a h @@sZddlmZmZddlZddlmZddlmZmZddl m Z ddl m Z ddl mZdd lmZdd lmZdd lmZd d lmZmZmZmZmZmZmZmZmZd dlm Z e!e"Z#GdddeZ$GdddeZ%GdddeZ&GdddeZ'Gddde Z(GdddeZ)GdddeZ*GdddeZ+Gdd d eZ,gd!Z-dS)")CallableOptionalN)Cache)PretrainedConfiglayer_type_validation)FlashAttentionKwargs)rope_config_validation)ALL_ATTENTION_FUNCTIONS)Unpack)logging)deprecate_kwarg) LlamaAttentionLlamaDecoderLayerLlamaForCausalLMLlamaForQuestionAnsweringLlamaForSequenceClassificationLlamaForTokenClassificationLlamaPreTrainedModelapply_rotary_pos_embeager_attention_forward) Qwen2ModelcsfeZdZdZdZdgZddddddddZdgdgfd d gd gfd gd gfd Zd fdd ZZ S)! SmolLM3Configa This is the configuration class to store the configuration of a [`SmolLM3Model`]. It is used to instantiate a SmolLM3 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 SmolLM3 3B. e.g. [HuggingFaceTB/SmolLM3-3B](https://huggingface.co/HuggingFaceTB/SmolLM3-3B) 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 128256): Vocabulary size of the SmolLM3 model. Defines the number of different tokens that can be represented by the `inputs_ids` passed when calling [`SmolLM3Model`] hidden_size (`int`, *optional*, defaults to 2048): Dimension of the hidden representations. intermediate_size (`int`, *optional*, defaults to 11008): Dimension of the MLP representations. num_hidden_layers (`int`, *optional*, defaults to 36): Number of hidden layers in the Transformer encoder. num_attention_heads (`int`, *optional*, defaults to 16): Number of attention heads for each attention layer in the Transformer encoder. num_key_value_heads (`int`, *optional*, defaults to 4): 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 checkout [this paper](https://huggingface.co/papers/2305.13245). If it is not specified, will default to `16`. 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 32768): The maximum sequence length that this model might ever be used with. 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*, defaults to 128004): The id of the padding token. bos_token_id (`int`, *optional*, defaults to 128000): The id of the beginning of sentence token. eos_token_id (`int`, *optional*, defaults to 128001): The id of the end of sentence token. rope_theta (`float`, *optional*, defaults to 2000000.0): 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', 'llama3'], 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 'llama3'. 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 'llama3'. Scaling factor applied to low frequency components of the RoPE `high_freq_factor` (`float`, *optional*): Only used with 'llama3'. Scaling factor applied to high frequency components of the RoPE use_sliding_window (`bool`, *optional*, defaults to `False`): Whether to use sliding window attention. sliding_window (`int`, *optional*): Sliding window attention (SWA) window size. If not specified, will default to `None`. no_rope_layers (`List[int]`, *optional*): List with at least the same length as the number of layers in the model. A `1` at an index position indicates that the corresponding layer will use RoPE, while a `0` indicates that it's a NoPE layer. no_rope_layer_interval (`int`, *optional*, defaults to 4): If `no_rope_layers` is `None`, it will be created using a NoPE layer every `no_rope_layer_interval` layers. layer_types (`list`, *optional*): Attention pattern for each layer. Automatically computed based on sliding window and NoPE settings. 