modeling_recast_llama.py

# filename: recastmlp_llama_model.py
from .configuration_recast_llama import RECAST8b_llama
from transformers import PreTrainedModel
import math
import torch
import torch.nn as nn
import torch.nn.functional as F
from typing import Optional, Tuple, Union, List
from transformers import AutoConfig
from transformers.utils import logging
from transformers.cache_utils import Cache, StaticCache
from transformers.modeling_outputs import CausalLMOutputWithPast
from transformers.generation import GenerationMixin
from transformers.modeling_attn_mask_utils import AttentionMaskConverter
from transformers.models.llama.modeling_llama import (
    LlamaDecoderLayer,
    LlamaRotaryEmbedding,
    LlamaRMSNorm,
    apply_rotary_pos_emb,
    repeat_kv,
)
from transformers.modeling_outputs import BaseModelOutputWithPast
import copy

logger = logging.get_logger(__name__)


class MLPTemplateBank(nn.Module):
    def __init__(self, config, coef_rows, coef_columns):
        super().__init__()
        self.hidden_size = config.hidden_size
        self.intermediate_size = config.intermediate_size
        self.coef_shape = (coef_rows, coef_columns)

        assert coef_columns is not None, "coef_columns must not be None"

        # Ensure divisibility for proper reshaping
        assert (
            self.hidden_size * self.intermediate_size
        ) % coef_rows == 0, f"hidden_size * intermediate_size ({self.hidden_size * self.intermediate_size}) must be divisible by coef_rows ({coef_rows})"

        template_size = self.hidden_size * self.intermediate_size // coef_rows

        self.up_templates = nn.Parameter(torch.randn(coef_columns, template_size))
        self.gate_templates = nn.Parameter(torch.randn(coef_columns, template_size))

        # Better initialization
        nn.init.xavier_uniform_(self.up_templates)
        nn.init.xavier_uniform_(self.gate_templates)

    def forward(self, up_coeffs, gate_coeffs):
        # Compute chunked weights
        up_chunks = torch.matmul(up_coeffs, self.up_templates)
        gate_chunks = torch.matmul(gate_coeffs, self.gate_templates)

        # Reshape to final weight matrices
        up_weights = up_chunks.reshape(self.intermediate_size, self.hidden_size)
        gate_weights = gate_chunks.reshape(self.intermediate_size, self.hidden_size)

        return up_weights, gate_weights


class SharedLlamaMLP(nn.Module):
    def __init__(self, config, bank):
        super().__init__()
        self.config = config
        self.bank = bank
        self.hidden_size = config.hidden_size
        self.intermediate_size = config.intermediate_size
        self.down_proj = nn.Linear(
            config.intermediate_size, config.hidden_size, bias=False
        )

        # Initialize coefficients with proper shapes
        self.up_coefficients = nn.Parameter(torch.randn(bank.coef_shape))
        self.gate_coefficients = nn.Parameter(torch.randn(bank.coef_shape))

        # Initialize with small random values instead of ones, then orthogonalize
        nn.init.orthogonal_(self.up_coefficients)
        nn.init.orthogonal_(self.gate_coefficients)

        if config.mlp_bias:
            self.gate_bias = nn.Parameter(torch.zeros(self.intermediate_size))
            self.up_bias = nn.Parameter(torch.zeros(self.intermediate_size))
        else:
            self.register_parameter("gate_bias", None)
            self.register_parameter("up_bias", None)

        self.act_fn = F.silu

    def forward(self, x):
        # Generate weights using template bank
        up_weights, gate_weights = self.bank(
            self.up_coefficients, self.gate_coefficients  # Fixed order
        )
        # Match dtype
        target_dtype = x.dtype
        up_weights = up_weights.to(target_dtype)
        gate_weights = gate_weights.to(target_dtype)
        # Apply SwiGLU: SiLU(gate * x) * up * x
        hidden_states = self.act_fn(
            F.linear(x, gate_weights, self.gate_bias)
        ) * F.linear(x, up_weights, self.up_bias)
        output = self.down_proj(hidden_states)

        return output


class AttTemplateBank(nn.Module):
    def __init__(self, config, coef_rows, coef_columns):
        super().__init__()
        self.hidden_size = config.hidden_size
        self.num_heads = config.num_attention_heads
        self.head_dim = config.hidden_size // config.num_attention_heads
        self.num_key_value_heads = getattr(
            config, "num_key_value_heads", config.num_attention_heads
        )
        self.kv_dim = self.num_key_value_heads * self.head_dim
        self.coef_shape = (coef_rows, coef_columns)

