perf: batch decode stream callbacks into one dispatch per step

- add BatchedStreamCallback sink type: TaskManager resolves a decode step's (task_id, token) events under one lock and delivers each sink a single list instead of one call per token
- keep the plain Callable[[str]] callback contract: per-token callbacks still receive one call per event, and invoke_callback/cancel_task wrap single events for batched sinks
- collect aborted, text, and finish STOP events in the scheduler decode loop and dispatch once per step instead of once per token
- register one _ResultSink per generate call (replacing per-task closures) so GenerateResult takes its lock and wakes waiters once per step, with late-bind replay for tasks that start decoding before add_task returns their id
- apply GenerateResult batches under a single condition hold via append_batch; append delegates to it
- update engine test fakes to the batched contract and add coverage for event grouping, single-event dispatch, cancel STOP, and late-bind replay

Benchmark: NVIDIA L20 (idle), CUDA 12.8, torch 2.11.0+cu128, 1.2B bf16 checkpoint, prompt 512, 256 greedy tokens, CUDA graph on, serving-level decode, 3 trials
- batch 32: 7.808 -> 7.506 ms/token (4098 -> 4263 batch tok/s, +4.0%)
- batch 1/8: unchanged within noise (3.768 -> 3.797 / 4.699 -> 4.607 ms/token)
- full suite: 896 passed

astrai/inference/__init__.py

@@ -17,12 +17,20 @@ from astrai.inference.network import get_app, run_server
 from astrai.inference.runtime.executor import Executor
 from astrai.inference.runtime.sample import sample
 from astrai.inference.scheduler import InferenceScheduler
-from astrai.inference.task import STOP, GenerationResult, Task, TaskManager, TaskStatus
+from astrai.inference.task import (
+    STOP,
+    BatchedStreamCallback,
+    GenerationResult,
+    Task,
+    TaskManager,
+    TaskStatus,
+)
 
 __all__ = [
     "InferenceEngine",
     "build_engine",
     "InferenceScheduler",
+    "BatchedStreamCallback",
     "GenerationResult",
     "Executor",
     "STOP",

astrai/inference/engine.py

@@ -13,7 +13,7 @@ import torch.nn as nn
 from astrai.extension import ATTN_BACKEND, AttentionBackend, get_backend
 from astrai.inference.cache import PagePool
 from astrai.inference.scheduler import InferenceScheduler
-from astrai.inference.task import STOP
+from astrai.inference.task import STOP, BatchedStreamCallback
 from astrai.model import AutoModel
 from astrai.tokenize import AutoTokenizer
 
@@ -33,15 +33,27 @@ class GenerateResult:
         self._total = count
 
     def append(self, token: str, idx: int = 0):
+        self.append_batch([(idx, token)])
+
+    def append_batch(self, items: List[Tuple[int, Any]]) -> None:
+        """Append multiple ``(idx, token)`` events under one lock/notify.
+
+        Batched counterpart to :meth:`append` for per-step delivery: state
+        updates for every event happen under a single condition hold and
+        waiters are woken once per batch instead of once per token.
+        """
+        if not items:
+            return
         with self._cond:
-            self.tokens.append((idx, token))
-            if token is not STOP:
-                self.results[idx] += token
-            else:
-                if not self._done[idx]:
-                    self._done[idx] = True
-                    self._completed += 1
-                    self._cond.notify_all()
+            for idx, token in items:
+                self.tokens.append((idx, token))
+                if token is STOP:
+                    if not self._done[idx]:
+                        self._done[idx] = True
+                        self._completed += 1
+                        self._cond.notify_all()
+                else:
+                    self.results[idx] += token
             self._event.set()
 
     def pop_all(self) -> List[Tuple[int, str]]:
@@ -69,6 +81,47 @@ class GenerateResult:
             return self.results.copy()
 
