model.py

import gc
import glob
import os

import torch
import torch.distributed as dist
import torch.nn.functional as F
from loguru import logger
from safetensors import safe_open

from lightx2v.models.networks.wan.infer.feature_caching.transformer_infer import (
    WanTransformerInferAdaCaching,
    WanTransformerInferCustomCaching,
    WanTransformerInferDualBlock,
    WanTransformerInferDynamicBlock,
    WanTransformerInferFirstBlock,
    WanTransformerInferMagCaching,
    WanTransformerInferTaylorCaching,
    WanTransformerInferTeaCaching,
)
from lightx2v.models.networks.wan.infer.offload.transformer_infer import (
    WanOffloadTransformerInfer,
)
from lightx2v.models.networks.wan.infer.post_infer import WanPostInfer
from lightx2v.models.networks.wan.infer.pre_infer import WanPreInfer
from lightx2v.models.networks.wan.infer.transformer_infer import (
    WanTransformerInfer,
)
from lightx2v.models.networks.wan.weights.pre_weights import WanPreWeights
from lightx2v.models.networks.wan.weights.transformer_weights import (
    WanTransformerWeights,
)
from lightx2v.utils.custom_compiler import CompiledMethodsMixin, compiled_method
from lightx2v.utils.envs import *
from lightx2v.utils.ggml_tensor import load_gguf_sd_ckpt
from lightx2v.utils.utils import *


class WanModel(CompiledMethodsMixin):
    pre_weight_class = WanPreWeights
    transformer_weight_class = WanTransformerWeights

    def __init__(self, model_path, config, device, model_type="wan2.1"):
        super().__init__()
        self.model_path = model_path
        self.config = config
        self.cpu_offload = self.config.get("cpu_offload", False)
        self.offload_granularity = self.config.get("offload_granularity", "block")
        self.model_type = model_type
        self.remove_keys = []
        self.lazy_load = self.config.get("lazy_load", False)
        if self.lazy_load:
            self.remove_keys.extend(["blocks."])
        if self.config["seq_parallel"]:
            self.seq_p_group = self.config.get("device_mesh").get_group(mesh_dim="seq_p")
        else:
            self.seq_p_group = None

        self.clean_cuda_cache = self.config.get("clean_cuda_cache", False)
        self.dit_quantized = self.config.get("dit_quantized", False)
        if self.dit_quantized:
            assert self.config.get("dit_quant_scheme", "Default") in [
                "Default-Force-FP32",
                "fp8-vllm",
                "int8-vllm",
                "fp8-q8f",
                "int8-q8f",
                "fp8-b128-deepgemm",
                "fp8-sgl",
                "int8-sgl",
                "int8-torchao",
                "nvfp4",
                "mxfp4",
                "mxfp6-mxfp8",
                "mxfp8",
                "int8-tmo",
                "gguf-Q8_0",
                "gguf-Q6_K",
                "gguf-Q5_K_S",
                "gguf-Q5_K_M",
                "gguf-Q5_0",
                "gguf-Q5_1",
                "gguf-Q4_K_S",
                "gguf-Q4_K_M",
                "gguf-Q4_0",
                "gguf-Q4_1",
                "gguf-Q3_K_S",
                "gguf-Q3_K_M",
            ]
        self.device = device
        self._init_infer_class()
        self._init_weights()
        self._init_infer()

    def _init_infer_class(self):
        self.pre_infer_class = WanPreInfer
        self.post_infer_class = WanPostInfer

        if self.config["feature_caching"] == "NoCaching":
            self.transformer_infer_class = WanTransformerInfer if not self.cpu_offload else WanOffloadTransformerInfer
        elif self.config["feature_caching"] == "Tea":
            self.transformer_infer_class = WanTransformerInferTeaCaching
        elif self.config["feature_caching"] == "TaylorSeer":
            self.transformer_infer_class = WanTransformerInferTaylorCaching
        elif self.config["feature_caching"] == "Ada":
            self.transformer_infer_class = WanTransformerInferAdaCaching
        elif self.config["feature_caching"] == "Custom":
            self.transformer_infer_class = WanTransformerInferCustomCaching
        elif self.config["feature_caching"] == "FirstBlock":
            self.transformer_infer_class = WanTransformerInferFirstBlock
        elif self.config["feature_caching"] == "DualBlock":
            self.transformer_infer_class = WanTransformerInferDualBlock
        elif self.config["feature_caching"] == "DynamicBlock":
            self.transformer_infer_class = WanTransformerInferDynamicBlock
        elif self.config["feature_caching"] == "Mag":
            self.transformer_infer_class = WanTransformerInferMagCaching
        else:
            raise NotImplementedError(f"Unsupported feature_caching type: {self.config['feature_caching']}")

    def _should_load_weights(self):
        """Determine if current rank should load weights from disk."""
        if self.config.get("device_mesh") is None:
            # Single GPU mode
            return True
        elif dist.is_initialized():
            if self.config.get("load_from_rank0", False):
                # Multi-GPU mode, only rank 0 loads
                if dist.get_rank() == 0:
                    logger.info(f"Loading weights from {self.model_path}")
                    return True
            else:
                return True
        return False

    def _should_init_empty_model(self):
        if self.config.get("lora_configs") and self.config["lora_configs"]:
            if self.model_type in ["wan2.1"]:
                return True
            if self.model_type in ["wan2.2_moe_high_noise"]:
                for lora_config in self.config["lora_configs"]:
                    if lora_config["name"] == "high_noise_model":
                        return True
            if self.model_type in ["wan2.2_moe_low_noise"]:
                for lora_config in self.config["lora_configs"]:
                    if lora_config["name"] == "low_noise_model":
                        return True
        return False

    def _load_safetensor_to_dict(self, file_path, unified_dtype, sensitive_layer):
        remove_keys = self.remove_keys if hasattr(self, "remove_keys") else []

        if self.device.type != "cpu" and dist.is_initialized():
            device = dist.get_rank()
        else:
            device = str(self.device)

        with safe_open(file_path, framework="pt", device=device) as f:
            return {
                key: (f.get_tensor(key).to(GET_DTYPE()) if unified_dtype or all(s not in key for s in sensitive_layer) else f.get_tensor(key).to(GET_SENSITIVE_DTYPE()))
                for key in f.keys()
                if not any(remove_key in key for remove_key in remove_keys)
            }

    def _load_ckpt(self, unified_dtype, sensitive_layer):
        if self.config.get("dit_original_ckpt", None):
            safetensors_path = self.config["dit_original_ckpt"]
        else:
            safetensors_path = self.model_path

        if os.path.isdir(safetensors_path):
            safetensors_files = glob.glob(os.path.join(safetensors_path, "*.safetensors"))
        else:
            safetensors_files = [safetensors_path]

        if self.lazy_load:
            assert len(safetensors_files) == 1, "Only support single safetensors file in lazy load mode"
            self.lazy_load_path = safetensors_files[0]

        weight_dict = {}
        for file_path in safetensors_files:
            if self.config.get("adapter_model_path", None) is not None:
                if self.config["adapter_model_path"] == file_path:
                    continue
            logger.info(f"Loading weights from {file_path}")
            file_weights = self._load_safetensor_to_dict(file_path, unified_dtype, sensitive_layer)
            weight_dict.update(file_weights)

        return weight_dict

    def _load_quant_ckpt(self, unified_dtype, sensitive_layer):
        remove_keys = self.remove_keys if hasattr(self, "remove_keys") else []

        if self.config.get("dit_quantized_ckpt", None):
            safetensors_path = self.config["dit_quantized_ckpt"]
        else:
            safetensors_path = self.model_path

        if "gguf" in self.config.get("dit_quant_scheme", ""):
            gguf_path = ""
            if os.path.isdir(safetensors_path):
                gguf_type = self.config.get("dit_quant_scheme").replace("gguf-", "")
                gguf_files = list(filter(lambda x: gguf_type in x, glob.glob(os.path.join(safetensors_path, "*.gguf"))))
                gguf_path = gguf_files[0]
            else:
                gguf_path = safetensors_path
            weight_dict = self._load_gguf_ckpt(gguf_path)
            return weight_dict

