Skip to content

Run frontier MoE models on hardware you already own — pure C, zero deps, experts streamed from disk. Tiny engine, immense model. 🐦

C

fix(test): keep the generated payload as bytes, never as str

JustVugg9278a215e277
diff --git a/c/tests/test_inkling_prefix_serve.py b/c/tests/test_inkling_prefix_serve.py
index c4d921ab..17e29eca 100644
--- a/c/tests/test_inkling_prefix_serve.py
+++ b/c/tests/test_inkling_prefix_serve.py
@@ -90,23 +90,31 @@ class Engine:
raise RuntimeError("engine never reported READY")
def ask(self, rid, prompt):
- payload = prompt.encode()
- header = f"SUBMIT {rid} 0 {len(payload)} {MAXTOK} 0 1\n".encode()
- self.p.stdin.write(header + payload + b"\n")
+ """Returns the generated payload as BYTES.
+
+ Never as str. A random-init model emits arbitrary byte sequences that
+ are mostly not valid UTF-8, so decoding with errors="replace" turns them
+ into U+FFFD and re-encoding yields entirely different bytes. Feeding
+ that back as the next turn's prefix made the prompt diverge from the
+ state the engine actually held -- a fault in the harness that looked
+ exactly like a fault in the engine.
+ """
+ assert isinstance(prompt, bytes), "prompts stay bytes end to end"
+ header = f"SUBMIT {rid} 0 {len(prompt)} {MAXTOK} 0 1\n".encode()
+ self.p.stdin.write(header + prompt + b"\n")
self.p.stdin.flush()
chunks = []
while True:
line = self.p.stdout.readline()
if not line:
raise RuntimeError("engine closed mid-request")
- text = line.decode(errors="replace").rstrip("\n")
+ text = line.decode("latin-1").rstrip("\n")
kind = text.split(" ", 1)[0]
if kind == "DATA":
- chunks.append(self.p.stdout.read(int(text.split()[2]))
- .decode(errors="replace"))
+ chunks.append(self.p.stdout.read(int(text.split()[2])))
self.p.stdout.readline() # the newline after the payload
elif kind in ("DONE", "END"):
- return "".join(chunks)
+ return b"".join(chunks)
elif kind == "ERROR":
raise RuntimeError(f"engine error: {text}")
@@ -145,10 +153,11 @@ class InklingPrefixServeTest(unittest.TestCase):
def test_reused_prefix_yields_identical_tokens(self):
warm = Engine()
- first = warm.ask("1", "The capital of France is")
+ opening = b"The capital of France is"
+ first = warm.ask("1", opening)
# Turn 2 EXTENDS what the state already holds — the shape a chat client
# produces when it resends the transcript with a new question appended.
- second_prompt = "The capital of France is" + first + " and the capital of Spain is"
+ second_prompt = opening + first + b" and the capital of Spain is"
reused = warm.ask("2", second_prompt)
log = warm.close()
@@ -168,12 +177,12 @@ class InklingPrefixServeTest(unittest.TestCase):
"""The rejection path matters as much as the reuse path: a prompt that
shares no prefix must start over, not splice itself onto stale state."""
eng = Engine()
- eng.ask("1", "The capital of France is")
- diverged = eng.ask("2", "Completely different opening text here")
+ eng.ask("1", b"The capital of France is")
+ diverged = eng.ask("2", b"Completely different opening text here")
log = eng.close()
cold = Engine(log_prefix=False)
- fresh = cold.ask("1", "Completely different opening text here")
+ fresh = cold.ask("1", b"Completely different opening text here")
cold.close()
self.assertEqual(diverged, fresh,