 
+class _ResultSink(BatchedStreamCallback):
+    """Batched stream channel from the scheduler into one GenerateResult.
+
+    Registered as the ``stream_callback`` for every task of a single
+    ``generate`` call, so the scheduler's one dispatch per decode step
+    maps to one ``append_batch`` (one lock, one waiter wake). A task can
+    start decoding the moment ``add_task`` returns — before the engine
+    learns its id — so events for ids not yet bound are buffered and
+    replayed on ``bind``.
+    """
+
+    def __init__(self, result: GenerateResult):
+        self._result = result
+        self._lock = threading.Lock()
+        self._index_of: Dict[str, int] = {}
+        self._pending: List[Tuple[str, Any]] = []
+
+    def bind(self, task_id: str, idx: int) -> None:
+        with self._lock:
+            self._index_of[task_id] = idx
+            replay = [(idx, token) for tid, token in self._pending if tid == task_id]
+            if replay:
+                self._pending = [
+                    (tid, token) for tid, token in self._pending if tid != task_id
+                ]
+        if replay:
+            self._result.append_batch(replay)
+
+    def __call__(self, events: List[Tuple[str, Any]]) -> None:
+        with self._lock:
+            items: List[Tuple[int, Any]] = []
+            for tid, token in events:
+                idx = self._index_of.get(tid)
+                if idx is None:
+                    self._pending.append((tid, token))
+                else:
+                    items.append((idx, token))
+        if items:
+            self._result.append_batch(items)
+
+
 class InferenceEngine:
     """Unified inference engine backed by continuous-batching scheduler."""
 
@@ -147,6 +200,7 @@ class InferenceEngine:
     ) -> AsyncGenerator[str, None]:
         request_backend = get_backend(use_default=False)
         result = GenerateResult()
+        sink = _ResultSink(result)
         task_id = self.scheduler.add_task(
             prompt=prompt,
             max_tokens=max_tokens,
@@ -156,8 +210,9 @@ class InferenceEngine:
             frequency_penalty=frequency_penalty,
             rep_window=rep_window,
             backend=request_backend,
-            stream_callback=result.append,
+            stream_callback=sink,
         )
+        sink.bind(task_id, 0)
 
         async def _agen():
             finished = False
@@ -191,8 +246,10 @@ class InferenceEngine:
         n = len(prompts)
         request_backend = get_backend(use_default=False)
         result = GenerateResult(count=n)
-        task_ids = [
-            self.scheduler.add_task(
+        sink = _ResultSink(result)
+        task_ids = []
+        for i, p in enumerate(prompts):
+            task_id = self.scheduler.add_task(
                 prompt=p,
                 max_tokens=max_tokens,
                 temperature=temperature,
@@ -201,10 +258,10 @@ class InferenceEngine:
                 frequency_penalty=frequency_penalty,
                 rep_window=rep_window,
                 backend=request_backend,
-                stream_callback=lambda token, idx=i: result.append(token, idx),
+                stream_callback=sink,
             )
-            for i, p in enumerate(prompts)
-        ]
+            sink.bind(task_id, i)
+            task_ids.append(task_id)
 
         if not stream:
             try:

astrai/inference/scheduler.py

@@ -264,17 +264,19 @@ class InferenceScheduler:
 
                     decoded, aborted = self._stepper.step(active)
 
-                    for t in aborted:
-                        self._task_mgr.invoke_callback(t.task_id, STOP)
-
+                    # One dispatch per step: batch-aware sinks take their
+                    # lock (and wake waiters) once instead of once per token.
+                    events: List[Tuple[str, Any]] = [(t.task_id, STOP) for t in aborted]
                     for t in decoded:
                         if t.status == TaskStatus.ABORTED:
                             continue
                         new_text = t.decode_new_token(self._task_mgr.tokenizer)
                         if new_text:
-                            self._task_mgr.invoke_callback(t.task_id, new_text)
+                            events.append((t.task_id, new_text))
                         if t.is_finished(stop_ids):
-                            self._task_mgr.invoke_callback(t.task_id, STOP)
+                            events.append((t.task_id, STOP))
+                    if events:
+                        self._task_mgr.invoke_callbacks(events)
 
         except Exception as e:
             self._stop_event.set()

astrai/inference/task.py

@@ -1,6 +1,7 @@
 import threading
 import time
 import uuid
+from abc import ABC, abstractmethod
 from collections import deque
 from dataclasses import dataclass
 from enum import Enum
@@ -145,6 +146,22 @@ class Task:
         return False
 
