sarm_llama.py

import torch
import torch.nn as nn

from typing import List, Optional, Union, Tuple
from transformers import LlamaConfig
from transformers.cache_utils import Cache, DynamicCache, StaticCache
from transformers.utils import logging
from transformers.modeling_outputs import (
    SequenceClassifierOutputWithPast,
    BaseModelOutputWithPast
)
from transformers.models.llama.modeling_llama import (
    LlamaDecoderLayer, 
    LlamaRMSNorm, 
    LlamaRotaryEmbedding,
    LlamaPreTrainedModel
)
from transformers.modeling_attn_mask_utils import AttentionMaskConverter

# Local
from sae import TopkSAE, pre_process, Normalized_MSE_loss, Masked_Normalized_MSE_loss


logger = logging.get_logger(__name__)
#==========================================================================================================================================================================
#==========================================================================================================================================================================
class MyLlamaModel(LlamaPreTrainedModel):
    def __init__(
            self, 
            config: LlamaConfig, 
            hidden_state_source_layer: int=None
    ):
        if hidden_state_source_layer==None:
            # default 1/2
            hidden_state_source_layer = int(config.num_hidden_layers/2)
            
        super().__init__(config)
        self.hidden_state_source_layer = hidden_state_source_layer
        self.padding_idx = config.pad_token_id
        self.vocab_size = config.vocab_size

        self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size, self.padding_idx)
        self.layers = nn.ModuleList(
            [LlamaDecoderLayer(config, layer_idx) for layer_idx in range(hidden_state_source_layer)]
        )
        self.norm = LlamaRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
        self.rotary_emb = LlamaRotaryEmbedding(config=config)
        self.gradient_checkpointing = False
        if getattr(config, "pretraining_tp", 1) != 1:
            logger.warn("`pretraining_tp` is deprecated, please use `model.tensor_parallel` instead.")

        # Initialize weights and apply final processing
        self.post_init()

    def get_input_embeddings(self):
        return self.embed_tokens

    def set_input_embeddings(self, value):
        self.embed_tokens = value

    def forward(
        self,
        input_ids: torch.LongTensor = None,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        past_key_values: Optional[Union[Cache, List[torch.FloatTensor]]] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        use_cache: Optional[bool] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
        cache_position: Optional[torch.LongTensor] = None,
    ) -> Union[Tuple, BaseModelOutputWithPast]:
        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
        output_hidden_states = (
            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
        )
        use_cache = use_cache if use_cache is not None else self.config.use_cache
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        if (input_ids is None) ^ (inputs_embeds is not None):
            raise ValueError("You must specify exactly one of input_ids or inputs_embeds")

        if self.gradient_checkpointing and self.training and use_cache:
            logger.warning_once(
                "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`."
            )
            use_cache = False

        if inputs_embeds is None:
            inputs_embeds = self.embed_tokens(input_ids)

        # kept for BC (non `Cache` `past_key_values` inputs)
        return_legacy_cache = False
        if use_cache and not isinstance(past_key_values, Cache):
            return_legacy_cache = True
            if past_key_values is None:
                past_key_values = DynamicCache()
            else:
                past_key_values = DynamicCache.from_legacy_cache(past_key_values)
                logger.warning_once(
                    "We detected that you are passing `past_key_values` as a tuple of tuples. This is deprecated and "
                    "will be removed in v4.47. Please convert your cache or use an appropriate `Cache` class "
                    "(https://huggingface.co/docs/transformers/kv_cache#legacy-cache-format)"
                )

        if cache_position is None:
            past_seen_tokens = past_key_values.get_seq_length() if past_key_values is not None else 0
            cache_position = torch.arange(
                past_seen_tokens, past_seen_tokens + inputs_embeds.shape[1], device=inputs_embeds.device
            )
        if position_ids is None:
            position_ids = cache_position.unsqueeze(0)

        causal_mask = self._update_causal_mask(
            attention_mask, inputs_embeds, cache_position, past_key_values, output_attentions
        )
        hidden_states = inputs_embeds

        # create position embeddings to be shared across the decoder layers
        position_embeddings = self.rotary_emb(hidden_states, position_ids)

