[BugFix] fix some typos found by typos. (#16314)
Signed-off-by: yihong0618 <zouzou0208@gmail.com>
This commit is contained in:
@@ -83,8 +83,8 @@ spda_o = scaled_dot_product_attention(
|
||||
return spda_o @ W_O
|
||||
|
||||
NOTE: in the actual code,
|
||||
`kv_b_proj` is [W_UK; W_UV] concatnated per head
|
||||
`q_b_proj` is [W_UQ; W_QR] concatnated per head
|
||||
`kv_b_proj` is [W_UK; W_UV] concatenated per head
|
||||
`q_b_proj` is [W_UQ; W_QR] concatenated per head
|
||||
`out_proj` is W_O
|
||||
|
||||
|
||||
|
||||
@@ -326,7 +326,7 @@ class WorkerProc:
|
||||
logger.debug("Worker interrupted.")
|
||||
|
||||
except Exception:
|
||||
# worker_busy_loop sends exceptions exceptons to Executor
|
||||
# worker_busy_loop sends exceptions to Executor
|
||||
# for shutdown, but if there is an error in startup or an
|
||||
# error with IPC itself, we need to alert the parent.
|
||||
psutil.Process().parent().send_signal(signal.SIGUSR1)
|
||||
|
||||
@@ -998,7 +998,7 @@ class GPUModelRunner(LoRAModelRunnerMixin):
|
||||
) -> Union[ModelRunnerOutput, torch.Tensor]:
|
||||
self._update_states(scheduler_output)
|
||||
if not scheduler_output.total_num_scheduled_tokens:
|
||||
# Return empty ModelRunnerOuptut if there's no work to do.
|
||||
# Return empty ModelRunnerOutput if there's no work to do.
|
||||
return EMPTY_MODEL_RUNNER_OUTPUT
|
||||
|
||||
if self.is_multimodal_model:
|
||||
|
||||
@@ -652,7 +652,7 @@ class TPUModelRunner:
|
||||
# Update cached state
|
||||
self._update_states(scheduler_output)
|
||||
if not scheduler_output.total_num_scheduled_tokens:
|
||||
# Return empty ModelRunnerOuptut if there's no work to do.
|
||||
# Return empty ModelRunnerOutput if there's no work to do.
|
||||
return EMPTY_MODEL_RUNNER_OUTPUT
|
||||
|
||||
if self.is_multimodal_model:
|
||||
|
||||
Reference in New Issue
Block a user