o Gi3@sddlmZddlZddlZddlmZmZddlm Z m Z ddl m Z ddl mZe eZeGd d d e ZGd d d eeZdS) ) dataclassN) ConfigMixinregister_to_config) BaseOutputlogging) randn_tensor)SchedulerMixinc@seZdZUdZejed<dS)$FlowMatchHeunDiscreteSchedulerOutputaJ Output class for the scheduler's `step` function output. Args: prev_sample (`torch.FloatTensor` of shape `(batch_size, num_channels, height, width)` for images): Computed sample `(x_{t-1})` of previous timestep. `prev_sample` should be used as next model input in the denoising loop. prev_sampleN)__name__ __module__ __qualname____doc__torch FloatTensor__annotations__rrl/home/ubuntu/.local/lib/python3.10/site-packages/diffusers/schedulers/scheduling_flow_match_heun_discrete.pyr s  r c@sjeZdZdZgZdZe  d4dedefddZ e d d Z e d d Z d5defddZ dejdeejBdejdejfddZdedefddZ d6dedeejBddfddZ d6deejBdejdBdefd d!ZdeejBddfd"d#Ze defd$d%Zd&d&ed'ddd(fd)ejdeejBdejd*ed+ed,ed-ed.ejdBd/edeeBfd0d1Zdefd2d3ZdS)7FlowMatchHeunDiscreteSchedulera Heun scheduler. This model inherits from [`SchedulerMixin`] and [`ConfigMixin`]. Check the superclass documentation for the generic methods the library implements for all schedulers such as loading and saving. Args: num_train_timesteps (`int`, defaults to 1000): The number of diffusion steps to train the model. shift (`float`, defaults to 1.0): The shift value for the timestep schedule. r?num_train_timestepsshiftcCstjd||tjdddd}t|jtjd}||}||d|d|}|||_d|_d|_ |d|_ |j d |_ |j d |_ dS)Nr )dtypecpur)nplinspacefloat32copyr from_numpyto timesteps _step_index _begin_indexsigmasitem sigma_min sigma_max)selfrrr$r'rrr__init__<s"  z'FlowMatchHeunDiscreteScheduler.__init__cC|jS)zg The index counter for current timestep. It will increase 1 after each scheduler step. )r%r+rrr step_indexQz)FlowMatchHeunDiscreteScheduler.step_indexcCr-)zq The index for the first timestep. It should be set from pipeline with `set_begin_index` method. r&r.rrr begin_indexXr0z*FlowMatchHeunDiscreteScheduler.begin_indexrr2cCs ||_dS)z Sets the begin index for the scheduler. This function should be run from pipeline before the inference. Args: begin_index (`int`, defaults to `0`): The begin index for the scheduler. Nr1)r+r2rrrset_begin_index`s z.FlowMatchHeunDiscreteScheduler.set_begin_indexsampletimestepnoisereturncCs8|jdur |||j|j}||d||}|S)a Forward process in flow-matching Args: sample (`torch.FloatTensor`): The input sample. timestep (`float` or `torch.FloatTensor`): The current timestep in the diffusion chain. noise (`torch.FloatTensor`): The noise tensor. Returns: `torch.FloatTensor`: A scaled input sample. Nr)r/_init_step_indexr')r+r4r5r6sigmarrr scale_noisejs   z*FlowMatchHeunDiscreteScheduler.scale_noiser9cCs ||jjSNconfigr)r+r9rrr _sigma_to_ts z*FlowMatchHeunDiscreteScheduler._sigma_to_tNnum_inference_stepsdevicecCs ||_t||j||j|}||jj}|jj|d|jjd|}t |j t j |d}||jj}t |dd|dddg}|j |d|_t |t jd|jdg}t |dd|ddd|ddg|_d|_d|_d|_d|_dS)a Sets the discrete timesteps used for the diffusion chain (to be run before inference). Args: num_inference_steps (`int`): The number of diffusion steps used when generating samples with a pre-trained model. device (`str` or `torch.device`, *optional*): The device to which the timesteps should be moved to. If `None`, the timesteps are not moved. r )rr@Nr)r@r)r?rrr>r*r)r=rrrr"r#r catrepeat_interleaver$zerosr@r'prev_derivativedtr%r&)r+r?r@r$r'rrr