modeling_internvl_chat.py

try:
    from peft import LoraConfig, get_peft_model
except ImportError:
    LoraConfig = None
# --------------------------------------------------------
# InternVL
# Copyright (c) 2024 OpenGVLab
# Licensed under The MIT License [see LICENSE for details]
# --------------------------------------------------------

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

import torch
import torch.utils.checkpoint
import transformers
from torch import nn
from torch.nn import CrossEntropyLoss
from transformers import GenerationConfig
from transformers.modeling_outputs import CausalLMOutputWithPast, MoeCausalLMOutputWithPast
from transformers.modeling_utils import PreTrainedModel
from transformers.utils import logging
from transformers import AutoModelForCausalLM
from transformers.models.gpt_oss.modeling_gpt_oss import load_balancing_loss_func

from .configuration_internvl_chat import InternVLChatConfig
from .conversation import get_conv_template
from .modeling_intern_vit import InternVisionModel, has_flash_attn

logger = logging.get_logger(__name__)


def version_cmp(v1, v2, op='eq'):
    import operator

    from packaging import version
    op_func = getattr(operator, op)
    return op_func(version.parse(v1), version.parse(v2))


class InternVLChatModel(PreTrainedModel):
    config_class = InternVLChatConfig
    main_input_name = 'pixel_values'
    base_model_prefix = 'language_model'
    _supports_flash_attn_2 = True
    supports_gradient_checkpointing = True
    accepts_loss_kwargs = False
    _no_split_modules = [
        "InternVisionModel",
        "GptOssDecoderLayer",
        # "Qwen3DecoderLayer",
        # "LlamaDecoderLayer",
        # "Qwen2DecoderLayer",
        # "InternLM2DecoderLayer",
    ]

    # support transformers 4.51.+
    _tp_plan = ''

    def __init__(self, config: InternVLChatConfig, vision_model=None, language_model=None, use_flash_attn=True):
        super().__init__(config)

        assert version_cmp(transformers.__version__, '4.37.0', 'ge')
        image_size = config.force_image_size or config.vision_config.image_size
        patch_size = config.vision_config.patch_size
        self.patch_size = patch_size
        self.select_layer = config.select_layer
        self.template = config.template
        self.num_image_token = int((image_size // patch_size) ** 2 * (config.downsample_ratio ** 2))
        self.downsample_ratio = config.downsample_ratio
        self.ps_version = config.ps_version
        use_flash_attn = use_flash_attn if has_flash_attn else False
        config.vision_config.use_flash_attn = True if use_flash_attn else False
        # config.llm_config._attn_implementation = 'flash_attention_2' if use_flash_attn else 'eager'

        logger.info(f'num_image_token: {self.num_image_token}')
        logger.info(f'ps_version: {self.ps_version}')
        if vision_model is not None:
            self.vision_model = vision_model
        else:
            self.vision_model = InternVisionModel(config.vision_config)

        if language_model is not None:
            self.language_model = language_model
        else:
            self.language_model = AutoModelForCausalLM.from_config(config.llm_config)
            logger.info(f"language_model type: {type(self.language_model)}")

        vit_hidden_size = config.vision_config.hidden_size
        llm_hidden_size = config.llm_config.hidden_size

        self.mlp1 = nn.Sequential(
            nn.LayerNorm(vit_hidden_size * int(1 / self.downsample_ratio) ** 2),
            nn.Linear(vit_hidden_size * int(1 / self.downsample_ratio) ** 2, llm_hidden_size),
            nn.GELU(),
            nn.Linear(llm_hidden_size, llm_hidden_size)
        ).to(torch.bfloat16)

        self.num_samples = 0
        self.img_context_token_id = None
        self.conv_template = get_conv_template(self.template)
        self.system_message = self.conv_template.system_message