        if os.path.isdir(safetensors_path):
            safetensors_files = glob.glob(os.path.join(safetensors_path, "*.safetensors"))
        else:
            safetensors_files = [safetensors_path]
            safetensors_path = os.path.dirname(safetensors_path)

        if self.lazy_load:
            assert len(safetensors_files) == 1, "Only support single safetensors file in lazy load mode"
            self.lazy_load_path = safetensors_files[0]

        weight_dict = {}
        for safetensor_path in safetensors_files:
            if self.config.get("adapter_model_path", None) is not None:
                if self.config["adapter_model_path"] == safetensor_path:
                    continue

            with safe_open(safetensor_path, framework="pt") as f:
                logger.info(f"Loading weights from {safetensor_path}")
                for k in f.keys():
                    if any(remove_key in k for remove_key in remove_keys):
                        continue
                    if f.get_tensor(k).dtype in [
                        torch.float16,
                        torch.bfloat16,
                        torch.float,
                    ]:
                        if unified_dtype or all(s not in k for s in sensitive_layer):
                            weight_dict[k] = f.get_tensor(k).to(GET_DTYPE()).to(self.device)
                        else:
                            weight_dict[k] = f.get_tensor(k).to(GET_SENSITIVE_DTYPE()).to(self.device)
                    else:
                        weight_dict[k] = f.get_tensor(k).to(self.device)

        if self.config.get("dit_quant_scheme", "Default") == "nvfp4":
            calib_path = os.path.join(safetensors_path, "calib.pt")
            logger.info(f"[CALIB] Loaded calibration data from: {calib_path}")
            calib_data = torch.load(calib_path, map_location="cpu")
            for k, v in calib_data["absmax"].items():
                weight_dict[k.replace(".weight", ".input_absmax")] = v.to(self.device)

        return weight_dict

    def _load_gguf_ckpt(self, gguf_path):
        state_dict = load_gguf_sd_ckpt(gguf_path, to_device=self.device)
        return state_dict

    def _init_weights(self, weight_dict=None):
        unified_dtype = GET_DTYPE() == GET_SENSITIVE_DTYPE()
        # Some layers run with float32 to achieve high accuracy
        sensitive_layer = {
            "norm",
            "embedding",
            "modulation",
            "time",
            "img_emb.proj.0",
            "img_emb.proj.4",
            "before_proj",  # vace
            "after_proj",  # vace
        }

        if weight_dict is None:
            is_weight_loader = self._should_load_weights()
            if is_weight_loader:
                if not self.dit_quantized:
                    # Load original weights
                    weight_dict = self._load_ckpt(unified_dtype, sensitive_layer)
                else:
                    # Load quantized weights
                    weight_dict = self._load_quant_ckpt(unified_dtype, sensitive_layer)

            if self.config.get("device_mesh") is not None and self.config.get("load_from_rank0", False):
                weight_dict = self._load_weights_from_rank0(weight_dict, is_weight_loader)

            if hasattr(self, "adapter_weights_dict"):
                weight_dict.update(self.adapter_weights_dict)

            self.original_weight_dict = weight_dict
        else:
            self.original_weight_dict = weight_dict

        # Initialize weight containers
        self.pre_weight = self.pre_weight_class(self.config)
        if self.lazy_load:
            self.transformer_weights = self.transformer_weight_class(self.config, self.lazy_load_path)
        else:
            self.transformer_weights = self.transformer_weight_class(self.config)
        if not self._should_init_empty_model():
            self._apply_weights()

    def _apply_weights(self, weight_dict=None):
        if weight_dict is not None:
            self.original_weight_dict = weight_dict
            del weight_dict
            gc.collect()
        # Load weights into containers
        self.pre_weight.load(self.original_weight_dict)
        self.transformer_weights.load(self.original_weight_dict)

        del self.original_weight_dict
        torch.cuda.empty_cache()
        gc.collect()

