Add --threads flag for llama.cpp

README.md

@@ -176,24 +176,31 @@ Optionally, you can use the following command-line flags:
 | Flag             | Description |
 |------------------|-------------|
 | `-h`, `--help`   | show this help message and exit |
-| `--model MODEL`  | Name of the model to load by default. |
-| `--lora LORA`    | Name of the LoRA to apply to the model by default. |
 | `--notebook`     | Launch the web UI in notebook mode, where the output is written to the same text box as the input. |
 | `--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 `img_bot.png` or `img_bot.jpg` exists in the same folder as server.py, this image will be used as the bot's profile picture. Similarly, `img_me.png` or `img_me.jpg` will be used as your profile picture. |
+| `--model MODEL`  | Name of the model to load by default. |
+| `--lora LORA`    | Name of the LoRA to apply to the model by default. |
+|  `--model-dir MODEL_DIR`                    | Path to directory with all the models |
+|  `--lora-dir LORA_DIR`                      | Path to directory with all the loras |
+| `--no-stream`    | Don't stream the text output in real time. |
+| `--settings SETTINGS_FILE` | Load the default interface settings from this json file. See `settings-template.json` for an example. If you create a file called `settings.json`, this file will be loaded by default without the need to use the `--settings` flag.|
+|  `--extensions EXTENSIONS [EXTENSIONS ...]` | The list of extensions to load. If you want to load more than one extension, write the names separated by spaces. |
+|  `--verbose`                                | Print the prompts to the terminal. |
 | `--cpu`          | Use the CPU to generate text.|
+| `--auto-devices` | Automatically split the model across the available GPU(s) and CPU.|
+|  `--gpu-memory GPU_MEMORY [GPU_MEMORY ...]` |  Maxmimum GPU memory in GiB to be allocated per GPU. Example: `--gpu-memory 10` for a single GPU, `--gpu-memory 10 5` for two GPUs. You can also set values in MiB like `--gpu-memory 3500MiB`. |
+| `--cpu-memory CPU_MEMORY` | Maximum CPU memory in GiB to allocate for offloaded weights. Must be an integer number. Defaults to 99.|
+| `--disk`         | If the model is too large for your GPU(s) and CPU combined, send the remaining layers to the disk. |
+| `--disk-cache-dir DISK_CACHE_DIR` | Directory to save the disk cache to. Defaults to `cache/`. |
 | `--load-in-8bit` | Load the model with 8-bit precision.|
+| `--bf16`         | Load the model with bfloat16 precision. Requires NVIDIA Ampere GPU. |
+| `--no-cache`     | Set `use_cache` to False while generating text. This reduces the VRAM usage a bit with a performance cost. |
+| `--threads`     | Number of threads to use in llama.cpp. |
 | `--wbits WBITS`            | GPTQ: Load a pre-quantized model with specified precision in bits. 2, 3, 4 and 8 are supported. |
 | `--model_type MODEL_TYPE`  | GPTQ: Model type of pre-quantized model. Currently LLaMA, OPT, and GPT-J are supported. |
 | `--groupsize GROUPSIZE`    | GPTQ: Group size. |
 | `--pre_layer PRE_LAYER`    | GPTQ: The number of layers to preload. |
-| `--bf16`         | Load the model with bfloat16 precision. Requires NVIDIA Ampere GPU. |
-| `--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-cache-dir DISK_CACHE_DIR` | Directory to save the disk cache to. Defaults to `cache/`. |
-|  `--gpu-memory GPU_MEMORY [GPU_MEMORY ...]` |  Maxmimum GPU memory in GiB to be allocated per GPU. Example: `--gpu-memory 10` for a single GPU, `--gpu-memory 10 5` for two GPUs. You can also set values in MiB like `--gpu-memory 3500MiB`. |
-| `--cpu-memory CPU_MEMORY` | Maximum CPU memory in GiB to allocate for offloaded weights. Must be an integer number. Defaults to 99.|
-| `--no-cache`     | Set `use_cache` to False while generating text. This reduces the VRAM usage a bit with a performance cost. |
 | `--flexgen`      | Enable the use of FlexGen offloading. |
 |  `--percent PERCENT [PERCENT ...]` |  FlexGen: allocation percentages. Must be 6 numbers separated by spaces (default: 0, 100, 100, 0, 100, 0). |
 |  `--compress-weight` |  FlexGen: Whether to compress weight (default: False).|
@@ -203,12 +210,6 @@ Optionally, you can use the following command-line flags:
 | `--local_rank LOCAL_RANK` | DeepSpeed: Optional argument for distributed setups. |
 |  `--rwkv-strategy RWKV_STRATEGY` |    RWKV: The strategy to use while loading the model. Examples: "cpu fp32", "cuda fp16", "cuda fp16i8". |
 |  `--rwkv-cuda-on` |   RWKV: Compile the CUDA kernel for better performance. |
-| `--no-stream`    | Don't stream the text output in real time. |
-| `--settings SETTINGS_FILE` | Load the default interface settings from this json file. See `settings-template.json` for an example. If you create a file called `settings.json`, this file will be loaded by default without the need to use the `--settings` flag.|
-|  `--extensions EXTENSIONS [EXTENSIONS ...]` | The list of extensions to load. If you want to load more than one extension, write the names separated by spaces. |
-|  `--model-dir MODEL_DIR`                    | Path to directory with all the models |
-|  `--lora-dir LORA_DIR`                      | Path to directory with all the loras |
-|  `--verbose`                                | Print the prompts to the terminal. |
 |  `--listen`                                 | Make the web UI reachable from your local network. |
 |  `--listen-port LISTEN_PORT`                | The listening port that the server will use. |
 |  `--share`                                  | Create a public URL. This is useful for running the web UI on Google Colab or similar. |