        # Ensure divisibility
        assert (
            self.hidden_size * self.hidden_size
        ) % coef_rows == 0, "Q projection size must be divisible by coef_rows"
        assert (
            self.kv_dim * self.hidden_size
        ) % coef_rows == 0, "K/V projection size must be divisible by coef_rows"

        # Create templates for Q, K, V
        self.q_templates = nn.Parameter(
            torch.randn(coef_columns, self.hidden_size * self.hidden_size // coef_rows)
        )
        self.k_templates = nn.Parameter(
            torch.randn(coef_columns, self.kv_dim * self.hidden_size // coef_rows)
        )
        self.v_templates = nn.Parameter(
            torch.randn(coef_columns, self.kv_dim * self.hidden_size // coef_rows)
        )

        # Initialize templates
        nn.init.xavier_uniform_(self.q_templates)
        nn.init.xavier_uniform_(self.k_templates)
        nn.init.xavier_uniform_(self.v_templates)

    def forward(self, q_coeffs, k_coeffs, v_coeffs):
        # Compute chunked weights
        q_chunks = torch.matmul(q_coeffs, self.q_templates)
        k_chunks = torch.matmul(k_coeffs, self.k_templates)
        v_chunks = torch.matmul(v_coeffs, self.v_templates)

        # Reshape to final weight matrices
        q_weights = q_chunks.reshape(self.hidden_size, self.hidden_size)
        k_weights = k_chunks.reshape(self.kv_dim, self.hidden_size)
        v_weights = v_chunks.reshape(self.kv_dim, self.hidden_size)

        return q_weights, k_weights, v_weights


class SharedLlamaAttention(nn.Module):
    def __init__(
        self,
        config,
        layer_idx: Optional[int] = None,
        bank: Optional[AttTemplateBank] = None,
    ):
        super().__init__()
        self.config = config
        self.bank = bank
        self.layer_idx = layer_idx
        self.attention_dropout = config.attention_dropout
        self.hidden_size = config.hidden_size
        self.num_heads = config.num_attention_heads
        self.head_dim = self.hidden_size // self.num_heads
        self.num_key_value_heads = getattr(
            config, "num_key_value_heads", config.num_attention_heads
        )
        self.num_key_value_groups = self.num_heads // self.num_key_value_heads
        self.max_position_embeddings = config.max_position_embeddings
        self.rope_theta = getattr(config, "rope_theta", 10000.0)
        self.is_causal = True

        self.o_proj = nn.Linear(
            self.hidden_size,
            self.hidden_size,
            bias=getattr(config, "attention_bias", False),
        )
        self.rotary_emb = LlamaRotaryEmbedding(config=self.config)

        # Initialize coefficients with proper shapes
        self.q_coefficients = nn.Parameter(torch.randn(bank.coef_shape))
        self.k_coefficients = nn.Parameter(torch.randn(bank.coef_shape))
        self.v_coefficients = nn.Parameter(torch.randn(bank.coef_shape))

        # Initialize with small random values
        nn.init.orthogonal_(self.q_coefficients)
        nn.init.orthogonal_(self.k_coefficients)
        nn.init.orthogonal_(self.v_coefficients)

    def forward(
        self,
        hidden_states,
        attention_mask=None,
        past_key_value=None,
        cache_position=None,
        position_embeddings=None,
        position_ids=None,
        output_attentions=False,
        use_cache=False,
        **kwargs,
    ):
        bsz, q_len, _ = hidden_states.size()

        # Generate weights using template bank
        q_weights, k_weights, v_weights = self.bank(
            self.q_coefficients, self.k_coefficients, self.v_coefficients
        )
        target_dtype = hidden_states.dtype
        q_weights = q_weights.to(target_dtype)
        k_weights = k_weights.to(target_dtype)
        v_weights = v_weights.to(target_dtype)

        # Apply projections
        query_states = F.linear(hidden_states, q_weights)
        key_states = F.linear(hidden_states, k_weights)
        value_states = F.linear(hidden_states, v_weights)

