Mirrors/text-generation-webui
1
0
Fork 0
mirror of https://github.com/oobabooga/text-generation-webui.git synced 2024-12-24 13:28:59 +01:00
Code Issues Packages Projects Releases Wiki Activity

Implement sending layers to disk with --disk (#10)

Browse Source
This commit is contained in:
oobabooga 2023-01-19 11:09:24 -03:00
parent 1ce95ee817
commit 7ace04864a
2 changed files with 7 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
README.md
View File

@ -130,8 +130,9 @@ Optionally, you can use the following command-line flags:
| `--chat` | Launch the web UI in chat mode.| | `--chat` | Launch the web UI in chat mode.|
| `--cai-chat` | Launch the web UI in chat mode with a style similar to Character.AI's. If the file profile.png or profile.jpg exists in the same folder as server.py, this image will be used as the bot's profile picture. | | `--cai-chat` | Launch the web UI in chat mode with a style similar to Character.AI's. If the file profile.png or profile.jpg exists in the same folder as server.py, this image will be used as the bot's profile picture. |
| `--cpu` | Use the CPU to generate text.| | `--cpu` | Use the CPU to generate text.|
| `--auto-devices` | Automatically split the model across the available GPU(s) and CPU.|
| `--load-in-8bit` | Load the model with 8-bit precision.| | `--load-in-8bit` | Load the model with 8-bit precision.|
| `--auto-devices` | Automatically split the model across the available GPU(s) and CPU.|
| `--disk` | If the model is too large for your GPU(s) and CPU combined, send the remaining layers to the disk. |
| `--max-gpu-memory MAX_GPU_MEMORY` | Maximum memory in GiB to allocate to the GPU when loading the model. This is useful if you get out of memory errors while trying to generate text. Must be an integer number. | | `--max-gpu-memory MAX_GPU_MEMORY` | Maximum memory in GiB to allocate to the GPU when loading the model. This is useful if you get out of memory errors while trying to generate text. Must be an integer number. |
| `--no-listen` | Make the web UI unreachable from your local network.| | `--no-listen` | Make the web UI unreachable from your local network.|
| `--no-stream` | Don't stream the text output in real time. This slightly improves the text generation performance.| | `--no-stream` | Don't stream the text output in real time. This slightly improves the text generation performance.|

6
server.py
View File

@ -21,8 +21,9 @@ parser.add_argument('--notebook', action='store_true', help='Launch the web UI i
parser.add_argument('--chat', action='store_true', help='Launch the web UI in chat mode.') parser.add_argument('--chat', action='store_true', help='Launch the web UI in chat mode.')
parser.add_argument('--cai-chat', action='store_true', help='Launch the web UI in chat mode with a style similar to Character.AI\'s. If the file profile.png or profile.jpg exists in the same folder as server.py, this image will be used as the bot\'s profile picture.') parser.add_argument('--cai-chat', action='store_true', help='Launch the web UI in chat mode with a style similar to Character.AI\'s. If the file profile.png or profile.jpg exists in the same folder as server.py, this image will be used as the bot\'s profile picture.')
parser.add_argument('--cpu', action='store_true', help='Use the CPU to generate text.') parser.add_argument('--cpu', action='store_true', help='Use the CPU to generate text.')
parser.add_argument('--auto-devices', action='store_true', help='Automatically split the model across the available GPU(s) and CPU.')
parser.add_argument('--load-in-8bit', action='store_true', help='Load the model with 8-bit precision.') parser.add_argument('--load-in-8bit', action='store_true', help='Load the model with 8-bit precision.')
parser.add_argument('--auto-devices', action='store_true', help='Automatically split the model across the available GPU(s) and CPU.')
parser.add_argument('--disk', action='store_true', help='If the model is too large for your GPU(s) and CPU combined, send the remaining layers to the disk.')
parser.add_argument('--max-gpu-memory', type=int, help='Maximum memory in GiB to allocate to the GPU when loading the model. This is useful if you get out of memory errors while trying to generate text. Must be an integer number.') parser.add_argument('--max-gpu-memory', type=int, help='Maximum memory in GiB to allocate to the GPU when loading the model. This is useful if you get out of memory errors while trying to generate text. Must be an integer number.')
parser.add_argument('--no-listen', action='store_true', help='Make the web UI unreachable from your local network.') parser.add_argument('--no-listen', action='store_true', help='Make the web UI unreachable from your local network.')
parser.add_argument('--no-stream', action='store_true', help='Don\'t stream the text output in real time. This slightly improves the text generation performance.') parser.add_argument('--no-stream', action='store_true', help='Don\'t stream the text output in real time. This slightly improves the text generation performance.')
@ -81,6 +82,8 @@ def load_model(model_name):
settings.append("device_map='auto'") settings.append("device_map='auto'")
if args.max_gpu_memory is not None: if args.max_gpu_memory is not None:
settings.append(f"max_memory={{0: '{args.max_gpu_memory}GiB', 'cpu': '99GiB'}}") settings.append(f"max_memory={{0: '{args.max_gpu_memory}GiB', 'cpu': '99GiB'}}")
if args.disk:
settings.append("offload_folder='cache'")
if args.load_in_8bit: if args.load_in_8bit:
settings.append("load_in_8bit=True") settings.append("load_in_8bit=True")
else: else:
@ -167,6 +170,7 @@ def generate_reply(question, tokens, inference_settings, selected_model, eos_tok
# Generate the reply 1 token at a time # Generate the reply 1 token at a time
else: else:
yield formatted_outputs(question, model_name)
input_ids = encode(question, 1) input_ids = encode(question, 1)
preset = preset.replace('max_new_tokens=tokens', 'max_new_tokens=1') preset = preset.replace('max_new_tokens=tokens', 'max_new_tokens=1')
for i in range(tokens): for i in range(tokens):