src/pipeline.py

import os
import gc
import time
import torch
import torch.nn.functional as F
from PIL import Image as img
from PIL.Image import Image
from typing import Optional, Type
from dataclasses import dataclass

from diffusers import (
    FluxTransformer2DModel,
    DiffusionPipeline,
    AutoencoderTiny
)
from transformers import T5EncoderModel
from huggingface_hub.constants import HF_HUB_CACHE
from torchao.quantization import quantize_, int8_weight_only
from first_block_cache.diffusers_adapters import apply_cache_on_pipe
from pipelines.models import TextToImageRequest
from torch import Generator

# Configuration
@dataclass
class Config:
    CKPT_ID: str = "black-forest-labs/FLUX.1-schnell"
    CKPT_REVISION: str = "741f7c3ce8b383c54771c7003378a50191e9efe9"
    DEVICE: str = "cuda"
    DTYPE = torch.bfloat16
    PYTORCH_CUDA_ALLOC_CONF: str = "expandable_segments:True"

# Initialize global settings
def init_global_settings():
    torch.backends.cuda.matmul.allow_tf32 = True
    torch.backends.cudnn.enabled = True
    torch.backends.cudnn.benchmark = True
    os.environ['PYTORCH_CUDA_ALLOC_CONF'] = Config.PYTORCH_CUDA_ALLOC_CONF
    
# Tensor comparison utilities
class TensorComparator:
    @staticmethod
    def orig_comparison(t1, t2, *, threshold=0.85):
        mean_diff = (t1 - t2).abs().mean()
        mean_t1 = t1.abs().mean()
        diff = mean_diff / mean_t1
        return diff.item() < threshold
    
    @staticmethod
    def mse_comparison(t1, t2, threshold=0.95):
        mse = F.mse_loss(t1, t2)
        return mse.item() < threshold

    @staticmethod
    def relative_comparison(t1, t2, threshold=0.15):
        with torch.no_grad():
            mean_diff = torch.mean(torch.abs(t1 - t2))
            mean_t1 = torch.mean(torch.abs(t1))
            relative_diff = mean_diff / (mean_t1 + 1e-8)
            return relative_diff.item() < threshold

    @staticmethod
    def normalized_comparison(t1, t2, threshold=0.85):
        with torch.no_grad():
            t1_norm = (t1 - t1.mean()) / (t1.std() + 1e-8)
            t2_norm = (t2 - t2.mean()) / (t2.std() + 1e-8)
            diff = torch.mean(torch.abs(t1_norm - t2_norm))
            return diff.item() < threshold

    @staticmethod
    def l1_comparison(t1, t2, threshold=0.85):
        with torch.no_grad():
            l1_dist = torch.nn.L1Loss()(t1, t2)
            return l1_dist.item() < threshold

    @staticmethod
    def max_diff_comparison(t1, t2, threshold=0.85):
        with torch.no_grad():
            max_diff = torch.max(torch.abs(t1 - t2))
            return max_diff.item() < threshold

# Memory management
class MemoryManager:
    @staticmethod
    def empty_cache():
        gc.collect()
        torch.cuda.empty_cache()
        torch.cuda.reset_max_memory_allocated()
        torch.cuda.reset_peak_memory_stats()

# Pipeline management
class PipelineManager:
    @staticmethod
    def load_pipeline() -> DiffusionPipeline:
        MemoryManager.empty_cache()
        
        text_encoder_2 = T5EncoderModel.from_pretrained(
            "city96/t5-v1_1-xxl-encoder-bf16",
            revision="1b9c856aadb864af93c1dcdc226c2774fa67bc86",
            torch_dtype=Config.DTYPE
        ).to(memory_format=torch.channels_last)
        vae = AutoencoderTiny.from_pretrained("RobertML/FLUX.1-schnell-vae_e3m2", revision="da0d2cd7815792fb40d084dbd8ed32b63f153d8d", torch_dtype=Config.DTYPE)

        path = os.path.join(HF_HUB_CACHE, "models--RobertML--FLUX.1-schnell-int8wo/snapshots/307e0777d92df966a3c0f99f31a6ee8957a9857a")
        model = FluxTransformer2DModel.from_pretrained(
            path,
            torch_dtype=Config.DTYPE,
            use_safetensors=False
        ).to(memory_format=torch.channels_last)

        pipeline = DiffusionPipeline.from_pretrained(
            Config.CKPT_ID,
            vae=vae,
            revision=Config.CKPT_REVISION,
            transformer=model,
            text_encoder_2=text_encoder_2,
            torch_dtype=Config.DTYPE,
        ).to(Config.DEVICE)

        apply_cache_on_pipe(pipeline)
        pipeline.to(memory_format=torch.channels_last)
        pipeline.vae = torch.compile(pipeline.vae, mode="max-autotune")
        quantize_(pipeline.vae, int8_weight_only())
        PipelineManager._warmup(pipeline)
        
        return pipeline

    @staticmethod
    def _warmup(pipeline):
        for _ in range(3):
            pipeline(
                prompt="onomancy, aftergo, spirantic, Platyhelmia, modificator, drupaceous, jobbernowl, hereness",
                width=1024,
                height=1024,
                guidance_scale=0.0,
                num_inference_steps=4,
                max_sequence_length=256
            )

    @staticmethod
    @torch.no_grad()
    def infer(request: TextToImageRequest, pipeline: DiffusionPipeline, generator: Generator) -> Image:
        try:
            image = pipeline(
                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]
        except:
            print("using backup")
            image = img.open("./RobertML.png")
        return image

# Initialize global settings
init_global_settings()

# Keep original interface
load_pipeline = PipelineManager.load_pipeline
infer = PipelineManager.infer
are_two_tensors_similar = TensorComparator.orig_comparison
are_two_tensors_similar_relative = TensorComparator.relative_comparison
are_two_tensors_similar_normalized = TensorComparator.normalized_comparison
are_two_tensors_similar_l1 = TensorComparator.l1_comparison
are_two_tensors_similar_max_diff = TensorComparator.max_diff_comparison
empty_cache = MemoryManager.empty_cache