modules/llamacpp_model.py

@@ -1,5 +1,8 @@
+import multiprocessing
+
 import llamacpp
 
+from modules import shared
 from modules.callbacks import Iteratorize
 
 
@@ -29,6 +32,7 @@ class LlamaCppModel:
     def from_pretrained(self, path):
         params = llamacpp.InferenceParams()
         params.path_model = str(path)
+        params.n_threads = shared.args.threads or multiprocessing.cpu_count() // 2
 
         _model = llamacpp.LlamaInference(params)
 

modules/shared.py

@@ -68,51 +68,68 @@ def str2bool(v):
         raise argparse.ArgumentTypeError('Boolean value expected.')
 
 parser = argparse.ArgumentParser(formatter_class=lambda prog: argparse.HelpFormatter(prog,max_help_position=54))
-parser.add_argument('--model', type=str, help='Name of the model to load by default.')
+
-parser.add_argument('--lora', type=str, help='Name of the LoRA to apply to the model by default.')
+# Basic settings
 parser.add_argument('--notebook', action='store_true', help='Launch the web UI in notebook mode, where the output is written to the same text box as the input.')
 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 img_bot.png or img_bot.jpg exists in the same folder as server.py, this image will be used as the bot\'s profile picture. Similarly, img_me.png or img_me.jpg will be used as your profile picture.')
-parser.add_argument('--cpu', action='store_true', help='Use the CPU to generate text.')
+parser.add_argument('--model', type=str, help='Name of the model to load by default.')
-parser.add_argument('--load-in-8bit', action='store_true', help='Load the model with 8-bit precision.')
+parser.add_argument('--lora', type=str, help='Name of the LoRA to apply to the model by default.')
+parser.add_argument("--model-dir", type=str, default='models/', help="Path to directory with all the models")
+parser.add_argument("--lora-dir", type=str, default='loras/', help="Path to directory with all the loras")
+parser.add_argument('--no-stream', action='store_true', help='Don\'t stream the text output in real time.')
+parser.add_argument('--settings', type=str, help='Load the default interface settings from this json file. See settings-template.json for an example. If you create a file called settings.json, this file will be loaded by default without the need to use the --settings flag.')
+parser.add_argument('--extensions', type=str, nargs="+", help='The list of extensions to load. If you want to load more than one extension, write the names separated by spaces.')
+parser.add_argument('--verbose', action='store_true', help='Print the prompts to the terminal.')
 
-parser.add_argument('--gptq-bits', type=int, default=0, help='DEPRECATED: use --wbits instead.')
+# Accelerate/transformers
-parser.add_argument('--gptq-model-type', type=str, help='DEPRECATED: use --model_type instead.')
+parser.add_argument('--cpu', action='store_true', help='Use the CPU to generate text.')
-parser.add_argument('--gptq-pre-layer', type=int, default=0, help='DEPRECATED: use --pre_layer instead.')
+parser.add_argument('--auto-devices', action='store_true', help='Automatically split the model across the available GPU(s) and CPU.')
+parser.add_argument('--gpu-memory', type=str, nargs="+", help='Maxmimum GPU memory in GiB to be allocated per GPU. Example: --gpu-memory 10 for a single GPU, --gpu-memory 10 5 for two GPUs.')
+parser.add_argument('--cpu-memory', type=str, help='Maximum CPU memory in GiB to allocate for offloaded weights. Must be an integer number. Defaults to 99.')
+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-cache-dir', type=str, default="cache", help='Directory to save the disk cache to. Defaults to "cache".')
+parser.add_argument('--load-in-8bit', action='store_true', help='Load the model with 8-bit precision.')
+parser.add_argument('--bf16', action='store_true', help='Load the model with bfloat16 precision. Requires NVIDIA Ampere GPU.')
+parser.add_argument('--no-cache', action='store_true', help='Set use_cache to False while generating text. This reduces the VRAM usage a bit at a performance cost.')
+
+# llama.cpp
+parser.add_argument('--threads', type=int, default=0, help='Number of threads to use in llama.cpp.')
+
+# GPTQ
 parser.add_argument('--wbits', type=int, default=0, help='GPTQ: Load a pre-quantized model with specified precision in bits. 2, 3, 4 and 8 are supported.')
 parser.add_argument('--model_type', type=str, help='GPTQ: Model type of pre-quantized model. Currently LLaMA, OPT, and GPT-J are supported.')
 parser.add_argument('--groupsize', type=int, default=-1, help='GPTQ: Group size.')
 parser.add_argument('--pre_layer', type=int, default=0, help='GPTQ: The number of layers to preload.')
+parser.add_argument('--gptq-bits', type=int, default=0, help='DEPRECATED: use --wbits instead.')
+parser.add_argument('--gptq-model-type', type=str, help='DEPRECATED: use --model_type instead.')
+parser.add_argument('--gptq-pre-layer', type=int, default=0, help='DEPRECATED: use --pre_layer instead.')
 
