Merge branch 'main' into llm-server-int-test

nod-ai · Nov 5, 2024 · 5876c37 · 5876c37
2 parents ee675ce + 85220d1
commit 5876c37
Show file tree

Hide file tree

Showing 3 changed files with 7 additions and 4 deletions.
diff --git a/sharktank/sharktank/serving_poc/framework/session.py b/sharktank/sharktank/serving_poc/framework/session.py
@@ -319,7 +319,7 @@ def __init__(self, session: DeviceSession, index: int = 0):
         self._semaphore = session.device.create_semaphore(0)
         self._step = 0
 
-    def execute_sequential(self, command_buffers: list[HalCommandBuffer]):
+    def execute_sequential(self, command_buffer: HalCommandBuffer):
         """Executes a list of command buffers at the current step, advancing to the
         next.
         """
@@ -329,7 +329,7 @@ def execute_sequential(self, command_buffers: list[HalCommandBuffer]):
             self._step = next_step
         sem = self._semaphore
         self._device.queue_execute(
-            command_buffers, [(sem, current_step)], [(sem, next_step)]
+            command_buffer, [(sem, current_step)], [(sem, next_step)]
         )
 
     def current_fence(self) -> HalFence:

diff --git a/sharktank/sharktank/serving_poc/llm/impl/service_v1.py b/sharktank/sharktank/serving_poc/llm/impl/service_v1.py
@@ -340,7 +340,7 @@ async def prefill(self) -> TimelineGuarded[HalBufferView]:
 
         # Perform h2d transfers.
         cb.end()
-        work_queue.execute_sequential([cb])
+        work_queue.execute_sequential(cb)
 
         # Inputs:
         #   token_ids
@@ -468,7 +468,7 @@ async def decode(self) -> TimelineGuarded[HalBufferView]:
 
         # Perform h2d transfers.
         cb.end()
-        work_queue.execute_sequential([cb])
+        work_queue.execute_sequential(cb)
 
         # Inputs:
         #   token_ids

diff --git a/shortfin/docs/reference.rst b/shortfin/docs/reference.rst
@@ -16,7 +16,9 @@ Array
 .. autoclass:: base_array
 .. autoclass:: device_array
     :members:
+.. autoclass:: RandomGenerator
 
+.. autofunction:: _shortfin_default.lib.array.fill_randn
 .. autofunction:: _shortfin_default.lib.array.argmax
 
 Local
@@ -52,6 +54,7 @@ AMD GPU
 ^^^^^^^
 .. automodule:: _shortfin_default.lib.local.amdgpu
 .. autoclass:: SystemBuilder
+  :members:
 .. autoclass:: AMDGPUDevice
 
 Host