Remove LingBot-specific attention env

Signed-off-by: hsliuustc0106 <liuhongsheng4@huawei.com>

recipes/Robbyant/LingBot-Video.md

@@ -45,7 +45,7 @@ resolution/frame count if your target workload runs out of memory.
 #### Offline T2V
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 LINGBOT_QWEN_ATTN_IMPLEMENTATION=sdpa \
+CUDA_VISIBLE_DEVICES=0 \
 python examples/offline_inference/text_to_video/text_to_video_lingbot.py \
   --model robbyant/lingbot-video-dense-1.3b \
   --prompt "a robotic arm picks up a red block" \
@@ -63,7 +63,7 @@ python examples/offline_inference/text_to_video/text_to_video_lingbot.py \
 #### Online Serving
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 LINGBOT_QWEN_ATTN_IMPLEMENTATION=sdpa \
+CUDA_VISIBLE_DEVICES=0 \
 vllm serve robbyant/lingbot-video-dense-1.3b \
   --omni \
   --model-class-name LingBotVideoPipeline \

vllm_omni/diffusion/models/lingbot_video/pipeline_lingbot_video.py

@@ -269,16 +269,10 @@ class LingBotVideoPipeline(nn.Module, ProgressBarMixin, SupportsComponentDiscove
             torch_dtype=transformer_dtype,
             local_files_only=local_files_only,
         ).to(self.device)
-        attn_implementation = os.environ.get(
-            "LINGBOT_QWEN_ATTN_IMPLEMENTATION",
-            str(model_config.get("qwen_attn_implementation", "sdpa")),
-        )
         text_encoder_kwargs: dict[str, Any] = {
             "dtype": text_encoder_dtype,
             "local_files_only": local_files_only,
         }
-        if attn_implementation:
-            text_encoder_kwargs["attn_implementation"] = attn_implementation
         self.text_encoder = Qwen3VLForConditionalGeneration.from_pretrained(
             model,
             subfolder=text_encoder_subfolder,