    def wrap_llm_lora(self, r=128, lora_alpha=256, lora_dropout=0.05):
        # Patched in for native LoRA at multi-rank experiments (MP-010)
        llm_arch_name = self.config.llm_config.architectures[0] if hasattr(self.config, 'llm_config') else self.language_model.config.architectures[0]
        if llm_arch_name == 'InternLM2ForCausalLM':
            target_modules = ['attention.wqkv', 'attention.wo', 'feed_forward.w1', 'feed_forward.w2', 'feed_forward.w3']
        elif llm_arch_name == 'Phi3ForCausalLM':
            target_modules = ['mlp.down_proj', 'mlp.gate_up_proj', 'self_attn.o_proj', 'self_attn.qkv_proj']
        elif llm_arch_name in ['Qwen2ForCausalLM', 'Qwen3ForCausalLM', 'LlamaForCausalLM']:
            target_modules = ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj', 'self_attn.o_proj',
                              'mlp.gate_proj', 'mlp.down_proj', 'mlp.up_proj']
        else:
            raise NotImplementedError
        lora_config = LoraConfig(
            r=r,
            target_modules=target_modules,
            lora_alpha=lora_alpha,
            lora_dropout=lora_dropout,
            task_type='CAUSAL_LM'
        )
        self.language_model = get_peft_model(self.language_model, lora_config)
        self.language_model.enable_input_require_grads()
        self.language_model.print_trainable_parameters()

    def forward(
            self,
            pixel_values: torch.FloatTensor,
            input_ids: torch.LongTensor = None,
            attention_mask: Optional[torch.Tensor] = None,
            position_ids: Optional[torch.LongTensor] = None,
            image_flags: Optional[torch.LongTensor] = None,
            past_key_values: Optional[List[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,
            statistics: Optional[torch.LongTensor] = None,
            cu_seqlens: Optional[torch.LongTensor] = None,
            loss_weight: Optional[List] = None,
            **kwargs,
    ) -> Union[Tuple, CausalLMOutputWithPast]:
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        image_flags = image_flags.squeeze(-1)
        input_embeds = self.language_model.get_input_embeddings()(input_ids).clone()

        vit_embeds = self.extract_feature(pixel_values)
        vit_embeds = vit_embeds[image_flags == 1]
        vit_batch_size = pixel_values.shape[0]

        B, N, C = input_embeds.shape
        input_embeds = input_embeds.reshape(B * N, C)

        if torch.distributed.is_initialized() and torch.distributed.get_rank() == 0:
            # print(f'dynamic ViT batch size: {vit_batch_size}, images per sample: {vit_batch_size / B}, dynamic token length: {N}')
            if statistics is not None:
                num_samples, num_effective_tokens, num_padding_tokens, num_padding_images = statistics.tolist()
                self.num_samples += num_samples
                print(f'total_samples={self.num_samples}, {num_samples=}, {num_padding_tokens=}, {num_padding_images=}, {num_effective_tokens=}')

        ignore = False
        input_ids = input_ids.reshape(B * N)
        selected = (input_ids == self.img_context_token_id)
        try:
            input_embeds[selected] = input_embeds[selected] * 0.0 + vit_embeds.reshape(-1, C)
        except Exception as e:
            vit_embeds = vit_embeds.reshape(-1, C)
            print(f'warning: {e}, input_embeds[selected].shape={input_embeds[selected].shape}, '
                  f'vit_embeds.shape={vit_embeds.shape}')
            n_token = selected.sum()
            input_embeds[selected] = input_embeds[selected] * 0.0 + vit_embeds[:n_token]
            ignore = True

        input_embeds = input_embeds.reshape(B, N, C)

        # Patched: robust layer access (handle PEFT-wrapped + non-wrapped)
        _lm = self.language_model
        _layers = None
        for path_fn in (
            lambda: _lm.base_model.model.model.layers,  # PEFT-wrapped
            lambda: _lm.model.layers,                    # unwrapped Qwen3ForCausalLM
            lambda: _lm.model.model.layers,              # other wrapper
        ):
            try:
                _layers = path_fn()
                break
            except (AttributeError, TypeError):
                continue
        if _layers is None:
            raise AttributeError(f"Cannot find transformer layers in {type(_lm).__name__}")
        for layer in _layers:
            layer.self_attn.cu_seqlens = cu_seqlens