 
+class BatchedStreamCallback(ABC):
+    """Stream sink that receives a whole scheduler step's events in one call.
+
+    The scheduling loop dispatches once per decode step: every
+    ``(task_id, token)`` event routed to the same sink object is delivered
+    as a single list, so batch-aware consumers take their lock and wake
+    waiters once per step instead of once per token. Plain per-token
+    callbacks keep the ``Callable[[str], None]`` contract.
+    """
+
+    @abstractmethod
+    def __call__(self, events: List[Tuple[str, Any]]) -> None:
+        """Consume ``[(task_id, token), ...]`` produced by one decode step."""
+        raise NotImplementedError
+
+
 class TaskManager:
     """Thread-safe task queues and lifecycle transitions (no page ops)."""
 
@@ -256,7 +273,10 @@ class TaskManager:
                 immediate = [task]
 
         if cancelled and callback is not None:
-            callback(STOP)
+            if isinstance(callback, BatchedStreamCallback):
+                callback([(task_id, STOP)])
+            else:
+                callback(STOP)
         return immediate, cancelled
 
     def remove_task(self, task_id: str) -> List[Task]:
@@ -264,10 +284,40 @@ class TaskManager:
         immediate, _ = self.cancel_task(task_id)
         return immediate
 
-    def invoke_callback(self, task_id: str, token: str):
+    def invoke_callback(self, task_id: str, token: Any):
         with self._lock:
             cb = self._callbacks.get(task_id)
-        if cb:
+        if isinstance(cb, BatchedStreamCallback):
+            cb([(task_id, token)])
+        elif cb:
+            cb(token)
+
+    def invoke_callbacks(self, events: List[Tuple[str, Any]]) -> None:
+        """Dispatch one decode step's ``(task_id, token)`` events.
+
+        Callbacks resolve under a single lock acquisition; events aimed at
+        the same batched sink are delivered as one list (one consumer-side
+        lock/notify per step), while plain per-token callbacks receive one
+        call per event.
+        """
+        grouped: Dict[int, Tuple[BatchedStreamCallback, List[Any]]] = {}
+        plain: List[Tuple[Callable[[str], None], Any]] = []
+        with self._lock:
+            for task_id, token in events:
+                cb = self._callbacks.get(task_id)
+                if cb is None:
+                    continue
+                if isinstance(cb, BatchedStreamCallback):
+                    entry = grouped.get(id(cb))
+                    if entry is None:
+                        grouped[id(cb)] = (cb, [(task_id, token)])
+                    else:
+                        entry[1].append((task_id, token))
+                else:
+                    plain.append((cb, token))
+        for cb, batch in grouped.values():
+            cb(batch)
+        for cb, token in plain:
             cb(token)
 
     def get_stats(self) -> Dict[str, Any]:

tests/inference/test_engine.py

@@ -1,6 +1,7 @@
 """Unit tests for GenerateResult accumulator and InferenceEngine.generate()."""
 
 import asyncio
+import itertools
 import threading
 from unittest.mock import MagicMock, patch
 
@@ -8,7 +9,12 @@ import pytest
 
 from astrai.extension import TorchNativeBackend, attn_backend
 from astrai.inference import STOP
-from astrai.inference.engine import GenerateResult, InferenceEngine, build_engine
+from astrai.inference.engine import (
+    GenerateResult,
+    InferenceEngine,
+    _ResultSink,
+    build_engine,
+)
 from tests.helpers import FakeTokenizer, make_model
 
 
@@ -50,6 +56,33 @@ def test_result_stop_does_not_double_count():
     assert r._completed == 1
 
 
+def test_result_append_batch_updates_state_in_one_commit():
+    r = GenerateResult(count=2)
+    r.append_batch([(0, "he"), (1, "wo"), (0, "llo"), (1, "rld")])
+    r.append_batch([(0, STOP), (1, STOP)])
+    assert r.results == ["hello", "world"]
+    assert r._completed == 2
+    assert r.pop_all() == [
+        (0, "he"),
+        (1, "wo"),
+        (0, "llo"),
+        (1, "rld"),
+        (0, STOP),
+        (1, STOP),
+    ]
+
+
+def test_result_sink_replays_events_arriving_before_bind():
+    r = GenerateResult(count=1)
+    sink = _ResultSink(r)
+    sink([("t0", "he")])  # task id not bound yet: buffered, not applied
+    assert r.results == [""]
+    sink.bind("t0", 0)
+    sink([("t0", "llo"), ("t0", STOP)])
+    assert r.results == ["hello"]
+    assert r._completed == 1
+
+
 def test_result_pop_all_returns_and_clears():
     r = GenerateResult(count=2)
     r.append("a", 0)
@@ -118,8 +151,8 @@ def test_engine_generate_non_streaming_single():
 
         def fake_add(prompt, **kw):
             cb = kw["stream_callback"]
-            cb("response")
-            cb(STOP)
+            cb([("task-1", "response"), ("task-1", STOP)])
+            return "task-1"
 
         instance.add_task.side_effect = fake_add
         instance.remove_task.return_value = []
@@ -136,6 +169,7 @@ def test_engine_generate_streaming_yields_tokens():
 
     def capture_cb(prompt, **kw):
         callbacks_saved.append(kw.get("stream_callback"))
+        return "task-0"
 
     with patch("astrai.inference.engine.InferenceScheduler") as MockSched:
         instance = MockSched.return_value
@@ -146,9 +180,9 @@ def test_engine_generate_streaming_yields_tokens():
         gen = eng.generate("hello", stream=True)
 
         cb = callbacks_saved[0]
-        cb("t1")
-        cb("t2")
-        cb(STOP)
+        cb([("task-0", "t1")])
+        cb([("task-0", "t2")])
+        cb([("task-0", STOP)])
 
         tokens = list(gen)
         assert tokens == ["t1", "t2"]
@@ -169,7 +203,7 @@ def test_engine_stream_close_cancels_unfinished_task():
         engine = InferenceEngine(mock_model, mock_tokenizer, max_batch_size=1)
         stream = engine.generate("hello", stream=True)
 
-        callbacks_saved[0]("t1")
+        callbacks_saved[0]([("task-1", "t1")])
         assert next(stream) == "t1"
         stream.close()
 
@@ -182,6 +216,7 @@ def test_engine_generate_async_yields_tokens_until_stop():
 
     def capture_cb(prompt, **kw):
         callbacks_saved.append(kw.get("stream_callback"))
+        return "task-0"
 
     with patch("astrai.inference.engine.InferenceScheduler") as MockSched:
         instance = MockSched.return_value
@@ -198,9 +233,9 @@ def test_engine_generate_async_yields_tokens_until_stop():
             return out
 
         cb = callbacks_saved[0]
-        cb("t1")
-        cb("t2")
-        cb(STOP)
+        cb([("task-0", "t1")])
+        cb([("task-0", "t2")])
+        cb([("task-0", STOP)])
 
         assert asyncio.run(collect()) == ["t1", "t2"]
 
@@ -219,7 +254,7 @@ def test_engine_async_close_cancels_unfinished_task():
 
         engine = InferenceEngine(mock_model, mock_tokenizer, max_batch_size=1)
         stream = engine.generate_async("hello")
-        callbacks_saved[0]("t1")
+        callbacks_saved[0]([("task-1", "t1")])
 
         async def consume_then_close():
             assert await anext(stream) == "t1"
@@ -233,20 +268,25 @@ def test_engine_async_close_cancels_unfinished_task():
 def test_engine_generate_non_streaming_batch():
     mock_model, mock_tokenizer = _make_engine_mocks(decode="r")
 