-parser.add_argument('--bf16', action='store_true', help='Load the model with bfloat16 precision. Requires NVIDIA Ampere GPU.')
+# FlexGen
-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-cache-dir', type=str, default="cache", help='Directory to save the disk cache to. Defaults to "cache".')
-parser.add_argument('--gpu-memory', type=str, nargs="+", help='Maxmimum GPU memory in GiB to be allocated per GPU. Example: --gpu-memory 10 for a single GPU, --gpu-memory 10 5 for two GPUs.')
-parser.add_argument('--cpu-memory', type=str, help='Maximum CPU memory in GiB to allocate for offloaded weights. Must be an integer number. Defaults to 99.')
-parser.add_argument('--no-cache', action='store_true', help='Set use_cache to False while generating text. This reduces the VRAM usage a bit at a performance cost.')
 parser.add_argument('--flexgen', action='store_true', help='Enable the use of FlexGen offloading.')
 parser.add_argument('--percent', type=int, nargs="+", default=[0, 100, 100, 0, 100, 0], help='FlexGen: allocation percentages. Must be 6 numbers separated by spaces (default: 0, 100, 100, 0, 100, 0).')
 parser.add_argument("--compress-weight", action="store_true", help="FlexGen: activate weight compression.")
 parser.add_argument("--pin-weight", type=str2bool, nargs="?", const=True, default=True, help="FlexGen: whether to pin weights (setting this to False reduces CPU memory by 20%%).")
+
+# DeepSpeed
 parser.add_argument('--deepspeed', action='store_true', help='Enable the use of DeepSpeed ZeRO-3 for inference via the Transformers integration.')
 parser.add_argument('--nvme-offload-dir', type=str, help='DeepSpeed: Directory to use for ZeRO-3 NVME offloading.')
 parser.add_argument('--local_rank', type=int, default=0, help='DeepSpeed: Optional argument for distributed setups.')
+
+# RWKV
 parser.add_argument('--rwkv-strategy', type=str, default=None, help='RWKV: The strategy to use while loading the model. Examples: "cpu fp32", "cuda fp16", "cuda fp16i8".')
 parser.add_argument('--rwkv-cuda-on', action='store_true', help='RWKV: Compile the CUDA kernel for better performance.')
-parser.add_argument('--no-stream', action='store_true', help='Don\'t stream the text output in real time.')
+
-parser.add_argument('--settings', type=str, help='Load the default interface settings from this json file. See settings-template.json for an example. If you create a file called settings.json, this file will be loaded by default without the need to use the --settings flag.')
+# Gradio
-parser.add_argument('--extensions', type=str, nargs="+", help='The list of extensions to load. If you want to load more than one extension, write the names separated by spaces.')
-parser.add_argument("--model-dir", type=str, default='models/', help="Path to directory with all the models")
-parser.add_argument("--lora-dir", type=str, default='loras/', help="Path to directory with all the loras")
-parser.add_argument('--verbose', action='store_true', help='Print the prompts to the terminal.')
 parser.add_argument('--listen', action='store_true', help='Make the web UI reachable from your local network.')
 parser.add_argument('--listen-port', type=int, help='The listening port that the server will use.')
 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('--auto-launch', action='store_true', default=False, help='Open the web UI in the default browser upon launch.')
 parser.add_argument("--gradio-auth-path", type=str, help='Set the gradio authentication file path. The file should contain one or more user:password pairs in this format: "u1:p1,u2:p2,u3:p3"', default=None)
+
 args = parser.parse_args()
 
+
 # Provisional, this will be deleted later
 deprecated_dict = {'gptq_bits': ['wbits', 0], 'gptq_model_type': ['model_type', None], 'gptq_pre_layer': ['prelayer', 0]}
 for k in deprecated_dict: