diff --git a/comfy/ipex_hijacks/__init__.py b/comfy/ipex_hijacks/__init__.py new file mode 100644 index 000000000..84eb1af63 --- /dev/null +++ b/comfy/ipex_hijacks/__init__.py @@ -0,0 +1,169 @@ +import os +import sys +import contextlib +import torch +import intel_extension_for_pytorch as ipex # pylint: disable=import-error, unused-import +from .hijacks import ipex_hijacks + +# pylint: disable=protected-access, missing-function-docstring, line-too-long + +def ipex_init(): # pylint: disable=too-many-statements + try: + # Replace cuda with xpu: + torch.cuda.current_device = torch.xpu.current_device + torch.cuda.current_stream = torch.xpu.current_stream + torch.cuda.device = torch.xpu.device + torch.cuda.device_count = torch.xpu.device_count + torch.cuda.device_of = torch.xpu.device_of + torch.cuda.get_device_name = torch.xpu.get_device_name + torch.cuda.get_device_properties = torch.xpu.get_device_properties + torch.cuda.init = torch.xpu.init + torch.cuda.is_available = torch.xpu.is_available + torch.cuda.is_initialized = torch.xpu.is_initialized + torch.cuda.is_current_stream_capturing = lambda: False + torch.cuda.set_device = torch.xpu.set_device + torch.cuda.stream = torch.xpu.stream + torch.cuda.synchronize = torch.xpu.synchronize + torch.cuda.Event = torch.xpu.Event + torch.cuda.Stream = torch.xpu.Stream + torch.cuda.FloatTensor = torch.xpu.FloatTensor + torch.Tensor.cuda = torch.Tensor.xpu + torch.Tensor.is_cuda = torch.Tensor.is_xpu + torch.UntypedStorage.cuda = torch.UntypedStorage.xpu + torch.cuda._initialization_lock = torch.xpu.lazy_init._initialization_lock + torch.cuda._initialized = torch.xpu.lazy_init._initialized + torch.cuda._lazy_seed_tracker = torch.xpu.lazy_init._lazy_seed_tracker + torch.cuda._queued_calls = torch.xpu.lazy_init._queued_calls + torch.cuda._tls = torch.xpu.lazy_init._tls + torch.cuda.threading = torch.xpu.lazy_init.threading + torch.cuda.traceback = torch.xpu.lazy_init.traceback + torch.cuda.Optional = torch.xpu.Optional + torch.cuda.__cached__ = torch.xpu.__cached__ + torch.cuda.__loader__ = torch.xpu.__loader__ + torch.cuda.ComplexFloatStorage = torch.xpu.ComplexFloatStorage + torch.cuda.Tuple = torch.xpu.Tuple + torch.cuda.streams = torch.xpu.streams + torch.cuda._lazy_new = torch.xpu._lazy_new + torch.cuda.FloatStorage = torch.xpu.FloatStorage + torch.cuda.Any = torch.xpu.Any + torch.cuda.__doc__ = torch.xpu.__doc__ + torch.cuda.default_generators = torch.xpu.default_generators + torch.cuda.HalfTensor = torch.xpu.HalfTensor + torch.cuda._get_device_index = torch.xpu._get_device_index + torch.cuda.__path__ = torch.xpu.__path__ + torch.cuda.Device = torch.xpu.Device + torch.cuda.IntTensor = torch.xpu.IntTensor + torch.cuda.ByteStorage = torch.xpu.ByteStorage + torch.cuda.set_stream = torch.xpu.set_stream + torch.cuda.BoolStorage = torch.xpu.BoolStorage + torch.cuda.os = torch.xpu.os + torch.cuda.torch = torch.xpu.torch + torch.cuda.BFloat16Storage = torch.xpu.BFloat16Storage + torch.cuda.Union = torch.xpu.Union + torch.cuda.DoubleTensor = torch.xpu.DoubleTensor + torch.cuda.ShortTensor = torch.xpu.ShortTensor + torch.cuda.LongTensor = torch.xpu.LongTensor + torch.cuda.IntStorage = torch.xpu.IntStorage + torch.cuda.LongStorage = torch.xpu.LongStorage + torch.cuda.__annotations__ = torch.xpu.__annotations__ + torch.cuda.__package__ = torch.xpu.__package__ + torch.cuda.__builtins__ = torch.xpu.__builtins__ + torch.cuda.CharTensor = torch.xpu.CharTensor + torch.cuda.List = torch.xpu.List + torch.cuda._lazy_init = torch.xpu._lazy_init + torch.cuda.BFloat16Tensor = torch.xpu.BFloat16Tensor + torch.cuda.DoubleStorage = torch.xpu.DoubleStorage + torch.cuda.ByteTensor = torch.xpu.ByteTensor + torch.cuda.StreamContext = torch.xpu.StreamContext + torch.cuda.ComplexDoubleStorage = torch.xpu.ComplexDoubleStorage + torch.cuda.ShortStorage = torch.xpu.ShortStorage + torch.cuda._lazy_call = torch.xpu._lazy_call + torch.cuda.HalfStorage = torch.xpu.HalfStorage + torch.cuda.random = torch.xpu.random + torch.cuda._device = torch.xpu._device + torch.cuda.classproperty = torch.xpu.classproperty + torch.cuda.__name__ = torch.xpu.__name__ + torch.cuda._device_t = torch.xpu._device_t + torch.cuda.warnings = torch.xpu.warnings + torch.cuda.__spec__ = torch.xpu.__spec__ + torch.cuda.BoolTensor = torch.xpu.BoolTensor + torch.cuda.CharStorage = torch.xpu.CharStorage + torch.cuda.__file__ = torch.xpu.__file__ + torch.cuda._is_in_bad_fork = torch.xpu.lazy_init._is_in_bad_fork + # torch.cuda.is_current_stream_capturing = torch.xpu.is_current_stream_capturing + + # Memory: + torch.cuda.memory = torch.xpu.memory + if 'linux' in sys.platform and "WSL2" in os.popen("uname -a").read(): + torch.xpu.empty_cache = lambda: None + torch.cuda.empty_cache = torch.xpu.empty_cache + torch.cuda.memory_stats = torch.xpu.memory_stats + torch.cuda.memory_summary = torch.xpu.memory_summary + torch.cuda.memory_snapshot = torch.xpu.memory_snapshot + torch.cuda.memory_allocated = torch.xpu.memory_allocated + torch.cuda.max_memory_allocated = torch.xpu.max_memory_allocated + torch.cuda.memory_reserved = torch.xpu.memory_reserved + torch.cuda.memory_cached = torch.xpu.memory_reserved + torch.cuda.max_memory_reserved = torch.xpu.max_memory_reserved + torch.cuda.max_memory_cached = torch.xpu.max_memory_reserved + torch.cuda.reset_peak_memory_stats = torch.xpu.reset_peak_memory_stats + torch.cuda.reset_max_memory_cached = torch.xpu.reset_peak_memory_stats + torch.cuda.reset_max_memory_allocated = torch.xpu.reset_peak_memory_stats + torch.cuda.memory_stats_as_nested_dict = torch.xpu.memory_stats_as_nested_dict + torch.cuda.reset_accumulated_memory_stats = torch.xpu.reset_accumulated_memory_stats + + # RNG: + torch.cuda.get_rng_state = torch.xpu.get_rng_state + torch.cuda.get_rng_state_all = torch.xpu.get_rng_state_all + torch.cuda.set_rng_state = torch.xpu.set_rng_state + torch.cuda.set_rng_state_all = torch.xpu.set_rng_state_all + torch.cuda.manual_seed = torch.xpu.manual_seed + torch.cuda.manual_seed_all = torch.xpu.manual_seed_all + torch.cuda.seed = torch.xpu.seed + torch.cuda.seed_all = torch.xpu.seed_all + torch.cuda.initial_seed = torch.xpu.initial_seed + + # AMP: + torch.cuda.amp = torch.xpu.amp + if not hasattr(torch.cuda.amp, "common"): + torch.cuda.amp.common = contextlib.nullcontext() + torch.cuda.amp.common.amp_definitely_not_available = lambda: False + try: + torch.cuda.amp.GradScaler = torch.xpu.amp.GradScaler + except Exception: # pylint: disable=broad-exception-caught + try: + from .gradscaler import gradscaler_init # pylint: disable=import-outside-toplevel, import-error + gradscaler_init() + torch.cuda.amp.GradScaler = torch.xpu.amp.GradScaler + except Exception: # pylint: disable=broad-exception-caught + torch.cuda.amp.GradScaler = ipex.cpu.autocast._grad_scaler.GradScaler + + # C + torch._C._cuda_getCurrentRawStream = ipex._C._getCurrentStream + ipex._C._DeviceProperties.major = 2023 + ipex._C._DeviceProperties.minor = 2 + + # Fix functions with ipex: + torch.cuda.mem_get_info = lambda device=None: [(torch.xpu.get_device_properties(device).total_memory - torch.xpu.memory_reserved(device)), torch.xpu.get_device_properties(device).total_memory] + torch._utils._get_available_device_type = lambda: "xpu" + torch.has_cuda = True + torch.cuda.has_half = True + torch.cuda.is_bf16_supported = lambda *args, **kwargs: True + torch.cuda.is_fp16_supported = lambda *args, **kwargs: True + torch.version.cuda = "11.7" + torch.cuda.get_device_capability = lambda *args, **kwargs: [11,7] + torch.cuda.get_device_properties.major = 11 + torch.cuda.get_device_properties.minor = 7 + torch.cuda.ipc_collect = lambda *args, **kwargs: None + torch.cuda.utilization = lambda *args, **kwargs: 0 + + ipex_hijacks() + if not torch.xpu.has_fp64_dtype(): + try: + from .diffusers import ipex_diffusers + ipex_diffusers() + except Exception: # pylint: disable=broad-exception-caught + pass + except Exception as e: + return False, e + return True, None \ No newline at end of file diff --git a/comfy/ipex_hijacks/attention.py b/comfy/ipex_hijacks/attention.py new file mode 100644 index 000000000..9f8e59c55 --- /dev/null +++ b/comfy/ipex_hijacks/attention.py @@ -0,0 +1,157 @@ +import os +import torch +import intel_extension_for_pytorch as ipex # pylint: disable=import-error, unused-import +from functools import cache + +# pylint: disable=protected-access, missing-function-docstring, line-too-long + +# ARC GPUs can't allocate more than 4GB to a single block so we slice the attetion layers + +sdpa_slice_trigger_rate = float(os.environ.get('IPEX_SDPA_SLICE_TRIGGER_RATE', 6)) +attention_slice_rate = float(os.environ.get('IPEX_ATTENTION_SLICE_RATE', 4)) + +# Find something divisible with the input_tokens +@cache +def find_slice_size(slice_size, slice_block_size): + while (slice_size * slice_block_size) > attention_slice_rate: + slice_size = slice_size // 2 + if slice_size <= 1: + slice_size = 1 + break + return slice_size + +# Find slice sizes for SDPA +@cache +def find_sdpa_slice_sizes(query_shape, query_element_size): + if len(query_shape) == 3: + batch_size_attention, query_tokens, shape_three = query_shape + shape_four = 1 + else: + batch_size_attention, query_tokens, shape_three, shape_four = query_shape + + slice_block_size = query_tokens * shape_three * shape_four / 1024 / 1024 * query_element_size + block_size = batch_size_attention * slice_block_size + + split_slice_size = batch_size_attention + split_2_slice_size = query_tokens + split_3_slice_size = shape_three + + do_split = False + do_split_2 = False + do_split_3 = False + + if block_size > sdpa_slice_trigger_rate: + do_split = True + split_slice_size = find_slice_size(split_slice_size, slice_block_size) + if split_slice_size * slice_block_size > attention_slice_rate: + slice_block_size_2 = split_slice_size * shape_three * shape_four / 1024 / 1024 * query_element_size + do_split_2 = True + split_2_slice_size = find_slice_size(split_2_slice_size, slice_block_size_2) + if split_2_slice_size * slice_block_size_2 > attention_slice_rate: + slice_block_size_3 = split_slice_size * split_2_slice_size * shape_four / 1024 / 1024 * query_element_size + do_split_3 = True + split_3_slice_size = find_slice_size(split_3_slice_size, slice_block_size_3) + + return do_split, do_split_2, do_split_3, split_slice_size, split_2_slice_size, split_3_slice_size + +# Find slice sizes for BMM +@cache +def find_bmm_slice_sizes(input_shape, input_element_size, mat2_shape): + batch_size_attention, input_tokens, mat2_atten_shape = input_shape[0], input_shape[1], mat2_shape[2] + slice_block_size = input_tokens * mat2_atten_shape / 1024 / 1024 * input_element_size + block_size = batch_size_attention * slice_block_size + + split_slice_size = batch_size_attention + split_2_slice_size = input_tokens + + do_split = False + do_split_2 = False + + if block_size > attention_slice_rate: + do_split = True + split_slice_size = find_slice_size(split_slice_size, slice_block_size) + if split_slice_size * slice_block_size > attention_slice_rate: + slice_block_size_2 = split_slice_size * mat2_atten_shape / 1024 / 1024 * input_element_size + do_split_2 = True + split_2_slice_size = find_slice_size(split_2_slice_size, slice_block_size_2) + + return do_split, do_split_2, split_slice_size, split_2_slice_size + + +original_torch_bmm = torch.bmm +def torch_bmm_32_bit(input, mat2, *, out=None): + do_split, do_split_2, split_slice_size, split_2_slice_size = find_bmm_slice_sizes(input.shape, input.element_size(), mat2.shape) + + # Slice BMM + if do_split: + batch_size_attention, input_tokens = input.shape[0], input.shape[1] + hidden_states = torch.zeros(input.shape[0], input.shape[1], mat2.shape[2], device=input.device, dtype=input.dtype) + for i in range(batch_size_attention // split_slice_size): + start_idx = i * split_slice_size + end_idx = (i + 1) * split_slice_size + if do_split_2: + for i2 in range(input_tokens // split_2_slice_size): # pylint: disable=invalid-name + start_idx_2 = i2 * split_2_slice_size + end_idx_2 = (i2 + 1) * split_2_slice_size + hidden_states[start_idx:end_idx, start_idx_2:end_idx_2] = original_torch_bmm( + input[start_idx:end_idx, start_idx_2:end_idx_2], + mat2[start_idx:end_idx, start_idx_2:end_idx_2], + out=out + ) + else: + hidden_states[start_idx:end_idx] = original_torch_bmm( + input[start_idx:end_idx], + mat2[start_idx:end_idx], + out=out + ) + else: + return original_torch_bmm(input, mat2, out=out) + return hidden_states + +original_scaled_dot_product_attention = torch.nn.functional.scaled_dot_product_attention +def scaled_dot_product_attention_32_bit(query, key, value, attn_mask=None, dropout_p=0.0, is_causal=False): + do_split, do_split_2, do_split_3, split_slice_size, split_2_slice_size, split_3_slice_size = find_sdpa_slice_sizes(query.shape, query.element_size()) + + # Slice SDPA + if do_split: + batch_size_attention, query_tokens, shape_three = query.shape[0], query.shape[1], query.shape[2] + hidden_states = torch.zeros(query.shape, device=query.device, dtype=query.dtype) + for i in range(batch_size_attention // split_slice_size): + start_idx = i * split_slice_size + end_idx = (i + 1) * split_slice_size + if do_split_2: + for i2 in range(query_tokens // split_2_slice_size): # pylint: disable=invalid-name + start_idx_2 = i2 * split_2_slice_size + end_idx_2 = (i2 + 1) * split_2_slice_size + if do_split_3: + for i3 in range(shape_three // split_3_slice_size): # pylint: disable=invalid-name + start_idx_3 = i3 * split_3_slice_size + end_idx_3 = (i3 + 1) * split_3_slice_size + hidden_states[start_idx:end_idx, start_idx_2:end_idx_2, start_idx_3:end_idx_3] = original_scaled_dot_product_attention( + query[start_idx:end_idx, start_idx_2:end_idx_2, start_idx_3:end_idx_3], + key[start_idx:end_idx, start_idx_2:end_idx_2, start_idx_3:end_idx_3], + value[start_idx:end_idx, start_idx_2:end_idx_2, start_idx_3:end_idx_3], + attn_mask=attn_mask[start_idx:end_idx, start_idx_2:end_idx_2, start_idx_3:end_idx_3] if attn_mask is not None else attn_mask, + dropout_p=dropout_p, is_causal=is_causal + ) + else: + hidden_states[start_idx:end_idx, start_idx_2:end_idx_2] = original_scaled_dot_product_attention( + query[start_idx:end_idx, start_idx_2:end_idx_2], + key[start_idx:end_idx, start_idx_2:end_idx_2], + value[start_idx:end_idx, start_idx_2:end_idx_2], + attn_mask=attn_mask[start_idx:end_idx, start_idx_2:end_idx_2] if attn_mask is not None else attn_mask, + dropout_p=dropout_p, is_causal=is_causal + ) + else: + hidden_states[start_idx:end_idx] = original_scaled_dot_product_attention( + query[start_idx:end_idx], + key[start_idx:end_idx], + value[start_idx:end_idx], + attn_mask=attn_mask[start_idx:end_idx] if attn_mask is not None else attn_mask, + dropout_p=dropout_p, is_causal=is_causal + ) + else: + return original_scaled_dot_product_attention( + query, key, value, attn_mask=attn_mask, dropout_p=dropout_p, is_causal=is_causal + ) + return hidden_states \ No newline at end of file diff --git a/comfy/ipex_hijacks/diffusers.py b/comfy/ipex_hijacks/diffusers.py new file mode 100644 index 000000000..d8afbb823 --- /dev/null +++ b/comfy/ipex_hijacks/diffusers.py @@ -0,0 +1,272 @@ +## code is copied directly from sdnext fork +## https://github.com/vladmandic/automatic/blob/master/modules/intel/ipex/ + +import torch +import diffusers #0.21.1 # pylint: disable=import-error +from diffusers.models.attention_processor import Attention +import torch.nn.functional as F +# pylint: disable=protected-access, missing-function-docstring, line-too-long + + +class AttnProcessor2_0: + r""" + Processor for implementing scaled dot-product attention (enabled by default if you're using PyTorch 2.0). + """ + + def __init__(self): + if not hasattr(F, "scaled_dot_product_attention"): + raise ImportError("AttnProcessor2_0 requires PyTorch 2.0, to use it, please upgrade PyTorch to 2.0.") + + def __call__( + self, + attn: Attention, + hidden_states, + encoder_hidden_states=None, + attention_mask=None, + temb=None, + ): + residual = hidden_states + + if attn.spatial_norm is not None: + hidden_states = attn.spatial_norm(hidden_states, temb) + + input_ndim = hidden_states.ndim + + if input_ndim == 4: + batch_size, channel, height, width = hidden_states.shape + hidden_states = hidden_states.view(batch_size, channel, height * width).transpose(1, 2) + + batch_size, sequence_length, _ = ( + hidden_states.shape if encoder_hidden_states is None else encoder_hidden_states.shape + ) + + if attention_mask is not None: + attention_mask = attn.prepare_attention_mask(attention_mask, sequence_length, batch_size) + # scaled_dot_product_attention expects attention_mask shape to be + # (batch, heads, source_length, target_length) + attention_mask = attention_mask.view(batch_size, attn.heads, -1, attention_mask.shape[-1]) + + if attn.group_norm is not None: + hidden_states = attn.group_norm(hidden_states.transpose(1, 2)).transpose(1, 2) + + query = attn.to_q(hidden_states) + + if encoder_hidden_states is None: + encoder_hidden_states = hidden_states + elif attn.norm_cross: + encoder_hidden_states = attn.norm_encoder_hidden_states(encoder_hidden_states) + + key = attn.to_k(encoder_hidden_states) + value = attn.to_v(encoder_hidden_states) + + inner_dim = key.shape[-1] + head_dim = inner_dim // attn.heads + + query = query.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2) + + key = key.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2) + value = value.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2) + + #ARC GPUs can't allocate more than 4GB to a single block, Slice it: + shape_one, batch_size_attention, query_tokens, shape_four = query.shape + block_multiply = 2.4 if query.dtype == torch.float32 else 1.2 + block_size = (shape_one * batch_size_attention * query_tokens * shape_four) / 1024 * block_multiply #MB + split_slice_size = batch_size_attention + if block_size >= 1000: + do_split = True + #Find something divisible with the shape_one + sanity_check = 0 + while ((shape_one * split_slice_size * query_tokens * shape_four) / 1024 * block_multiply) > 4000: + split_slice_size = split_slice_size // 2 + if split_slice_size <= 1: + split_slice_size = 1 + break + sanity_check = sanity_check + 1 + if sanity_check >= 128: + break + else: + do_split = False + + split_block_size = (shape_one * split_slice_size * query_tokens * shape_four) / 1024 * block_multiply #MB + split_2_slice_size = query_tokens + if split_block_size >= 1000: + do_split_2 = True + #Find something divisible with the batch_size_attention + sanity_check = 0 + while ((shape_one * split_slice_size * split_2_slice_size * shape_four) / 1024 * block_multiply) > 4000: + split_2_slice_size = split_2_slice_size // 2 + if split_2_slice_size <= 1: + split_2_slice_size = 1 + break + sanity_check = sanity_check + 1 + if sanity_check >= 128: + break + else: + do_split_2 = False + + if do_split: + hidden_states = torch.zeros(query.shape, device=query.device, dtype=query.dtype) + for i in range(batch_size_attention // split_slice_size): + start_idx = i * split_slice_size + end_idx = (i + 1) * split_slice_size + if do_split_2: + for i2 in range(query_tokens // split_2_slice_size): + start_idx_2 = i2 * split_2_slice_size + end_idx_2 = (i2 + 1) * split_2_slice_size + + query_slice = query[:, start_idx:end_idx, start_idx_2:end_idx_2] + key_slice = key[:, start_idx:end_idx, start_idx_2:end_idx_2] + attn_mask_slice = attention_mask[:, start_idx:end_idx, start_idx_2:end_idx_2] if attention_mask is not None else None + + attn_slice = F.scaled_dot_product_attention( + query_slice, key_slice, value[:, start_idx:end_idx, start_idx_2:end_idx_2], + attn_mask=attn_mask_slice, dropout_p=0.0, is_causal=False + ) + hidden_states[:, start_idx:end_idx, start_idx_2:end_idx_2] = attn_slice + else: + query_slice = query[:, start_idx:end_idx] + key_slice = key[:, start_idx:end_idx] + attn_mask_slice = attention_mask[:, start_idx:end_idx] if attention_mask is not None else None + + attn_slice = F.scaled_dot_product_attention( + query_slice, key_slice, value[:, start_idx:end_idx], + attn_mask=attn_mask_slice, dropout_p=0.0, is_causal=False + ) + hidden_states[:, start_idx:end_idx] = attn_slice + else: + hidden_states = F.scaled_dot_product_attention( + query, key, value, attn_mask=attention_mask, dropout_p=0.0, is_causal=False + ) + + hidden_states = hidden_states.transpose(1, 2).reshape(batch_size, -1, attn.heads * head_dim) + hidden_states = hidden_states.to(query.dtype) + + # linear proj + hidden_states = attn.to_out[0](hidden_states) + # dropout + hidden_states = attn.to_out[1](hidden_states) + + if input_ndim == 4: + hidden_states = hidden_states.transpose(-1, -2).reshape(batch_size, channel, height, width) + + if attn.residual_connection: + hidden_states = hidden_states + residual + + hidden_states = hidden_states / attn.rescale_output_factor + + return hidden_states + + +class SlicedAttnProcessor: # pylint: disable=too-few-public-methods + r""" + Processor for implementing sliced attention. + Args: + slice_size (`int`, *optional*): + The number of steps to compute attention. Uses as many slices as `attention_head_dim // slice_size`, and + `attention_head_dim` must be a multiple of the `slice_size`. + """ + + def __init__(self, slice_size): + self.slice_size = slice_size + + def __call__(self, attn: Attention, hidden_states, encoder_hidden_states=None, attention_mask=None): # pylint: disable=too-many-statements, too-many-locals, too-many-branches + residual = hidden_states + + input_ndim = hidden_states.ndim + + if input_ndim == 4: + batch_size, channel, height, width = hidden_states.shape + hidden_states = hidden_states.view(batch_size, channel, height * width).transpose(1, 2) + + batch_size, sequence_length, _ = ( + hidden_states.shape if encoder_hidden_states is None else encoder_hidden_states.shape + ) + attention_mask = attn.prepare_attention_mask(attention_mask, sequence_length, batch_size) + + if attn.group_norm is not None: + hidden_states = attn.group_norm(hidden_states.transpose(1, 2)).transpose(1, 2) + + query = attn.to_q(hidden_states) + dim = query.shape[-1] + query = attn.head_to_batch_dim(query) + + if encoder_hidden_states is None: + encoder_hidden_states = hidden_states + elif attn.norm_cross: + encoder_hidden_states = attn.norm_encoder_hidden_states(encoder_hidden_states) + + key = attn.to_k(encoder_hidden_states) + value = attn.to_v(encoder_hidden_states) + key = attn.head_to_batch_dim(key) + value = attn.head_to_batch_dim(value) + + batch_size_attention, query_tokens, shape_three = query.shape + hidden_states = torch.zeros( + (batch_size_attention, query_tokens, dim // attn.heads), device=query.device, dtype=query.dtype + ) + + #ARC GPUs can't allocate more than 4GB to a single block, Slice it: + block_multiply = query.element_size() + slice_block_size = self.slice_size * shape_three / 1024 / 1024 * block_multiply + block_size = query_tokens * slice_block_size + split_2_slice_size = query_tokens + if block_size > 4: + do_split_2 = True + #Find something divisible with the query_tokens + while (split_2_slice_size * slice_block_size) > 4: + split_2_slice_size = split_2_slice_size // 2 + if split_2_slice_size <= 1: + split_2_slice_size = 1 + break + else: + do_split_2 = False + + for i in range(batch_size_attention // self.slice_size): + start_idx = i * self.slice_size + end_idx = (i + 1) * self.slice_size + + if do_split_2: + for i2 in range(query_tokens // split_2_slice_size): # pylint: disable=invalid-name + start_idx_2 = i2 * split_2_slice_size + end_idx_2 = (i2 + 1) * split_2_slice_size + + query_slice = query[start_idx:end_idx, start_idx_2:end_idx_2] + key_slice = key[start_idx:end_idx, start_idx_2:end_idx_2] + attn_mask_slice = attention_mask[start_idx:end_idx, start_idx_2:end_idx_2] if attention_mask is not None else None + + attn_slice = attn.get_attention_scores(query_slice, key_slice, attn_mask_slice) + attn_slice = torch.bmm(attn_slice, value[start_idx:end_idx, start_idx_2:end_idx_2]) + + hidden_states[start_idx:end_idx, start_idx_2:end_idx_2] = attn_slice + else: + query_slice = query[start_idx:end_idx] + key_slice = key[start_idx:end_idx] + attn_mask_slice = attention_mask[start_idx:end_idx] if attention_mask is not None else None + + attn_slice = attn.get_attention_scores(query_slice, key_slice, attn_mask_slice) + + attn_slice = torch.bmm(attn_slice, value[start_idx:end_idx]) + + hidden_states[start_idx:end_idx] = attn_slice + + hidden_states = attn.batch_to_head_dim(hidden_states) + + # linear proj + hidden_states = attn.to_out[0](hidden_states) + # dropout + hidden_states = attn.to_out[1](hidden_states) + + if input_ndim == 4: + hidden_states = hidden_states.transpose(-1, -2).reshape(batch_size, channel, height, width) + + if attn.residual_connection: + hidden_states = hidden_states + residual + + hidden_states = hidden_states / attn.rescale_output_factor + + return hidden_states + +def ipex_diffusers(): + #ARC GPUs can't allocate more than 4GB to a single block: + diffusers.models.attention_processor.SlicedAttnProcessor = SlicedAttnProcessor + #diffusers.models.attention_processor.AttnProcessor2_0 = AttnProcessor2_0 \ No newline at end of file diff --git a/comfy/ipex_hijacks/gradscaler.py b/comfy/ipex_hijacks/gradscaler.py new file mode 100644 index 000000000..3c265ddb3 --- /dev/null +++ b/comfy/ipex_hijacks/gradscaler.py @@ -0,0 +1,179 @@ +from collections import defaultdict +import torch +import intel_extension_for_pytorch as ipex # pylint: disable=import-error, unused-import +import intel_extension_for_pytorch._C as core # pylint: disable=import-error, unused-import + +# pylint: disable=protected-access, missing-function-docstring, line-too-long + +OptState = ipex.cpu.autocast._grad_scaler.OptState +_MultiDeviceReplicator = ipex.cpu.autocast._grad_scaler._MultiDeviceReplicator +_refresh_per_optimizer_state = ipex.cpu.autocast._grad_scaler._refresh_per_optimizer_state + +def _unscale_grads_(self, optimizer, inv_scale, found_inf, allow_fp16): # pylint: disable=unused-argument + per_device_inv_scale = _MultiDeviceReplicator(inv_scale) + per_device_found_inf = _MultiDeviceReplicator(found_inf) + + # To set up _amp_foreach_non_finite_check_and_unscale_, split grads by device and dtype. + # There could be hundreds of grads, so we'd like to iterate through them just once. + # However, we don't know their devices or dtypes in advance. + + # https://stackoverflow.com/questions/5029934/defaultdict-of-defaultdict + # Google says mypy struggles with defaultdicts type annotations. + per_device_and_dtype_grads = defaultdict(lambda: defaultdict(list)) # type: ignore[var-annotated] + # sync grad to master weight + if hasattr(optimizer, "sync_grad"): + optimizer.sync_grad() + with torch.no_grad(): + for group in optimizer.param_groups: + for param in group["params"]: + if param.grad is None: + continue + if (not allow_fp16) and param.grad.dtype == torch.float16: + raise ValueError("Attempting to unscale FP16 gradients.") + if param.grad.is_sparse: + # is_coalesced() == False means the sparse grad has values with duplicate indices. + # coalesce() deduplicates indices and adds all values that have the same index. + # For scaled fp16 values, there's a good chance coalescing will cause overflow, + # so we should check the coalesced _values(). + if param.grad.dtype is torch.float16: + param.grad = param.grad.coalesce() + to_unscale = param.grad._values() + else: + to_unscale = param.grad + + # -: is there a way to split by device and dtype without appending in the inner loop? + to_unscale = to_unscale.to("cpu") + per_device_and_dtype_grads[to_unscale.device][ + to_unscale.dtype + ].append(to_unscale) + + for _, per_dtype_grads in per_device_and_dtype_grads.items(): + for grads in per_dtype_grads.values(): + core._amp_foreach_non_finite_check_and_unscale_( + grads, + per_device_found_inf.get("cpu"), + per_device_inv_scale.get("cpu"), + ) + + return per_device_found_inf._per_device_tensors + +def unscale_(self, optimizer): + """ + Divides ("unscales") the optimizer's gradient tensors by the scale factor. + :meth:`unscale_` is optional, serving cases where you need to + :ref:`modify or inspect gradients` + between the backward pass(es) and :meth:`step`. + If :meth:`unscale_` is not called explicitly, gradients will be unscaled automatically during :meth:`step`. + Simple example, using :meth:`unscale_` to enable clipping of unscaled gradients:: + ... + scaler.scale(loss).backward() + scaler.unscale_(optimizer) + torch.nn.utils.clip_grad_norm_(model.parameters(), max_norm) + scaler.step(optimizer) + scaler.update() + Args: + optimizer (torch.optim.Optimizer): Optimizer that owns the gradients to be unscaled. + .. warning:: + :meth:`unscale_` should only be called once per optimizer per :meth:`step` call, + and only after all gradients for that optimizer's assigned parameters have been accumulated. + Calling :meth:`unscale_` twice for a given optimizer between each :meth:`step` triggers a RuntimeError. + .. warning:: + :meth:`unscale_` may unscale sparse gradients out of place, replacing the ``.grad`` attribute. + """ + if not self._enabled: + return + + self._check_scale_growth_tracker("unscale_") + + optimizer_state = self._per_optimizer_states[id(optimizer)] + + if optimizer_state["stage"] is OptState.UNSCALED: # pylint: disable=no-else-raise + raise RuntimeError( + "unscale_() has already been called on this optimizer since the last update()." + ) + elif optimizer_state["stage"] is OptState.STEPPED: + raise RuntimeError("unscale_() is being called after step().") + + # FP32 division can be imprecise for certain compile options, so we carry out the reciprocal in FP64. + assert self._scale is not None + inv_scale = self._scale.to("cpu").double().reciprocal().float().to(self._scale.device) + found_inf = torch.full( + (1,), 0.0, dtype=torch.float32, device=self._scale.device + ) + + optimizer_state["found_inf_per_device"] = self._unscale_grads_( + optimizer, inv_scale, found_inf, False + ) + optimizer_state["stage"] = OptState.UNSCALED + +def update(self, new_scale=None): + """ + Updates the scale factor. + If any optimizer steps were skipped the scale is multiplied by ``backoff_factor`` + to reduce it. If ``growth_interval`` unskipped iterations occurred consecutively, + the scale is multiplied by ``growth_factor`` to increase it. + Passing ``new_scale`` sets the new scale value manually. (``new_scale`` is not + used directly, it's used to fill GradScaler's internal scale tensor. So if + ``new_scale`` was a tensor, later in-place changes to that tensor will not further + affect the scale GradScaler uses internally.) + Args: + new_scale (float or :class:`torch.FloatTensor`, optional, default=None): New scale factor. + .. warning:: + :meth:`update` should only be called at the end of the iteration, after ``scaler.step(optimizer)`` has + been invoked for all optimizers used this iteration. + """ + if not self._enabled: + return + + _scale, _growth_tracker = self._check_scale_growth_tracker("update") + + if new_scale is not None: + # Accept a new user-defined scale. + if isinstance(new_scale, float): + self._scale.fill_(new_scale) # type: ignore[union-attr] + else: + reason = "new_scale should be a float or a 1-element torch.FloatTensor with requires_grad=False." + assert isinstance(new_scale, torch.FloatTensor), reason # type: ignore[attr-defined] + assert new_scale.numel() == 1, reason + assert new_scale.requires_grad is False, reason + self._scale.copy_(new_scale) # type: ignore[union-attr] + else: + # Consume shared inf/nan data collected from optimizers to update the scale. + # If all found_inf tensors are on the same device as self._scale, this operation is asynchronous. + found_infs = [ + found_inf.to(device="cpu", non_blocking=True) + for state in self._per_optimizer_states.values() + for found_inf in state["found_inf_per_device"].values() + ] + + assert len(found_infs) > 0, "No inf checks were recorded prior to update." + + found_inf_combined = found_infs[0] + if len(found_infs) > 1: + for i in range(1, len(found_infs)): + found_inf_combined += found_infs[i] + + to_device = _scale.device + _scale = _scale.to("cpu") + _growth_tracker = _growth_tracker.to("cpu") + + core._amp_update_scale_( + _scale, + _growth_tracker, + found_inf_combined, + self._growth_factor, + self._backoff_factor, + self._growth_interval, + ) + + _scale = _scale.to(to_device) + _growth_tracker = _growth_tracker.to(to_device) + # To prepare for next iteration, clear the data collected from optimizers this iteration. + self._per_optimizer_states = defaultdict(_refresh_per_optimizer_state) + +def gradscaler_init(): + torch.xpu.amp.GradScaler = ipex.cpu.autocast._grad_scaler.GradScaler + torch.xpu.amp.GradScaler._unscale_grads_ = _unscale_grads_ + torch.xpu.amp.GradScaler.unscale_ = unscale_ + torch.xpu.amp.GradScaler.update = update + return torch.xpu.amp.GradScaler \ No newline at end of file diff --git a/comfy/ipex_hijacks/hijacks.py b/comfy/ipex_hijacks/hijacks.py new file mode 100644 index 000000000..3fecc8181 --- /dev/null +++ b/comfy/ipex_hijacks/hijacks.py @@ -0,0 +1,228 @@ +import contextlib +import torch +import intel_extension_for_pytorch as ipex # pylint: disable=import-error, unused-import +from .sd_hijack_utils import CondFunc + +# pylint: disable=protected-access, missing-function-docstring, line-too-long, unnecessary-lambda, no-else-return + +def _shutdown_workers(self): + if torch.utils.data._utils is None or torch.utils.data._utils.python_exit_status is True or torch.utils.data._utils.python_exit_status is None: + return + if hasattr(self, "_shutdown") and not self._shutdown: + self._shutdown = True + try: + if hasattr(self, '_pin_memory_thread'): + self._pin_memory_thread_done_event.set() + self._worker_result_queue.put((None, None)) + self._pin_memory_thread.join() + self._worker_result_queue.cancel_join_thread() + self._worker_result_queue.close() + self._workers_done_event.set() + for worker_id in range(len(self._workers)): + if self._persistent_workers or self._workers_status[worker_id]: + self._mark_worker_as_unavailable(worker_id, shutdown=True) + for w in self._workers: # pylint: disable=invalid-name + w.join(timeout=torch.utils.data._utils.MP_STATUS_CHECK_INTERVAL) + for q in self._index_queues: # pylint: disable=invalid-name + q.cancel_join_thread() + q.close() + finally: + if self._worker_pids_set: + torch.utils.data._utils.signal_handling._remove_worker_pids(id(self)) + self._worker_pids_set = False + for w in self._workers: # pylint: disable=invalid-name + if w.is_alive(): + w.terminate() + +class DummyDataParallel(torch.nn.Module): # pylint: disable=missing-class-docstring, unused-argument, too-few-public-methods + def __new__(cls, module, device_ids=None, output_device=None, dim=0): # pylint: disable=unused-argument + if isinstance(device_ids, list) and len(device_ids) > 1: + print("IPEX backend doesn't support DataParallel on multiple XPU devices") + return module.to("xpu") + +def return_null_context(*args, **kwargs): # pylint: disable=unused-argument + return contextlib.nullcontext() + +def check_device(device): + return bool((isinstance(device, torch.device) and device.type == "cuda") or (isinstance(device, str) and "cuda" in device) or isinstance(device, int)) + +def return_xpu(device): + return f"xpu:{device.split(':')[-1]}" if isinstance(device, str) and ":" in device else f"xpu:{device}" if isinstance(device, int) else torch.device("xpu") if isinstance(device, torch.device) else "xpu" + +def ipex_no_cuda(orig_func, *args, **kwargs): + torch.cuda.is_available = lambda: False + orig_func(*args, **kwargs) + torch.cuda.is_available = torch.xpu.is_available + +original_autocast = torch.autocast +def ipex_autocast(*args, **kwargs): + if len(args) > 0 and args[0] == "cuda" or args[0] == "xpu": + if "dtype" in kwargs: + return original_autocast("xpu", *args[1:], **kwargs) + else: + return original_autocast("xpu", *args[1:], dtype=torch.float16, **kwargs) + else: + return original_autocast(*args, **kwargs) + +# Embedding BF16 +original_torch_cat = torch.cat +def torch_cat(tensor, *args, **kwargs): + if len(tensor) == 3 and (tensor[0].dtype != tensor[1].dtype or tensor[2].dtype != tensor[1].dtype): + return original_torch_cat([tensor[0].to(tensor[1].dtype), tensor[1], tensor[2].to(tensor[1].dtype)], *args, **kwargs) + else: + return original_torch_cat(tensor, *args, **kwargs) + +# Latent antialias: +original_interpolate = torch.nn.functional.interpolate +def interpolate(tensor, size=None, scale_factor=None, mode='nearest', align_corners=None, recompute_scale_factor=None, antialias=False): # pylint: disable=too-many-arguments + if antialias or align_corners is not None: + return_device = tensor.device + return_dtype = tensor.dtype + return original_interpolate(tensor.to("cpu", dtype=torch.float32), size=size, scale_factor=scale_factor, mode=mode, + align_corners=align_corners, recompute_scale_factor=recompute_scale_factor, antialias=antialias).to(return_device, dtype=return_dtype) + else: + return original_interpolate(tensor, size=size, scale_factor=scale_factor, mode=mode, + align_corners=align_corners, recompute_scale_factor=recompute_scale_factor, antialias=antialias) + +original_linalg_solve = torch.linalg.solve +def linalg_solve(A, B, *args, **kwargs): # pylint: disable=invalid-name + if A.device != torch.device("cpu") or B.device != torch.device("cpu"): + return_device = A.device + return original_linalg_solve(A.to("cpu"), B.to("cpu"), *args, **kwargs).to(return_device) + else: + return original_linalg_solve(A, B, *args, **kwargs) + +if torch.xpu.has_fp64_dtype(): + original_torch_bmm = torch.bmm + original_scaled_dot_product_attention = torch.nn.functional.scaled_dot_product_attention +else: + # 64 bit attention workarounds for Alchemist: + try: + from .attention import torch_bmm_32_bit as original_torch_bmm + from .attention import scaled_dot_product_attention_32_bit as original_scaled_dot_product_attention + except Exception: # pylint: disable=broad-exception-caught + original_torch_bmm = torch.bmm + original_scaled_dot_product_attention = torch.nn.functional.scaled_dot_product_attention + +# dtype errors: +def torch_bmm(input, mat2, *, out=None): + if input.dtype != mat2.dtype: + mat2 = mat2.to(input.dtype) + return original_torch_bmm(input, mat2, out=out) + +def scaled_dot_product_attention(query, key, value, attn_mask=None, dropout_p=0.0, is_causal=False): + if query.dtype != key.dtype: + key = key.to(dtype=query.dtype) + if query.dtype != value.dtype: + value = value.to(dtype=query.dtype) + return original_scaled_dot_product_attention(query, key, value, attn_mask=attn_mask, dropout_p=dropout_p, is_causal=is_causal) + +@property +def is_cuda(self): + return self.device.type == 'xpu' + +def ipex_hijacks(): + CondFunc('torch.tensor', + lambda orig_func, *args, device=None, **kwargs: orig_func(*args, device=return_xpu(device), **kwargs), + lambda orig_func, *args, device=None, **kwargs: check_device(device)) + CondFunc('torch.Tensor.to', + lambda orig_func, self, device=None, *args, **kwargs: orig_func(self, return_xpu(device), *args, **kwargs), + lambda orig_func, self, device=None, *args, **kwargs: check_device(device)) + CondFunc('torch.Tensor.cuda', + lambda orig_func, self, device=None, *args, **kwargs: orig_func(self, return_xpu(device), *args, **kwargs), + lambda orig_func, self, device=None, *args, **kwargs: check_device(device)) + CondFunc('torch.UntypedStorage.__init__', + lambda orig_func, *args, device=None, **kwargs: orig_func(*args, device=return_xpu(device), **kwargs), + lambda orig_func, *args, device=None, **kwargs: check_device(device)) + CondFunc('torch.UntypedStorage.cuda', + lambda orig_func, self, device=None, *args, **kwargs: orig_func(self, return_xpu(device), *args, **kwargs), + lambda orig_func, self, device=None, *args, **kwargs: check_device(device)) + CondFunc('torch.empty', + lambda orig_func, *args, device=None, **kwargs: orig_func(*args, device=return_xpu(device), **kwargs), + lambda orig_func, *args, device=None, **kwargs: check_device(device)) + CondFunc('torch.randn', + lambda orig_func, *args, device=None, **kwargs: orig_func(*args, device=return_xpu(device), **kwargs), + lambda orig_func, *args, device=None, **kwargs: check_device(device)) + CondFunc('torch.ones', + lambda orig_func, *args, device=None, **kwargs: orig_func(*args, device=return_xpu(device), **kwargs), + lambda orig_func, *args, device=None, **kwargs: check_device(device)) + CondFunc('torch.zeros', + lambda orig_func, *args, device=None, **kwargs: orig_func(*args, device=return_xpu(device), **kwargs), + lambda orig_func, *args, device=None, **kwargs: check_device(device)) + CondFunc('torch.linspace', + lambda orig_func, *args, device=None, **kwargs: orig_func(*args, device=return_xpu(device), **kwargs), + lambda orig_func, *args, device=None, **kwargs: check_device(device)) + CondFunc('torch.load', + lambda orig_func, f, map_location=None, pickle_module=None, *, weights_only=False, mmap=None, **kwargs: + orig_func(f, map_location=return_xpu(map_location), pickle_module=pickle_module, weights_only=weights_only, mmap=mmap, **kwargs), + lambda orig_func, f, map_location=None, pickle_module=None, *, weights_only=False, mmap=None, **kwargs: check_device(map_location)) + if hasattr(torch.xpu, "Generator"): + CondFunc('torch.Generator', + lambda orig_func, device=None: torch.xpu.Generator(return_xpu(device)), + lambda orig_func, device=None: device is not None and device != torch.device("cpu") and device != "cpu") + else: + CondFunc('torch.Generator', + lambda orig_func, device=None: orig_func(return_xpu(device)), + lambda orig_func, device=None: check_device(device)) + + # TiledVAE and ControlNet: + CondFunc('torch.batch_norm', + lambda orig_func, input, weight, bias, *args, **kwargs: orig_func(input, + weight if weight is not None else torch.ones(input.size()[1], device=input.device), + bias if bias is not None else torch.zeros(input.size()[1], device=input.device), *args, **kwargs), + lambda orig_func, input, *args, **kwargs: input.device != torch.device("cpu")) + CondFunc('torch.instance_norm', + lambda orig_func, input, weight, bias, *args, **kwargs: orig_func(input, + weight if weight is not None else torch.ones(input.size()[1], device=input.device), + bias if bias is not None else torch.zeros(input.size()[1], device=input.device), *args, **kwargs), + lambda orig_func, input, *args, **kwargs: input.device != torch.device("cpu")) + + # Functions with dtype errors: + CondFunc('torch.nn.modules.GroupNorm.forward', + lambda orig_func, self, input: orig_func(self, input.to(self.weight.data.dtype)), + lambda orig_func, self, input: input.dtype != self.weight.data.dtype) + # Training: + CondFunc('torch.nn.modules.linear.Linear.forward', + lambda orig_func, self, input: orig_func(self, input.to(self.weight.data.dtype)), + lambda orig_func, self, input: input.dtype != self.weight.data.dtype) + CondFunc('torch.nn.modules.conv.Conv2d.forward', + lambda orig_func, self, input: orig_func(self, input.to(self.weight.data.dtype)), + lambda orig_func, self, input: input.dtype != self.weight.data.dtype) + # BF16: + CondFunc('torch.nn.functional.layer_norm', + lambda orig_func, input, normalized_shape=None, weight=None, *args, **kwargs: + orig_func(input.to(weight.data.dtype), normalized_shape, weight, *args, **kwargs), + lambda orig_func, input, normalized_shape=None, weight=None, *args, **kwargs: + weight is not None and input.dtype != weight.data.dtype) + # SwinIR BF16: + CondFunc('torch.nn.functional.pad', + lambda orig_func, input, pad, mode='constant', value=None: orig_func(input.to(torch.float32), pad, mode=mode, value=value).to(dtype=torch.bfloat16), + lambda orig_func, input, pad, mode='constant', value=None: mode == 'reflect' and input.dtype == torch.bfloat16) + + # Diffusers Float64 (Alchemist GPUs doesn't support 64 bit): + if not torch.xpu.has_fp64_dtype(): + CondFunc('torch.from_numpy', + lambda orig_func, ndarray: orig_func(ndarray.astype('float32')), + lambda orig_func, ndarray: ndarray.dtype == float) + + # Broken functions when torch.cuda.is_available is True: + # Pin Memory: + CondFunc('torch.utils.data.dataloader._BaseDataLoaderIter.__init__', + lambda orig_func, *args, **kwargs: ipex_no_cuda(orig_func, *args, **kwargs), + lambda orig_func, *args, **kwargs: True) + + # Functions that make compile mad with CondFunc: + torch.nn.DataParallel = DummyDataParallel + torch.utils.data.dataloader._MultiProcessingDataLoaderIter._shutdown_workers = _shutdown_workers + + torch.autocast = ipex_autocast + torch.backends.cuda.sdp_kernel = return_null_context + torch.UntypedStorage.is_cuda = is_cuda + + torch.nn.functional.interpolate = interpolate + if hasattr(torch.xpu, "Generator"): + torch.linalg.solve = linalg_solve + + torch.bmm = torch_bmm + torch.cat = torch_cat + torch.nn.functional.scaled_dot_product_attention = scaled_dot_product_attention \ No newline at end of file diff --git a/comfy/ipex_hijacks/sd_hijack_utils.py b/comfy/ipex_hijacks/sd_hijack_utils.py new file mode 100644 index 000000000..b1edf9fb0 --- /dev/null +++ b/comfy/ipex_hijacks/sd_hijack_utils.py @@ -0,0 +1,28 @@ +import importlib + +class CondFunc: + def __new__(cls, orig_func, sub_func, cond_func): + self = super(CondFunc, cls).__new__(cls) + if isinstance(orig_func, str): + func_path = orig_func.split('.') + for i in range(len(func_path)-1, -1, -1): + try: + resolved_obj = importlib.import_module('.'.join(func_path[:i])) + break + except ImportError: + pass + for attr_name in func_path[i:-1]: + resolved_obj = getattr(resolved_obj, attr_name) + orig_func = getattr(resolved_obj, func_path[-1]) + setattr(resolved_obj, func_path[-1], lambda *args, **kwargs: self(*args, **kwargs)) + self.__init__(orig_func, sub_func, cond_func) + return lambda *args, **kwargs: self(*args, **kwargs) + def __init__(self, orig_func, sub_func, cond_func): + self.__orig_func = orig_func + self.__sub_func = sub_func + self.__cond_func = cond_func + def __call__(self, *args, **kwargs): + if not self.__cond_func or self.__cond_func(self.__orig_func, *args, **kwargs): + return self.__sub_func(self.__orig_func, *args, **kwargs) + else: + return self.__orig_func(*args, **kwargs) \ No newline at end of file diff --git a/comfy/ipex_hijacks/stable_diffusion_ipex.py b/comfy/ipex_hijacks/stable_diffusion_ipex.py new file mode 100644 index 000000000..ee9461470 --- /dev/null +++ b/comfy/ipex_hijacks/stable_diffusion_ipex.py @@ -0,0 +1,841 @@ +# Copyright 2023 The HuggingFace Team. All rights reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import inspect +from typing import Any, Callable, Dict, List, Optional, Union + +import intel_extension_for_pytorch as ipex +import torch +from packaging import version +from transformers import CLIPFeatureExtractor, CLIPTextModel, CLIPTokenizer + +from diffusers.configuration_utils import FrozenDict +from diffusers.loaders import LoraLoaderMixin, TextualInversionLoaderMixin +from diffusers.models import AutoencoderKL, UNet2DConditionModel +from diffusers.pipelines.pipeline_utils import DiffusionPipeline +from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput +from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker +from diffusers.schedulers import KarrasDiffusionSchedulers +from diffusers.utils import ( + deprecate, + is_accelerate_available, + is_accelerate_version, + logging, + replace_example_docstring, +) +from diffusers.utils.torch_utils import randn_tensor + + +logger = logging.get_logger(__name__) # pylint: disable=invalid-name + +EXAMPLE_DOC_STRING = """ + Examples: + ```py + >>> import torch + >>> from diffusers import StableDiffusionPipeline + >>> pipe = DiffusionPipeline.from_pretrained("runwayml/stable-diffusion-v1-5", custom_pipeline="stable_diffusion_ipex") + >>> # For Float32 + >>> pipe.prepare_for_ipex(prompt, dtype=torch.float32, height=512, width=512) #value of image height/width should be consistent with the pipeline inference + >>> # For BFloat16 + >>> pipe.prepare_for_ipex(prompt, dtype=torch.bfloat16, height=512, width=512) #value of image height/width should be consistent with the pipeline inference + >>> prompt = "a photo of an astronaut riding a horse on mars" + >>> # For Float32 + >>> image = pipe(prompt, num_inference_steps=num_inference_steps, height=512, width=512).images[0] #value of image height/width should be consistent with 'prepare_for_ipex()' + >>> # For BFloat16 + >>> with torch.cpu.amp.autocast(enabled=True, dtype=torch.bfloat16): + >>> image = pipe(prompt, num_inference_steps=num_inference_steps, height=512, width=512).images[0] #value of image height/width should be consistent with 'prepare_for_ipex()' + ``` +""" + + +class StableDiffusionIPEXPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraLoaderMixin): + r""" + Pipeline for text-to-image generation using Stable Diffusion on IPEX. + This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods the + library implements for all the pipelines (such as downloading or saving, running on a particular device, etc.) + Args: + vae ([`AutoencoderKL`]): + Variational Auto-Encoder (VAE) Model to encode and decode images to and from latent representations. + text_encoder ([`CLIPTextModel`]): + Frozen text-encoder. Stable Diffusion uses the text portion of + [CLIP](https://huggingface.co/docs/transformers/model_doc/clip#transformers.CLIPTextModel), specifically + the [clip-vit-large-patch14](https://huggingface.co/openai/clip-vit-large-patch14) variant. + tokenizer (`CLIPTokenizer`): + Tokenizer of class + [CLIPTokenizer](https://huggingface.co/docs/transformers/v4.21.0/en/model_doc/clip#transformers.CLIPTokenizer). + unet ([`UNet2DConditionModel`]): Conditional U-Net architecture to denoise the encoded image latents. + scheduler ([`SchedulerMixin`]): + A scheduler to be used in combination with `unet` to denoise the encoded image latents. Can be one of + [`DDIMScheduler`], [`LMSDiscreteScheduler`], or [`PNDMScheduler`]. + safety_checker ([`StableDiffusionSafetyChecker`]): + Classification module that estimates whether generated images could be considered offensive or harmful. + Please, refer to the [model card](https://huggingface.co/runwayml/stable-diffusion-v1-5) for details. + feature_extractor ([`CLIPFeatureExtractor`]): + Model that extracts features from generated images to be used as inputs for the `safety_checker`. + """ + + _optional_components = ["safety_checker", "feature_extractor"] + + def __init__( + self, + vae: AutoencoderKL, + text_encoder: CLIPTextModel, + tokenizer: CLIPTokenizer, + unet: UNet2DConditionModel, + scheduler: KarrasDiffusionSchedulers, + safety_checker: StableDiffusionSafetyChecker, + feature_extractor: CLIPFeatureExtractor, + requires_safety_checker: bool = True, + ): + super().__init__() + + if hasattr(scheduler.config, "steps_offset") and scheduler.config.steps_offset != 1: + deprecation_message = ( + f"The configuration file of this scheduler: {scheduler} is outdated. `steps_offset`" + f" should be set to 1 instead of {scheduler.config.steps_offset}. Please make sure " + "to update the config accordingly as leaving `steps_offset` might led to incorrect results" + " in future versions. If you have downloaded this checkpoint from the Hugging Face Hub," + " it would be very nice if you could open a Pull request for the `scheduler/scheduler_config.json`" + " file" + ) + deprecate("steps_offset!