update 8 gpus infer config and scripts (#281)

configs/seko_talk/seko_talk_12_fp8_dist_fixed_shape_8gpus_1s.json

+{
+    "infer_steps": 4,
+    "target_fps": 16,
+    "video_duration": 12,
+    "audio_sr": 16000,
+    "target_video_length": 17,
+    "prev_frame_length": 1,
+    "resize_mode": "fixed_shape",
+    "fixed_shape": [480, 480],
+    "self_attn_1_type": "sage_attn2",
+    "cross_attn_1_type": "sage_attn2",
+    "cross_attn_2_type": "sage_attn2",
+    "seed": 42,
+    "sample_guide_scale": 1.0,
+    "sample_shift": 5,
+    "enable_cfg": false,
+    "cpu_offload": false,
+    "use_31_block": false,
+    "parallel": {
+        "seq_p_size": 8,
+        "seq_p_attn_type": "ulysses"
+    },
+    "mm_config": {
+        "mm_type": "W-fp8-channel-sym-A-fp8-channel-sym-dynamic-Vllm"
+    },
+    "adapter_quantized": true,
+    "adapter_quant_scheme": "fp8",
+    "t5_quantized": true,
+    "t5_quant_scheme": "fp8"
+}

scripts/seko_talk/run_seko_talk_12_fp8_dist_fixed_shape_8gpus_1s.sh

+#!/bin/bash
+
+lightx2v_path=
+model_path=
+
+export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
+
+# set environment variables
+source ${lightx2v_path}/scripts/base/base.sh
+
+export PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True
+
+export ENABLE_GRAPH_MODE=false
+export SENSITIVE_LAYER_DTYPE=None
+
+torchrun --nproc-per-node 8 -m lightx2v.infer \
+--model_cls seko_talk \
+--task i2v \
+--model_path $model_path \
+--config_json ${lightx2v_path}/configs/seko_talk/seko_talk_12_fp8_dist_fixed_shape_8gpus_1s.json \
+--prompt  "The video features a old lady is saying something and knitting a sweater." \
+--negative_prompt 色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走 \
+--image_path ${lightx2v_path}/assets/inputs/audio/15.png \
+--audio_path ${lightx2v_path}/assets/inputs/audio/15.wav \
+--save_video_path ${lightx2v_path}/save_results/output_lightx2v_seko_talk.mp4