image-match/python/py/Lib/site-packages/transformers/models/stablelm/modeling_stablelm.py

# Copyright 2024 EleutherAI and the HuggingFace Inc. team. All rights reserved.
#
# This code is based on EleutherAI's GPT-NeoX library and the GPT-NeoX
# and OPT implementations in this library. It has been modified from its
# original forms to accommodate minor architectural differences compared
# to GPT-NeoX and OPT used by the Meta AI team that trained the model.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""PyTorch StableLM model."""

from collections.abc import Callable
from typing import Optional

import torch
from torch import nn

from ...activations import ACT2FN
from ...cache_utils import Cache, DynamicCache
from ...generation import GenerationMixin
from ...masking_utils import create_causal_mask
from ...modeling_layers import (
    GenericForSequenceClassification,
    GenericForTokenClassification,
    GradientCheckpointingLayer,
)
from ...modeling_outputs import (
    BaseModelOutputWithPast,
    CausalLMOutputWithPast,
)
from ...modeling_rope_utils import (
    ROPE_INIT_FUNCTIONS,
    dynamic_rope_update,
)
from ...modeling_utils import ALL_ATTENTION_FUNCTIONS, PreTrainedModel
from ...processing_utils import Unpack
from ...utils import TransformersKwargs, auto_docstring, can_return_tuple, logging
from ...utils.generic import maybe_autocast, merge_with_config_defaults
from ...utils.output_capturing import capture_outputs
from .configuration_stablelm import StableLmConfig


logger = logging.get_logger(__name__)


# Copied from transformers.models.llama.modeling_llama.LlamaRotaryEmbedding with Llama->StableLm
class StableLmRotaryEmbedding(nn.Module):
    inv_freq: torch.Tensor  # fix linting for `register_buffer`

    def __init__(self, config: StableLmConfig, device=None):
        super().__init__()
        self.max_seq_len_cached = config.max_position_embeddings
        self.original_max_seq_len = config.max_position_embeddings

        self.config = config

        self.rope_type = self.config.rope_parameters["rope_type"]
        rope_init_fn: Callable = self.compute_default_rope_parameters
        if self.rope_type != "default":
            rope_init_fn = ROPE_INIT_FUNCTIONS[self.rope_type]
        inv_freq, self.attention_scaling = rope_init_fn(self.config, device)

        self.register_buffer("inv_freq", inv_freq, persistent=False)
        self.register_buffer("original_inv_freq", inv_freq.clone(), persistent=False)

    @staticmethod
    # Ignore copy
    def compute_default_rope_parameters(
        config: StableLmConfig | None = None,
        device: Optional["torch.device"] = None,
        seq_len: int | None = None,
    ) -> tuple["torch.Tensor", float]:
        """
        Computes the inverse frequencies according to the original RoPE implementation
        Args:
            config ([`~transformers.PreTrainedConfig`]):
                The model configuration.
            device (`torch.device`):
                The device to use for initialization of the inverse frequencies.
            seq_len (`int`, *optional*):
                The current sequence length. Unused for this type of RoPE.
        Returns:
            Tuple of (`torch.Tensor`, `float`), containing the inverse frequencies for the RoPE embeddings and the
            post-processing scaling factor applied to the computed cos/sin (unused in this type of RoPE).
        """
        base = config.rope_parameters["rope_theta"]
        partial_rotary_factor = config.rope_parameters.get("partial_rotary_factor", 1.0)
        head_dim = getattr(config, "head_dim", None) or config.hidden_size // config.num_attention_heads
        dim = int(head_dim * partial_rotary_factor)

        attention_factor = 1.0  # Unused in this type of RoPE

        # Compute the inverse frequencies
        inv_freq = 1.0 / (
            base ** (torch.arange(0, dim, 2, dtype=torch.int64).to(device=device, dtype=torch.float) / dim)
        )
        return inv_freq, attention_factor

