A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 10, 17:45 UTC
[Docs] Improve malformed exception caused by backslash line continuations (#31694)
Signed-off-by: maang <maang_h@163.com> Signed-off-by: maang <55082429+maang-h@users.noreply.github.com> Co-authored-by: Cyrus Leung <cyrus.tl.leung@gmail.com> Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
maang committed
Jan 6, 2026 at 09:51 UTC
d386ab141273041693455d33651a3cba6a12e8d1
7 files changed
+14
-17
vllm/config/compilation.py
+3
-3
@@ -839,9 +839,9 @@ class CompilationConfig:
839
"""
840
if self.mode is None:
841
raise ValueError(
842
- "No compilation mode is set. This method should only be \
843
- called via vllm config where the level is set if none is \
844
- provided."
842
+ "No compilation mode is set. This method should only be "
843
+ "called via vllm config where the level is set if none is "
844
+ "provided."
845
)
846
if self.mode == CompilationMode.NONE:
847
raise ValueError("No compilation mode is set.")
vllm/model_executor/layers/fused_moe/rocm_aiter_fused_moe.py
+1
-2
@@ -231,8 +231,7 @@ def rocm_aiter_fused_experts(
231
# w8a8 block-scaled
232
if quant_config.block_shape is not None and quant_config.use_fp8_w8a8:
233
assert not apply_router_weight_on_input, (
234
- "apply_router_weight_on_input is\
235
- not supported for block scaled moe"
234
+ "apply_router_weight_on_input is not supported for block scaled moe"
235
)
236
assert quant_config.w1_scale is not None
237
assert quant_config.w2_scale is not None
vllm/model_executor/model_loader/weight_utils.py
+2
-2
@@ -681,8 +681,8 @@ def safetensors_weights_iterator(
681
# instead we reconstruct the subclasses here before returning
682
if not torchao_version_at_least("0.15.0"):
683
raise ValueError(
684
- "Please use torchao version >= 0.15.0 \
685
- to load torchao safetensors checkpoint"
684
+ "Please use torchao version >= 0.15.0 "
685
+ "to load torchao safetensors checkpoint"
686
)
687
from torchao.prototype.safetensors.safetensors_support import (
688
unflatten_tensor_state_dict,
vllm/model_executor/models/llama.py
+2
-2
@@ -201,8 +201,8 @@ class LlamaAttention(nn.Module):
201
# This is a target model, use layer_idx directly
202
effective_layer_idx = layer_idx
203
assert effective_layer_idx < len(layer_types), (
204
- f"effective_layer_idx: {effective_layer_idx} \
205
- is out of bounds for layer_types: {layer_types}"
204
+ f"effective_layer_idx: {effective_layer_idx} "
205
+ f"is out of bounds for layer_types: {layer_types}"
206
)
207
208
is_sliding = layer_types[effective_layer_idx] == "sliding_attention"
vllm/model_executor/models/phi4mm.py
+3
-4
@@ -428,14 +428,13 @@ class Phi4MMImageEncoder(nn.Module):
428
output_imgs.append(torch.cat([sub_img, self.glb_GN, glb_img], dim=1))
429
else:
430
raise NotImplementedError(
431
- f'hd_transform_order = {self.hd_transform_order}, "\
432
- "not implemented'
431
+ f"hd_transform_order = {self.hd_transform_order}, not implemented"
432
)
433
434
# temp_len = int((h*w+1)*144 + 1 + (h+1)*12)
435
assert temp_len == output_imgs[-1].shape[1], (
437
- f'temp_len: {temp_len}, output_imgs[-1].shape[1]: "\
438
- "{output_imgs[-1].shape[1]}'
436
+ f"temp_len: {temp_len}, output_imgs[-1].shape[1]: "
437
+ f"{output_imgs[-1].shape[1]}"
438
)
439
440
output_len.append(temp_len)
vllm/multimodal/registry.py
+2
-2
@@ -367,8 +367,8 @@ class MultiModalRegistry:
367
# than whisper.
368
return 0
369
assert len(max_tokens) == 1, (
370
- "Encoder-decoder models are expected \
371
- to implement the multimodal interface with at most one modality."
370
+ "Encoder-decoder models are expected "
371
+ "to implement the multimodal interface with at most one modality."
372
)
373
374
first_modality = next(iter(max_tokens))
vllm/utils/argparse_utils.py
+1
-2
@@ -399,8 +399,7 @@ class FlexibleArgumentParser(ArgumentParser):
399
index = args.index("--config")
400
if index == len(args) - 1:
401
raise ValueError(
402
- "No config file specified! \
403
- Please check your command-line arguments."
402
+ "No config file specified! Please check your command-line arguments."
403
)
404
405
file_path = args[index + 1]