modeling_ashish_ocr.py

"""AshishOCR model implementation."""

import math
from typing import List, Optional, Tuple, Union

import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.nn import CrossEntropyLoss

from transformers.activations import ACT2FN
from transformers.cache_utils import Cache, DynamicCache
from transformers.modeling_outputs import (
    BaseModelOutputWithPast,
    CausalLMOutputWithPast,
)
from transformers.modeling_utils import PreTrainedModel
from transformers.utils import logging

from .configuration_ashish_ocr import AshishOcrConfig, AshishOcrTextConfig, AshishOcrVisionConfig


logger = logging.get_logger(__name__)


class AshishOcrRMSNorm(nn.Module):
    def __init__(self, hidden_size, eps=1e-6):
        super().__init__()
        self.weight = nn.Parameter(torch.ones(hidden_size))
        self.variance_epsilon = eps

    def forward(self, hidden_states):
        input_dtype = hidden_states.dtype
        hidden_states = hidden_states.to(torch.float32)
        variance = hidden_states.pow(2).mean(-1, keepdim=True)
        hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)
        return self.weight * hidden_states.to(input_dtype)


class AshishOcrRotaryEmbedding(nn.Module):
    def __init__(self, dim, max_position_embeddings=131072, base=10000, device=None):
        super().__init__()
        self.dim = dim
        self.max_position_embeddings = max_position_embeddings
        self.base = base
        inv_freq = 1.0 / (self.base ** (torch.arange(0, self.dim, 2, dtype=torch.float32, device=device) / self.dim))
        self.register_buffer("inv_freq", inv_freq, persistent=False)

    @torch.no_grad()
    def forward(self, x, position_ids):
        inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1)
        position_ids_expanded = position_ids[:, None, :].float()
        freqs = (inv_freq_expanded @ position_ids_expanded).transpose(1, 2)
        emb = torch.cat((freqs, freqs), dim=-1)
        cos = emb.cos()
        sin = emb.sin()
        return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)


def rotate_half(x):
    x1 = x[..., : x.shape[-1] // 2]
    x2 = x[..., x.shape[-1] // 2 :]
    return torch.cat((-x2, x1), dim=-1)


def apply_rotary_pos_emb(q, k, cos, sin, unsqueeze_dim=1):
    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


class AshishOcrMLP(nn.Module):
    def __init__(self, config):
        super().__init__()
        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):
        return self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))


class AshishOcrAttention(nn.Module):
    def __init__(self, config: AshishOcrTextConfig, layer_idx: int):
        super().__init__()
        self.config = config
        self.layer_idx = layer_idx
        self.hidden_size = config.hidden_size
        self.num_heads = config.num_attention_heads
        self.head_dim = config.head_dim
        self.num_key_value_heads = config.num_key_value_heads
        self.num_key_value_groups = self.num_heads // self.num_key_value_heads
        self.attention_dropout = config.attention_dropout

        self.q_proj = nn.Linear(self.hidden_size, self.num_heads * self.head_dim, bias=config.attention_bias)
        self.k_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=config.attention_bias)
        self.v_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=config.attention_bias)
        self.o_proj = nn.Linear(self.num_heads * self.head_dim, self.hidden_size, bias=False)

        self.rotary_emb = AshishOcrRotaryEmbedding(
            self.head_dim,
            max_position_embeddings=config.max_position_embeddings,
        )

    def forward(
        self,
        hidden_states: torch.Tensor,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        past_key_value: Optional[Cache] = None,
        output_attentions: bool = False,
        use_cache: bool = False,
    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Cache]]:
        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)

        cos, sin = self.rotary_emb(value_states, position_ids)
        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)

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

        # Repeat k/v heads for grouped query attention
        key_states = key_states.repeat_interleave(self.num_key_value_groups, dim=1)
        value_states = value_states.repeat_interleave(self.num_key_value_groups, dim=1)

        attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) / math.sqrt(self.head_dim)

        if attention_mask is not None:
            attn_weights = attn_weights + attention_mask

        attn_weights = F.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query_states.dtype)
        attn_weights = F.dropout(attn_weights, p=self.attention_dropout, training=self.training)
        attn_output = torch.matmul(attn_weights, value_states)

