CLip L, Clip H, textual inversion

[ObsidianStack]

Guys, I'm struggling in how to load a custom negatives embed... but I don't have a single clue in how to load the .safetensor file since the normal textual inversion loading doesn't work. this is my code.

        images = pipe(
            prompt=positive_prompt,
            negative_prompt=negative_prompt,
            num_inference_steps=num_inference_steps,
            guidance_scale=guidance_scale,
            num_images_per_prompt=num_images,
            generator=generator,
            denoising_end=1.0,
            height=batch_singleimg_height,
            width=batch_singleimg_width,
            custom_pipeline   = "lpw_stable_diffusion_xl"
        ).images

should I add something? import something? also i don't have clip_g , clip_l and when I try to load the .safetensor file it returns the clip_g\l missing error.