    def _load_weights_from_rank0(self, weight_dict, is_weight_loader):
        logger.info("Loading distributed weights")
        global_src_rank = 0
        target_device = "cpu" if self.cpu_offload else "cuda"

        if is_weight_loader:
            meta_dict = {}
            for key, tensor in weight_dict.items():
                meta_dict[key] = {"shape": tensor.shape, "dtype": tensor.dtype}

            obj_list = [meta_dict]
            dist.broadcast_object_list(obj_list, src=global_src_rank)
            synced_meta_dict = obj_list[0]
        else:
            obj_list = [None]
            dist.broadcast_object_list(obj_list, src=global_src_rank)
            synced_meta_dict = obj_list[0]

        distributed_weight_dict = {}
        for key, meta in synced_meta_dict.items():
            distributed_weight_dict[key] = torch.empty(meta["shape"], dtype=meta["dtype"], device=target_device)

        if target_device == "cuda":
            dist.barrier(device_ids=[torch.cuda.current_device()])

        for key in sorted(synced_meta_dict.keys()):
            if is_weight_loader:
                distributed_weight_dict[key].copy_(weight_dict[key], non_blocking=True)

            if target_device == "cpu":
                if is_weight_loader:
                    gpu_tensor = distributed_weight_dict[key].cuda()
                    dist.broadcast(gpu_tensor, src=global_src_rank)
                    distributed_weight_dict[key].copy_(gpu_tensor.cpu(), non_blocking=True)
                    del gpu_tensor
                    torch.cuda.empty_cache()
                else:
                    gpu_tensor = torch.empty_like(distributed_weight_dict[key], device="cuda")
                    dist.broadcast(gpu_tensor, src=global_src_rank)
                    distributed_weight_dict[key].copy_(gpu_tensor.cpu(), non_blocking=True)
                    del gpu_tensor
                    torch.cuda.empty_cache()

                if distributed_weight_dict[key].is_pinned():
                    distributed_weight_dict[key].copy_(distributed_weight_dict[key], non_blocking=True)
            else:
                dist.broadcast(distributed_weight_dict[key], src=global_src_rank)

        if target_device == "cuda":
            torch.cuda.synchronize()
        else:
            for tensor in distributed_weight_dict.values():
                if tensor.is_pinned():
                    tensor.copy_(tensor, non_blocking=False)

        logger.info(f"Weights distributed across {dist.get_world_size()} devices on {target_device}")

        return distributed_weight_dict

    def _init_infer(self):
        self.pre_infer = self.pre_infer_class(self.config)
        self.post_infer = self.post_infer_class(self.config)
        self.transformer_infer = self.transformer_infer_class(self.config)
        if hasattr(self.transformer_infer, "offload_manager"):
            self.transformer_infer.offload_manager.init_cuda_buffer(self.transformer_weights.offload_block_cuda_buffers, self.transformer_weights.offload_phase_cuda_buffers)
            if self.lazy_load:
                self.transformer_infer.offload_manager.init_cpu_buffer(self.transformer_weights.offload_block_cpu_buffers, self.transformer_weights.offload_phase_cpu_buffers)

    def set_scheduler(self, scheduler):
        self.scheduler = scheduler
        self.pre_infer.set_scheduler(scheduler)
        self.post_infer.set_scheduler(scheduler)
        self.transformer_infer.set_scheduler(scheduler)

    def to_cpu(self):
        self.pre_weight.to_cpu()
        self.transformer_weights.to_cpu()

    def to_cuda(self):
        self.pre_weight.to_cuda()
        self.transformer_weights.to_cuda()

