From 1cefb5d788f5e1a5b59fd57394ed93cbe71f0d86 Mon Sep 17 00:00:00 2001 From: Cyril Vallez Date: Wed, 9 Jul 2025 15:46:53 +0200 Subject: [PATCH] [modular] Allow method with the same name in case of @property decorator (#39308) * fix * add example * fix * Update modular_model_converter.py --- .../configuration_duplicated_method.py | 216 ++++++++++++++++++ .../modeling_multimodal2.py | 1 + .../modeling_my_new_model2.py | 3 +- .../modeling_new_task_model.py | 1 + .../modular-transformers/modeling_super.py | 3 +- .../modular_duplicated_method.py | 11 + utils/modular_model_converter.py | 37 ++- 7 files changed, 264 insertions(+), 8 deletions(-) create mode 100644 examples/modular-transformers/configuration_duplicated_method.py create mode 100644 examples/modular-transformers/modular_duplicated_method.py diff --git a/examples/modular-transformers/configuration_duplicated_method.py b/examples/modular-transformers/configuration_duplicated_method.py new file mode 100644 index 0000000000..ffe4079d4d --- /dev/null +++ b/examples/modular-transformers/configuration_duplicated_method.py @@ -0,0 +1,216 @@ +# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨 +# This file was automatically generated from examples/modular-transformers/modular_duplicated_method.py. +# Do NOT edit this file manually as any edits will be overwritten by the generation of +# the file from the modular. If any change should be done, please apply the change to the +# modular_duplicated_method.py file directly. One of our CI enforces this. +# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨 + +from ...configuration_utils import PretrainedConfig +from ...modeling_rope_utils import rope_config_validation + + +class DuplicatedMethodConfig(PretrainedConfig): + r""" + This is the configuration class to store the configuration of a [`DuplicatedMethodModel`]. It is used to instantiate an DuplicatedMethod + model according to the specified arguments, defining the model architecture. Instantiating a configuration with the + defaults will yield a similar configuration to that of the DuplicatedMethod-7B. + e.g. [meta-duplicated_method/DuplicatedMethod-2-7b-hf](https://huggingface.co/meta-duplicated_method/DuplicatedMethod-2-7b-hf) + + Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the + documentation from [`PretrainedConfig`] for more information. + + + Args: + vocab_size (`int`, *optional*, defaults to 32000): + Vocabulary size of the DuplicatedMethod model. Defines the number of different tokens that can be represented by the + `inputs_ids` passed when calling [`DuplicatedMethodModel`] + hidden_size (`int`, *optional*, defaults to 4096): + Dimension of the hidden representations. + intermediate_size (`int`, *optional*, defaults to 11008): + Dimension of the MLP representations. + num_hidden_layers (`int`, *optional*, defaults to 32): + Number of hidden layers in the Transformer decoder. + num_attention_heads (`int`, *optional*, defaults to 32): + Number of attention heads for each attention layer in the Transformer decoder. + num_key_value_heads (`int`, *optional*): + This is the number of key_value heads that should be used to implement Grouped Query Attention. If + `num_key_value_heads=num_attention_heads`, the model will use Multi Head Attention (MHA), if + `num_key_value_heads=1` the model will use Multi Query Attention (MQA) otherwise GQA is used. When + converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed + by meanpooling all the original heads within that group. For more details, check out [this + paper](https://huggingface.co/papers/2305.13245). If it is not specified, will default to + `num_attention_heads`. + hidden_act (`str` or `function`, *optional*, defaults to `"silu"`): + The non-linear activation function (function or string) in the decoder. + max_position_embeddings (`int`, *optional*, defaults to 2048): + The maximum sequence length that this model might ever be used with. DuplicatedMethod 1 supports up to 2048 tokens, + DuplicatedMethod 2 up to 4096, CodeLlama up to 16384. + initializer_range (`float`, *optional*, defaults to 0.02): + The standard deviation of the truncated_normal_initializer for initializing all weight matrices. + rms_norm_eps (`float`, *optional*, defaults to 1e-06): + The epsilon used by the rms normalization layers. + use_cache (`bool`, *optional*, defaults to `True`): + Whether or not the model should return