clean up pull

ollama/engine.py

@@ -22,7 +22,6 @@ def generate(model, prompt, models_home=".", llms={}, *args, **kwargs):
     llm = load(model, models_home=models_home, llms=llms)
 
     prompt = ollama.prompt.template(model, prompt)
-
     if "max_tokens" not in kwargs:
         kwargs.update({"max_tokens": 16384})
 
@@ -39,11 +38,10 @@ def generate(model, prompt, models_home=".", llms={}, *args, **kwargs):
 def load(model, models_home=".", llms={}):
     llm = llms.get(model, None)
     if not llm:
-        model_path = {
-            name: path for name, path in ollama.model.models(models_home)
-        }.get(model, None)
-
-        if not model_path:
+        stored_model_path = os.path.join(models_home, model, ".bin")
+        if os.path.exists(stored_model_path):
+            model_path = stored_model_path
+        else:
             # try loading this as a path to a model, rather than a model name
             model_path = os.path.abspath(model)
 

ollama/model.py

@@ -16,26 +16,18 @@ def models(models_home='.', *args, **kwargs):
                 yield base
 
 
-def pull(model, models_home='.', *args, **kwargs):
-    url = model
-    if not validators.url(url) and not url.startswith('huggingface.co'):
-        # this may just be a local model location
-        if model in models(models_home):
-            return model
-        # see if we have this model in our directory
-        response = requests.get(models_endpoint_url)
-        response.raise_for_status()
-        directory = response.json()
-        for model_info in directory:
-            if model_info.get('name') == model:
-                url = f"https://{model_info.get('url')}"
-                break
-        if not validators.url(url):
-            raise Exception(f'Unknown model {model}')
+# get the url of the model from our curated directory
+def get_url_from_directory(model):
+    response = requests.get(models_endpoint_url)
+    response.raise_for_status()
+    directory = response.json()
+    for model_info in directory:
+        if model_info.get('name') == model:
+            return model_info.get('url')
+    return model
 
-    if not (url.startswith('http://') or url.startswith('https://')):
-        url = f'https://{url}'
 
+def download_from_repo(url, models_home='.'):
     parts = urlsplit(url)
     path_parts = parts.path.split('/tree/')
 
@@ -47,7 +39,6 @@ def pull(model, models_home='.', *args, **kwargs):
 
     location = location.strip('/')
 
-    # Reconstruct the URL
     download_url = urlunsplit(
         (
             'https',
@@ -57,13 +48,15 @@ def pull(model, models_home='.', *args, **kwargs):
             parts.fragment,
         )
     )
-
     response = requests.get(download_url)
-    response.raise_for_status()  # Raises stored HTTPError, if one occurred
-
+    response.raise_for_status()
     json_response = response.json()
 
-    # get the last bin file we find, this is probably the most up to date
+    download_url, file_size = find_bin_file(json_response, location, branch)
+    return download_file(download_url, models_home, location, file_size)
+
+
+def find_bin_file(json_response, location, branch):
     download_url = None
     file_size = 0
     for file_info in json_response:
@@ -77,27 +70,25 @@ def pull(model, models_home='.', *args, **kwargs):
     if download_url is None:
         raise Exception('No model found')
 
-    local_filename = os.path.join(models_home, os.path.basename(url)) + '.bin'
+    return download_url, file_size
 
-    # Check if file already exists
-    first_byte = 0
-    if os.path.exists(local_filename):
-        # TODO: check if the file is the same SHA
-        first_byte = os.path.getsize(local_filename)
+
+def download_file(download_url, models_home, location, file_size):
+    local_filename = os.path.join(models_home, os.path.basename(location)) + '.bin'
+
+    first_byte = (
+        os.path.getsize(local_filename) if os.path.exists(local_filename) else 0
+    )
 
     if first_byte >= file_size:
         return local_filename
 
-    print(f'Pulling {model}...')
+    print(f'Pulling {os.path.basename(location)}...')
 
-    # If file size is non-zero, resume download
-    if first_byte != 0:
-        header = {'Range': f'bytes={first_byte}-'}
-    else:
-        header = {}
+    header = {'Range': f'bytes={first_byte}-'} if first_byte != 0 else {}
 
     response = requests.get(download_url, headers=header, stream=True)
-    response.raise_for_status()  # Raises stored HTTPError, if one occurred
+    response.raise_for_status()
 
     total_size = int(response.headers.get('content-length', 0))
 
@@ -115,3 +106,26 @@ def pull(model, models_home='.', *args, **kwargs):
             bar.update(size)
 
     return local_filename
+
+
+def pull(model, models_home='.', *args, **kwargs):
+    if os.path.exists(model):
+        # a file on the filesystem is being specified
+        return model
+    # check the remote model location and see if it needs to be downloaded
+    url = model
+    if not validators.url(url) and not url.startswith('huggingface.co'):
+        url = get_url_from_directory(model)
+
+    if not (url.startswith('http://') or url.startswith('https://')):
+        url = f'https://{url}'
+
+    if not validators.url(url):
+        if model in models(models_home):
+            # the model is already downloaded, and specified by name
+            return model
+        raise Exception(f'Unknown model {model}')
+
+    local_filename = download_from_repo(url, models_home)
+
+    return local_filename

ollama/prompt.py

@@ -1,4 +1,4 @@
-from os import path
+import os
 from difflib import SequenceMatcher
 from jinja2 import Environment, PackageLoader
 
@@ -9,8 +9,8 @@ def template(model, prompt):
 
     environment = Environment(loader=PackageLoader(__name__, 'templates'))
     for template in environment.list_templates():
-        base, _ = path.splitext(template)
-        ratio = SequenceMatcher(None, model.lower(), base).ratio()
+        base, _ = os.path.splitext(template)
+        ratio = SequenceMatcher(None, os.path.basename(model.lower()), base).ratio()
         if ratio > best_ratio:
             best_ratio = ratio
             best_template = template