        outputs = self.language_model(
            inputs_embeds=input_embeds,
            attention_mask=attention_mask,
            position_ids=position_ids,
            past_key_values=past_key_values,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
            **kwargs,
        )
        logits = outputs.logits

        loss = None
        aux_loss = None
        if labels is not None and loss_weight is not None:
            loss_weight = torch.tensor(loss_weight, dtype=torch.float32, device=labels.device)
            # Shift so that tokens < n predict n
            shift_logits = logits[..., :-1, :].contiguous()
            shift_labels = labels[..., 1:].contiguous()
            shift_weights = loss_weight[..., 1:].contiguous()
            # Flatten the tokens
            loss_fct = CrossEntropyLoss(reduction='none')
            shift_logits = shift_logits.view(-1, self.language_model.config.vocab_size)
            shift_labels = shift_labels.view(-1)
            shift_weights = shift_weights.view(-1)
            # Enable model parallelism
            shift_labels = shift_labels.to(shift_logits.device)
            shift_weights = shift_weights.to(shift_logits.device)
            loss = loss_fct(shift_logits, shift_labels)

            shift_weights_sum = shift_weights.sum()
            torch.distributed.all_reduce(shift_weights_sum, op=torch.distributed.ReduceOp.AVG)

            loss = loss * shift_weights
            loss = loss.sum() / shift_weights_sum

        elif labels is not None:
            # Shift so that tokens < n predict n
            shift_logits = logits[..., :-1, :].contiguous()
            shift_labels = labels[..., 1:].contiguous()

            if (shift_labels == -100).all():
                ignore = True
                shift_labels = shift_labels * 0

            # Flatten the tokens
            loss_fct = CrossEntropyLoss(reduction='none')
            shift_logits = shift_logits.view(-1, self.language_model.config.vocab_size)
            shift_labels = shift_labels.view(-1)
            # Enable model parallelism
            shift_labels = shift_labels.to(shift_logits.device)
            loss = loss_fct(shift_logits, shift_labels)

            loss_weight = (labels != -100).sum(dim=-1).float()
            loss_weight = 1 / loss_weight.sqrt()
            loss_weight = torch.where(labels != -100, loss_weight.unsqueeze(1), 0.0)

            shift_weights = loss_weight[..., 1:].contiguous()
            shift_weights = shift_weights.view(-1)
            shift_weights = shift_weights.to(shift_logits.device)
            shift_weights_sum = shift_weights.sum()
            torch.distributed.all_reduce(shift_weights_sum, op=torch.distributed.ReduceOp.AVG)

            loss = loss * shift_weights
            loss = loss.sum() / shift_weights_sum

            # debug_input = self.tokenizer.batch_decode(input_ids.reshape(B, N), skip_special_tokens=False)[0]
            # debug_label = self.tokenizer.batch_decode(torch.where(labels >= 0, labels, self.tokenizer.pad_token_id), skip_special_tokens=False)[0]

            # debug_input = debug_input.replace("<IMG_CONTEXT>", "")
            # debug_input = debug_input.replace(self.tokenizer.pad_token, "")
            # debug_label = debug_label.replace(self.tokenizer.pad_token, "")

            # if torch.distributed.get_rank() == 0:
            #     print(
            #         f'[Debug]\n'
            #         f'input ({input_ids.reshape(B, N).shape}): {debug_input}\n'
            #         f'label ({labels.shape}): {debug_label}\n'
            #         f'pad_token: {self.tokenizer.pad_token}\n'
            #         f'[/Debug]\n\n'
            #     )

        if getattr(outputs, 'router_logits', None) is not None:
            aux_loss = load_balancing_loss_func(
                outputs.router_logits,
                self.language_model.num_experts,
                self.language_model.num_experts_per_tok,
                attention_mask,
            )

            if loss is not None:
                if torch.distributed.get_rank() == 0:
                    print(f"[Debug] lm_loss: {loss.item()}, aux_loss: {aux_loss.item()},  weighted_aux_loss: {self.language_model.router_aux_loss_coef * aux_loss.item()}")

                loss = loss + self.language_model.router_aux_loss_coef * aux_loss.to(loss.device)

        if ignore and loss is not None:
            print(f"[Debug] ignore curr loss")
            loss = loss * 0.0

