How to use safe_getattr method in Pytest

Best Python code snippet using pytest

transformer_lm.py

Source:transformer_lm.py

...228    def build_model(cls, args, task):229        """Build a new model instance."""230        if args.decoder_layers_to_keep:231            args.decoder_layers = len(args.decoder_layers_to_keep.split(","))232        if safe_getattr(args, "max_target_positions", None) is None:233            args.max_target_positions = safe_getattr(234                args, "tokens_per_sample", DEFAULT_MAX_TARGET_POSITIONS235            )236        if args.character_embeddings:237            embed_tokens = CharacterTokenEmbedder(238                task.source_dictionary,239                eval(args.character_filters),240                args.character_embedding_dim,241                args.decoder_embed_dim,242                args.char_embedder_highway_layers,243            )244        elif args.adaptive_input:245            embed_tokens = AdaptiveInput(246                len(task.source_dictionary),247                task.source_dictionary.pad(),248                args.decoder_input_dim,249                args.adaptive_input_factor,250                args.decoder_embed_dim,251                options.eval_str_list(args.adaptive_input_cutoff, type=int),252                args.quant_noise_pq,253                args.quant_noise_pq_block_size,254            )255        else:256            embed_tokens = cls.build_embedding(257                args, task.source_dictionary, args.decoder_input_dim258            )259        if args.tie_adaptive_weights:260            assert args.adaptive_input261            assert args.adaptive_input_factor == args.adaptive_softmax_factor262            assert (263                args.adaptive_softmax_cutoff == args.adaptive_input_cutoff264            ), "{} != {}".format(265                args.adaptive_softmax_cutoff, args.adaptive_input_cutoff266            )267            assert args.decoder_input_dim == args.decoder_output_dim268        decoder = TransformerDecoder(269            args, task.target_dictionary, embed_tokens, no_encoder_attn=True270        )271        return cls(decoder)272    @classmethod273    def build_embedding(cls, args, dictionary, embed_dim, path=None):274        embed_tokens = Embedding(len(dictionary), embed_dim, dictionary.pad())275        return embed_tokens276def base_lm_architecture(args):277    # backward compatibility for older model checkpoints278    if safe_hasattr(args, "no_tie_adaptive_proj"):279        # previous models defined --no-tie-adaptive-proj, so use the existence of280        # that option to determine if this is an "old" model checkpoint281        args.no_decoder_final_norm = True  # old models always set this to True282        if args.no_tie_adaptive_proj is False:283            args.tie_adaptive_proj = True284    if safe_hasattr(args, "decoder_final_norm"):285        args.no_decoder_final_norm = not args.decoder_final_norm286    args.dropout = safe_getattr(args, "dropout", 0.1)287    args.attention_dropout = safe_getattr(args, "attention_dropout", 0.0)288    args.decoder_embed_dim = safe_getattr(args, "decoder_embed_dim", 512)289    args.decoder_ffn_embed_dim = safe_getattr(args, "decoder_ffn_embed_dim", 2048)290    args.decoder_layers = safe_getattr(args, "decoder_layers", 6)291    args.decoder_attention_heads = safe_getattr(args, "decoder_attention_heads", 8)292    args.adaptive_softmax_cutoff = safe_getattr(args, "adaptive_softmax_cutoff", None)293    args.adaptive_softmax_dropout = safe_getattr(args, "adaptive_softmax_dropout", 0)294    args.adaptive_softmax_factor = safe_getattr(args, "adaptive_softmax_factor", 4)295    args.decoder_learned_pos = safe_getattr(args, "decoder_learned_pos", False)296    args.activation_fn = safe_getattr(args, "activation_fn", "relu")297    args.decoder_layerdrop = safe_getattr(args, "decoder_layerdrop", 0)298    args.decoder_layers_to_keep = safe_getattr(args, "decoder_layers_to_keep", None)299    args.quant_noise_pq = safe_getattr(args, "quant_noise_pq", 0)300    args.quant_noise_pq_block_size = safe_getattr(args, "quant_noise_pq_block_size", 8)301    args.quant_noise_scalar = safe_getattr(args, "quant_noise_scalar", 0)302    args.base_layers = safe_getattr(args, "base_layers", 0)303    args.base_sublayers = safe_getattr(args, "base_sublayers", 1)304    args.base_shuffle = safe_getattr(args, "base_shuffle", False)305    args.add_bos_token = safe_getattr(args, "add_bos_token", False)306    args.no_token_positional_embeddings = safe_getattr(307        args, "no_token_positional_embeddings", False308    )309    args.share_decoder_input_output_embed = safe_getattr(310        args, "share_decoder_input_output_embed", False311    )312    args.character_embeddings = safe_getattr(args, "character_embeddings", False)313    args.decoder_output_dim = safe_getattr(314        args, "decoder_output_dim", args.decoder_embed_dim315    )316    args.decoder_input_dim = safe_getattr(args, "decoder_input_dim", args.decoder_embed_dim)317    # Model training is not stable without this318    args.decoder_normalize_before = True319    args.no_decoder_final_norm = safe_getattr(args, "no_decoder_final_norm", False)320    args.adaptive_input = safe_getattr(args, "adaptive_input", False)321    args.adaptive_input_factor = safe_getattr(args, "adaptive_input_factor", 4)322    args.adaptive_input_cutoff = safe_getattr(args, "adaptive_input_cutoff", None)323    args.tie_adaptive_weights = safe_getattr(args, "tie_adaptive_weights", False)324    args.tie_adaptive_proj = safe_getattr(args, "tie_adaptive_proj", False)325    args.no_scale_embedding = safe_getattr(args, "no_scale_embedding", False)326    args.layernorm_embedding = safe_getattr(args, "layernorm_embedding", False)327    args.checkpoint_activations = safe_getattr(args, "checkpoint_activations", False)328    args.offload_activations = safe_getattr(args, "offload_activations", False)329    if args.offload_activations:330        args.checkpoint_activations = True331@register_model_architecture("transformer_lm", "transformer_lm_big")332def transformer_lm_big(args):333    args.decoder_layers = safe_getattr(args, "decoder_layers", 12)334    args.decoder_embed_dim = safe_getattr(args, "decoder_embed_dim", 1024)335    args.decoder_ffn_embed_dim = safe_getattr(args, "decoder_ffn_embed_dim", 4096)336    args.decoder_attention_heads = safe_getattr(args, "decoder_attention_heads", 16)337    base_lm_architecture(args)338@register_model_architecture("transformer_lm", "transformer_lm_wiki103")339@register_model_architecture("transformer_lm", "transformer_lm_baevski_wiki103")340def transformer_lm_baevski_wiki103(args):341    args.decoder_layers = safe_getattr(args, "decoder_layers", 16)342    args.decoder_attention_heads = safe_getattr(args, "decoder_attention_heads", 8)343    args.dropout = safe_getattr(args, "dropout", 0.3)344    args.adaptive_input = safe_getattr(args, "adaptive_input", True)345    args.tie_adaptive_weights = safe_getattr(args, "tie_adaptive_weights", True)346    args.adaptive_input_cutoff = safe_getattr(args, "adaptive_input_cutoff", "20000,60000")347    args.adaptive_softmax_cutoff = safe_getattr(348        args, "adaptive_softmax_cutoff", "20000,60000"349    )350    args.adaptive_softmax_dropout = safe_getattr(args, "adaptive_softmax_dropout", 0.2)351    args.attention_dropout = safe_getattr(args, "attention_dropout", 0.1)352    args.activation_dropout = safe_getattr(args, "activation_dropout", 0.1)353    args.no_decoder_final_norm = safe_getattr(args, "no_decoder_final_norm", True)354    args.tie_adaptive_proj = safe_getattr(args, "tie_adaptive_proj", True)355    