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

        if not output_attentions:
            attn_weights = None

        return attn_output, attn_weights, past_key_value


class AshishOcrDecoderLayer(nn.Module):
    def __init__(self, config: AshishOcrTextConfig, layer_idx: int):
        super().__init__()
        self.hidden_size = config.hidden_size
        self.self_attn = AshishOcrAttention(config, layer_idx)
        self.mlp = AshishOcrMLP(config)
        self.input_layernorm = AshishOcrRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
        self.post_attention_layernorm = AshishOcrRMSNorm(config.hidden_size, eps=config.rms_norm_eps)

    def forward(
        self,
        hidden_states: torch.Tensor,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        past_key_value: Optional[Cache] = None,
        output_attentions: bool = False,
        use_cache: bool = False,
    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Cache]]:
        residual = hidden_states
        hidden_states = self.input_layernorm(hidden_states)

        hidden_states, self_attn_weights, present_key_value = self.self_attn(
            hidden_states=hidden_states,
            attention_mask=attention_mask,
            position_ids=position_ids,
            past_key_value=past_key_value,
            output_attentions=output_attentions,
            use_cache=use_cache,
        )
        hidden_states = residual + hidden_states

        residual = hidden_states
        hidden_states = self.post_attention_layernorm(hidden_states)
        hidden_states = self.mlp(hidden_states)
        hidden_states = residual + hidden_states

        outputs = (hidden_states,)

        if output_attentions:
            outputs += (self_attn_weights,)

        if use_cache:
            outputs += (present_key_value,)

        return outputs


# ==================== Vision Encoder ====================

class AshishOcrVisionMLP(nn.Module):
    def __init__(self, config: AshishOcrVisionConfig):
        super().__init__()
        self.fc1 = nn.Linear(config.hidden_size, config.intermediate_size)
        self.act = ACT2FN[config.hidden_act]
        self.fc2 = nn.Linear(config.intermediate_size, config.hidden_size)

    def forward(self, hidden_states):
        hidden_states = self.fc1(hidden_states)
        hidden_states = self.act(hidden_states)
        hidden_states = self.fc2(hidden_states)
        return hidden_states


class AshishOcrVisionAttention(nn.Module):
    def __init__(self, config: AshishOcrVisionConfig):
        super().__init__()
        self.num_heads = config.num_heads
        self.head_dim = config.hidden_size // config.num_heads

        self.qkv = nn.Linear(config.hidden_size, 3 * config.hidden_size, bias=config.attention_bias)
        self.proj = nn.Linear(config.hidden_size, config.hidden_size, bias=config.attention_bias)

    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
        bsz, seq_len, _ = hidden_states.size()

        qkv = self.qkv(hidden_states)
        qkv = qkv.reshape(bsz, seq_len, 3, self.num_heads, self.head_dim).permute(2, 0, 3, 1, 4)
        q, k, v = qkv.unbind(0)

        attn_weights = torch.matmul(q, k.transpose(-2, -1)) / math.sqrt(self.head_dim)
        attn_weights = F.softmax(attn_weights, dim=-1)
        attn_output = torch.matmul(attn_weights, v)

        attn_output = attn_output.transpose(1, 2).reshape(bsz, seq_len, -1)
        attn_output = self.proj(attn_output)
        return attn_output


class AshishOcrVisionBlock(nn.Module):
    def __init__(self, config: AshishOcrVisionConfig):
        super().__init__()
        self.norm1 = AshishOcrRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
        self.attn = AshishOcrVisionAttention(config)
        self.norm2 = AshishOcrRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
        self.mlp = AshishOcrVisionMLP(config)

    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
        hidden_states = hidden_states + self.attn(self.norm1(hidden_states))
        hidden_states = hidden_states + self.mlp(self.norm2(hidden_states))
        return hidden_states


class AshishOcrPatchEmbed(nn.Module):
    def __init__(self, config: AshishOcrVisionConfig):
        super().__init__()
        self.patch_size = config.patch_size
        self.temporal_patch_size = config.temporal_patch_size
        self.proj = nn.Conv3d(
            3,
            config.hidden_size,
            kernel_size=(config.temporal_patch_size, config.patch_size, config.patch_size),
            stride=(config.temporal_patch_size, config.patch_size, config.patch_size),
            bias=False,
        )