        # Reshape for multi-head attention
        query_states = query_states.view(
            bsz, q_len, self.num_heads, self.head_dim
        ).transpose(1, 2)
        key_states = key_states.view(
            bsz, q_len, self.num_key_value_heads, self.head_dim
        ).transpose(1, 2)
        value_states = value_states.view(
            bsz, q_len, self.num_key_value_heads, self.head_dim
        ).transpose(1, 2)

        # Apply rotary embeddings
        if position_embeddings is None:
            cos, sin = self.rotary_emb(value_states, position_ids)
        else:
            cos, sin = position_embeddings

        query_states, key_states = apply_rotary_pos_emb(
            query_states, key_states, cos, sin
        )

        # Handle past key values
        if past_key_value is not None:
            cache_kwargs = {"sin": sin, "cos": cos, "cache_position": cache_position}
            key_states, value_states = past_key_value.update(
                key_states, value_states, self.layer_idx, cache_kwargs
            )

        # Repeat key/value for grouped query attention

        key_states = repeat_kv(key_states, self.num_key_value_groups)
        value_states = repeat_kv(value_states, self.num_key_value_groups)

        # ============ CRITICAL CHANGE: Use SDPA instead of manual attention ============
        causal_mask = attention_mask
        if attention_mask is not None:
            causal_mask = causal_mask[:, :, :, : key_states.shape[-2]]

        # Make contiguous for SDPA
        if query_states.device.type == "cuda" and causal_mask is not None:
            query_states = query_states.contiguous()
            key_states = key_states.contiguous()
            value_states = value_states.contiguous()

        # Determine is_causal for SDPA
        is_causal = True if causal_mask is None and q_len > 1 else False

        # Use PyTorch's SDPA (same as LlamaSdpaAttention)
        attn_output = torch.nn.functional.scaled_dot_product_attention(
            query_states,
            key_states,
            value_states,
            attn_mask=causal_mask,
            dropout_p=self.attention_dropout if self.training else 0.0,
            is_causal=is_causal,
        )
        # ============================================================================

        attn_output = attn_output.transpose(1, 2).contiguous()
        attn_output = attn_output.reshape(bsz, q_len, -1)
        attn_output = self.o_proj(attn_output)

        return attn_output, None, past_key_value


def fixed_cross_entropy(
    source,
    target,
    num_items_in_batch: int = None,
    ignore_index: int = -100,
    **kwargs,
):
    reduction = "sum" if num_items_in_batch is not None else "mean"
    loss = nn.functional.cross_entropy(
        source, target, ignore_index=ignore_index, reduction=reduction
    )
    if reduction == "sum":
        loss = loss / num_items_in_batch
    return loss


class RECAST8b_llamaModel(PreTrainedModel):
    config_class = RECAST8b_llama
    base_model_prefix = "llama"
    supports_gradient_checkpointing = True
    _no_split_modules = ["LlamaDecoderLayer"]  # Add this
    _skip_keys_device_placement = "past_key_values"  # Add this

    def __init__(self, config):
        super().__init__(config)
        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
        )

        original_config = AutoConfig.from_pretrained(
            "meta-llama/Llama-3.1-8b", trust_remote_code=True
        )
        self.rotary_emb = LlamaRotaryEmbedding(
            config=original_config,
        )

        # Create template banks first
        self.mlp_banks = []
        self.attn_banks = []
        layers_per_group = config.num_hidden_layers // config.num_groups
        # Explicitly calculate coef_width if not provided in config
        if hasattr(config, "coef_width") and config.coef_width is not None:
            coef_width = config.coef_width
        else:
            coef_width = config.coef_height * layers_per_group
        config.coef_width = coef_width
        print(
            f"Model config: num_groups={config.num_groups}, layers_per_group={layers_per_group}, K={config.k}"
        )
        print(f"Coefficient shape: ({config.coef_height}, {config.coef_width})")

        mlp_banks = nn.ModuleList(
            [
                MLPTemplateBank(
                    config=config, coef_rows=config.coef_height, coef_columns=coef_width
                )
                for _ in range(config.num_groups)
            ]
        )

