Mirrors/text-generation-webui
1
0
Fork 0
mirror of https://github.com/oobabooga/text-generation-webui.git synced 2024-12-25 05:48:55 +01:00
Code Issues Packages Projects Releases Wiki Activity

Better parameter naming

Browse Source
This commit is contained in:
oobabooga 2023-01-20 23:45:16 -03:00
parent c0f2367b54
commit faaafe7c0e
2 changed files with 11 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
README.md
View File

@ -139,10 +139,10 @@ Optionally, you can use the following command-line flags:
| `--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.| | `--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. | | `--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. | | `--gpu-memory 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-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.|
| `--settings SETTINGS_FILE` | Load the default interface settings from this json file. See `settings-template.json` for an example.| | `--settings SETTINGS_FILE` | Load the default interface settings from this json file. See `settings-template.json` for an example.|
| `--no-listen` | Make the web UI unreachable from your local network.| | `--listen` | Make the web UI reachable from your local network.|
| `--share` | Create a public URL. This is useful for running the web UI on Google Colab or similar. | | `--share` | Create a public URL. This is useful for running the web UI on Google Colab or similar. |
Out of memory errors? [Check this guide](https://github.com/oobabooga/text-generation-webui/wiki/Low-VRAM-guide). Out of memory errors? [Check this guide](https://github.com/oobabooga/text-generation-webui/wiki/Low-VRAM-guide).

20
server.py
View File

@ -26,10 +26,10 @@ parser.add_argument('--cpu', action='store_true', help='Use the CPU to generate
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('--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('--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('--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-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.')
parser.add_argument('--settings', type=str, help='Load the default interface settings from this json file. See settings-template.json for an example.') parser.add_argument('--settings', type=str, help='Load the default interface settings from this json file. See settings-template.json for an example.')
parser.add_argument('--no-listen', action='store_true', help='Make the web UI unreachable from your local network.') parser.add_argument('--listen', action='store_true', help='Make the web UI reachable from your local network.')
parser.add_argument('--share', action='store_true', help='Create a public URL. This is useful for running the web UI on Google Colab or similar.') parser.add_argument('--share', action='store_true', help='Create a public URL. This is useful for running the web UI on Google Colab or similar.')
args = parser.parse_args() args = parser.parse_args()
@ -71,7 +71,7 @@ def load_model(model_name):
t0 = time.time() t0 = time.time()
# Default settings # Default settings
if not (args.cpu or args.load_in_8bit or args.auto_devices or args.disk or args.max_gpu_memory is not None): if not (args.cpu or args.load_in_8bit or args.auto_devices or args.disk or args.gpu_memory is not None):
if Path(f"torch-dumps/{model_name}.pt").exists(): if Path(f"torch-dumps/{model_name}.pt").exists():
print("Loading in .pt format...") print("Loading in .pt format...")
model = torch.load(Path(f"torch-dumps/{model_name}.pt")) model = torch.load(Path(f"torch-dumps/{model_name}.pt"))
@ -89,8 +89,8 @@ def load_model(model_name):
settings.append("torch_dtype=torch.float32") settings.append("torch_dtype=torch.float32")
else: else:
settings.append("device_map='auto'") settings.append("device_map='auto'")
if args.max_gpu_memory is not None: if args.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.gpu_memory}GiB', 'cpu': '99GiB'}}")
if args.disk: if args.disk:
settings.append("offload_folder='cache'") settings.append("offload_folder='cache'")
if args.load_in_8bit: if args.load_in_8bit:
@ -160,8 +160,7 @@ def generate_reply(question, tokens, inference_settings, selected_model, eos_tok
if selected_model != model_name: if selected_model != model_name:
model_name = selected_model model_name = selected_model
model = None model = tokenizer = None
tokenizer = None
if not args.cpu: if not args.cpu:
gc.collect() gc.collect()
torch.cuda.empty_cache() torch.cuda.empty_cache()
@ -418,7 +417,6 @@ if args.chat or args.cai_chat:
else: else:
upload.upload(lambda : history, [], [display1]) upload.upload(lambda : history, [], [display1])
elif args.notebook: elif args.notebook:
with gr.Blocks(css=css, analytics_enabled=False) as interface: with gr.Blocks(css=css, analytics_enabled=False) as interface:
gr.Markdown(description) gr.Markdown(description)
@ -471,7 +469,7 @@ else:
stop.click(None, None, None, cancels=[gen_event, gen_event2, cont_event]) stop.click(None, None, None, cancels=[gen_event, gen_event2, cont_event])
interface.queue() interface.queue()
if args.no_listen: if args.listen:
interface.launch(share=args.share)
else:
interface.launch(share=args.share, server_name="0.0.0.0") interface.launch(share=args.share, server_name="0.0.0.0")
else:
interface.launch(share=args.share)