    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
        # hidden_states: (B, C, T, H, W)
        hidden_states = self.proj(hidden_states)
        hidden_states = hidden_states.flatten(2).transpose(1, 2)  # (B, N, D)
        return hidden_states


class AshishOcrPatchMerger(nn.Module):
    def __init__(self, config: AshishOcrVisionConfig):
        super().__init__()
        self.hidden_size = config.hidden_size
        self.out_hidden_size = config.out_hidden_size
        self.spatial_merge_size = config.spatial_merge_size

        self.mlp = nn.Sequential(
            AshishOcrRMSNorm(config.hidden_size * config.spatial_merge_size ** 2, eps=config.rms_norm_eps),
            nn.Linear(config.hidden_size * config.spatial_merge_size ** 2, config.out_hidden_size, bias=False),
            nn.GELU(),
            nn.Linear(config.out_hidden_size, config.out_hidden_size, bias=False),
        )

    def forward(self, hidden_states: torch.Tensor, grid_thw: torch.Tensor) -> torch.Tensor:
        # Merge spatial patches
        batch_size = hidden_states.shape[0]
        merged_states = []

        for b in range(batch_size):
            t, h, w = grid_thw[b].tolist() if grid_thw.dim() > 1 else grid_thw.tolist()
            states = hidden_states[b, :t*h*w]
            states = states.view(t, h, w, -1)

            # Merge spatial patches
            h_new = h // self.spatial_merge_size
            w_new = w // self.spatial_merge_size
            states = states.view(t, h_new, self.spatial_merge_size, w_new, self.spatial_merge_size, -1)
            states = states.permute(0, 1, 3, 2, 4, 5).contiguous()
            states = states.view(t * h_new * w_new, -1)
            merged_states.append(states)

        hidden_states = torch.stack(merged_states, dim=0)
        hidden_states = self.mlp(hidden_states)
        return hidden_states


class AshishOcrVisionEncoder(nn.Module):
    def __init__(self, config: AshishOcrVisionConfig):
        super().__init__()
        self.config = config
        self.patch_embed = AshishOcrPatchEmbed(config)
        self.blocks = nn.ModuleList([AshishOcrVisionBlock(config) for _ in range(config.depth)])
        self.merger = AshishOcrPatchMerger(config)

    def forward(
        self,
        pixel_values: torch.Tensor,
        grid_thw: Optional[torch.Tensor] = None,
    ) -> torch.Tensor:
        hidden_states = self.patch_embed(pixel_values)

        for block in self.blocks:
            hidden_states = block(hidden_states)

        if grid_thw is not None:
            hidden_states = self.merger(hidden_states, grid_thw)

        return hidden_states


# ==================== Main Model ====================

class AshishOcrPreTrainedModel(PreTrainedModel):
    config_class = AshishOcrConfig
    base_model_prefix = "model"
    supports_gradient_checkpointing = True
    _no_split_modules = ["AshishOcrDecoderLayer", "AshishOcrVisionBlock"]

    def _init_weights(self, module):
        std = self.config.text_config.initializer_range if hasattr(self.config, 'text_config') else 0.02
        if isinstance(module, nn.Linear):
            module.weight.data.normal_(mean=0.0, std=std)
            if module.bias is not None:
                module.bias.data.zero_()
        elif isinstance(module, nn.Embedding):
            module.weight.data.normal_(mean=0.0, std=std)


class AshishOcrTextModel(AshishOcrPreTrainedModel):
    def __init__(self, config: AshishOcrTextConfig):
        super().__init__(config)
        self.config = config
        self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size)
        self.layers = nn.ModuleList(
            [AshishOcrDecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
        )
        self.norm = AshishOcrRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
        self.gradient_checkpointing = False
        self.post_init()

    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[Cache] = 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,
    ) -> 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 inputs_embeds is None:
            inputs_embeds = self.embed_tokens(input_ids)

        batch_size, seq_length = inputs_embeds.shape[:2]

        if position_ids is None:
            position_ids = torch.arange(seq_length, device=inputs_embeds.device).unsqueeze(0)

        if past_key_values is None:
            past_key_values = DynamicCache()

