diff --git a/scripts/minimax_h3/run_minimax_h3_fl2av.sh b/scripts/minimax_h3/run_minimax_h3_fl2av.sh index eeaca59e0..2f4025978 100755 --- a/scripts/minimax_h3/run_minimax_h3_fl2av.sh +++ b/scripts/minimax_h3/run_minimax_h3_fl2av.sh @@ -9,7 +9,7 @@ export CUDA_VISIBLE_DEVICES=0 # set environment variables source ${lightx2v_path}/scripts/base/base.sh export DTYPE=BF16 -export SENSITIVE_LAYER_DTYPE=FP32 +export SENSITIVE_LAYER_DTYPE=BF16 python -m lightx2v.infer \ --model_cls minimax_h3 \ diff --git a/scripts/minimax_h3/run_minimax_h3_i2av.sh b/scripts/minimax_h3/run_minimax_h3_i2av.sh index b4edd1c40..2489f163e 100755 --- a/scripts/minimax_h3/run_minimax_h3_i2av.sh +++ b/scripts/minimax_h3/run_minimax_h3_i2av.sh @@ -9,7 +9,7 @@ export CUDA_VISIBLE_DEVICES=0 # set environment variables source ${lightx2v_path}/scripts/base/base.sh export DTYPE=BF16 -export SENSITIVE_LAYER_DTYPE=FP32 +export SENSITIVE_LAYER_DTYPE=BF16 python -m lightx2v.infer \ --model_cls minimax_h3 \ diff --git a/scripts/minimax_h3/run_minimax_h3_l2av.sh b/scripts/minimax_h3/run_minimax_h3_l2av.sh index 37d72f246..ba9df6a3d 100755 --- a/scripts/minimax_h3/run_minimax_h3_l2av.sh +++ b/scripts/minimax_h3/run_minimax_h3_l2av.sh @@ -9,7 +9,7 @@ export CUDA_VISIBLE_DEVICES=0 # set environment variables source ${lightx2v_path}/scripts/base/base.sh export DTYPE=BF16 -export SENSITIVE_LAYER_DTYPE=FP32 +export SENSITIVE_LAYER_DTYPE=BF16 python -m lightx2v.infer \ --model_cls minimax_h3 \ diff --git a/scripts/minimax_h3/run_minimax_h3_ref2av.sh b/scripts/minimax_h3/run_minimax_h3_ref2av.sh index 474cdaeca..0bc111c51 100755 --- a/scripts/minimax_h3/run_minimax_h3_ref2av.sh +++ b/scripts/minimax_h3/run_minimax_h3_ref2av.sh @@ -9,7 +9,7 @@ export CUDA_VISIBLE_DEVICES=0 # set environment variables source ${lightx2v_path}/scripts/base/base.sh export DTYPE=BF16 -export SENSITIVE_LAYER_DTYPE=FP32 +export SENSITIVE_LAYER_DTYPE=BF16 python -m lightx2v.infer \ --model_cls minimax_h3 \ diff --git a/scripts/minimax_h3/run_minimax_h3_t2av.sh b/scripts/minimax_h3/run_minimax_h3_t2av.sh index 9f4ce9904..c00b51901 100755 --- a/scripts/minimax_h3/run_minimax_h3_t2av.sh +++ b/scripts/minimax_h3/run_minimax_h3_t2av.sh @@ -9,7 +9,7 @@ export CUDA_VISIBLE_DEVICES=0 # set environment variables source ${lightx2v_path}/scripts/base/base.sh export DTYPE=BF16 -export SENSITIVE_LAYER_DTYPE=FP32 +export SENSITIVE_LAYER_DTYPE=BF16 prompt='integrated_multimodal_description: [Shot 1] Cinematic low-angle tracking shot following a stylish woman from behind as she strolls down a bustling post-rain Tokyo street. The asphalt is completely wet, acting as a black mirror that perfectly reflects the dense canopy of overhead neon signs—warm pink lanterns, icy cyan katakana signage, and giant animated billboards playing silently. She walks slightly to the left of frame, revealing the back of her sleek black leather jacket, which glistens with specular highlights, and the hem of a flowing crimson dress that swirls around her calves. Black heeled boots splash subtly in shallow puddles, and a black leather purse hangs from her shoulder. The camera slowly pushes forward and gently rises, while out-of-focus pedestrians in modern clothing cross the frame, adding life. [Shot 2] At 00:03.500, a sharp cut to a medium profile shot from her right side, camera dollying