transformer_lm_big(args)356@register_model_architecture("transformer_lm", "transformer_lm_gbw")357@register_model_architecture("transformer_lm", "transformer_lm_baevski_gbw")358def transformer_lm_baevski_gbw(args):359    args.decoder_embed_dim = safe_getattr(args, "decoder_embed_dim", 512)360    args.dropout = safe_getattr(args, "dropout", 0.1)361    args.attention_dropout = safe_getattr(args, "attention_dropout", 0.1)362    args.no_decoder_final_norm = safe_getattr(args, "no_decoder_final_norm", True)363    transformer_lm_big(args)364@register_model_architecture("transformer_lm", "transformer_lm_gpt")365def transformer_lm_gpt(args):366    args.decoder_embed_dim = safe_getattr(args, "decoder_embed_dim", 768)367    args.decoder_ffn_embed_dim = safe_getattr(args, "decoder_ffn_embed_dim", 3072)368    args.decoder_layers = safe_getattr(args, "decoder_layers", 12)369    args.decoder_attention_heads = safe_getattr(args, "decoder_attention_heads", 12)370    args.dropout = safe_getattr(args, "dropout", 0.1)371    args.attention_dropout = safe_getattr(args, "attention_dropout", 0.1)372    args.activation_fn = safe_getattr(args, "activation_fn", "gelu")373    base_lm_architecture(args)374@register_model_architecture("transformer_lm", "transformer_lm_gpt2_small")375def transformer_lm_gpt2_small(args):376    args.decoder_embed_dim = safe_getattr(args, "decoder_embed_dim", 1024)377    args.decoder_ffn_embed_dim = safe_getattr(args, "decoder_ffn_embed_dim", 4096)378    args.decoder_layers = safe_getattr(args, "decoder_layers", 24)379    args.decoder_attention_heads = safe_getattr(args, "decoder_attention_heads", 16)380    args.dropout = safe_getattr(args, "dropout", 0.1)381    args.attention_dropout = safe_getattr(args, "attention_dropout", 0.1)382    args.activation_fn = safe_getattr(args, "activation_fn", "gelu")383    base_lm_architecture(args)384@register_model_architecture("transformer_lm", "transformer_lm_gpt2_tiny")385def transformer_lm_gpt2_tiny(args):386    args.decoder_embed_dim = safe_getattr(args, "decoder_embed_dim", 64)387    args.decoder_ffn_embed_dim = safe_getattr(args, "decoder_ffn_embed_dim", 64)388    args.decoder_layers = safe_getattr(args, "decoder_layers", 2)389    args.decoder_attention_heads = safe_getattr(args, "decoder_attention_heads", 1)390    args.dropout = safe_getattr(args, "dropout", 0.1)391    args.attention_dropout = safe_getattr(args, "attention_dropout", 0.1)392    args.activation_fn = safe_getattr(args, "activation_fn", "gelu")393    base_lm_architecture(args)394@register_model_architecture("transformer_lm", "transformer_lm_gpt2_medium")395def transformer_lm_gpt2_medium(args):396    args.decoder_embed_dim = safe_getattr(args, "decoder_embed_dim", 1280)397    args.decoder_ffn_embed_dim = safe_getattr(args, "decoder_ffn_embed_dim", 5120)398    args.decoder_layers = safe_getattr(args, "decoder_layers", 36)399    args.decoder_attention_heads = safe_getattr(args, "decoder_attention_heads", 20)400    args.dropout = safe_getattr(args, "dropout", 0.1)401    args.attention_dropout = safe_getattr(args, "attention_dropout", 0.1)402    args.activation_fn = safe_getattr(args, "activation_fn", "gelu")403    base_lm_architecture(args)404@register_model_architecture("transformer_lm", "transformer_lm_gpt2_big")405def transformer_lm_gpt2_big(args):406    args.decoder_embed_dim = safe_getattr(args, "decoder_embed_dim", 1600)407    args.decoder_ffn_embed_dim = safe_getattr(args, "decoder_ffn_embed_dim", 6400)408    args.decoder_layers = safe_getattr(args, "decoder_layers", 48)409    args.decoder_attention_heads = safe_getattr(args, "decoder_attention_heads", 25)410    args.dropout = safe_getattr(args, "dropout", 0.1)411    args.attention_dropout = safe_getattr(args, "attention_dropout", 0.1)412    args.activation_fn = safe_getattr(args, "activation_fn", "gelu")413    base_lm_architecture(args)414def base_gpt3_architecture(args):415    