    @torch.no_grad()
    @dynamic_rope_update  # power user: used with advanced RoPE types (e.g. dynamic rope)
    def forward(self, x, position_ids):
        inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1).to(x.device)
        position_ids_expanded = position_ids[:, None, :].float()

        device_type = x.device.type if isinstance(x.device.type, str) and x.device.type != "mps" else "cpu"
        with maybe_autocast(device_type=device_type, enabled=False):  # Force float32
            freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).transpose(1, 2)
            emb = torch.cat((freqs, freqs), dim=-1)
            cos = emb.cos() * self.attention_scaling
            sin = emb.sin() * self.attention_scaling

        return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)


# Copied from transformers.models.llama.modeling_llama.rotate_half
def rotate_half(x):
    """Rotates half the hidden dims of the input."""
    x1 = x[..., : x.shape[-1] // 2]
    x2 = x[..., x.shape[-1] // 2 :]
    return torch.cat((-x2, x1), dim=-1)


# Copied from transformers.models.llama.modeling_llama.apply_rotary_pos_emb
def apply_rotary_pos_emb(q, k, cos, sin, unsqueeze_dim=1):
    """Applies Rotary Position Embedding to the query and key tensors.

    Args:
        q (`torch.Tensor`): The query tensor.
        k (`torch.Tensor`): The key tensor.
        cos (`torch.Tensor`): The cosine part of the rotary embedding.
        sin (`torch.Tensor`): The sine part of the rotary embedding.
        unsqueeze_dim (`int`, *optional*, defaults to 1):
            The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and
            sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note
            that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and
            k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes
            cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have
            the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.
    Returns:
        `tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding.
    """
    cos = cos.unsqueeze(unsqueeze_dim)
    sin = sin.unsqueeze(unsqueeze_dim)
    q_embed = (q * cos) + (rotate_half(q) * sin)
    k_embed = (k * cos) + (rotate_half(k) * sin)
    return q_embed, k_embed


# Copied from transformers.models.mistral.modeling_mistral.MistralMLP with Mistral->StableLm
class StableLmMLP(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.config = config
        self.hidden_size = config.hidden_size
        self.intermediate_size = config.intermediate_size
        self.gate_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False)
        self.up_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False)
        self.down_proj = nn.Linear(self.intermediate_size, self.hidden_size, bias=False)
        self.act_fn = ACT2FN[config.hidden_act]

    def forward(self, x):
        down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))
        return down_proj


class StableLmLayerNormPerHead(nn.Module):
    def __init__(self, dim, num_heads, eps=1e-5, bias=False):
        super().__init__()
        self.dim = dim
        self.num_heads = num_heads
        self.norms = nn.ModuleList([nn.LayerNorm(dim, eps=eps, bias=bias) for _ in range(self.num_heads)])

    def forward(self, hidden_states: torch.Tensor):
        # Split along the num_heads axis to get per-head inputs
        # [batch_size, num_heads, seq_len, head_dim] -> [batch_size, 1, seq_len, head_dim] * num_heads
        states_per_heads = torch.split(hidden_states, 1, dim=1)
        # Normalize and merge the heads back together
        return torch.cat([norm(hidden_states) for norm, hidden_states in zip(self.norms, states_per_heads)], dim=1)


# Copied from transformers.models.llama.modeling_llama.repeat_kv
def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
    """
    This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch,
    num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim)
    """
    batch, num_key_value_heads, slen, head_dim = hidden_states.shape
    if n_rep == 1:
        return hidden_states
    hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)
    return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)