    @torch.no_grad()
    def infer(self, inputs):
        if self.cpu_offload:
            if self.offload_granularity == "model" and self.scheduler.step_index == 0 and "wan2.2_moe" not in self.config["model_cls"]:
                self.to_cuda()
            elif self.offload_granularity != "model":
                self.pre_weight.to_cuda()
                self.transformer_weights.non_block_weights_to_cuda()

        if self.config["enable_cfg"]:
            if self.config["cfg_parallel"]:
                # ==================== CFG Parallel Processing ====================
                cfg_p_group = self.config["device_mesh"].get_group(mesh_dim="cfg_p")
                assert dist.get_world_size(cfg_p_group) == 2, "cfg_p_world_size must be equal to 2"
                cfg_p_rank = dist.get_rank(cfg_p_group)

                if cfg_p_rank == 0:
                    noise_pred = self._infer_cond_uncond(inputs, infer_condition=True)
                else:
                    noise_pred = self._infer_cond_uncond(inputs, infer_condition=False)

                noise_pred_list = [torch.zeros_like(noise_pred) for _ in range(2)]
                dist.all_gather(noise_pred_list, noise_pred, group=cfg_p_group)
                noise_pred_cond = noise_pred_list[0]  # cfg_p_rank == 0
                noise_pred_uncond = noise_pred_list[1]  # cfg_p_rank == 1
            else:
                # ==================== CFG Processing ====================
                noise_pred_cond = self._infer_cond_uncond(inputs, infer_condition=True)
                noise_pred_uncond = self._infer_cond_uncond(inputs, infer_condition=False)

            self.scheduler.noise_pred = noise_pred_uncond + self.scheduler.sample_guide_scale * (noise_pred_cond - noise_pred_uncond)
        else:
            # ==================== No CFG ====================
            self.scheduler.noise_pred = self._infer_cond_uncond(inputs, infer_condition=True)

        if self.cpu_offload:
            if self.offload_granularity == "model" and self.scheduler.step_index == self.scheduler.infer_steps - 1 and "wan2.2_moe" not in self.config["model_cls"]:
                self.to_cpu()
            elif self.offload_granularity != "model":
                self.pre_weight.to_cpu()
                self.transformer_weights.non_block_weights_to_cpu()

    @compiled_method()
    @torch.no_grad()
    def _infer_cond_uncond(self, inputs, infer_condition=True):
        self.scheduler.infer_condition = infer_condition

        pre_infer_out = self.pre_infer.infer(self.pre_weight, inputs)

        if self.config["seq_parallel"]:
            pre_infer_out = self._seq_parallel_pre_process(pre_infer_out)

        x = self.transformer_infer.infer(self.transformer_weights, pre_infer_out)

        if self.config["seq_parallel"]:
            x = self._seq_parallel_post_process(x)

        noise_pred = self.post_infer.infer(x, pre_infer_out)[0]

        if self.clean_cuda_cache:
            del x, pre_infer_out
            torch.cuda.empty_cache()

        return noise_pred

    @torch.no_grad()
    def _seq_parallel_pre_process(self, pre_infer_out):
        x = pre_infer_out.x
        world_size = dist.get_world_size(self.seq_p_group)
        cur_rank = dist.get_rank(self.seq_p_group)

        padding_size = (world_size - (x.shape[0] % world_size)) % world_size
        if padding_size > 0:
            x = F.pad(x, (0, 0, 0, padding_size))

        pre_infer_out.x = torch.chunk(x, world_size, dim=0)[cur_rank]

        if self.config["model_cls"] in ["wan2.2", "wan2.2_audio"] and self.config["task"] in ["i2v", "s2v"]:
            embed, embed0 = pre_infer_out.embed, pre_infer_out.embed0

            padding_size = (world_size - (embed.shape[0] % world_size)) % world_size
            if padding_size > 0:
                embed = F.pad(embed, (0, 0, 0, padding_size))
                embed0 = F.pad(embed0, (0, 0, 0, 0, 0, padding_size))

            pre_infer_out.embed = torch.chunk(embed, world_size, dim=0)[cur_rank]
            pre_infer_out.embed0 = torch.chunk(embed0, world_size, dim=0)[cur_rank]

        return pre_infer_out

    @torch.no_grad()
    def _seq_parallel_post_process(self, x):
        world_size = dist.get_world_size(self.seq_p_group)
        gathered_x = [torch.empty_like(x) for _ in range(world_size)]
        dist.all_gather(gathered_x, x, group=self.seq_p_group)
        combined_output = torch.cat(gathered_x, dim=0)
        return combined_output