args.decoder_input_dim = args.decoder_embed_dim416    args.decoder_output_dim = args.decoder_embed_dim417    args.decoder_ffn_embed_dim = safe_getattr(args, "decoder_ffn_embed_dim", args.decoder_embed_dim * 4)418    # GPT-3 used learned positional embeddings, rather than sinusoidal419    args.decoder_learned_pos = safe_getattr(args, "decoder_learned_pos", True)420    args.dropout = safe_getattr(args, "dropout", 0.0)421    args.attention_dropout = safe_getattr(args, "attention_dropout", 0.0)422    args.activation_fn = safe_getattr(args, "activation_fn", "gelu")423    args.share_decoder_input_output_embed = True424    base_lm_architecture(args)425@register_model_architecture("transformer_lm", "transformer_lm_gpt3_small")426def transformer_lm_gpt3_small(args):427    # 125M params428    args.decoder_layers = safe_getattr(args, "decoder_layers", 12)429    args.decoder_embed_dim = safe_getattr(args, "decoder_embed_dim", 768)430    args.decoder_attention_heads = safe_getattr(args, "decoder_attention_heads", 12)431    base_gpt3_architecture(args)432@register_model_architecture("transformer_lm", "transformer_lm_gpt3_medium")433def transformer_lm_gpt3_medium(args):434    # 350M params435    args.decoder_layers = safe_getattr(args, "decoder_layers", 24)436    args.decoder_embed_dim = safe_getattr(args, "decoder_embed_dim", 1024)437    args.decoder_attention_heads = safe_getattr(args, "decoder_attention_heads", 16)438    base_gpt3_architecture(args)439@register_model_architecture("transformer_lm", "transformer_lm_gpt3_large")440def transformer_lm_gpt3_large(args):441    # 760M params442    args.decoder_layers = safe_getattr(args, "decoder_layers", 24)443    args.decoder_embed_dim = safe_getattr(args, "decoder_embed_dim", 1536)444    args.decoder_attention_heads = safe_getattr(args, "decoder_attention_heads", 16)445    base_gpt3_architecture(args)446@register_model_architecture("transformer_lm", "transformer_lm_gpt3_xl")447def transformer_lm_gpt3_xl(args):448    # 1.3B params449    args.decoder_layers = safe_getattr(args, "decoder_layers", 24)450    args.decoder_embed_dim = safe_getattr(args, "decoder_embed_dim", 2048)451    args.decoder_attention_heads = safe_getattr(args, "decoder_attention_heads", 32)452    base_gpt3_architecture(args)453@register_model_architecture("transformer_lm", "transformer_lm_gpt3_2_7")454def transformer_lm_gpt3_2_7(args):455    # 2.7B params456    args.decoder_layers = safe_getattr(args, "decoder_layers", 32)457    args.decoder_embed_dim = safe_getattr(args, "decoder_embed_dim", 2560)458    args.decoder_attention_heads = safe_getattr(args, "decoder_attention_heads", 32)459    base_gpt3_architecture(args)460@register_model_architecture("transformer_lm", "transformer_lm_gpt3_6_7")461def transformer_lm_gpt3_6_7(args):462    # 6.7B params463    args.decoder_layers = safe_getattr(args, "decoder_layers", 32)464    args.decoder_embed_dim = safe_getattr(args, "decoder_embed_dim", 4096)465    args.decoder_attention_heads = safe_getattr(args, "decoder_attention_heads", 32)466    base_gpt3_architecture(args)467@register_model_architecture("transformer_lm", "transformer_lm_gpt3_13")468def transformer_lm_gpt3_13(args):469    # 13B params470    args.decoder_layers = safe_getattr(args, "decoder_layers", 40)471    args.decoder_embed_dim = safe_getattr(args, "decoder_embed_dim", 5120)472    args.decoder_attention_heads = safe_getattr(args, "decoder_attention_heads", 40)473    base_gpt3_architecture(args)474@register_model_architecture("transformer_lm", "transformer_lm_gpt3_175")475def transformer_lm_gpt3_175(args):476    # 175B params477    args.decoder_layers = safe_getattr(args, "decoder_layers", 96)478    args.decoder_embed_dim = safe_getattr(args, "decoder_embed_dim", 12288)479    args.decoder_attention_heads = safe_getattr(args, "decoder_attention_heads", 96)...