sideways in perfect sync. She comes into clear view: oversized black sunglasses perched on her nose reflect a giant LED screen across the street, with purple and blue animations gliding across the lenses. Her bold matte red lipstick stands out against fair skin, and a hint of a confident smile plays on her lips. The sharp tailoring of her jacket catches rim light, and her stride is poised and rhythmic. The background is a bokeh of neon blur, while the wet ground distorts the red dress’s reflection into abstract color streaks. A subtle handheld camera shake increases immediacy. [Shot 3] At 00:06.800, a stylized slow-motion frontal medium close-up as she walks directly toward the lens, which pulls back. Time stretches—she casually removes her sunglasses in one smooth motion, revealing sharp winged eyeliner and a piercing gaze that locks directly with the viewer. A shaft of hot pink neon light sweeps across her cheekbones, then she slides the glasses back on with a soft click. The camera then racks focus from her face to the endless corridor of neon-lit street behind her as she walks past, dissolving into a blur of vibrant city lights. overall_soundscape: Rich city atmosphere on wet streets: a constant damp hiss of car tires rolling through water in the distance, the resonant electrical hum and faint crackle of neon transformers overhead, a muffled J-pop bassline leaking from a nearby record store, layers of pedestrian chatter and soft laughter in Japanese, and in the foreground, the crisp, wet footsteps of her heeled boots striking the mirrored asphalt, with occasional tiny splashes. When she removes her sunglasses, a delicate, intimate "click" of the frame folding is audible, momentarily cutting through the noise. diff --git a/scripts/minimax_h3/run_minimax_h3_t2av_sp.sh b/scripts/minimax_h3/run_minimax_h3_t2av_sp.sh index 8fffabb2f..8de457ca5 100755 --- a/scripts/minimax_h3/run_minimax_h3_t2av_sp.sh +++ b/scripts/minimax_h3/run_minimax_h3_t2av_sp.sh @@ -9,7 +9,7 @@ export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 # set environment variables source ${lightx2v_path}/scripts/base/base.sh export DTYPE=BF16 -export SENSITIVE_LAYER_DTYPE=FP32 +export SENSITIVE_LAYER_DTYPE=BF16 torchrun --standalone --nproc_per_node=8 -m lightx2v.infer \ --model_cls minimax_h3 \ diff --git a/scripts/minimax_h3/run_minimax_h3_t2av_tp.sh b/scripts/minimax_h3/run_minimax_h3_t2av_tp.sh index 5313b011d..021e6b480 100755 --- a/scripts/minimax_h3/run_minimax_h3_t2av_tp.sh +++ b/scripts/minimax_h3/run_minimax_h3_t2av_tp.sh @@ -7,7 +7,7 @@ export CUDA_VISIBLE_DEVICES=0,1 source ${lightx2v_path}/scripts/base/base.sh export DTYPE=BF16 -export SENSITIVE_LAYER_DTYPE=FP32 +export SENSITIVE_LAYER_DTYPE=BF16 torchrun --standalone --nproc_per_node=2 -m lightx2v.infer \ --model_cls minimax_h3 \ diff --git a/scripts/minimax_h3/run_minimax_h3_t2av_tp_sp.sh b/scripts/minimax_h3/run_minimax_h3_t2av_tp_sp.sh index 41bfe9c9b..0bcdb7acc 100755 --- a/scripts/minimax_h3/run_minimax_h3_t2av_tp_sp.sh +++ b/scripts/minimax_h3/run_minimax_h3_t2av_tp_sp.sh @@ -7,7 +7,7 @@ export CUDA_VISIBLE_DEVICES=0,1,2,3 source ${lightx2v_path}/scripts/base/base.sh export DTYPE=BF16 -export SENSITIVE_LAYER_DTYPE=FP32 +export SENSITIVE_LAYER_DTYPE=BF16 torchrun --standalone --nproc_per_node=4 -m lightx2v.infer \ --model_cls minimax_h3 \