        attn_banks = nn.ModuleList(
            [
                AttTemplateBank(
                    config=config, coef_rows=config.coef_height, coef_columns=coef_width
                )
                for _ in range(config.num_groups)
            ]
        )
        self.mlp_banks = mlp_banks
        self.attn_banks = attn_banks
        # Create layers using LlamaDecoderLayer but replace MLPs
        self.layers = nn.ModuleList()
        for layer_idx in range(config.num_hidden_layers):
            # Create standard LlamaDecoderLayer
            decoder_layer = LlamaDecoderLayer(config, layer_idx)
            # Replace its MLP with our SharedLlamaMLP
            group_idx = layer_idx // layers_per_group
            decoder_layer.mlp = SharedLlamaMLP(
                config=config,
                bank=self.mlp_banks[group_idx],
            )
            decoder_layer.self_attn = SharedLlamaAttention(
                config=config,
                layer_idx=layer_idx,
                bank=self.attn_banks[group_idx],
            )

            self.layers.append(decoder_layer)

        self.norm = LlamaRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
        self.gradient_checkpointing = False

    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,
        **flash_attn_kwargs,
    ) -> 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)
        # Set up cache position if not provided
        if cache_position is None:
            past_seen_tokens = (
                0
                if past_key_values is None
                else (
                    past_key_values.get_seq_length()
                    if isinstance(past_key_values, Cache)
                    else past_key_values[0][0].size(-2) if past_key_values else 0
                )
            )
            cache_position = torch.arange(
                past_seen_tokens,
                past_seen_tokens + inputs_embeds.shape[1],
                device=inputs_embeds.device,
            )
        # Create position embeddings to be shared across the decoder layers
        # Set up position IDs if not provided
        if position_ids is None:
            position_ids = cache_position.unsqueeze(0)
        # Get updated causal mask
        causal_mask = self._update_causal_mask(
            attention_mask,
            inputs_embeds,
            cache_position,
            past_key_values,
            output_attentions,
        )
        hidden_states = inputs_embeds
        position_embeddings = self.rotary_emb(hidden_states, position_ids)

        # Initialize outputs
        all_hidden_states = () if output_hidden_states else None
        all_self_attns = () if output_attentions else None
        next_decoder_cache = None

        # Process through layers
        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,
                    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,
                    position_embeddings=position_embeddings,
                    **flash_attn_kwargs,
                )

            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],)

        # Final layer norm
        hidden_states = self.norm(hidden_states)

        # Add last hidden state
        if output_hidden_states:
            all_hidden_states += (hidden_states,)

        next_cache = next_decoder_cache if use_cache else None

        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,
        )

    @classmethod
    def from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs):
        if isinstance(
            pretrained_model_name_or_path, str
        ) and pretrained_model_name_or_path.endswith(".pt"):
            print("Loading from local checkpoint")
            # Load from local checkpoint
            config = kwargs.get("config", None)
            if config is None:
                config = AutoConfig.from_pretrained(
                    pretrained_model_name_or_path, trust_remote_code=True
                )

            model = cls(config)
            checkpoint = torch.load(pretrained_model_name_or_path, map_location="cpu")
            state_dict = checkpoint["model_state_dict"]
            logger.info(
                f"Loaded checkpoint from epoch {checkpoint.get('epoch')} with loss {checkpoint.get('loss')}"
            )

            missing_keys, unexpected_keys = model.load_state_dict(
                state_dict, strict=False
            )

            if len(missing_keys) > 0:
                logger.warning(f"Missing keys: {missing_keys}")
            if len(unexpected_keys) > 0:
                logger.warning(f"Unexpected keys: {unexpected_keys}")

            return model
        else:
            print("Loading from hub")
            # Load from hub using parent's from_pretrained
            return super().from_pretrained(
                pretrained_model_name_or_path, *model_args, **kwargs
            )

    def get_input_embeddings(self):
        return self.embed_tokens

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

    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,
    ):
        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


class RECAST8b_LlamaForCausalLM(PreTrainedModel, GenerationMixin):
    _tied_weights_keys = ["lm_head.weight"]
    _tp_plan = {"lm_head": "colwise_rep"}
    config_class = RECAST8b_llama
    base_model_prefix = "llama"
    supports_gradient_checkpointing = True
    _no_split_modules = ["LlamaDecoderLayer"]  # Add this
    _skip_keys_device_placement = "past_key_values"  # Add this

    def __init__(self, config):
        super().__init__(config)
        self.model = RECAST8b_llamaModel(config)
        self.vocab_size = config.vocab_size
        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, 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 get_output_embeddings(self):
        return self.lm_head

