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

Add no_flash_attn option

Browse Source
This commit is contained in:
oobabooga 2023-11-02 08:19:42 -07:00
parent aaf726dbfb
commit 77abd9b69b
4 changed files with 4 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
README.md
View File

@ -336,6 +336,7 @@ Optionally, you can use the following command-line flags:
|`--gpu-split` | Comma-separated list of VRAM (in GB) to use per GPU device for model layers. Example: 20,7,7. | |`--gpu-split` | Comma-separated list of VRAM (in GB) to use per GPU device for model layers. Example: 20,7,7. |
|`--max_seq_len MAX_SEQ_LEN` | Maximum sequence length. | |`--max_seq_len MAX_SEQ_LEN` | Maximum sequence length. |
|`--cfg-cache` | ExLlama_HF: Create an additional cache for CFG negative prompts. Necessary to use CFG with that loader, but not necessary for CFG with base ExLlama. | |`--cfg-cache` | ExLlama_HF: Create an additional cache for CFG negative prompts. Necessary to use CFG with that loader, but not necessary for CFG with base ExLlama. |
|`--no_flash_attn` | Force flash-attention to not be used. |
#### AutoGPTQ #### AutoGPTQ

1
modules/exllamav2.py
View File

@ -46,6 +46,7 @@ class Exllamav2Model:
config.max_seq_len = shared.args.max_seq_len config.max_seq_len = shared.args.max_seq_len
config.scale_pos_emb = shared.args.compress_pos_emb config.scale_pos_emb = shared.args.compress_pos_emb
config.scale_alpha_value = shared.args.alpha_value config.scale_alpha_value = shared.args.alpha_value
config.no_flash_attn = shared.args.no_flash_attn
model = ExLlamaV2(config) model = ExLlamaV2(config)

1
modules/exllamav2_hf.py
View File

@ -152,5 +152,6 @@ class Exllamav2HF(PreTrainedModel):
config.max_seq_len = shared.args.max_seq_len config.max_seq_len = shared.args.max_seq_len
config.scale_pos_emb = shared.args.compress_pos_emb config.scale_pos_emb = shared.args.compress_pos_emb
config.scale_alpha_value = shared.args.alpha_value config.scale_alpha_value = shared.args.alpha_value
config.no_flash_attn = shared.args.no_flash_attn
return Exllamav2HF(config) return Exllamav2HF(config)

1
modules/shared.py
View File

@ -117,6 +117,7 @@ parser.add_argument('--cache-capacity', type=str, help='Maximum cache capacity (
parser.add_argument('--gpu-split', type=str, help='Comma-separated list of VRAM (in GB) to use per GPU device for model layers. Example: 20,7,7.') parser.add_argument('--gpu-split', type=str, help='Comma-separated list of VRAM (in GB) to use per GPU device for model layers. Example: 20,7,7.')
parser.add_argument('--max_seq_len', type=int, default=2048, help='Maximum sequence length.') parser.add_argument('--max_seq_len', type=int, default=2048, help='Maximum sequence length.')
parser.add_argument('--cfg-cache', action='store_true', help='ExLlama_HF: Create an additional cache for CFG negative prompts. Necessary to use CFG with that loader, but not necessary for CFG with base ExLlama.') parser.add_argument('--cfg-cache', action='store_true', help='ExLlama_HF: Create an additional cache for CFG negative prompts. Necessary to use CFG with that loader, but not necessary for CFG with base ExLlama.')
parser.add_argument('--no_flash_attn', action='store_true', help='Force flash-attention to not be used.')
# AutoGPTQ # AutoGPTQ
parser.add_argument('--triton', action='store_true', help='Use triton.') parser.add_argument('--triton', action='store_true', help='Use triton.')