the last key/values attentions (not used by all models). Only + relevant if `config.is_decoder=True`. + pad_token_id (`int`, *optional*): + Padding token id. + bos_token_id (`int`, *optional*, defaults to 1): + Beginning of stream token id. + eos_token_id (`int`, *optional*, defaults to 2): + End of stream token id. + pretraining_tp (`int`, *optional*, defaults to 1): + Experimental feature. Tensor parallelism rank used during pretraining. Please refer to [this + document](https://huggingface.co/docs/transformers/main/perf_train_gpu_many#tensor-parallelism) to + understand more about it. This value is necessary to ensure exact reproducibility of the pretraining + results. Please refer to [this issue](https://github.com/pytorch/pytorch/issues/76232). + tie_word_embeddings (`bool`, *optional*, defaults to `False`): + Whether to tie weight embeddings + rope_theta (`float`, *optional*, defaults to 10000.0): + The base period of the RoPE embeddings. + rope_scaling (`Dict`, *optional*): + Dictionary containing the scaling configuration for the RoPE embeddings. NOTE: if you apply new rope type + and you expect the model to work on longer `max_position_embeddings`, we recommend you to update this value + accordingly. + Expected contents: + `rope_type` (`str`): + The sub-variant of RoPE to use. Can be one of ['default', 'linear', 'dynamic', 'yarn', 'longrope', + 'duplicated_method3'], with 'default' being the original RoPE implementation. + `factor` (`float`, *optional*): + Used with all rope types except 'default'. The scaling factor to apply to the RoPE embeddings. In + most scaling types, a `factor` of x will enable the model to handle sequences of length x * + original maximum pre-trained length. + `original_max_position_embeddings` (`int`, *optional*): + Used with 'dynamic', 'longrope' and 'duplicated_method3'. The original max position embeddings used during + pretraining. + `attention_factor` (`float`, *optional*): + Used with 'yarn' and 'longrope'. The scaling factor to be applied on the attention + computation. If unspecified, it defaults to value recommended by the implementation, using the + `factor` field to infer the suggested value. + `beta_fast` (`float`, *optional*): + Only used with 'yarn'. Parameter to set the boundary for extrapolation (only) in the linear + ramp function. If unspecified, it defaults to 32. + `beta_slow` (`float`, *optional*): + Only used with 'yarn'. Parameter to set the boundary for interpolation (only) in the linear + ramp function. If unspecified, it defaults to 1. + `short_factor` (`list[float]`, *optional*): + Only used with 'longrope'. The scaling factor to be applied to short contexts (< + `original_max_position_embeddings`). Must be a list of numbers with the same length as the hidden + size divided by the number of attention heads divided by 2 + `long_factor` (`list[float]`, *optional*): + Only used with 'longrope'. The scaling factor to be applied to long contexts (< + `original_max_position_embeddings`). Must be a list of numbers with the same length as the hidden + size divided by the number of attention heads divided by 2 + `low_freq_factor` (`float`, *optional*): + Only used with 'duplicated_method3'. Scaling factor applied to low frequency components of the RoPE + `high_freq_factor` (`float`, *optional*): + Only used with 'duplicated_method3'. Scaling factor applied to high frequency components of the RoPE + attention_bias (`bool`, *optional*, defaults to `False`): + Whether to use a bias in the query, key, value and output projection layers during self-attention. + attention_dropout (`float`, *optional*, defaults to 0.0): + The dropout ratio for the attention probabilities. + mlp_bias (`bool`, *optional*, defaults to `False`): + Whether to use a bias in up_proj, down_proj and gate_proj layers in the MLP layers. + head_dim (`int`, *optional*): + The attention head dimension. If None, it will default to hidden_size // num_attention_heads + + ```python + >>> from transformers import DuplicatedMethodModel, DuplicatedMethodConfig + + >>> # Initializing a DuplicatedMethod duplicated_method-7b style configuration + >>> configuration = DuplicatedMethodConfig() + + >>> # Initializing a model from the duplicated_method-7b style configuration + >>> model = DuplicatedMethodModel(configuration) + + >>> # Accessing the model configuration + >>> configuration = model.config + ```""" + + model_type = "duplicated_method" + keys_to_ignore_at_inference = ["past_key_values"] + # Default tensor parallel plan for base model `DuplicatedMethodModel` + base_model_tp_plan = { + "layers.