        # decoder layers
        all_hidden_states = () if output_hidden_states else None
        all_self_attns = () if output_attentions else None
        next_decoder_cache = None
        

        for decoder_layer in self.layers:
            if output_hidden_states:
                all_hidden_states += (hidden_states,)

            if self.gradient_checkpointing and self.training:
                layer_outputs = self._gradient_checkpointing_func(
                    decoder_layer.__call__,
                    hidden_states,
                    causal_mask,
                    position_ids,
                    past_key_values,
                    output_attentions,
                    use_cache,
                    cache_position,
                    position_embeddings,
                )
            else:
                layer_outputs = decoder_layer(
                    hidden_states,
                    attention_mask=causal_mask,
                    position_ids=position_ids,
                    past_key_value=past_key_values,
                    output_attentions=output_attentions,
                    use_cache=use_cache,
                    cache_position=cache_position,
                    position_embeddings=position_embeddings,
                )

            hidden_states = layer_outputs[0]

            if use_cache:
                next_decoder_cache = layer_outputs[2 if output_attentions else 1]

            if output_attentions:
                all_self_attns += (layer_outputs[1],)

        # hidden_states = self.norm(hidden_states)

        # add hidden states from the last decoder layer
        if output_hidden_states:
            all_hidden_states += (hidden_states,)

        next_cache = next_decoder_cache if use_cache else None
        if return_legacy_cache:
            next_cache = next_cache.to_legacy_cache()

        if not return_dict:
            return tuple(v for v in [hidden_states, next_cache, all_hidden_states, all_self_attns] if v is not None)
        return BaseModelOutputWithPast(
            last_hidden_state=hidden_states,
            past_key_values=next_cache,
            hidden_states=all_hidden_states,
            attentions=all_self_attns,
        )
    
    def _update_causal_mask(
        self,
        attention_mask: torch.Tensor,
        input_tensor: torch.Tensor,
        cache_position: torch.Tensor,
        past_key_values: Cache,
        output_attentions: bool,
    ):
        if self.config._attn_implementation == "flash_attention_2":
            if attention_mask is not None and 0.0 in attention_mask:
                return attention_mask
            return None

        # For SDPA, when possible, we will rely on its `is_causal` argument instead of its `attn_mask` argument, in
        # order to dispatch on Flash Attention 2. This feature is not compatible with static cache, as SDPA will fail
        # to infer the attention mask.
        past_seen_tokens = past_key_values.get_seq_length() if past_key_values is not None else 0
        using_static_cache = isinstance(past_key_values, StaticCache)

        # When output attentions is True, sdpa implementation's forward method calls the eager implementation's forward
        if self.config._attn_implementation == "sdpa" and not using_static_cache and not output_attentions:
            if AttentionMaskConverter._ignore_causal_mask_sdpa(
                attention_mask,
                inputs_embeds=input_tensor,
                past_key_values_length=past_seen_tokens,
                is_training=self.training,
            ):
                return None

        dtype, device = input_tensor.dtype, input_tensor.device
        sequence_length = input_tensor.shape[1]
        if using_static_cache:
            target_length = past_key_values.get_max_cache_shape()
        else:
            target_length = (
                attention_mask.shape[-1]
                if isinstance(attention_mask, torch.Tensor)
                else past_seen_tokens + sequence_length + 1
            )

        # In case the provided `attention` mask is 2D, we generate a causal mask here (4D).
        causal_mask = self._prepare_4d_causal_attention_mask_with_cache_position(
            attention_mask,
            sequence_length=sequence_length,
            target_length=target_length,
            dtype=dtype,
            device=device,
            cache_position=cache_position,
            batch_size=input_tensor.shape[0],
        )

        if (
            self.config._attn_implementation == "sdpa"
            and attention_mask is not None
            and attention_mask.device.type == "cuda"
            and not output_attentions
        ):
            # Attend to all tokens in fully masked rows in the causal_mask, for example the relevant first rows when
            # using left padding. This is required by F.scaled_dot_product_attention memory-efficient attention path.
            # Details: https://github.com/pytorch/pytorch/issues/110213
            min_dtype = torch.finfo(dtype).min
            causal_mask = AttentionMaskConverter._unmask_unattended(causal_mask, min_dtype)

        return causal_mask

    @staticmethod
    def _prepare_4d_causal_attention_mask_with_cache_position(
        attention_mask: torch.Tensor,
        sequence_length: int,
        target_length: int,
        dtype: torch.dtype,
        device: torch.device,
        cache_position: torch.Tensor,
        batch_size: int,
        **kwargs,
    ):
        """
        Creates a causal 4D mask of shape `(batch_size, 1, query_length, key_value_length)` from a 2D mask of shape
        `(batch_size, key_value_length)`, or if the input `attention_mask` is already 4D, do nothing.

