Merge branch 'upstream' into concedo_experimental

# Conflicts:
#	.github/workflows/build.yml
#	.github/workflows/server.yml
#	CMakeLists.txt
#	cmake/build-info.cmake
#	examples/run/CMakeLists.txt
#	examples/run/run.cpp
#	examples/simple-chat/simple-chat.cpp
#	tests/CMakeLists.txt
#	tests/test-backend-ops.cpp
#	tests/test-sampling.cpp

examples/server/tests/unit/test_completion.py

@@ -1,4 +1,5 @@
 import pytest
+import requests
 import time
 from openai import OpenAI
 from utils import *
@@ -405,3 +406,23 @@ def test_n_probs_post_sampling():
             assert "bytes" in prob and type(prob["bytes"]) == list
         # because the test model usually output token with either 100% or 0% probability, we need to check all the top_probs
         assert any(prob["prob"] == 1.0 for prob in tok["top_probs"])
+
+
+def test_cancel_request():
+    global server
+    server.n_ctx = 4096
+    server.n_predict = -1
+    server.n_slots = 1
+    server.server_slots = True
+    server.start()
+    # send a request that will take a long time, but cancel it before it finishes
+    try:
+        server.make_request("POST", "/completion", data={
+            "prompt": "I believe the meaning of life is",
+        }, timeout=0.1)
+    except requests.exceptions.ReadTimeout:
+        pass # expected
+    # make sure the slot is free
+    time.sleep(1) # wait for HTTP_POLLING_SECONDS
+    res = server.make_request("GET", "/slots")
+    assert res.body[0]["is_processing"] == False

examples/server/tests/utils.py

@@ -26,6 +26,9 @@ from re import RegexFlag
 import wget
 
 
+DEFAULT_HTTP_TIMEOUT = 10 if "LLAMA_SANITIZE" not in os.environ else 30
+
+
 class ServerResponse:
     headers: dict
     status_code: int
@@ -88,7 +91,7 @@ class ServerProcess:
         if "PORT" in os.environ:
             self.server_port = int(os.environ["PORT"])
 
-    def start(self, timeout_seconds: int = 10) -> None:
+    def start(self, timeout_seconds: int | None = DEFAULT_HTTP_TIMEOUT) -> None:
         if "LLAMA_SERVER_BIN_PATH" in os.environ:
             server_path = os.environ["LLAMA_SERVER_BIN_PATH"]
         elif os.name == "nt":
@@ -219,17 +222,18 @@ class ServerProcess:
         path: str,
         data: dict | Any | None = None,
         headers: dict | None = None,
+        timeout: float | None = None,
     ) -> ServerResponse:
         url = f"http://{self.server_host}:{self.server_port}{path}"
         parse_body = False
         if method == "GET":
-            response = requests.get(url, headers=headers)
+            response = requests.get(url, headers=headers, timeout=timeout)
             parse_body = True
         elif method == "POST":
-            response = requests.post(url, headers=headers, json=data)
+            response = requests.post(url, headers=headers, json=data, timeout=timeout)
             parse_body = True
         elif method == "OPTIONS":
-            response = requests.options(url, headers=headers)
+            response = requests.options(url, headers=headers, timeout=timeout)
         else:
             raise ValueError(f"Unimplemented method: {method}")
         result = ServerResponse()