# Copied from transformers.models.llama.modeling_llama.eager_attention_forward
def eager_attention_forward(
    module: nn.Module,
    query: torch.Tensor,
    key: torch.Tensor,
    value: torch.Tensor,
    attention_mask: torch.Tensor | None,
    scaling: float,
    dropout: float = 0.0,
    **kwargs: Unpack[TransformersKwargs],
):
    key_states = repeat_kv(key, module.num_key_value_groups)
    value_states = repeat_kv(value, module.num_key_value_groups)

    attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling
    if attention_mask is not None:
        attn_weights = attn_weights + attention_mask

    attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)
    attn_weights = nn.functional.dropout(attn_weights, p=dropout, training=module.training)
    attn_output = torch.matmul(attn_weights, value_states)
    attn_output = attn_output.transpose(1, 2).contiguous()

    return attn_output, attn_weights


class StableLmAttention(nn.Module):
    """Multi-headed attention from 'Attention Is All You Need' paper"""

    def __init__(self, config: StableLmConfig, layer_idx: int | None = None):
        super().__init__()
        self.config = config
        self.layer_idx = layer_idx
        if layer_idx is None:
            logger.warning_once(
                f"Instantiating {self.__class__.__name__} without passing a `layer_idx` is not recommended and will "
                "lead to errors during the forward call if caching is used. Please make sure to provide a `layer_idx` "
                "when creating this class."
            )

        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 = config.num_key_value_heads
        self.num_key_value_groups = self.num_heads // self.num_key_value_heads

        self.rotary_ndims = int(self.head_dim * config.rope_parameters["partial_rotary_factor"])
        self.is_causal = True
        self.scaling = self.head_dim**-0.5

        if (self.head_dim * self.num_heads) != self.hidden_size:
            raise ValueError(
                f"hidden_size must be divisible by num_heads (got `hidden_size`: {self.hidden_size}"
                f" and `num_heads`: {self.num_heads})."
            )
        self.q_proj = nn.Linear(self.hidden_size, self.num_heads * self.head_dim, bias=config.use_qkv_bias)
        self.k_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=config.use_qkv_bias)
        self.v_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=config.use_qkv_bias)
        self.o_proj = nn.Linear(self.hidden_size, self.hidden_size, bias=False)

        self.qk_layernorm = config.qk_layernorm
        if self.qk_layernorm:
            self.q_layernorm = StableLmLayerNormPerHead(self.head_dim, self.num_heads, eps=config.layer_norm_eps)
            self.k_layernorm = StableLmLayerNormPerHead(
                self.head_dim, self.num_key_value_heads, eps=config.layer_norm_eps
            )

        self.attention_dropout = config.attention_dropout

    def forward(
        self,
        hidden_states: torch.Tensor,
        attention_mask: torch.Tensor | None = None,
        position_ids: torch.LongTensor | None = None,
        past_key_values: Cache | None = None,
        output_attentions: bool = False,
        use_cache: bool = False,
        position_embeddings: tuple[torch.Tensor, torch.Tensor] | None = None,
        **kwargs,
    ) -> tuple[torch.Tensor, torch.Tensor | None, tuple[torch.Tensor] | None]:
        bsz, q_len, _ = hidden_states.size()

        query_states = self.q_proj(hidden_states)
        key_states = self.k_proj(hidden_states)
        value_states = self.v_proj(hidden_states)

        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)

        if self.qk_layernorm:
            query_states = self.q_layernorm(query_states)
            key_states = self.k_layernorm(key_states)

        cos, sin = position_embeddings
        query_rot, query_pass = (
            query_states[..., : self.rotary_ndims],
            query_states[..., self.rotary_ndims :],
        )
        key_rot, key_pass = (
            key_states[..., : self.rotary_ndims],
            key_states[..., self.rotary_ndims :],
        )
        # [batch_size, seq_length, num_heads, head_dim // config.partial_rotary_factor]
        query_rot, key_rot = apply_rotary_pos_emb(query_rot, key_rot, cos, sin)