        Args:
            attention_mask (`torch.Tensor`):
                A 2D attention mask of shape `(batch_size, key_value_length)` or a 4D attention mask of shape
                `(batch_size, 1, query_length, key_value_length)`.
            sequence_length (`int`):
                The sequence length being processed.
            target_length (`int`):
                The target length: when generating with static cache, the mask should be as long as the static cache,
                to account for the 0 padding, the part of the cache that is not filled yet.
            dtype (`torch.dtype`):
                The dtype to use for the 4D attention mask.
            device (`torch.device`):
                The device to plcae the 4D attention mask on.
            cache_position (`torch.Tensor`):
                Indices depicting the position of the input sequence tokens in the sequence.
            batch_size (`torch.Tensor`):
                Batch size.
        """
        if attention_mask is not None and attention_mask.dim() == 4:
            # In this case we assume that the mask comes already in inverted form and requires no inversion or slicing.
            causal_mask = attention_mask
        else:
            min_dtype = torch.finfo(dtype).min
            causal_mask = torch.full(
                (sequence_length, target_length), fill_value=min_dtype, dtype=dtype, device=device
            )
            if sequence_length != 1:
                causal_mask = torch.triu(causal_mask, diagonal=1)
            causal_mask *= torch.arange(target_length, device=device) > cache_position.reshape(-1, 1)
            causal_mask = causal_mask[None, None, :, :].expand(batch_size, 1, -1, -1)
            if attention_mask is not None:
                causal_mask = causal_mask.clone()  # copy to contiguous memory for in-place edit
                mask_length = attention_mask.shape[-1]
                padding_mask = causal_mask[:, :, :, :mask_length] + attention_mask[:, None, None, :]
                padding_mask = padding_mask == 0
                causal_mask[:, :, :, :mask_length] = causal_mask[:, :, :, :mask_length].masked_fill(
                    padding_mask, min_dtype
                )

        return causal_mask




#==========================================================================================================================================================================
#============================================               从LlamaForSequenceClassification为原型，修改为SAE4RM的形式            =============================================
#==========================================================================================================================================================================


class LlamaSARM(LlamaPreTrainedModel):
    def __init__(
            self, config, sae_hidden_state_source_layer, sae_latent_size, sae_k, 
            sae_use_sequence_level=False,
            sarm_use_topk=False, 
            sarm_train_mode=1
    ):
        super().__init__(config)
        self.num_labels = config.num_labels
        self.model = MyLlamaModel(config, hidden_state_source_layer=sae_hidden_state_source_layer)
        
        self.sae_use_sequence_level = sae_use_sequence_level
        self.sarm_use_topk = sarm_use_topk
        self.sarm_train_mode = sarm_train_mode

        self.score = nn.Linear(sae_latent_size, self.num_labels, bias=False)
        self.sae = TopkSAE(hidden_size=self.model.config.hidden_size, latent_size=sae_latent_size, k=sae_k)

        if self.sarm_train_mode==0:
            for p in self.model.parameters():
                p.requires_grad_(False)
        if self.sarm_train_mode==0 or self.sarm_train_mode==1:
            for p in self.sae.parameters():
                p.requires_grad_(False)

        # Initialize weights and apply final processing
        self.post_init()
 

    def get_input_embeddings(self):
        return self.model.embed_tokens

    def set_input_embeddings(self, value):
        self.model.embed_tokens = value

    
    def forward(
        self,
        input_ids: Optional[torch.LongTensor] = None,
        attention_mask: Optional[torch.Tensor] = None,
        assistant_masks: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        past_key_values: Optional[Union[Cache, List[torch.FloatTensor]]] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        labels: Optional[torch.LongTensor] = None,
        use_cache: Optional[bool] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
    ) -> Union[Tuple, SequenceClassifierOutputWithPast]:
        r"""
        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        transformer_outputs = self.model(
            input_ids,
            attention_mask=attention_mask,
            position_ids=position_ids,
            past_key_values=past_key_values,
            inputs_embeds=inputs_embeds,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )
        hidden_states = transformer_outputs[0]


        h, _, _ = pre_process(hidden_states)
        sae_features = self.sae.pre_acts(h)
        if self.sarm_use_topk:
            sae_features = self.sae.get_latents(sae_features)


        logits = self.score(sae_features)

