o  hG@shddlZddlZddlmZddlmZddlmZmZm Z m Z m Z ddl Z ddl m Z ddlmZddlmZdd lmZdd lmZdd lmZdd lmZdd lmZmZddlmZereddlZerddlm Z ddl!m"Z"ddl#m$Z$ddl%m&Z&ddl'm(Z(ddl)m*Z*ddl+m,Z,ddl-m.Z.ddl/m0Z0m1Z1ddl2m3Z3e4e5Z6GdddeZ7dS)N)deepcopy)Path) TYPE_CHECKINGAnyCallableOptionalUnion)nn)FullyShardedDataParallel)Dataset)GenerationConfig)is_deepspeed_zero3_enabled)is_fsdp_managed_module)Trainer)is_datasets_availablelogging)deprecate_kwarg)IterableDataset) DataCollator)FeatureExtractionMixin)BaseImageProcessor)PreTrainedModel)ProcessorMixin)PreTrainedTokenizerBase)TrainerCallback)EvalPredictionPredictionOutput)TrainingArgumentscseZdZeddddd            d1ded ejfd d d ed deeeddfdeeee e effdeeddee gd fdee dee dge fdee dde ejjejjjfdee ejejgejfffdd Zedee efdefdd Z   !d2deed"ee e d#e de e efffd$d% Z  &d3d'ed"ee e d#e dd(ffd)d* Z d4dejd+e e eejeffd,ed"ee e de eeeejeejff fd-d. Zd/d0ZZS)5Seq2SeqTrainer tokenizerprocessing_classz5.0.0T)new_nameversionraise_if_both_namesNNNmodelrargsr data_collatorr train_datasetrzdatasets.Dataset eval_dataset)rrrr model_initcompute_loss_funccompute_metricsr callbacksr optimizerspreprocess_logits_for_metricsc sNtj||||||||| | | | d |jjdur%||jj} | |j_dSdS)N) r&r'r(r)r*r!r+r,r-r.r/r0)super__init__r'generation_configload_generation_configr&)selfr&r'r(r)r*r!r+r,r-r.r/r0 gen_config __class__p/var/www/html/construction_image-detection-poc/venv/lib/python3.10/site-packages/transformers/trainer_seq2seq.pyr27s$  zSeq2SeqTrainer.__init__gen_config_argreturnc Cst|tr t|}n%t|trt|n|}d}|r"|j}|j}n|r'n|}t ||}z.t j dd }| Wdn1sEwYt |dkr[ttdd|DW|Styq}ztdt|d}~ww)aW Loads a `~generation.GenerationConfig` from the `Seq2SeqTrainingArguments.generation_config` arguments. Args: gen_config_arg (`str` or [`~generation.GenerationConfig]`): `Seq2SeqTrainingArguments.generation_config` argument. Returns: A `~generation.GenerationConfig`. NT)recordrcSsg|]}|jqSr9)message).0wr9r9r: sz9Seq2SeqTrainer.load_generation_config..zThe loaded generation config instance is invalid -- `GenerationConfig.validate()` throws warnings and/or exceptions. Fix these issues to train your model. Thrown during validation: ) isinstancer rstrris_filenameparentis_dirfrom_pretrainedwarningscatch_warningsvalidatelen ValueError)r;r6pretrained_model_nameconfig_file_namecaught_warningsexcr9r9r:r4^s6     z%Seq2SeqTrainer.load_generation_configeval ignore_keysmetric_key_prefixc |}|ddur|ddur|jjdur|jj|d<|ddur1|jjdur1|jj|d<|jj|_||_t j |||dS)a Run evaluation and returns metrics. The calling script will be responsible for providing a method to compute metrics, as they are task-dependent (pass it to the init `compute_metrics` argument). You can also subclass and override this method to inject custom behavior. Args: eval_dataset (`Dataset`, *optional*): Pass a dataset if you wish to override `self.eval_dataset`. If it is an [`~datasets.Dataset`], columns not accepted by the `model.forward()` method are automatically removed. It must implement the `__len__` method. ignore_keys (`List[str]`, *optional*): A list of keys in the output of your model (if it is a dictionary) that should be ignored when gathering predictions. metric_key_prefix (`str`, *optional*, defaults to `"eval"`): An optional prefix to be used as the metrics key prefix. For example the metrics "bleu" will be named "eval_bleu" if the prefix is `"eval"` (default) max_length (`int`, *optional*): The maximum target length to use when predicting with the generate method. num_beams (`int`, *optional*): Number of beams for beam search that will be used when predicting with the generate method. 1 means no beam search. gen_kwargs: Additional `generate` specific kwargs. Returns: A dictionary containing the evaluation loss and the potential metrics computed from the predictions. The dictionary also contains the epoch number which comes from the training state. max_lengthNmax_new_tokens num_beamsrSrT) copygetr'generation_max_lengthgeneration_num_beams acceleratorgathergather_function _gen_kwargsr1evaluate)r5r*rSrT gen_kwargsr7r9r:rbs'    zSeq2SeqTrainer.evaluatetest test_datasetrc rU)a& Run prediction and returns predictions and potential metrics. Depending on the dataset and your use case, your test dataset may contain