我单卡推理生视频要50分钟，多卡推理时候，报错 flash-attn和diffusers 版本不兼容， flash-attn版本是2.6.3，diffusers 版本是0.36.0。我现在需要把flash-attn版本进行降级，但是我找不到对应python=3.13版本的flash-atten #22
Description
export CP_SIZE=2
export CUDA_VISIBLE_DEVICES=0,1
export CKPT_PATH=/data3/MOVA/MOVA-720p
torchrun
--nproc_per_node=$CP_SIZE
scripts/inference_single.py
--ckpt_path $CKPT_PATH
--cp_size $CP_SIZE
--height 1280
--width 720
--num_inference_steps 20
--cfg_scale 5.0 --sigma_shift 5.0
--prompt "深夜，一条昏暗漫长的医院走廊，荧光灯忽明忽暗地闪烁着，营造出压抑恐怖的氛围。冷蓝绿色调、深邃的阴影和微妙的体积雾充斥着整个空间。
画面中，左侧是女医生阿伦（手中拿着病历夹），右侧是女护士艾玛（头戴护士帽，身穿蓝色手术服）。两人面对面站立，全程保持着紧张的眼神交流。镜头在说话者面部的紧凑特写（聚焦表情和嘴唇动作）与展示两人全身的中景之间交替切换。
艾玛（右侧的护士）微微仰起头，眼神闪烁不定，紧张不安地看着对方，用颤抖的声音说话："你终于来了……"
阿伦（左侧的医生）立即回应："为什么这么说？"
走廊的灯光继续微妙地闪烁。
艾玛（右侧的护士）低下头整理手中的病历，手指微微颤抖，用轻柔但不安的声音说："今晚的夜班将会非常漫长。"
阿伦（左侧的医生）询问道，完美的唇形同步，自然的口齿发音："医院以前发生过什么事？"
艾玛（右侧的护士）突然颤抖起来，紧张地环顾四周，用低沉而急切的声音说："有些事情不该问。"
阿伦（左侧的医生）回应道："好吧，那我们去工作吧。"
艾玛（右侧的护士）露出如释重负的表情，用略微放松但仍带恐惧的语气说："跟我来，先查一下病房。"
走廊尽头传来缓慢的手推车滚动声。
艾玛（右侧的护士）快步走向走廊尽头，边走边转身，紧张地说："304病房有异常。"
贯穿整个场景：低沉微妙的恐怖背景音乐，远处回荡的脚步声和医疗器械的碰撞声，人物表情充满恐惧与不安，眼睛频繁瞥向黑暗的角落。电影级恐怖风格，真实的光影效果，细腻的面部表情，高质量的细节，超真实的皮肤纹理和动态。没有任何文字，没有字幕，没有标题，没有叠加的文字、字母、标志、Logo或任何形式的屏幕文字。"
--negative_prompt "subtitle, caption, text, watermark, logo, timestamp, OSD, UI, scoreboard, channel icon, credits, opening titles, closing titles, lower third, on-screen text, any text, blurry white strip, static overlay, silent movie text cards, subtitle bar, karaoke lyrics, forced narrative, forced subtitles, burned-in text, hardsubs, softsubs, text box, speech bubble, scrolling text, ticker, copyright notice, disclaimer text, rating badge, TV-MA bug, Dolby logo, aspect ratio tag, metadata overlay, digital on-screen graphics, unexpected characters, glyphs, letters, numbers, symbols, three hands, extra hand, third hand, 3 hands, multiple hands, identity inconsistency, frame-by-frame facial drift, static image, blurry, out of focus, overexposed, underexposed, low contrast, washed out colors, excessive noise, grainy texture, poor lighting, flickering, motion blur, distorted proportions, unnatural skin tones, deformed facial features, asymmetrical face, missing facial features, extra limbs, disfigured hands, wrong hand count, artifacts around text, inconsistent perspective, camera shake, incorrect depth of field, background too sharp, background clutter, distracting reflections, harsh shadows, inconsistent lighting direction, color banding, cartoonish rendering, 3D CGI look, unrealistic materials, uncanny valley effect, incorrect ethnicity, wrong gender, exaggerated expressions, wrong gaze direction, mismatched lip sync, silent or muted audio, distorted voice, robotic voice, echo, background noise, off-sync audio, incorrect dialogue, added dialogue, repetitive speech, jittery movement, awkward pauses, incorrect timing, unnatural transitions, inconsistent framing, tilted camera, flat lighting, inconsistent tone, cinematic oversaturation, stylized filters, AI artifacts"
--ref_path "/data3/LTX-2/inputs/7.png"
--output_path "/data3/MOVA/MOVA-720p/outputs/1.mp4"
--seed 42
--offload cpu