xinference 0.13.0__py3-none-any.whl → 0.13.2__py3-none-any.whl

xinference/model/rerank/core.py

@@ -18,7 +18,7 @@ import os
 import uuid
 from collections import defaultdict
 from collections.abc import Sequence
-from typing import Dict, List, Optional, Tuple
+from typing import Dict, List, Literal, Optional, Tuple
 
 import numpy as np
 import torch
@@ -285,7 +285,12 @@ def cache(model_spec: RerankModelSpec):
 
 
 def create_rerank_model_instance(
-    subpool_addr: str, devices: List[str], model_uid: str, model_name: str, **kwargs
+    subpool_addr: str,
+    devices: List[str],
+    model_uid: str,
+    model_name: str,
+    download_hub: Optional[Literal["huggingface", "modelscope", "csghub"]] = None,
+    **kwargs,
 ) -> Tuple[RerankModel, RerankModelDescription]:
     from ..utils import download_from_modelscope
     from . import BUILTIN_RERANK_MODELS, MODELSCOPE_RERANK_MODELS
@@ -298,30 +303,24 @@ def create_rerank_model_instance(
             break
 
     if model_spec is None:
-        if download_from_modelscope():
-            if model_name in MODELSCOPE_RERANK_MODELS:
-                logger.debug(f"Rerank model {model_name} found in ModelScope.")
-                model_spec = MODELSCOPE_RERANK_MODELS[model_name]
-            else:
-                logger.debug(
-                    f"Rerank model {model_name} not found in ModelScope, "
-                    f"now try to download from huggingface."
-                )
-                if model_name in BUILTIN_RERANK_MODELS:
-                    model_spec = BUILTIN_RERANK_MODELS[model_name]
-                else:
-                    raise ValueError(
-                        f"Rerank model {model_name} not found, available"
-                        f"model list: {BUILTIN_RERANK_MODELS.keys()}"
-                    )
+        if download_hub == "huggingface" and model_name in BUILTIN_RERANK_MODELS:
+            logger.debug(f"Rerank model {model_name} found in Huggingface.")
+            model_spec = BUILTIN_RERANK_MODELS[model_name]
+        elif download_hub == "modelscope" and model_name in MODELSCOPE_RERANK_MODELS:
+            logger.debug(f"Rerank model {model_name} found in ModelScope.")
+            model_spec = MODELSCOPE_RERANK_MODELS[model_name]
+        elif download_from_modelscope() and model_name in MODELSCOPE_RERANK_MODELS:
+            logger.debug(f"Rerank model {model_name} found in ModelScope.")
+            model_spec = MODELSCOPE_RERANK_MODELS[model_name]
+        elif model_name in BUILTIN_RERANK_MODELS:
+            logger.debug(f"Rerank model {model_name} found in Huggingface.")
+            model_spec = BUILTIN_RERANK_MODELS[model_name]
         else:
-            if model_name in BUILTIN_RERANK_MODELS:
-                model_spec = BUILTIN_RERANK_MODELS[model_name]
-            else:
-                raise ValueError(
-                    f"Rerank model {model_name} not found, available"
-                    f"model list: {BUILTIN_RERANK_MODELS.keys()}"
-                )
+            raise ValueError(
+                f"Rerank model {model_name} not found, available"
+                f"Huggingface: {BUILTIN_RERANK_MODELS.keys()}"
+                f"ModelScope: {MODELSCOPE_RERANK_MODELS.keys()}"
+            )
 
     model_path = cache(model_spec)
     use_fp16 = kwargs.pop("use_fp16", False)

xinference/types.py

@@ -285,7 +285,6 @@ class LlamaCppModelConfig(TypedDict, total=False):
     vocab_only: bool
     use_mmap: bool
     use_mlock: bool
-    embedding: bool
     n_threads: Optional[int]
     n_batch: int
     last_n_tokens_size: int

xinference/web/ui/build/asset-manifest.json

@@ -1,14 +1,14 @@
 {
   "files": {
     "main.css": "./static/css/main.4bafd904.css",
-    "main.js": "./static/js/main.0fb6f3ab.js",
+    "main.js": "./static/js/main.95c1d652.js",
     "static/media/icon.webp": "./static/media/icon.4603d52c63041e5dfbfd.webp",
     "index.html": "./index.html",
     "main.4bafd904.css.map": "./static/css/main.4bafd904.css.map",
-    "main.0fb6f3ab.js.map": "./static/js/main.0fb6f3ab.js.map"
+    "main.95c1d652.js.map": "./static/js/main.95c1d652.js.map"
   },
   "entrypoints": [
     "static/css/main.4bafd904.css",
-    "static/js/main.0fb6f3ab.js"
+    "static/js/main.95c1d652.js"
   ]
 }

xinference/web/ui/build/index.html

@@ -1 +1 @@
-<!doctype html><html lang="en"><head><meta charset="utf-8"/><link rel="icon" href="./favicon.svg"/><meta name="viewport" content="width=device-width,initial-scale=1"/><meta name="theme-color" content="#000000"/><meta name="description" content="Web site created using create-react-app"/><link rel="apple-touch-icon" href="./logo192.png"/><link rel="manifest" href="./manifest.json"/><title>Xinference</title><script defer="defer" src="./static/js/main.0fb6f3ab.js"></script><link href="./static/css/main.4bafd904.css" rel="stylesheet"></head><body><noscript>You need to enable JavaScript to run this app.</noscript><div id="root"></div></body></html>
+<!doctype html><html lang="en"><head><meta charset="utf-8"/><link rel="icon" href="./favicon.svg"/><meta name="viewport" content="width=device-width,initial-scale=1"/><meta name="theme-color" content="#000000"/><meta name="description" content="Web site created using create-react-app"/><link rel="apple-touch-icon" href="./logo192.png"/><link rel="manifest" href="./manifest.json"/><title>Xinference</title><script defer="defer" src="./static/js/main.95c1d652.js"></script><link href="./static/css/main.4bafd904.css" rel="stylesheet"></head><body><noscript>You need to enable JavaScript to run this app.</noscript><div id="root"></div></body></html>