=1", "1.0.0", deprecation_message, standard_warn=False) + new_config = dict(scheduler.config) + new_config["steps_offset"] = 1 + scheduler._internal_dict = FrozenDict(new_config) + + if hasattr(scheduler.config, "clip_sample") and scheduler.config.clip_sample is True: + deprecation_message = ( + f"The configuration file of this scheduler: {scheduler} has not set the configuration `clip_sample`." + " `clip_sample` should be set to False in the configuration file. Please make sure to update the" + " config accordingly as not setting `clip_sample` in the config might lead to incorrect results in" + " future versions. If you have downloaded this checkpoint from the Hugging Face Hub, it would be very" + " nice if you could open a Pull request for the `scheduler/scheduler_config.json` file" + ) + deprecate("clip_sample not set", "1.0.0", deprecation_message, standard_warn=False) + new_config = dict(scheduler.config) + new_config["clip_sample"] = False + scheduler._internal_dict = FrozenDict(new_config) + + if safety_checker is None and requires_safety_checker: + logger.warning( + f"You have disabled the safety checker for {self.__class__} by passing `safety_checker=None`. Ensure" + " that you abide to the conditions of the Stable Diffusion license and do not expose unfiltered" + " results in services or applications open to the public. Both the diffusers team and Hugging Face" + " strongly recommend to keep the safety filter enabled in all public facing circumstances, disabling" + " it only for use-cases that involve analyzing network behavior or auditing its results. For more" + " information, please have a look at https://github.com/huggingface/diffusers/pull/254 ." + ) + + if safety_checker is not None and feature_extractor is None: + raise ValueError( + "Make sure to define a feature extractor when loading {self.__class__} if you want to use the safety" + " checker. If you do not want to use the safety checker, you can pass `'safety_checker=None'` instead." + ) + + is_unet_version_less_0_9_0 = hasattr(unet.config, "_diffusers_version") and version.parse( + version.parse(unet.config._diffusers_version).base_version + ) < version.parse("0.9.0.dev0") + is_unet_sample_size_less_64 = hasattr(unet.config, "sample_size") and unet.config.sample_size < 64 + if is_unet_version_less_0_9_0 and is_unet_sample_size_less_64: + deprecation_message = ( + "The configuration file of the unet has set the default `sample_size` to smaller than" + " 64 which seems highly unlikely. If your checkpoint is a fine-tuned version of any of the" + " following: \n- CompVis/stable-diffusion-v1-4 \n- CompVis/stable-diffusion-v1-3 \n-" + " CompVis/stable-diffusion-v1-2 \n- CompVis/stable-diffusion-v1-1 \n- runwayml/stable-diffusion-v1-5" + " \n- runwayml/stable-diffusion-inpainting \n you should change 'sample_size' to 64 in the" + " configuration file. Please make sure to update the config accordingly as leaving `sample_size=32`" + " in the config might lead to incorrect results in future versions. If you have downloaded this" + " checkpoint from the Hugging Face Hub, it would be very nice if you could open a Pull request for" + " the `unet/config.json` file" + ) + deprecate("sample_size<64", "1.0.0", deprecation_message, standard_warn=False) + new_config = dict(unet.config) + new_config["sample_size"] = 64 + unet._internal_dict = FrozenDict(new_config) + + self.register_modules( + vae=vae, + text_encoder=text_encoder, + tokenizer=tokenizer, + unet=unet, + scheduler=scheduler, + safety_checker=safety_checker, + feature_extractor=feature_extractor, + ) + self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) + self.register_to_config(requires_safety_checker=requires_safety_checker) + + def get_input_example(self, prompt, height=None, width=None, guidance_scale=7.5, num_images_per_prompt=1): + prompt_embeds = None + negative_prompt_embeds = None + negative_prompt = None + callback_steps = 1 + generator = None + latents = None + + # 0. Default height and width to unet + height = height or self.unet.config.sample_size * self.vae_scale_factor + width = width or self.unet.config.sample_size * self.vae_scale_factor + + # 1. Check inputs. Raise error if not correct + self.check_inputs( + prompt, height, width, callback_steps, negative_prompt, prompt_embeds, negative_prompt_embeds + ) + + # 2. Define call parameters + if prompt is not None and isinstance(prompt, str): + batch_size = 1 + elif prompt is not None and isinstance(prompt, list): + batch_size = len(prompt) + + device = "cpu" + # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2) + # of the Imagen paper: https://arxiv.org/pdf/2205.11487.pdf . `guidance_scale = 1` + # corresponds to doing no classifier free guidance. + do_classifier_free_guidance = guidance_scale > 1.0 + + # 3. Encode input prompt + prompt_embeds = self._encode_prompt( + prompt, + device, + num_images_per_prompt, + do_classifier_free_guidance, + negative_prompt, + prompt_embeds=prompt_embeds, + negative_prompt_embeds=negative_prompt_embeds, + ) + + # 5. Prepare latent variables + latents = self.prepare_latents( + batch_size * num_images_per_prompt, + self.unet.in_channels, + height, + width, + prompt_embeds.dtype, + device, + generator, + latents, + ) + dummy = torch.ones(1, dtype=torch.int32) + latent_model_input = torch.cat([latents] * 2) if do_classifier_free_guidance else latents + latent_model_input = self.scheduler.scale_model_input(latent_model_input, dummy) + + unet_input_example = (latent_model_input, dummy, prompt_embeds) + vae_decoder_input_example = latents + + return unet_input_example, vae_decoder_input_example + + def prepare_for_ipex(self, promt, dtype=torch.float32, height=None, width=None, guidance_scale=7.5): + self.unet = self.unet.to(memory_format=torch.channels_last) + self.vae.decoder = self.vae.decoder.to(memory_format=torch.channels_last) + self.text_encoder = self.text_encoder.to(memory_format=torch.channels_last) + if self.safety_checker is not None: + self.safety_checker = self.safety_checker.to(memory_format=torch.channels_last) + + unet_input_example, vae_decoder_input_example = self.get_input_example(promt, height, width, guidance_scale) + + # optimize with ipex + if dtype == torch.bfloat16: + self.unet = ipex.optimize(self.unet.eval(), dtype=torch.bfloat16, inplace=True) + self.vae.decoder = ipex.optimize(self.vae.decoder.eval(), dtype=torch.bfloat16, inplace=True) + self.text_encoder = ipex.optimize(self.text_encoder.eval(), dtype=torch.bfloat16, inplace=True) + if self.safety_checker is not None: + self.safety_checker = ipex.optimize(self.safety_checker.eval(), dtype=torch.bfloat16, inplace=True) + elif dtype == torch.float32: + self.unet = ipex.optimize( + self.unet.eval(), + dtype=torch.float32, + inplace=True, + weights_prepack=True, + auto_kernel_selection=False, + ) + self.vae.decoder = ipex.optimize( + self.vae.decoder.eval(), + dtype=torch.float32, + inplace=True, + weights_prepack=True, + auto_kernel_selection=False, + ) + self.text_encoder = ipex.optimize( + self.text_encoder.eval(), + dtype=torch.float32, + inplace=True, + weights_prepack=True, + auto_kernel_selection=False, + ) + if self.safety_checker is not None: + self.safety_checker = ipex.optimize( + self.safety_checker.eval(), + dtype=torch.float32, + inplace=True, + weights_prepack=True, + auto_kernel_selection=False, + ) + else: + raise ValueError(" The value of 'dtype' should be 'torch.bfloat16' or 'torch.float32' !") + + # trace unet model to get better performance on IPEX + with torch.cpu.amp.autocast(enabled=dtype == torch.bfloat16), torch.no_grad(): + unet_trace_model = torch.jit.trace(self.unet, unet_input_example, check_trace=False, strict=False) + unet_trace_model = torch.jit.freeze(unet_trace_model) + self.unet.forward = unet_trace_model.forward + + # trace vae.decoder model to get better performance on IPEX + with torch.cpu.amp.autocast(enabled=dtype == torch.bfloat16), torch.no_grad(): + ave_decoder_trace_model = torch.jit.trace( + self.vae.decoder, vae_decoder_input_example, check_trace=False, strict=False + ) + ave_decoder_trace_model = torch.jit.freeze(ave_decoder_trace_model) + self.vae.decoder.forward = ave_decoder_trace_model.forward + + def enable_vae_slicing(self): + r""" + Enable sliced VAE decoding. + When this option is enabled, the VAE will split the input tensor in slices to compute decoding in several + steps. This is useful to save some memory and allow larger batch sizes. + """ + self.vae.enable_slicing() + + def disable_vae_slicing(self): + r""" + Disable sliced VAE decoding. If `enable_vae_slicing` was previously invoked, this method will go back to + computing decoding in one step. + """ + self.vae.disable_slicing() + + def enable_vae_tiling(self): + r""" + Enable tiled VAE decoding. + When this option is enabled, the VAE will split the input tensor into tiles to compute decoding and encoding in + several steps. This is useful to save a large amount of memory and to allow the processing of larger images. + """ + self.vae.enable_tiling() + + def disable_vae_tiling(self): + r""" + Disable tiled VAE decoding. If `enable_vae_tiling` was previously invoked, this method will go back to + computing decoding in one step. + """ + self.vae.disable_tiling() + + def enable_sequential_cpu_offload(self, gpu_id=0): + r""" + Offloads all models to CPU using accelerate, significantly reducing memory usage. When called, unet, + text_encoder, vae and safety checker have their state dicts saved to CPU and then are moved to a + `torch.device('meta') and loaded to GPU only when their specific submodule has its `forward` method called. + Note that offloading happens on a submodule basis. Memory savings are higher than with + `enable_model_cpu_offload`, but performance is lower. + """ + if is_accelerate_available() and is_accelerate_version(">=", "0.14.0"): + from accelerate import cpu_offload + else: + raise ImportError("`enable_sequential_cpu_offload` requires `accelerate v0.14.0` or higher") + + device = torch.device(f"cuda:{gpu_id}") + + if self.device.type != "cpu": + self.to("cpu", silence_dtype_warnings=True) + torch.cuda.empty_cache() # otherwise we don't see the memory savings (but they probably exist) + + for cpu_offloaded_model in [self.unet, self.text_encoder, self.vae]: + cpu_offload(cpu_offloaded_model, device) + + if self.safety_checker is not None: + cpu_offload(self.safety_checker, execution_device=device, offload_buffers=True) + + def