labels. In that case, this method will also return metrics, like in `evaluate()`. Args: test_dataset (`Dataset`): Dataset to run the predictions on. If it is a [`~datasets.Dataset`], columns not accepted by the `model.forward()` method are automatically removed. Has to implement the method `__len__` ignore_keys (`List[str]`, *optional*): A list of keys in the output of your model (if it is a dictionary) that should be ignored when gathering predictions. metric_key_prefix (`str`, *optional*, defaults to `"eval"`): An optional prefix to be used as the metrics key prefix. For example the metrics "bleu" will be named "eval_bleu" if the prefix is `"eval"` (default) max_length (`int`, *optional*): The maximum target length to use when predicting with the generate method. num_beams (`int`, *optional*): Number of beams for beam search that will be used when predicting with the generate method. 1 means no beam search. gen_kwargs: Additional `generate` specific kwargs. If your predictions or labels have different sequence lengths (for instance because you're doing dynamic padding in a token classification task) the predictions will be padded (on the right) to allow for concatenation into one array. The padding index is -100. Returns: *NamedTuple* A namedtuple with the following keys: - predictions (`np.ndarray`): The predictions on `test_dataset`. - label_ids (`np.ndarray`, *optional*): The labels (if the dataset contained some). - metrics (`Dict[str, float]`, *optional*): The potential dictionary of metrics (if the dataset contained labels). rVNrWrXrY) rZr[r'r\r]r^r_r`rar1predict)r5rerSrTrcr7r9r:rfs/    zSeq2SeqTrainer.predictinputsprediction_loss_onlyc s|jjr|rtj||||dSd|v}||}t|dkr)t|dr)|j}d|vr8|ddur8| dd|vrG|ddurG| dt pNt |j }| d||d<|}d|vrvd |vrv|dj|d jkrvd d |D}t|j trt|j nt} | |j jdi||} Wdn1swY|j jjrd |j j_|j j} | jd | jkr|| | j} n| jdur| jd | jdkr|| | jd} tN|r"||di|} Wdn1swY|jdur|| |d } nt| t!r| dn| d } nd} Wdn 1s/wY|jj"r>| ddfS|rq|d}|jd | jkrV||| j}n| jdurp|jd | jdkrp||| jd}nd}| | |fS)a Perform an evaluation step on `model` using `inputs`. Subclass and override to inject custom behavior. Args: model (`nn.Module`): The model to evaluate. inputs (`Dict[str, Union[torch.Tensor, Any]]`): The inputs and targets of the model. The dictionary will be unpacked before being fed to the model. Most models expect the targets under the argument `labels`. Check your model's documentation for all accepted arguments. prediction_loss_only (`bool`): Whether or not to return the loss only. gen_kwargs: Additional `generate` specific kwargs. Return: Tuple[Optional[float], Optional[torch.Tensor], Optional[torch.Tensor]]: A tuple with the loss, logits and labels (each being optional). )rhrSlabelsrrarXNrV synced_gpusdecoder_input_idscSsi|] \}}|dvr||qS))rkdecoder_attention_maskr9)r?kvr9r9r: Bsz2Seq2SeqTrainer.prediction_step..Fr lossr9)#r'predict_with_generater1prediction_step_prepare_inputsrLhasattrrarZpoprrr&r[shapeitemsrBr summon_full_params contextlib nullcontextgenerater3_from_model_configrV_pad_tensors_to_max_lenrWtorchno_gradcompute_loss_context_managerlabel_smootherdetachmeandictrh)r5r&rgrhrSrc has_labelsdefault_synced_gpusgeneration_inputssummon_full_params_contextgenerated_tokensr6outputsrqrir7r9r:rssr           & " zSeq2SeqTrainer.prediction_stepcCs|jdurt|jdr|jjdur|jjn|jj}n|jjjdur'|jjj}ntd|tj|j d|f|j |j d}||ddd|j df<|S)N pad_token_idzSPad_token_id must be set in the configuration of the model, in order to pad tensorsr)dtypedevicerp) r!rur eos_token_idr&configrMronesrwrr)r5tensorrVr padded_tensorr9r9r:r~vs  z&Seq2SeqTrainer._pad_tensors_to_max_len) NNNNNNNNNNr%N)NNrR)Nrd)N) __name__ __module__ __qualname__rrr Modulerr rrCrlisttupleroptim Optimizer lr_schedulerLambdaLRTensorr2 staticmethodr r4floatrbrfrboolrsr~ __classcell__r9r9r7r:r6s     &2  ; E or)8rzrIrZrpathlibrtypingrrrrrrr torch.distributed.fsdpr torch.utils.datar generation.configuration_utilsr integrations.deepspeedrintegrations.fsdprtrainerrutilsrrutils.deprecationrdatasetsrdata.data_collatorrfeature_extraction_utilsrimage_processing_utilsrmodeling_utilsrprocessing_utilsrtokenization_utils_basertrainer_callbackr trainer_utilsrr training_argsr get_loggerrloggerrr9r9r9r:s<