        if input_ids is not None:
            batch_size = input_ids.shape[0]
        else:
            batch_size = inputs_embeds.shape[0]

        if self.config.pad_token_id is None and batch_size != 1:
            raise ValueError("Cannot handle batch sizes > 1 if no padding token is defined.")
        if self.config.pad_token_id is None:
            sequence_lengths = -1
        else:
            if input_ids is not None:
                # if no pad token found, use modulo instead of reverse indexing for ONNX compatibility
                sequence_lengths = torch.eq(input_ids, self.config.pad_token_id).int().argmax(-1) - 1
                sequence_lengths = sequence_lengths % input_ids.shape[-1]
                sequence_lengths = sequence_lengths.to(logits.device)
            else:
                sequence_lengths = -1
        # ensure last_token is <|eot_id|>
        assert ((input_ids[torch.arange(batch_size, device=logits.device), sequence_lengths]!=torch.ones(batch_size, device=logits.device)*128009).sum() == 0).item()
        
        # joint training
        rec_loss = None
        if self.sarm_train_mode==2:
            if not self.sarm_use_topk:
                sae_features_t = self.sae.get_latents(sae_features)
            h_hat = self.sae.decode(sae_features_t)
            rec_loss = Masked_Normalized_MSE_loss(h, h_hat, assistant_masks)
        elif self.sarm_train_mode==3 and not self.sae_use_sequence_level:
            h_d = h.detach()
            _, h_hat = self.sae(h_d)
            rec_loss = Masked_Normalized_MSE_loss(h_d, h_hat, assistant_masks)        
        elif self.sarm_train_mode==3 and self.sae_use_sequence_level:
            h_d = h.detach()
            sequence_lengths_t = sequence_lengths.view(-1,1,1)
            last_token_mask = torch.zeros([h_d.shape[0] ,1 ,h_d.shape[1]], device=h_d.device)
            last_token_mask.scatter_(-1, sequence_lengths_t, torch.ones_like(sequence_lengths_t, dtype=last_token_mask.dtype))
            
            # h_d -> (bs, seq_len, d), last_token_mask -> (bs, 1, seq_len)
            h_d = torch.matmul(last_token_mask.to(h_d.dtype), h_d) 
            
            _, h_hat = self.sae(h_d)
            rec_loss = Normalized_MSE_loss(h_d, h_hat)       


        pooled_logits = logits[torch.arange(batch_size, device=logits.device), sequence_lengths]


        loss = None
        if labels is not None:
            loss = self.loss_function(logits=logits, labels=labels, pooled_logits=pooled_logits, config=self.config)
        if rec_loss is not None:
            loss = rec_loss

        if not return_dict:
            output = (pooled_logits,) + transformer_outputs[1:]
            return ((loss,) + output) if loss is not None else output

        return SequenceClassifierOutputWithPast(
            loss=loss,
            logits=pooled_logits,
            past_key_values=transformer_outputs.past_key_values,
            hidden_states=transformer_outputs.hidden_states,
            attentions=transformer_outputs.attentions,
        )



#==========================================================================================================================================================================
#=================================               从LlamaForSequenceClassification为原型，可以放在任意层的score head(两层MLP)            ========================================
#==========================================================================================================================================================================
class LlamaBaseline(LlamaPreTrainedModel):
    def __init__(
            self, config, sae_hidden_state_source_layer, sae_latent_size
    ):
        super().__init__(config)
        self.num_labels = config.num_labels
        self.model = MyLlamaModel(config, hidden_state_source_layer=sae_hidden_state_source_layer)
        
        self.untrained_sae_encoder = nn.Linear(self.model.config.hidden_size, sae_latent_size)
        self.score = nn.Linear(sae_latent_size, self.num_labels, bias=False)
        
        # Initialize weights and apply final processing
        self.post_init()
 

    def get_input_embeddings(self):
        return self.model.embed_tokens

    def set_input_embeddings(self, value):
        self.model.embed_tokens = value

    
    def forward(
        self,
        input_ids: Optional[torch.LongTensor] = None,
        attention_mask: Optional[torch.Tensor] = None,
        assistant_masks: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        past_key_values: Optional[Union[Cache, List[torch.FloatTensor]]] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        labels: Optional[torch.LongTensor] = None,
        use_cache: Optional[bool] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
    ) -> Union[Tuple, SequenceClassifierOutputWithPast]:
        r"""
        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        transformer_outputs = self.model(
            input_ids,
            attention_mask=attention_mask,
            position_ids=position_ids,
            past_key_values=past_key_values,
            inputs_embeds=inputs_embeds,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )
        hidden_states = transformer_outputs[0]
        logits = self.score(self.untrained_sae_encoder(hidden_states))