enable_model_cpu_offload(self, gpu_id=0): + r""" + Offloads all models to CPU using accelerate, reducing memory usage with a low impact on performance. Compared + to `enable_sequential_cpu_offload`, this method moves one whole model at a time to the GPU when its `forward` + method is called, and the model remains in GPU until the next model runs. Memory savings are lower than with + `enable_sequential_cpu_offload`, but performance is much better due to the iterative execution of the `unet`. + """ + if is_accelerate_available() and is_accelerate_version(">=", "0.17.0.dev0"): + from accelerate import cpu_offload_with_hook + else: + raise ImportError("`enable_model_offload` requires `accelerate v0.17.0` or higher.") + + device = torch.device(f"cuda:{gpu_id}") + + if self.device.type != "cpu": + self.to("cpu", silence_dtype_warnings=True) + torch.cuda.empty_cache() # otherwise we don't see the memory savings (but they probably exist) + + hook = None + for cpu_offloaded_model in [self.text_encoder, self.unet, self.vae]: + _, hook = cpu_offload_with_hook(cpu_offloaded_model, device, prev_module_hook=hook) + + if self.safety_checker is not None: + _, hook = cpu_offload_with_hook(self.safety_checker, device, prev_module_hook=hook) + + # We'll offload the last model manually. + self.final_offload_hook = hook + + @property + def _execution_device(self): + r""" + Returns the device on which the pipeline's models will be executed. After calling + `pipeline.enable_sequential_cpu_offload()` the execution device can only be inferred from Accelerate's module + hooks. + """ + if not hasattr(self.unet, "_hf_hook"): + return self.device + for module in self.unet.modules(): + if ( + hasattr(module, "_hf_hook") + and hasattr(module._hf_hook, "execution_device") + and module._hf_hook.execution_device is not None + ): + return torch.device(module._hf_hook.execution_device) + return self.device + + def _encode_prompt( + self, + prompt, + device, + num_images_per_prompt, + do_classifier_free_guidance, + negative_prompt=None, + prompt_embeds: Optional[torch.FloatTensor] = None, + negative_prompt_embeds: Optional[torch.FloatTensor] = None, + ): + r""" + Encodes the prompt into text encoder hidden states. + Args: + prompt (`str` or `List[str]`, *optional*): + prompt to be encoded + device: (`torch.device`): + torch device + num_images_per_prompt (`int`): + number of images that should be generated per prompt + do_classifier_free_guidance (`bool`): + whether to use classifier free guidance or not + negative_prompt (`str` or `List[str]`, *optional*): + The prompt or prompts not to guide the image generation. If not defined, one has to pass + `negative_prompt_embeds`. instead. If not defined, one has to pass `negative_prompt_embeds`. instead. + Ignored when not using guidance (i.e., ignored if `guidance_scale` is less than `1`). + prompt_embeds (`torch.FloatTensor`, *optional*): + Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not + provided, text embeddings will be generated from `prompt` input argument. + negative_prompt_embeds (`torch.FloatTensor`, *optional*): + Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt + weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input + argument. + """ + if prompt is not None and isinstance(prompt, str): + batch_size = 1 + elif prompt is not None and isinstance(prompt, list): + batch_size = len(prompt) + else: + batch_size = prompt_embeds.shape[0] + + if prompt_embeds is None: + # textual inversion: procecss multi-vector tokens if necessary + if isinstance(self, TextualInversionLoaderMixin): + prompt = self.maybe_convert_prompt(prompt, self.tokenizer) + + text_inputs = self.tokenizer( + prompt, + padding="max_length", + max_length=self.tokenizer.model_max_length, + truncation=True, + return_tensors="pt", + ) + text_input_ids = text_inputs.input_ids + untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids + + if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal( + text_input_ids, untruncated_ids + ): + removed_text = self.tokenizer.batch_decode( + untruncated_ids[:, self.tokenizer.model_max_length - 1 : -1] + ) + logger.warning( + "The following part of your input was truncated because CLIP can only handle sequences up to" + f" {self.tokenizer.model_max_length} tokens: {removed_text}" + ) + + if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask: + attention_mask = text_inputs.attention_mask.to(device) + else: + attention_mask = None + + prompt_embeds = self.text_encoder( + text_input_ids.to(device), + attention_mask=attention_mask, + ) + prompt_embeds = prompt_embeds[0] + + prompt_embeds = prompt_embeds.to(dtype=self.text_encoder.dtype, device=device) + + bs_embed, seq_len, _ = prompt_embeds.shape + # duplicate text embeddings for each generation per prompt, using mps friendly method + prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1) + prompt_embeds = prompt_embeds.view(bs_embed * num_images_per_prompt, seq_len, -1) + + # get unconditional embeddings for classifier free guidance + if do_classifier_free_guidance and negative_prompt_embeds is None: + uncond_tokens: List[str] + if negative_prompt is None: + uncond_tokens = [""] * batch_size + elif type(prompt) is not type(negative_prompt): + raise TypeError( + f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !=" + f" {type(prompt)}." + ) + elif isinstance(negative_prompt, str): + uncond_tokens = [negative_prompt] + elif batch_size != len(negative_prompt): + raise ValueError( + f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:" + f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches" + " the batch size of `prompt`." + ) + else: + uncond_tokens = negative_prompt + + # textual inversion: procecss multi-vector tokens if necessary + if isinstance(self, TextualInversionLoaderMixin): + uncond_tokens = self.maybe_convert_prompt(uncond_tokens, self.tokenizer) + + max_length = prompt_embeds.shape[1] + uncond_input = self.tokenizer( + uncond_tokens, + padding="max_length", + max_length=max_length, + truncation=True, + return_tensors="pt", + ) + + if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask: + attention_mask = uncond_input.attention_mask.to(device) + else: + attention_mask = None + + negative_prompt_embeds = self.text_encoder( + uncond_input.input_ids.to(device), + attention_mask=attention_mask, + ) + negative_prompt_embeds = negative_prompt_embeds[0] + + if do_classifier_free_guidance: + # duplicate unconditional embeddings for each generation per prompt, using mps friendly method + seq_len = negative_prompt_embeds.shape[1] + + negative_prompt_embeds = negative_prompt_embeds.to(dtype=self.text_encoder.dtype, device=device) + + negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1) + negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1) + + # For classifier free guidance, we need to do two forward passes. + # Here we concatenate the unconditional and text embeddings into a single batch + # to avoid doing two forward passes + prompt_embeds = torch.cat([negative_prompt_embeds, prompt_embeds]) + + return prompt_embeds + + def run_safety_checker(self, image, device, dtype): + if self.safety_checker is not None: + safety_checker_input = self.feature_extractor(self.numpy_to_pil(image), return_tensors="pt").to(device) + image, has_nsfw_concept = self.safety_checker( + images=image, clip_input=safety_checker_input.pixel_values.to(dtype) + ) + else: + has_nsfw_concept = None + return image, has_nsfw_concept + + def decode_latents(self, latents): + latents = 1 / self.vae.config.scaling_factor * latents + image = self.vae.decode(latents).sample + image = (image / 2 + 0.5).clamp(0, 1) + # we always cast to float32 as this does not cause significant overhead and is compatible with bfloat16 + image = image.cpu().permute(0, 2, 3, 1).float().numpy() + return image + + def prepare_extra_step_kwargs(self, generator, eta): + # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature + # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers. + # eta corresponds to η in DDIM paper: https://arxiv.org/abs/2010.02502 + # and should be between [0, 1] + + accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys()) + extra_step_kwargs = {} + if accepts_eta: + extra_step_kwargs["eta"] = eta + + # check if the scheduler accepts generator + accepts_generator = "generator" in set(inspect.signature(self.scheduler.step).parameters.keys()) + if accepts_generator: + extra_step_kwargs["generator"] = generator + return extra_step_kwargs + + def check_inputs( + self, + prompt, + height, + width, + callback_steps, + negative_prompt=None, + prompt_embeds=None, + negative_prompt_embeds=None, + ): + if height % 8 != 0 or width % 8 != 0: + raise ValueError(f"`height` and `width` have to be divisible by 8 but are {height} and {width}.") + + if (callback_steps is None) or ( + callback_steps is not None and (not isinstance(callback_steps, int) or callback_steps <= 0) + ): + raise ValueError( + f"`callback_steps` has to be a positive integer but is {callback_steps} of type" + f" {type(callback_steps)}." + ) + + if prompt is not None and prompt_embeds is not None: + raise ValueError( + f"Cannot forward both `prompt`: {prompt} and `prompt_embeds`: {prompt_embeds}. Please make sure to" + " only forward one of the two." + ) + elif prompt is None and prompt_embeds is None: + raise ValueError( + "Provide either `prompt` or `prompt_embeds`. Cannot leave both `prompt` and `prompt_embeds` undefined." + ) + elif prompt is not None and (not isinstance(prompt, str) and not isinstance(prompt, list)): + raise ValueError(f"`prompt` has to be of type `str` or `list` but is {type(prompt)}") + + if negative_prompt is not None and negative_prompt_embeds is not None: + raise ValueError( + f"Cannot forward both `negative_prompt`: {negative_prompt} and `negative_prompt_embeds`:" + f" {negative_prompt_embeds}. Please make sure to only forward one of the two." + ) + + if prompt_embeds is not None and negative_prompt_embeds is not None: + if prompt_embeds.shape != negative_prompt_embeds.shape: + raise ValueError( + "`prompt_embeds` and `negative_prompt_embeds` must have the same shape when passed directly, but" + f" got: `prompt_embeds` {prompt_embeds.shape} != `negative_prompt_embeds`" + f" {negative_prompt_embeds.shape}." + ) + + def