    def set_output_embeddings(self, new_embeddings):
        self.lm_head = new_embeddings

    def set_decoder(self, decoder):
        self.model = decoder

    def get_decoder(self):
        return self.model

    def loss_function(
        self,
        logits,
        labels,
        vocab_size: int,
        num_items_in_batch: int = None,
        ignore_index: int = -100,
        **kwargs,
    ):
        # Upcast to float if we need to compute the loss to avoid potential precision issues
        logits = logits.float()
        # Shift so that tokens < n predict n
        shift_logits = logits[..., :-1, :].contiguous()
        shift_labels = labels[..., 1:].contiguous()
        # Flatten the tokens
        shift_logits = shift_logits.view(-1, vocab_size)
        shift_labels = shift_labels.view(-1)
        # Enable model parallelism
        shift_labels = shift_labels.to(shift_logits.device)
        loss = fixed_cross_entropy(
            shift_logits, shift_labels, num_items_in_batch, ignore_index, **kwargs
        )
        return loss

    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,
        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,
        cache_position: Optional[torch.LongTensor] = None,
        num_logits_to_keep: int = 0,
        **kwargs,
    ) -> Union[Tuple, CausalLMOutputWithPast]:
        """
        Args:
            labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
                Labels for computing the masked language modeling loss. Indices should be in
                `[0, ..., config.vocab_size]` or -100 (masked tokens).
            num_logits_to_keep (`int`, *optional*):
                Calculate logits for the last `num_logits_to_keep` tokens. If `0`, calculate all logits.
        """
        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
        )
        return_dict = (
            return_dict if return_dict is not None else self.config.use_return_dict
        )

        outputs = self.model(
            input_ids=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,
            cache_position=cache_position,
            **kwargs,
        )

        hidden_states = outputs[0]
        # Only compute necessary logits
        logits = self.lm_head(hidden_states[:, -num_logits_to_keep:, :])

        loss = None
        if labels is not None:
            # Calculate batch size for loss function
            num_items_in_batch = (
                input_ids.size(0) if input_ids is not None else inputs_embeds.size(0)
            )
            loss = self.loss_function(
                logits=logits,
                labels=labels,
                vocab_size=self.config.vocab_size,
                num_items_in_batch=num_items_in_batch,
                **kwargs,
            )

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

        return CausalLMOutputWithPast(
            loss=loss,
            logits=logits,
            past_key_values=outputs.past_key_values,
            hidden_states=outputs.hidden_states,
            attentions=outputs.attentions,
        )

    def prepare_inputs_for_generation(
        self,
        input_ids,
        past_key_values=None,
        attention_mask=None,
        inputs_embeds=None,
        **kwargs,
    ):
        if past_key_values:
            input_ids = input_ids[:, -1:]

        position_ids = kwargs.get("position_ids", None)
        if attention_mask is not None and position_ids is None:
            # create position_ids on the fly for batch generation
            position_ids = attention_mask.long().cumsum(-1) - 1
            position_ids.masked_fill_(attention_mask == 0, 1)
            if past_key_values:
                position_ids = position_ids[:, -1].unsqueeze(-1)

        # if `inputs_embeds` are passed, we only want to use them in the 1st generation step
        if inputs_embeds is not None and past_key_values is None:
            model_inputs = {"inputs_embeds": inputs_embeds}
        else:
            model_inputs = {"input_ids": input_ids}

        model_inputs.update(
            {
                "position_ids": position_ids,
                "past_key_values": past_key_values,
                "use_cache": kwargs.get("use_cache"),
                "attention_mask": attention_mask,
            }
        )
        return model_inputs

    @classmethod
    def from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs):
        if isinstance(
            pretrained_model_name_or_path, str
        ) and pretrained_model_name_or_path.endswith(".pt"):
            print("Loading from local checkpoint")
            config = kwargs.get("config", None)
            if config is None:
                config = AutoConfig.from_pretrained(
                    pretrained_model_name_or_path, trust_remote_code=True
                )
            model = torch.load(pretrained_model_name_or_path, map_location="cpu")

            return model
        else:
            print("Loading from hub")
            return super().from_pretrained(
                pretrained_model_name_or_path, *model_args, **kwargs
            )