大模型训练 个人定制LORA
等服务内容
24K 订阅
个人定制LoRA（低秩适配）服务中，用户可通过高效微调技术实现大模型的个性化定制，以下从服务流程、技术优势、应用场景、核心工具.
导出模型时选择通用格式（如Hugging Face模型格式），便于长期保存和跨平台迁移.
导出模型时选择通用格式（如Hugging Face模型格式），便于长期保存和跨平台迁移.
大模型训练 个人定制LORA
EasyWan22 FastMix
Wan 2.2 I2V-A14B, 6steps/generation. apache-2.0.使用 Q4_K_M 与 Geforce RTX 3060 12GB 配合使用。有关不同量化的参数,请参阅 HuggingFace。
EasyWan22 FastMix (Wan 2.2 I2V-A14B)
Wan 2.2 I2V-A14B, 6steps/generation. apache-2.0.
使用 Q4_K_M 与 Geforce RTX 3060 12GB 配合使用。
有关不同量化的参数,请参阅 HuggingFace。
万22-I2V-FastMix_v10-H
- wan2.2_i2v_high_noise_14B_fp16.safetensors
- Wan2.2-Lightning/Wan2.2-I2V-A14B-4steps-lora-rank64-Seko-V1/high_noise_model.safetenss
HighNoise
- Sampler
- scheduler:
lcm - shift:
5 - cfg:
1 - sigmas:
1, 0.94, 0.85, 0.73, 0.55, 0.28, 0 - steps:
6, start_step:0, end_step:3
- scheduler:
- Enhance-A-Video
- weight:
1
- weight:
- Fresca
- use_fresca:
true
- use_fresca:
- NAG
- Kijai default
- Additional Negative Prompt:
mouth moving, 嘴在动, talking, 说话, speaking, 讲话,
- EnhanceMotion: <lora:Wan2.2-Lightning/Wan2.2-I2V-A14B-4steps-lora-rank64-Seko-V1/high_noise_model.safetensors: -1 ~ 1>
Wan22-I2V-FastMix_v10-L
- wan2.2_i2v_low_noise_14B_fp16.safetensors
- lightx2v_14B_T2V_cfg_step_distill_lora_adaptive_rank_quantile_0.15_bf16.safetensors
- lightx2v_I2V_14B_480p_cfg_step_distill_rank256_bf16.safetensors
LowNoise
Generation
- Sampler
- scheduler:
lcm - shift:
5 - cfg:
1 - sigmas:
1, 0.94, 0.85, 0.73, 0.55, 0.28, 0 - steps:
6, start_step:3
- scheduler:
- Enhance-A-Video
- weight:
1
- weight:
- NAG
- Kijai default
- Additional Negative Prompt:
mouth moving, 嘴在动, talking, 说话, speaking, 讲话,
Refiner
SeedGacha SSR Video > Upscale x1.5 & Encode samples
- Sampler
- scheduler:
lcm - shift:
5 - cfg:
1 - sigmas:
1.0, 0.97, 0.94, 0.90, 0.85, 0.795, 0.73, 0.65, 0.55, 0.42, 0.28, 0.14, 0.0 - steps:
12, start_step:10 - add_noise_to_samples:
true
- scheduler:
- Enhance-A-Video
- weight:
1
- weight:
- NAG
- Kijai default
- Additional Negative Prompt:
mouth moving, 嘴在动, talking, 说话, speaking, 讲话,