        if input_ids is not None:
            batch_size = input_ids.shape[0]
        else:
            batch_size = inputs_embeds.shape[0]

        if self.config.pad_token_id is None and batch_size != 1:
            raise ValueError("Cannot handle batch sizes > 1 if no padding token is defined.")
        if self.config.pad_token_id is None:
            sequence_lengths = -1
        else:
            if input_ids is not None:
                # if no pad token found, use modulo instead of reverse indexing for ONNX compatibility
                sequence_lengths = torch.eq(input_ids, self.config.pad_token_id).int().argmax(-1) - 1
                sequence_lengths = sequence_lengths % input_ids.shape[-1]
                sequence_lengths = sequence_lengths.to(logits.device)
            else:
                sequence_lengths = -1

        pooled_logits = logits[torch.arange(batch_size, device=logits.device), sequence_lengths]

        loss = None
        if labels is not None:
            loss = self.loss_function(logits=logits, labels=labels, pooled_logits=pooled_logits, config=self.config)

        if not return_dict:
            output = (pooled_logits,) + transformer_outputs[1:]
            return ((loss,) + output) if loss is not None else output

        return SequenceClassifierOutputWithPast(
            loss=loss,
            logits=pooled_logits,
            past_key_values=transformer_outputs.past_key_values,
            hidden_states=transformer_outputs.hidden_states,
            attentions=transformer_outputs.attentions,
        )
    

class LlamaBaselineFrozen(LlamaPreTrainedModel):
    def __init__(
            self, config, sae_hidden_state_source_layer, sae_latent_size
    ):
        super().__init__(config)
        self.num_labels = config.num_labels
        self.model = MyLlamaModel(config, hidden_state_source_layer=sae_hidden_state_source_layer)
        
        self.untrained_sae_encoder = nn.Linear(self.model.config.hidden_size, sae_latent_size)
        self.score = nn.Linear(sae_latent_size, self.num_labels, bias=False)
        
        # Initialize weights and apply final processing
        self.post_init()
 

    def get_input_embeddings(self):
        return self.model.embed_tokens

    def set_input_embeddings(self, value):
        self.model.embed_tokens = value

    
    def forward(
        self,
        input_ids: Optional[torch.LongTensor] = None,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        past_key_values: Optional[Union[Cache, List[torch.FloatTensor]]] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        labels: Optional[torch.LongTensor] = None,
        use_cache: Optional[bool] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
    ) -> Union[Tuple, SequenceClassifierOutputWithPast]:
        r"""
        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        transformer_outputs = self.model(
            input_ids,
            attention_mask=attention_mask,
            position_ids=position_ids,
            past_key_values=past_key_values,
            inputs_embeds=inputs_embeds,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )
        hidden_states = transformer_outputs[0]
        logits = self.score(self.untrained_sae_encoder(hidden_states))

        if input_ids is not None:
            batch_size = input_ids.shape[0]
        else:
            batch_size = inputs_embeds.shape[0]

        if self.config.pad_token_id is None and batch_size != 1:
            raise ValueError("Cannot handle batch sizes > 1 if no padding token is defined.")
        if self.config.pad_token_id is None:
            sequence_lengths = -1
        else:
            if input_ids is not None:
                # if no pad token found, use modulo instead of reverse indexing for ONNX compatibility
                sequence_lengths = torch.eq(input_ids, self.config.pad_token_id).int().argmax(-1) - 1
                sequence_lengths = sequence_lengths % input_ids.shape[-1]
                sequence_lengths = sequence_lengths.to(logits.device)
            else:
                sequence_lengths = -1

        pooled_logits = logits[torch.arange(batch_size, device=logits.device), sequence_lengths]

        loss = None
        if labels is not None:
            loss = self.loss_function(logits=logits, labels=labels, pooled_logits=pooled_logits, config=self.config)

        if not return_dict:
            output = (pooled_logits,) + transformer_outputs[1:]
            return ((loss,) + output) if loss is not None else output

        return SequenceClassifierOutputWithPast(
            loss=loss,
            logits=pooled_logits,
            past_key_values=transformer_outputs.past_key_values,
            hidden_states=transformer_outputs.hidden_states,
            attentions=transformer_outputs.attentions,
        )