model.py

Source:model.py

...454        return self.lm_head(features, masked_tokens)455    def max_positions(self):456        """Maximum output length supported by the encoder."""457        return self.args.max_positions458def safe_getattr(obj, k, default=None):459    from omegaconf import OmegaConf460    if OmegaConf.is_config(obj):461        return obj[k] if k in obj and obj[k] is not None else default462    return getattr(obj, k, default)463@register_model_architecture("roberta", "roberta")464def base_architecture(args):465    args.encoder_layers = safe_getattr(args, "encoder_layers", 12)466    args.encoder_embed_dim = safe_getattr(args, "encoder_embed_dim", 768)467    args.encoder_ffn_embed_dim = safe_getattr(args, "encoder_ffn_embed_dim", 3072)468    args.encoder_attention_heads = safe_getattr(args, "encoder_attention_heads", 12)469    args.dropout = safe_getattr(args, "dropout", 0.1)470    args.attention_dropout = safe_getattr(args, "attention_dropout", 0.1)471    args.activation_dropout = safe_getattr(args, "activation_dropout", 0.0)472    args.pooler_dropout = safe_getattr(args, "pooler_dropout", 0.0)473    args.max_source_positions = safe_getattr(args, "max_positions", 512)474    args.no_token_positional_embeddings = safe_getattr(475        args, "no_token_positional_embeddings", False476    )477    # BERT has a few structural differences compared to the original Transformer478    args.encoder_learned_pos = safe_getattr(args, "encoder_learned_pos", True)479    args.layernorm_embedding = safe_getattr(args, "layernorm_embedding", True)480    args.no_scale_embedding = safe_getattr(args, "no_scale_embedding", True)481    args.activation_fn = safe_getattr(args, "activation_fn", "gelu")482    args.encoder_normalize_before = safe_getattr(args, "encoder_normalize_before", False)483    args.pooler_activation_fn = safe_getattr(args, "pooler_activation_fn", "tanh")484    args.untie_weights_roberta = safe_getattr(args, "untie_weights_roberta", False)485    # Adaptive input config486    args.adaptive_input = safe_getattr(args, "adaptive_input", False)487    # LayerDrop config488    args.encoder_layerdrop = safe_getattr(args, "encoder_layerdrop", 0.0)489    args.encoder_layers_to_keep = safe_getattr(args, "encoder_layers_to_keep", None)490    # Quantization noise config491    args.quant_noise_pq = safe_getattr(args, "quant_noise_pq", 0)492    args.quant_noise_pq_block_size = safe_getattr(args, "quant_noise_pq_block_size", 8)493    args.quant_noise_scalar = safe_getattr(args, "quant_noise_scalar", 0)494    # R4F config495    args.spectral_norm_classification_head = safe_getattr(496        args, "spectral_norm_classification_head", False497    )498@register_model_architecture("roberta", "roberta_prenorm")499def roberta_prenorm_architecture(args):500    args.layernorm_embedding = safe_getattr(args, "layernorm_embedding", False)501    args.encoder_normalize_before = safe_getattr(args, "encoder_normalize_before", True)502    base_architecture(args)503@register_model_architecture("roberta", "roberta_base")504def roberta_base_architecture(args):505    base_architecture(args)506@register_model_architecture("roberta", "roberta_large")507def roberta_large_architecture(args):508    args.encoder_layers = safe_getattr(args, "encoder_layers", 24)509    args.encoder_embed_dim = safe_getattr(args, "encoder_embed_dim", 1024)510    args.encoder_ffn_embed_dim = safe_getattr(args, "encoder_ffn_embed_dim", 4096)511    args.encoder_attention_heads = safe_getattr(args, "encoder_attention_heads", 16)512    base_architecture(args)513@register_model_architecture("roberta", "xlm")514def xlm_architecture(args):515    args.encoder_layers = safe_getattr(args, "encoder_layers", 16)516    args.encoder_embed_dim = safe_getattr(args, "encoder_embed_dim", 1280)517    args.encoder_ffn_embed_dim = safe_getattr(args, "encoder_ffn_embed_dim", 1280 * 4)518    args.encoder_attention_heads = safe_getattr(args, "encoder_attention_heads", 16)...