*.self_attn.q_proj": "colwise", + "layers.*.self_attn.k_proj": "colwise", + "layers.*.self_attn.v_proj": "colwise", + "layers.*.self_attn.o_proj": "rowwise", + "layers.*.mlp.gate_proj": "colwise", + "layers.*.mlp.up_proj": "colwise", + "layers.*.mlp.down_proj": "rowwise", + } + base_model_pp_plan = { + "embed_tokens": (["input_ids"], ["inputs_embeds"]), + "layers": (["hidden_states", "attention_mask"], ["hidden_states"]), + "norm": (["hidden_states"], ["hidden_states"]), + } + + def __init__( + self, + vocab_size=32000, + hidden_size=4096, + intermediate_size=11008, + num_hidden_layers=32, + num_attention_heads=32, + num_key_value_heads=None, + hidden_act="silu", + max_position_embeddings=2048, + initializer_range=0.02, + rms_norm_eps=1e-6, + use_cache=True, + pad_token_id=None, + bos_token_id=1, + eos_token_id=2, + pretraining_tp=1, + tie_word_embeddings=False, + rope_theta=10000.0, + rope_scaling=None, + attention_bias=False, + attention_dropout=0.0, + mlp_bias=False, + head_dim=None, + **kwargs, + ): + self.vocab_size = vocab_size + self.max_position_embeddings = max_position_embeddings + self.hidden_size = hidden_size + self.intermediate_size = intermediate_size + self.num_hidden_layers = num_hidden_layers + self.num_attention_heads = num_attention_heads + + # for backward compatibility + if num_key_value_heads is None: + num_key_value_heads = num_attention_heads + + self.num_key_value_heads = num_key_value_heads + self.hidden_act = hidden_act + self.initializer_range = initializer_range + self.rms_norm_eps = rms_norm_eps + self.pretraining_tp = pretraining_tp + self.use_cache = use_cache + self.rope_theta = rope_theta + self.rope_scaling = rope_scaling + self.attention_bias = attention_bias + self.attention_dropout = attention_dropout + self.mlp_bias = mlp_bias + self.head_dim = head_dim if head_dim is not None else self.hidden_size // self.num_attention_heads + # Validate the correctness of rotary position embeddings parameters + # BC: if there is a 'type' field, copy it it to 'rope_type'. + if self.rope_scaling is not None and "type" in self.rope_scaling: + self.rope_scaling["rope_type"] = self.rope_scaling["type"] + rope_config_validation(self) + + super().__init__( + pad_token_id=pad_token_id, + bos_token_id=bos_token_id, + eos_token_id=eos_token_id, + tie_word_embeddings=tie_word_embeddings, + **kwargs, + ) + + @property + def vocab_size(self): + return 45 + + @vocab_size.setter + def vocab_size(self, value): + self.vocab_size = value diff --git a/examples/modular-transformers/modeling_multimodal2.py b/examples/modular-transformers/modeling_multimodal2.py index 64264ca30b..d7592047fc 100644 --- a/examples/modular-transformers/modeling_multimodal2.py +++ b/examples/modular-transformers/modeling_multimodal2.py @@ -498,6 +498,7 @@ class Multimodal2VisionPreTrainedModel(PreTrainedModel): supports_gradient_checkpointing = True _supports_sdpa = True _supports_flash_attn_2 = True + _supports_flash_attn_3 = True _supports_flex_attn = True _supports_attention_backend = True diff --git a/examples/modular-transformers/modeling_my_new_model2.py b/examples/modular-transformers/modeling_my_new_model2.py index 58bce52275..b87977cf4c 100644 --- a/examples/modular-transformers/modeling_my_new_model2.py +++ b/examples/modular-transformers/modeling_my_new_model2.py @@ -65,7 +65,7 @@ class MyNewModel2RotaryEmbedding(nn.Module): def __init__(self, config: MyNewModel2Config, device=None): super().__init__() # BC: "rope_type" was originally "type" - if hasattr(config, "rope_scaling") and config.rope_scaling is not None: + if hasattr(config, "rope_scaling") and isinstance(config.rope_scaling, dict): self.rope_type = config.rope_scaling.get("rope_type", config.rope_scaling.get("type")) else: self.rope_type = "default" @@ -290,6 +290,7 @@ class MyNewModel2PreTrainedModel(PreTrainedModel): _no_split_modules = ["MyNewModel2DecoderLayer"] _skip_keys_device_placement = ["past_key_values"] _supports_flash_attn_2 = True + _supports_flash_attn_3 = True _supports_sdpa = True _supports_flex_attn = True _supports_cache_class = True diff --git a/examples/modular-transformers/modeling_new_task_model.py b/examples/modular-transformers/modeling_new_task_model.py