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. ```python >>> from transformers import SmolLM3Model, SmolLM3Config >>> # Initializing a SmolLM3 style configuration >>> configuration = SmolLM3Config() >>> # Initializing a model from the SmolLM3 style configuration >>> model = SmolLM3Model(configuration) >>> # Accessing the model configuration >>> configuration = model.config ```Zsmollm3past_key_valuesZcolwiseZrowwise)zlayers.*.self_attn.q_projzlayers.*.self_attn.k_projzlayers.*.self_attn.v_projzlayers.*.self_attn.o_projzlayers.*.mlp.gate_projzlayers.*.mlp.up_projzlayers.*.mlp.down_projZ input_idsZ inputs_embeds hidden_statesattention_mask)Z embed_tokensZlayersZnorm+$silu{Gz?ư>T>ANFc sVtjf| | |d|||_||_||_||_||_||_||_||_ ||_ |dur\|}||_ ||_ | |_ | |_| |_||_||_||_||_|durfddt|D|_n||_|_|durg}t|D]:}|j|}|r|dur|s|dq|dq||_t|j|jdurJd|jvrJ|jd|jd<t|dS)N) pad_token_id bos_token_id eos_token_idcs g|]}t|ddkqS)r)int).0 layer_idxno_rope_layer_intervalg/var/www/html/assistant/venv/lib/python3.9/site-packages/transformers/models/smollm3/modular_smollm3.py sz*SmolLM3Config.__init__..sliding_attentionZfull_attentiontypeZ rope_type)super__init__ vocab_sizemax_position_embeddingsmlp_bias hidden_sizeintermediate_sizenum_hidden_layersnum_attention_headsuse_sliding_windowsliding_windownum_key_value_heads hidden_actinitializer_range rms_norm_eps use_cache rope_theta rope_scalingattention_biasattention_dropoutrangeno_rope_layersr4append layer_typesrr )selfr<r?r@rArBrErFr=rGrHrIr,r-r.rJrKrCrDrOr4rQrLrMr>kwargsr2Zhas_rope __class__r3r6r;sZ        zSmolLM3Config.__init__)rrrr r!r"r#r$r%r&Tr'r(r)r*NFNNr"NFr+F) __name__ __module__ __qualname____doc__Z model_typeZkeys_to_ignore_at_inferenceZbase_model_tp_planZbase_model_pp_planr; __classcell__r5r5rTr6r-sPs   rc seZdZeedfdd Zeddddd eje ejejfe eje e e ej e ee eje ejfd d d ZZS) SmolLM3Attentionconfigr2cs>t|||j||_|jr4|j|dkr4|jnd|_dS)Nr8)r:r;rOuse_roperCrQrDrRr]r2rTr5r6r; s  zSmolLM3Attention.__init__Zpast_key_valuerz4.58)new_nameversionN)rposition_embeddingsrrcache_positionrSreturncKs*|jdd}g|d|jR}|||dd} |||dd} |||dd} |jr|\} } t| | | | \} } |durd|i}| | | |j |\} } t }|j j dkrt|j j }||| | | |f|jsdn|j|j|jd|\}}|jg|dR}||}||fS)Nr/rrceagerr+)ZdropoutscalingrD)shapeZhead_dimZq_projviewZ transposeZk_projZv_projr^rupdater2rr]Z_attn_implementationr ZtrainingrMrgrDZreshape contiguousZo_proj)rRrrbrrrcrSZ input_shapeZ hidden_shapeZ query_statesZ key_statesZ value_statescossinZ cache_kwargsZattention_interfaceZ attn_outputZ attn_weightsr5r5r6forwards<      zSmolLM3Attention.forward)NN)rVrWrXrr0r;r torchZTensortuplerrZ LongTensorr rrnrZr5r5rTr6r[ s r[cs$eZdZeedfdd ZZS)SmolLM3DecoderLayerr\cst|||j||_dS)N)r:r;rQZattention_typer_rTr5r6r;DszSmolLM3DecoderLayer.__init__)rVrWrXrr0r;rZr5r5rTr6rqCsrqc@s eZdZdS)SmolLM3PreTrainedModelNrVrWrXr5r5r5r6rrIsrrc@s eZdZdS) SmolLM3ModelNrsr5r5r5r6rtMsrtc@s eZdZdS)SmolLM3ForCausalLMNrsr5r5r5r6ruQsruc@s eZdZdS) SmolLM3ForSequenceClassificationNrsr5r5r5r6rvUsrvc@s eZdZdS)SmolLM3ForTokenClassificationNrsr5r5r5r6rwYsrwc@s eZdZdS)SmolLM3ForQuestionAnsweringNrsr5r5r5r6rx]srx)rrrrtrurvrwrx).typingrrroZ cache_utilsrZconfiguration_utilsrrZmodeling_flash_attention_utilsrZmodeling_rope_utilsr Zmodeling_utilsr Zprocessing_utilsr utilsr Zutils.deprecationr Zllama.modeling_llamarrrrrrrrrZqwen2.modeling_qwen2rZ get_loggerrVloggerrr[rqrrrtrurvrwrx__all__r5r5r5r6s.       ,  ^9