__init__.py

Source:__init__.py

...3import raven4from raven_aiohttp import AioHttpTransport5from sanic.handlers import ErrorHandler6from sanic import exceptions as sanic_exceptions7def safe_getattr(request, attr_name, default=None):8    # pylint:disable=bare-except9    try:10        return getattr(request, attr_name, default)11    except:12        return default13class SanicSentryErrorHandler(ErrorHandler):14    DEFAULT_EXCEPTIONS_TO_IGNORE = (sanic_exceptions.NotFound,)15    def __init__(self, dsn, exceptions_to_ignore=None, **sentry_kwargs):16        super(SanicSentryErrorHandler, self).__init__()17        self.exceptions_to_ignore = tuple(exceptions_to_ignore) if exceptions_to_ignore is not None else self.DEFAULT_EXCEPTIONS_TO_IGNORE18        # For sentry_kwargs see19        # https://docs.sentry.io/clients/python/advanced/#client-arguments20        self.sentry_client = raven.Client(dsn, transport=AioHttpTransport, **sentry_kwargs)21    def default(self, request, exception):22        if not isinstance(exception, self.exceptions_to_ignore):23            exc_info = (type(exception), exception, exception.__traceback__)24            extra = self._request_debug_info(request) if request else dict()25            self.sentry_client.captureException(exc_info, extra=extra)26        return super(SanicSentryErrorHandler, self).default(request, exception)27    def _request_debug_info(self, request):28        # pylint:disable=no-self-use29        return dict(30            url=safe_getattr(request, "url"),31            method=safe_getattr(request, "method"),32            headers=safe_getattr(request, "headers"),33            body=safe_getattr(request, "body"),34            query_string=safe_getattr(request, "query_string"),35        )36    def intercept_exception(self, function):37        """38        Decorator for Sanic exception views.39        You should use this decorator only if your exception handler returns its own response.40        If you're handler returns `None` the default exception handler will be called which41        means Sentry will be called twice.42        Example:43            >> @app.exception([Exception, ])44            >> @sentry_client.intercept_exception45            >> def handle_exception(request, exception):46            >>     pass47        """48        @wraps(function)...

inspect.py

Source:inspect.py

...8"""9def isdescriptor(x):10    """Check if the object is some kind of descriptor."""11    for item in '__get__', '__set__', '__delete__':12        if hasattr(safe_getattr(x, item, None), '__call__'):13            return True14    return False15def safe_getattr(obj, name, *defargs):16    """A getattr() that turns all exceptions into AttributeErrors."""17    try:18        return getattr(obj, name, *defargs)19    except Exception:20        # this is a catch-all for all the weird things that some modules do21        # with attribute access22        if defargs:23            return defargs[0]24        raise AttributeError(name)25def safe_getmembers(object, predicate=None):26    """A version of inspect.getmembers() that uses safe_getattr()."""27    results = []28    for key in dir(object):29        try:30            value = safe_getattr(object, key, None)31        except AttributeError:32            continue33        if not predicate or predicate(value):34            results.append((key, value))35    results.sort()...

Pytest Tutorial

Looking for an in-depth tutorial around pytest? LambdaTest covers the detailed pytest tutorial that has everything related to the pytest, from setting up the pytest framework to automation testing. Delve deeper into pytest testing by exploring advanced use cases like parallel testing, pytest fixtures, parameterization, executing multiple test cases from a single file, and more.

Chapters

What is pytest
Pytest installation: Want to start pytest from scratch? See how to install and configure pytest for Python automation testing.
Run first test with pytest framework: Follow this step-by-step tutorial to write and run your first pytest script.
Parallel testing with pytest: A hands-on guide to parallel testing with pytest to improve the scalability of your test automation.
Generate pytest reports: Reports make it easier to understand the results of pytest-based test runs. Learn how to generate pytest reports.
Pytest Parameterized tests: Create and run your pytest scripts while avoiding code duplication and increasing test coverage with parameterization.
Pytest Fixtures: Check out how to implement pytest fixtures for your end-to-end testing needs.
Execute Multiple Test Cases: Explore different scenarios for running multiple test cases in pytest from a single file.
Stop Test Suite after N Test Failures: See how to stop your test suite after n test failures in pytest using the @pytest.mark.incremental decorator and maxfail command-line option.