+    counter = itertools.count()
+    task_ids = []
+
+    def fake_add(prompt, **kw):
+        cb = kw["stream_callback"]
+        tid = f"task-{next(counter)}"
+        task_ids.append(tid)
+        cb([(tid, "r"), (tid, STOP)])
+        return tid
+
     with patch("astrai.inference.engine.InferenceScheduler") as MockSched:
         instance = MockSched.return_value
-
-        def fake_add(prompt, **kw):
-            cb = kw["stream_callback"]
-            cb("r")
-            cb(STOP)
-
         instance.add_task.side_effect = fake_add
         instance.remove_task.return_value = []
 
         eng = InferenceEngine(mock_model, mock_tokenizer, max_batch_size=2)
         results = eng.generate(["hello", "world"])
         assert results == ["r", "r"]
+        assert task_ids == ["task-0", "task-1"]
 
 
 def test_engine_generate_zero_max_tokens_returns_empty():
@@ -294,7 +334,7 @@ def test_generate_captures_calling_backend_context():
 
         def fake_add(prompt, **kwargs):
             captured.append(kwargs["backend"])
-            kwargs["stream_callback"](STOP)
+            kwargs["stream_callback"]([("task", STOP)])
             return "task"
 
         instance.add_task.side_effect = fake_add
@@ -328,9 +368,12 @@ def test_build_engine_passes_engine_kwargs_through():
     model, _ = make_model("cpu", max_position_embeddings=64)
     backend = TorchNativeBackend()
     with patch("astrai.inference.engine.InferenceScheduler") as MockSched:
-        MockSched.return_value.add_task.side_effect = lambda *args, **k: (
-            k["stream_callback"](STOP) or "task"
-        )
+
+        def fake_add(*args, **k):
+            k["stream_callback"]([("task", STOP)])
+            return "task"
+
+        MockSched.return_value.add_task.side_effect = fake_add
         engine = build_engine(
             model=model,
             tokenizer=FakeTokenizer(),

tests/inference/test_task.py

@@ -4,7 +4,23 @@ from unittest.mock import MagicMock
 
 import pytest
 
-from astrai.inference import STOP, Task, TaskManager, TaskStatus
+from astrai.inference import (
+    STOP,
+    BatchedStreamCallback,
+    Task,
+    TaskManager,
+    TaskStatus,
+)
+
+
+class RecordingSink(BatchedStreamCallback):
+    """Batch-aware callback capturing every dispatch as one batch."""
+
+    def __init__(self):
+        self.batches = []
+
+    def __call__(self, events):
+        self.batches.append(events)
 
 
 def _make_mock_tokenizer():
@@ -217,3 +233,46 @@ def test_task_manager_cancel_active_task_delivers_stop_callback():
     immediate, cancelled = tm.cancel_task(task_id)
     assert cancelled and immediate == []
     assert received == [STOP]
+
+
+def test_invoke_callbacks_batches_sink_events_and_keeps_plain_per_token():
+    tm = TaskManager(tokenizer=_make_mock_tokenizer())
+    plain = []
+    tid_plain = tm.add_task("plain", stream_callback=plain.append)
+    sink = RecordingSink()
+    tid_a = tm.add_task("sink a", stream_callback=sink)
+    tid_b = tm.add_task("sink b", stream_callback=sink)
+
+    tm.invoke_callbacks(
+        [
+            (tid_a, "x"),
+            (tid_plain, "p"),
+            (tid_b, "y"),
+            ("unknown-task", "dropped"),
+            (tid_a, STOP),
+        ]
+    )
+
+    assert plain == ["p"]
+    assert sink.batches == [[(tid_a, "x"), (tid_b, "y"), (tid_a, STOP)]]
+
+
+def test_invoke_callback_delivers_single_event_to_batched_sink():
+    tm = TaskManager(tokenizer=_make_mock_tokenizer())
+    sink = RecordingSink()
+    task_id = tm.add_task("test", stream_callback=sink)
+
+    tm.invoke_callback(task_id, STOP)
+
+    assert sink.batches == [[(task_id, STOP)]]
+
+
+def test_cancel_delivers_batched_stop_to_sink():
+    tm = TaskManager(tokenizer=_make_mock_tokenizer())
+    sink = RecordingSink()
+    task_id = tm.add_task("test", stream_callback=sink)
+
+    immediate, cancelled = tm.cancel_task(task_id)
+
+    assert cancelled
+    assert sink.batches == [[(task_id, STOP)]]