set_timestepss$     $0 z,FlowMatchHeunDiscreteScheduler.set_timestepsschedule_timestepscCs:|dur|j}||k}t|dkrdnd}||S)a Find the index of a given timestep in the timestep schedule. Args: timestep (`float` or `torch.FloatTensor`): The timestep value to find in the schedule. schedule_timesteps (`torch.FloatTensor`, *optional*): The timestep schedule to search in. If `None`, uses `self.timesteps`. Returns: `int`: The index of the timestep in the schedule. Nr r)r$nonzerolenr()r+r5rGindicesposrrrindex_for_timesteps   z1FlowMatchHeunDiscreteScheduler.index_for_timestepcCs@|jdurt|tjr||jj}|||_dS|j |_dSr;) r2 isinstancerTensorr#r$r@rLr%r&)r+r5rrrr8s   z/FlowMatchHeunDiscreteScheduler._init_step_indexcCs |jduS)zL Returns whether the scheduler is in the first-order state. N)rEr.rrrstate_in_first_orders z3FlowMatchHeunDiscreteScheduler.state_in_first_orderinfT model_outputs_churns_tmins_tmaxs_noise generator return_dictc Cst|tst|tjst|tjrtd|jdur|||tj }|j r7|j |j} |j |jd} n|j |jd} |j |j} || krO|kr]nn t |t |j ddnd} | | d} | dkrt|j|j|j|d}||}||| d| dd }|j r||| }||| }| | }||_||_||_n"||| }||| }d |j|}|j}|j}d|_d|_d|_|||}||j}|jd7_| s|fSt|d S) a[ Predict the sample from the previous timestep by reversing the SDE. This function propagates the diffusion process from the learned model outputs (most often the predicted noise). Args: model_output (`torch.FloatTensor`): The direct output from learned diffusion model. timestep (`float` or `torch.FloatTensor`): The current discrete timestep in the diffusion chain. sample (`torch.FloatTensor`): A current instance of a sample created by the diffusion process. s_churn (`float`): Stochasticity parameter that controls the amount of noise added during sampling. Higher values increase randomness. s_tmin (`float`): Minimum timestep threshold for applying stochasticity. Only timesteps above this value will have noise added. s_tmax (`float`): Maximum timestep threshold for applying stochasticity. Only timesteps below this value will have noise added. s_noise (`float`, defaults to 1.0): Scaling factor for noise added to the sample. generator (`torch.Generator`, *optional*): A random number generator. return_dict (`bool`): Whether or not to return a [`~schedulers.scheduling_flow_match_heun_discrete.FlowMatchHeunDiscreteSchedulerOutput`] tuple. Returns: [`~schedulers.scheduling_flow_match_heun_discrete.FlowMatchHeunDiscreteSchedulerOutput`] or `tuple`: If return_dict is `True`, [`~schedulers.scheduling_flow_match_heun_discrete.FlowMatchHeunDiscreteSchedulerOutput`] is returned, otherwise a tuple is returned where the first element is the sample tensor. zPassing integer indices (e.g. from `enumerate(timesteps)`) as timesteps to `FlowMatchHeunDiscreteScheduler.step()` is not supported. Make sure to pass one of the `scheduler.timesteps` as a timestep.Nr g4y?rPr)rr@rWrg?)r )rMintr IntTensor LongTensor ValueErrorr/r8r#r rOr'minrIrshaperr@rDrEr4r%r )r+rRr5r4rSrTrUrVrWrXr9 sigma_nextgamma sigma_hatr6epsdenoised derivativerEr rrrstepsb0       4        z#FlowMatchHeunDiscreteScheduler.stepcCs|jjSr;r<r.rrr__len__bsz&FlowMatchHeunDiscreteScheduler.__len__)rr)rr;)r rrr _compatiblesorderrrYfloatr,propertyr/r2r3rrr:r>strr@rFrLr8boolrO Generatorr tuplererfrrrrr+s       +     r) dataclassesrnumpyrrconfiguration_utilsrrutilsrrutils.torch_utilsrscheduling_utilsr get_loggerr loggerr rrrrrs