prepare_latents(self, batch_size, num_channels_latents, height, width, dtype, device, generator, latents=None): + shape = (batch_size, num_channels_latents, height // self.vae_scale_factor, width // self.vae_scale_factor) + if isinstance(generator, list) and len(generator) != batch_size: + raise ValueError( + f"You have passed a list of generators of length {len(generator)}, but requested an effective batch" + f" size of {batch_size}. Make sure the batch size matches the length of the generators." + ) + + if latents is None: + latents = randn_tensor(shape, generator=generator, device=device, dtype=dtype) + else: + latents = latents.to(device) + + # scale the initial noise by the standard deviation required by the scheduler + latents = latents * self.scheduler.init_noise_sigma + return latents + + @torch.no_grad() + @replace_example_docstring(EXAMPLE_DOC_STRING) + def __call__( + self, + prompt: Union[str, List[str]] = None, + height: Optional[int] = None, + width: Optional[int] = None, + num_inference_steps: int = 50, + guidance_scale: float = 7.5, + negative_prompt: Optional[Union[str, List[str]]] = None, + num_images_per_prompt: Optional[int] = 1, + eta: float = 0.0, + generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None, + latents: Optional[torch.FloatTensor] = None, + prompt_embeds: Optional[torch.FloatTensor] = None, + negative_prompt_embeds: Optional[torch.FloatTensor] = None, + output_type: Optional[str] = "pil", + return_dict: bool = True, + callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None, + callback_steps: int = 1, + cross_attention_kwargs: Optional[Dict[str, Any]] = None, + ): + r""" + Function invoked when calling the pipeline for generation. + Args: + prompt (`str` or `List[str]`, *optional*): + The prompt or prompts to guide the image generation. If not defined, one has to pass `prompt_embeds`. + instead. + height (`int`, *optional*, defaults to self.unet.config.sample_size * self.vae_scale_factor): + The height in pixels of the generated image. + width (`int`, *optional*, defaults to self.unet.config.sample_size * self.vae_scale_factor): + The width in pixels of the generated image. + num_inference_steps (`int`, *optional*, defaults to 50): + The number of denoising steps. More denoising steps usually lead to a higher quality image at the + expense of slower inference. + guidance_scale (`float`, *optional*, defaults to 7.5): + Guidance scale as defined in [Classifier-Free Diffusion Guidance](https://arxiv.org/abs/2207.12598). + `guidance_scale` is defined as `w` of equation 2. of [Imagen + Paper](https://arxiv.org/pdf/2205.11487.pdf). Guidance scale is enabled by setting `guidance_scale > + 1`. Higher guidance scale encourages to generate images that are closely linked to the text `prompt`, + usually at the expense of lower image quality. + negative_prompt (`str` or `List[str]`, *optional*): + The prompt or prompts not to guide the image generation. If not defined, one has to pass + `negative_prompt_embeds`. instead. If not defined, one has to pass `negative_prompt_embeds`. instead. + Ignored when not using guidance (i.e., ignored if `guidance_scale` is less than `1`). + num_images_per_prompt (`int`, *optional*, defaults to 1): + The number of images to generate per prompt. + eta (`float`, *optional*, defaults to 0.0): + Corresponds to parameter eta (η) in the DDIM paper: https://arxiv.org/abs/2010.02502. Only applies to + [`schedulers.DDIMScheduler`], will be ignored for others. + generator (`torch.Generator` or `List[torch.Generator]`, *optional*): + One or a list of [torch generator(s)](https://pytorch.org/docs/stable/generated/torch.Generator.html) + to make generation deterministic. + latents (`torch.FloatTensor`, *optional*): + Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image + generation. Can be used to tweak the same generation with different prompts. If not provided, a latents + tensor will ge generated by sampling using the supplied random `generator`. + prompt_embeds (`torch.FloatTensor`, *optional*): + Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not + provided, text embeddings will be generated from `prompt` input argument. + negative_prompt_embeds (`torch.FloatTensor`, *optional*): + Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt + weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input + argument. + output_type (`str`, *optional*, defaults to `"pil"`): + The output format of the generate image. Choose between + [PIL](https://pillow.readthedocs.io/en/stable/): `PIL.Image.Image` or `np.array`. + return_dict (`bool`, *optional*, defaults to `True`): + Whether or not to return a [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] instead of a + plain tuple. + callback (`Callable`, *optional*): + A function that will be called every `callback_steps` steps during inference. The function will be + called with the following arguments: `callback(step: int, timestep: int, latents: torch.FloatTensor)`. + callback_steps (`int`, *optional*, defaults to 1): + The frequency at which the `callback` function will be called. If not specified, the callback will be + called at every step. + cross_attention_kwargs (`dict`, *optional*): + A kwargs dictionary that if specified is passed along to the `AttnProcessor` as defined under + `self.processor` in + [diffusers.models.attention_processor](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py). + Examples: + Returns: + [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] or `tuple`: + [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] if `return_dict` is True, otherwise a `tuple. + When returning a tuple, the first element is a list with the generated images, and the second element is a + list of `bool`s denoting whether the corresponding generated image likely represents "not-safe-for-work" + (nsfw) content, according to the `safety_checker`. + """ + # 0. Default height and width to unet + height = height or self.unet.config.sample_size * self.vae_scale_factor + width = width or self.unet.config.sample_size * self.vae_scale_factor + + # 1. Check inputs. Raise error if not correct + self.check_inputs( + prompt, height, width, callback_steps, negative_prompt, prompt_embeds, negative_prompt_embeds + ) + + # 2. Define call parameters + if prompt is not None and isinstance(prompt, str): + batch_size = 1 + elif prompt is not None and isinstance(prompt, list): + batch_size = len(prompt) + else: + batch_size = prompt_embeds.shape[0] + + device = self._execution_device + # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2) + # of the Imagen paper: https://arxiv.org/pdf/2205.11487.pdf . `guidance_scale = 1` + # corresponds to doing no classifier free guidance. + do_classifier_free_guidance = guidance_scale > 1.0 + + # 3. Encode input prompt + prompt_embeds = self._encode_prompt( + prompt, + device, + num_images_per_prompt, + do_classifier_free_guidance, + negative_prompt, + prompt_embeds=prompt_embeds, + negative_prompt_embeds=negative_prompt_embeds, + ) + + # 4. Prepare timesteps + self.scheduler.set_timesteps(num_inference_steps, device=device) + timesteps = self.scheduler.timesteps + + # 5. Prepare latent variables + num_channels_latents = self.unet.in_channels + latents = self.prepare_latents( + batch_size * num_images_per_prompt, + num_channels_latents, + height, + width, + prompt_embeds.dtype, + device, + generator, + latents, + ) + + # 6. Prepare extra step kwargs. TODO: Logic should ideally just be moved out of the pipeline + extra_step_kwargs = self.prepare_extra_step_kwargs(generator, eta) + + # 7. Denoising loop + num_warmup_steps = len(timesteps) - num_inference_steps * self.scheduler.order + with self.progress_bar(total=num_inference_steps) as progress_bar: + for i, t in enumerate(timesteps): + # expand the latents if we are doing classifier free guidance + latent_model_input = torch.cat([latents] * 2) if do_classifier_free_guidance else latents + latent_model_input = self.scheduler.scale_model_input(latent_model_input, t) + + # predict the noise residual + noise_pred = self.unet(latent_model_input, t, encoder_hidden_states=prompt_embeds)["sample"] + + # perform guidance + if do_classifier_free_guidance: + noise_pred_uncond, noise_pred_text = noise_pred.chunk(2) + noise_pred = noise_pred_uncond + guidance_scale * (noise_pred_text - noise_pred_uncond) + + # compute the previous noisy sample x_t -> x_t-1 + latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs).prev_sample + + # call the callback, if provided + if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0): + progress_bar.update() + if callback is not None and i % callback_steps == 0: + step_idx = i // getattr(self.scheduler, "order", 1) + callback(step_idx, t, latents) + + if output_type == "latent": + image = latents + has_nsfw_concept = None + elif output_type == "pil": + # 8. Post-processing + image = self.decode_latents(latents) + + # 9. Run safety checker + image, has_nsfw_concept = self.run_safety_checker(image, device, prompt_embeds.dtype) + + # 10. Convert to PIL + image = self.numpy_to_pil(image) + else: + # 8. Post-processing + image = self.decode_latents(latents) + + # 9. Run safety checker + image, has_nsfw_concept = self.run_safety_checker(image, device, prompt_embeds.dtype) + + # Offload last model to CPU + if hasattr(self, "final_offload_hook") and self.final_offload_hook is not None: + self.final_offload_hook.offload() + + if not return_dict: + return (image, has_nsfw_concept) + + return StableDiffusionPipelineOutput(images=image, nsfw_content_detected=has_nsfw_concept) \ No newline at end of file diff --git a/requirements-ipex.txt b/requirements-ipex.txt new file mode 100644 index 000000000..74cc83a17 --- /dev/null +++ b/requirements-ipex.txt @@ -0,0 +1,13 @@ +torch @ https://github.com/Nuullll/intel-extension-for-pytorch/releases/download/v2.1.10%2Bxpu/torch-2.1.0a0+cxx11.abi-cp310-cp310-win_amd64.whl +torchsde +torchvision @ https://github.com/Nuullll/intel-extension-for-pytorch/releases/download/v2.1.10%2Bxpu/torchvision-0.16.0a0+cxx11.abi-cp310-cp310-win_amd64.whl +einops +transformers>=4.25.1 +safetensors>=0.3.0 +aiohttp +pyyaml +Pillow +scipy +tqdm +psutil +intel_extension_for_pytorch @ https://github.com/Nuullll/intel-extension-for-pytorch/releases/download/v2.1.10%2Bxpu/intel_extension_for_pytorch-2.1.10+xpu-cp310-cp310-win_amd64.whl