        # [batch_size, seq_length, num_heads, head_dim]
        query_states = torch.cat((query_rot, query_pass), dim=-1)
        key_states = torch.cat((key_rot, key_pass), dim=-1)

        if past_key_values is not None:
            key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)

        attention_interface: Callable = ALL_ATTENTION_FUNCTIONS.get_interface(
            self.config._attn_implementation, eager_attention_forward
        )

        attn_output, attn_weights = attention_interface(
            self,
            query_states,
            key_states,
            value_states,
            attention_mask,
            dropout=0.0 if not self.training else self.attention_dropout,
            scaling=self.scaling,
            position_ids=position_ids,  # pass `position_ids` for FA2
            **kwargs,
        )

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

        return attn_output, attn_weights


class StableLmDecoderLayer(GradientCheckpointingLayer):
    def __init__(self, config: StableLmConfig, layer_idx: int):
        super().__init__()
        self.use_parallel_residual = config.use_parallel_residual
        self.hidden_size = config.hidden_size
        self.self_attn = StableLmAttention(config, layer_idx=layer_idx)
        self.mlp = StableLmMLP(config)
        self.input_layernorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
        self.post_attention_layernorm = None
        if not self.use_parallel_residual:
            self.post_attention_layernorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
        self.dropout = nn.Dropout(config.hidden_dropout)

    def forward(
        self,
        hidden_states: torch.Tensor,
        attention_mask: torch.Tensor | None = None,
        position_ids: torch.LongTensor | None = None,
        past_key_values: Cache | None = None,
        use_cache: bool | None = False,
        position_embeddings: tuple[torch.Tensor, torch.Tensor] | None = None,
        **kwargs,
    ) -> torch.Tensor:
        residual = hidden_states

        hidden_states = self.input_layernorm(hidden_states)

        # Self Attention
        self_attn_output, _ = self.self_attn(
            hidden_states=hidden_states,
            attention_mask=attention_mask,
            position_ids=position_ids,
            past_key_values=past_key_values,
            use_cache=use_cache,
            position_embeddings=position_embeddings,
        )

        # copied from transformers.models.gpt_neox.modeling_gpt_neox.GPTNeoXLayer.forward
        if self.use_parallel_residual:
            # x = x + attn(ln1(x)) + mlp(ln1(x))
            # Fully Connected
            mlp_output = self.mlp(hidden_states)
            mlp_output = self.dropout(mlp_output)
            hidden_states = residual + self_attn_output + mlp_output
        else:
            # x = x + attn(ln1(x))
            # x = x + mlp(ln2(x))
            residual = residual + self_attn_output
            # Fully Connected
            mlp_output = self.mlp(self.post_attention_layernorm(residual))
            mlp_output = self.dropout(mlp_output)
            hidden_states = residual + mlp_output

        return hidden_states


@auto_docstring
class StableLmPreTrainedModel(PreTrainedModel):
    config: StableLmConfig
    base_model_prefix = "model"
    supports_gradient_checkpointing = True
    _no_split_modules = ["StableLmDecoderLayer"]
    _skip_keys_device_placement = "past_key_values"
    _supports_flash_attn = True
    _supports_sdpa = True
    _can_compile_fullgraph = True
    _can_record_outputs = {
        "hidden_states": StableLmDecoderLayer,
        "attentions": StableLmAttention,
    }


@auto_docstring
class StableLmModel(StableLmPreTrainedModel):
    """
    Transformer decoder consisting of *config.num_hidden_layers* layers. Each layer is a [`StableLmDecoderLayer`]

    Args:
        config: StableLmConfig
    """

    def __init__(self, config: StableLmConfig):
        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)
        self.layers = nn.ModuleList(
            [StableLmDecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
        )
        self.norm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)

        self._attn_implementation = config._attn_implementation
        self.gradient_checkpointing = False
        self.rotary_emb = StableLmRotaryEmbedding(config=self.config)