        # Create causal mask
        if attention_mask is None:
            attention_mask = torch.ones((batch_size, seq_length), device=inputs_embeds.device)

        causal_mask = self._prepare_attention_mask(attention_mask, seq_length, inputs_embeds.dtype, inputs_embeds.device)

        hidden_states = inputs_embeds
        all_hidden_states = () if output_hidden_states else None
        all_self_attns = () if output_attentions else None

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

            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,
            )
            hidden_states = layer_outputs[0]

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

        hidden_states = self.norm(hidden_states)

        if output_hidden_states:
            all_hidden_states += (hidden_states,)

        return BaseModelOutputWithPast(
            last_hidden_state=hidden_states,
            past_key_values=past_key_values if use_cache else None,
            hidden_states=all_hidden_states,
            attentions=all_self_attns,
        )

    def _prepare_attention_mask(self, attention_mask, seq_length, dtype, device):
        # Create causal mask
        causal_mask = torch.triu(torch.ones((seq_length, seq_length), device=device), diagonal=1)
        causal_mask = causal_mask.masked_fill(causal_mask == 1, float("-inf"))
        causal_mask = causal_mask.unsqueeze(0).unsqueeze(0)

        # Expand attention mask
        if attention_mask.dim() == 2:
            extended_mask = attention_mask[:, None, None, :]
            extended_mask = (1.0 - extended_mask) * float("-inf")
            causal_mask = causal_mask + extended_mask

        return causal_mask.to(dtype)


class AshishOcrForConditionalGeneration(AshishOcrPreTrainedModel):
    _tied_weights_keys = ["lm_head.weight"]

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

        self.visual = AshishOcrVisionEncoder(config.vision_config)
        self.model = AshishOcrTextModel(config.text_config)
        self.lm_head = nn.Linear(config.text_config.hidden_size, config.text_config.vocab_size, bias=False)

        self.image_token_id = config.image_token_id
        self.video_token_id = config.video_token_id

        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 forward(
        self,
        input_ids: Optional[torch.LongTensor] = None,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        past_key_values: Optional[Cache] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        pixel_values: Optional[torch.FloatTensor] = None,
        pixel_values_videos: Optional[torch.FloatTensor] = None,
        image_grid_thw: Optional[torch.LongTensor] = None,
        video_grid_thw: Optional[torch.LongTensor] = 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, CausalLMOutputWithPast]:
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

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

            # Process images if provided
            if pixel_values is not None:
                image_embeds = self.visual(pixel_values, image_grid_thw)
                image_mask = input_ids == self.image_token_id
                inputs_embeds = inputs_embeds.clone()
                inputs_embeds[image_mask] = image_embeds.view(-1, image_embeds.shape[-1])

            # Process videos if provided
            if pixel_values_videos is not None:
                video_embeds = self.visual(pixel_values_videos, video_grid_thw)
                video_mask = input_ids == self.video_token_id
                inputs_embeds = inputs_embeds.clone()
                inputs_embeds[video_mask] = video_embeds.view(-1, video_embeds.shape[-1])

        outputs = self.model(
            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 = outputs[0]
        logits = self.lm_head(hidden_states)
        logits = logits.float()

        loss = None
        if labels is not None:
            shift_logits = logits[..., :-1, :].contiguous()
            shift_labels = labels[..., 1:].contiguous()
            loss_fct = CrossEntropyLoss()
            shift_logits = shift_logits.view(-1, self.config.text_config.vocab_size)
            shift_labels = shift_labels.view(-1)
            shift_labels = shift_labels.to(shift_logits.device)
            loss = loss_fct(shift_logits, shift_labels)

        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,
        pixel_values=None,
        pixel_values_videos=None,
        image_grid_thw=None,
        video_grid_thw=None,
        **kwargs,
    ):
        if past_key_values is not None:
            input_ids = input_ids[:, -1:]

        model_inputs = {
            "input_ids": input_ids,
            "past_key_values": past_key_values,
            "attention_mask": attention_mask,
            "inputs_embeds": inputs_embeds,
            "pixel_values": pixel_values,
            "pixel_values_videos": pixel_values_videos,
            "image_grid_thw": image_grid_thw,
            "video_grid_thw": video_grid_thw,
        }
        return model_inputs