index c116b55d4d..5bcb2f4fb5 100644 --- a/examples/modular-transformers/modeling_new_task_model.py +++ b/examples/modular-transformers/modeling_new_task_model.py @@ -96,6 +96,7 @@ class NewTaskModelPreTrainedModel(PreTrainedModel): _supports_quantized_cache = True _supports_static_cache = True _supports_flash_attn_2 = True + _supports_flash_attn_3 = True _supports_sdpa = True _supports_flex_attn = True _supports_attention_backend = True diff --git a/examples/modular-transformers/modeling_super.py b/examples/modular-transformers/modeling_super.py index 6bc5faf78e..6f78eb972c 100644 --- a/examples/modular-transformers/modeling_super.py +++ b/examples/modular-transformers/modeling_super.py @@ -48,7 +48,7 @@ class SuperRotaryEmbedding(nn.Module): def __init__(self, config: SuperConfig, device=None): super().__init__() # BC: "rope_type" was originally "type" - if hasattr(config, "rope_scaling") and config.rope_scaling is not None: + if hasattr(config, "rope_scaling") and isinstance(config.rope_scaling, dict): self.rope_type = config.rope_scaling.get("rope_type", config.rope_scaling.get("type")) else: self.rope_type = "default" @@ -289,6 +289,7 @@ class SuperPreTrainedModel(PreTrainedModel): _no_split_modules = ["SuperDecoderLayer"] _skip_keys_device_placement = ["past_key_values"] _supports_flash_attn_2 = True + _supports_flash_attn_3 = True _supports_sdpa = True _supports_flex_attn = True _supports_cache_class = True diff --git a/examples/modular-transformers/modular_duplicated_method.py b/examples/modular-transformers/modular_duplicated_method.py new file mode 100644 index 0000000000..06d5e03437 --- /dev/null +++ b/examples/modular-transformers/modular_duplicated_method.py @@ -0,0 +1,11 @@ +from transformers.models.llama.configuration_llama import LlamaConfig + + +class DuplicatedMethodConfig(LlamaConfig): + @property + def vocab_size(self): + return 45 + + @vocab_size.setter + def vocab_size(self, value): + self.vocab_size = value diff --git a/utils/modular_model_converter.py b/utils/modular_model_converter.py index 2752c9b228..e138277b03 100644 --- a/utils/modular_model_converter.py +++ b/utils/modular_model_converter.py @@ -972,12 +972,37 @@ def replace_class_node( # Use all original modeling attributes, and potentially override some with values in the modular new_class_attributes = list({**original_modeling_class_attributes, **modular_class_attributes}.values()) - original_modeling_methods = { - node.name.value: node for node in original_modeling_node.body.body if m.matches(node, m.FunctionDef()) - } - modular_methods = { - node.name.value: node for node in modular_class_node.body.body if m.matches(node, m.FunctionDef()) - } + # Check class methods defined in the modular and associated modeling + original_modeling_methods = {} + for node in original_modeling_node.body.body: + if m.matches(node, m.FunctionDef()): + # Due to the @property and @name.setter decorators, methods can sometimes have the same name, so we need a way + # to separate them + if node.name.value in original_modeling_methods: + # If it's already present, and the decorator is @property, it means the node already added was the setter + if node.decorators[0].decorator.value == "property": + original_modeling_methods[f"{node.name.value}_setter"] = original_modeling_methods[node.name.value] + original_modeling_methods[node.name.value] = node + # In this case current node is the setter + else: + original_modeling_methods[f"{node.name.value}_setter"] = node + else: + original_modeling_methods[node.name.value] = node + modular_methods = {} + for node in modular_class_node.body.body: + if m.matches(node, m.FunctionDef()): + # Due to the @property and @name.setter decorators, methods can sometimes have the same name, so we need a way + # to separate them + if node.name.value in modular_methods: + # If it's already present, and the decorator is @property, it means the node already added was the setter + if node.decorators[0].decorator.value == "property": + modular_methods[f"{node.name.value}_setter"] = modular_methods[node.name.value] + modular_methods[node.name.value] = node + # In this case current node is the setter + else: + modular_methods[f"{node.name.value}_setter"] = node + else: + modular_methods[node.name.value] = node new_class_methods = [] # Iterate over the methods of the original modeling code, and add them to the list of methods to add