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

    @merge_with_config_defaults
    @capture_outputs
    @auto_docstring
    def forward(
        self,
        input_ids: torch.LongTensor | None = None,
        attention_mask: torch.Tensor | None = None,
        position_ids: torch.LongTensor | None = None,
        past_key_values: Cache | None = None,
        inputs_embeds: torch.FloatTensor | None = None,
        use_cache: bool | None = None,
        **kwargs: Unpack[TransformersKwargs],
    ) -> BaseModelOutputWithPast:
        if (input_ids is None) ^ (inputs_embeds is not None):
            raise ValueError("You must specify exactly one of input_ids or inputs_embeds")

        if use_cache and past_key_values is None:
            past_key_values = DynamicCache(config=self.config)

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

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

        causal_mask = create_causal_mask(
            config=self.config,
            inputs_embeds=inputs_embeds,
            attention_mask=attention_mask,
            past_key_values=past_key_values,
            position_ids=position_ids,
        )

        hidden_states = inputs_embeds
        position_embeddings = self.rotary_emb(hidden_states, position_ids=position_ids)

        for decoder_layer in self.layers:
            hidden_states = decoder_layer(
                hidden_states,
                attention_mask=causal_mask,
                position_ids=position_ids,
                past_key_values=past_key_values,
                use_cache=use_cache,
                position_embeddings=position_embeddings,
                **kwargs,
            )

        hidden_states = self.norm(hidden_states)

        return BaseModelOutputWithPast(
            last_hidden_state=hidden_states,
            past_key_values=past_key_values,
        )


# Copied from transformers.models.persimmon.modeling_persimmon.PersimmonForCausalLM with PERSIMMON->STABLELM,Persimmon->StableLm
class StableLmForCausalLM(StableLmPreTrainedModel, GenerationMixin):
    _tied_weights_keys = {"lm_head.weight": "model.embed_tokens.weight"}

    # Copied from transformers.models.llama.modeling_llama.LlamaForCausalLM.__init__ with LLAMA->STABLELM,Llama->StableLm
    def __init__(self, config):
        super().__init__(config)
        self.model = StableLmModel(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()

    @can_return_tuple
    @auto_docstring
    # Ignore copy
    def forward(
        self,
        input_ids: torch.LongTensor | None = None,
        attention_mask: torch.Tensor | None = None,
        position_ids: torch.LongTensor | None = None,
        past_key_values: Cache | None = None,
        inputs_embeds: torch.FloatTensor | None = None,
        labels: torch.LongTensor | None = None,
        use_cache: bool | None = None,
        logits_to_keep: int | torch.Tensor = 0,
        **kwargs: Unpack[TransformersKwargs],
    ) -> CausalLMOutputWithPast:
        r"""
        labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
            Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
            config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
            (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`.

        Example:

        ```python
        >>> from transformers import AutoTokenizer, StableLmForCausalLM

        >>> model = StableLmForCausalLM.from_pretrained("adept/persimmon-8b-base")
        >>> tokenizer = AutoTokenizer.from_pretrained("adept/persimmon-8b-base")

        >>> prompt = "human: Hey, what should I eat for dinner?"
        >>> inputs = tokenizer(prompt, return_tensors="pt")

        >>> # Generate
        >>> generate_ids = model.generate(inputs.input_ids, max_length=30)
        >>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
        'human: Hey, what should I eat for dinner?\n\ncat: 🐱\n\nhuman: 😐\n\n'
        ```"""
        outputs: BaseModelOutputWithPast = 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,
            **kwargs,
        )

        hidden_states = outputs.last_hidden_state
        slice_indices = slice(-logits_to_keep, None) if isinstance(logits_to_keep, int) else logits_to_keep
        logits = self.lm_head(hidden_states[:, slice_indices, :])

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

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


class StableLmForSequenceClassification(GenericForSequenceClassification, StableLmPreTrainedModel): ...


class StableLmForTokenClassification(GenericForTokenClassification, StableLmPreTrainedModel): ...


__all__ = [
    "StableLmForCausalLM",
    "StableLmModel",
    "StableLmPreTrainedModel",
    "StableLmForSequenceClassification",
    "StableLmForTokenClassification",
]