Sampling functions

Sampling functions for model inference.

Generation

source

get_batch_samples

 get_batch_samples (samples:int, auto_batch_size:int=512)

source

batched_sampling

 batched_sampling (pipeline:genQC.pipeline.pipeline.Pipeline,
                   cond_kwargs:dict[str,torch.Tensor], samples:int,
                   system_size:int, num_of_qubits:int, max_gates:int,
                   g:float=1.0, init_latents:Optional[torch.Tensor]=None,
                   no_bar:bool=True, unique:bool=False,
                   auto_batch_size:int=512, enable_params:bool=True,
                   reduce_spatial:bool=True,
                   return_predicted_x0:bool=False)

e.g. cond_kwargs.keys = {“c”, “micro_cond”, “negative_c”, “U”}

source

prepare_prompts

 prepare_prompts (pipeline:genQC.pipeline.pipeline.Pipeline,
                  prompt:Union[str,Sequence[str]],
                  negative_prompt:Union[str,Sequence[str],NoneType]=None)

Task specific sampling

source

generate_tensors

 generate_tensors (pipeline:genQC.pipeline.pipeline.Pipeline,
                   prompt:Union[str,Sequence[str]], samples:int,
                   system_size:int, num_of_qubits:int, max_gates:int,
                   g:float=1.0, init_latents:Optional[torch.Tensor]=None,
                   no_bar:bool=True, unique:bool=False,
                   auto_batch_size:int=512, enable_params:bool=False,
                   reduce_spatial:bool=True,
                   return_predicted_x0:bool=False,
                   negative_prompt:Union[str,Sequence[str],NoneType]=None,
                   micro_cond:Optional[torch.Tensor]=None)

source

generate_compilation_tensors

 generate_compilation_tensors (pipeline:genQC.pipeline.pipeline.Pipeline,
                               prompt:Union[str,Sequence[str]],
                               U:torch.Tensor, samples:int,
                               system_size:int, num_of_qubits:int,
                               max_gates:int, g:float=1.0,
                               tensor_prod_pad:bool=True,
                               init_latents:Optional[torch.Tensor]=None,
                               no_bar:bool=True, unique:bool=False,
                               auto_batch_size:int=512,
                               enable_params:bool=True,
                               reduce_spatial:bool=True,
                               return_predicted_x0:bool=False, negative_pr
                               ompt:Union[str,Sequence[str],NoneType]=None
                               , negative_u:Optional[torch.Tensor]=None,
                               micro_cond:Optional[torch.Tensor]=None)

*Samples tensor encodings from the DM for the given sample parameters.

What kind of unitary padding we have depends on what we used for model training, so it depends on the concrete model weights.*

Convertion

source

decode_tensors_to_backend

 decode_tensors_to_backend (simulator:genQC.platform.simulation.Simulator,
                            tokenizer:genQC.platform.tokenizer.base_tokeni
                            zer.BaseTokenizer, tensors:torch.Tensor,
                            params:Optional[torch.Tensor]=None,
                            silent:bool=True, n_jobs:int=1,
                            filter_errs:bool=True,
                            return_tensors:bool=False)
Back to top