument('--fused_mlp', action='store_true', help='(triton) Enable fused mlp.')
# FlexGen
parser.add_argument('--flexgen', action='store_true', help='Enable the use of FlexGen offloading.')