| import os |
| import gc |
| import torch |
| from torch import Generator |
| from PIL.Image import Image |
| from diffusers import AutoencoderKL, FluxPipeline |
| from diffusers.image_processor import VaeImageProcessor |
| from pipelines.models import TextToImageRequest |
| from transformers import T5EncoderModel |
| os.environ["PYTORCH_CUDA_ALLOC_CONF"] = "expandable_segments:False,garbage_collection_threshold:0.001" |
| torch.set_float32_matmul_precision("medium") |
| os.environ["TOKENIZERS_PARALLELISM"] = "True" |
| ckpt_id = "black-forest-labs/FLUX.1-schnell" |
| dtype = torch.bfloat16 |
| Pipeline = None |
| |
| torch.backends.cudnn.benchmark = True |
| torch.backends.cuda.matmul.allow_tf32 = True |
| torch.cuda.set_per_process_memory_fraction(0.999) |
|
|
| class BasicQuantization: |
| def __init__(self, bits=1): |
| self.bits = bits |
| self.qmin = -(2**(bits-1)) |
| self.qmax = 2**(bits-1) - 1 |
|
|
| def quantize_tensor(self, tensor): |
| scale = (tensor.max() - tensor.min()) / (self.qmax - self.qmin) |
| zero_point = self.qmin - torch.round(tensor.min() / scale) |
| qtensor = torch.round(tensor / scale + zero_point) |
| qtensor = torch.clamp(qtensor, self.qmin, self.qmax) |
| return (qtensor - zero_point) * scale, scale, zero_point |
|
|
| class ModelQuantization: |
| def __init__(self, model, bits=9): |
| self.model = model |
| self.quant = BasicQuantization(bits) |
|
|
| def quantize_model(self): |
| for name, module in self.model.named_modules(): |
| if isinstance(module, torch.nn.Linear): |
| if hasattr(module, 'weightML'): |
| quantized_weight, _, _ = self.quant.quantize_tensor(module.weight) |
| module.weight = torch.nn.Parameter(quantized_weight) |
| if hasattr(module, 'bias') and module.bias is not None: |
| quantized_bias, _, _ = self.quant.quantize_tensor(module.bias) |
| module.bias = torch.nn.Parameter(quantized_bias) |
|
|
| def empty_cache(): |
| gc.collect() |
| torch.cuda.empty_cache() |
| torch.cuda.reset_max_memory_allocated() |
| torch.cuda.reset_peak_memory_stats() |
|
|
| def load_pipeline() -> Pipeline: |
| empty_cache() |
| |
| |
| vae = AutoencoderKL.from_pretrained(ckpt_id, subfolder="vae", torch_dtype=dtype) |
| quantizer = ModelQuantization(vae) |
| quantizer.quantize_model() |
| |
| text_encoder_2 = T5EncoderModel.from_pretrained( |
| "city96/t5-v1_1-xxl-encoder-bf16", torch_dtype=torch.bfloat16 |
| ) |
|
|
| |
| pipeline = FluxPipeline.from_pretrained( |
| ckpt_id, |
| text_encoder_2=text_encoder_2, |
| vae=vae, |
| torch_dtype=dtype |
| ) |
|
|
|
|
| |
| for component in [pipeline.text_encoder, pipeline.text_encoder_2, pipeline.transformer, pipeline.vae]: |
| component.to(memory_format=torch.channels_last) |
|
|
| |
| pipeline.vae = torch.compile(pipeline.vae, fullgraph=True, dynamic=False, mode="max-autotune") |
| pipeline._exclude_from_cpu_offload = ["vae"] |
| pipeline.enable_sequential_cpu_offload() |
|
|
| |
| empty_cache() |
| for _ in range(3): |
| pipeline( |
| prompt="posteroexternal, eurythmical, inspection, semicotton, specification, Mercatorial, ethylate, misprint", |
| width=1024, |
| height=1024, |
| guidance_scale=0.0, |
| num_inference_steps=4, |
| max_sequence_length=256 |
| ) |
| |
| return pipeline |
|
|
| _inference_count = 0 |
|
|
| @torch.inference_mode() |
| def infer(request: TextToImageRequest, pipeline: Pipeline) -> Image: |
| global _inference_count |
| |
| |
| if _inference_count == 0: |
| empty_cache() |
| |
| |
| _inference_count += 1 |
| if _inference_count >= 4: |
| empty_cache() |
| _inference_count = 0 |
| |
| torch.cuda.reset_peak_memory_stats() |
| generator = Generator("cuda").manual_seed(request.seed) |
| return pipeline( |
| prompt=request.prompt, |
| generator=generator, |
| guidance_scale=0.0, |
| num_inference_steps=4, |
| max_sequence_length=256, |
| height=request.height, |
| width=request.width, |
| output_type="pil" |
| ).images[0] |
|
|