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

        if aux_loss is not None:
            return MoeCausalLMOutputWithPast(
                loss=loss,
                aux_loss=aux_loss,
                logits=logits,
                past_key_values=outputs.past_key_values,
                hidden_states=outputs.hidden_states,
                attentions=outputs.attentions,
            )

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

    def pixel_shuffle(self, x, scale_factor=0.5):
        n, w, h, c = x.size()
        # N, W, H, C --> N, W, H * scale, C // scale
        x = x.view(n, w, int(h * scale_factor), int(c / scale_factor))
        # N, W, H * scale, C // scale --> N, H * scale, W, C // scale
        x = x.permute(0, 2, 1, 3).contiguous()
        # N, H * scale, W, C // scale --> N, H * scale, W * scale, C // (scale ** 2)
        x = x.view(n, int(h * scale_factor), int(w * scale_factor),
                   int(c / (scale_factor * scale_factor)))
        if self.ps_version == 'v1':
            warnings.warn("In ps_version 'v1', the height and width have not been swapped back, "
                          'which results in a transposed image.')
        else:
            x = x.permute(0, 2, 1, 3).contiguous()
        return x

    def extract_feature(self, pixel_values):
        if self.select_layer == -1:
            vit_embeds = self.vision_model(
                pixel_values=pixel_values,
                output_hidden_states=False,
                return_dict=True).last_hidden_state
        else:
            vit_embeds = self.vision_model(
                pixel_values=pixel_values,
                output_hidden_states=True,
                return_dict=True).hidden_states[self.select_layer]
        vit_embeds = vit_embeds[:, 1:, :]

        h = w = int(vit_embeds.shape[1] ** 0.5)
        vit_embeds = vit_embeds.reshape(vit_embeds.shape[0], h, w, -1)
        vit_embeds = self.pixel_shuffle(vit_embeds, scale_factor=self.downsample_ratio)
        vit_embeds = vit_embeds.reshape(vit_embeds.shape[0], -1, vit_embeds.shape[-1])
        vit_embeds = self.mlp1(vit_embeds)
        return vit_embeds

    def batch_chat(self, tokenizer, pixel_values, questions, generation_config, num_patches_list=None,
                   history=None, return_history=False, IMG_START_TOKEN='<img>', IMG_END_TOKEN='</img>',
                   IMG_CONTEXT_TOKEN='<IMG_CONTEXT>', verbose=False, image_counts=None):
        if history is not None or return_history:
            print('Now multi-turn chat is not supported in batch_chat.')
            raise NotImplementedError

        if image_counts is not None:
            num_patches_list = image_counts
            print('Warning: `image_counts` is deprecated. Please use `num_patches_list` instead.')

        img_context_token_id = tokenizer.convert_tokens_to_ids(IMG_CONTEXT_TOKEN)
        self.img_context_token_id = img_context_token_id

        if verbose and pixel_values is not None:
            image_bs = pixel_values.shape[0]
            print(f'dynamic ViT batch size: {image_bs}')

        queries = []
        for idx, num_patches in enumerate(num_patches_list):
            question = questions[idx]
            if pixel_values is not None and '<image>' not in question:
                question = '<image>\n' + question
            template = get_conv_template(self.template)
            template.system_message = self.system_message
            template.append_message(template.roles[0], question)
            template.append_message(template.roles[1], None)
            query = template.get_prompt()

            image_tokens = IMG_START_TOKEN + IMG_CONTEXT_TOKEN * self.num_image_token * num_patches + IMG_END_TOKEN
            query = query.replace('<image>', image_tokens, 1)
            queries.append(query)

        tokenizer.padding_side = 'left'
        model_inputs = tokenizer(queries, return_tensors='pt', padding=True)
        input_ids = model_inputs['input_ids'].to(self.device)
        attention_mask = model_inputs['attention_mask'].to(self.device)

        sep = template.sep.strip() if template.sep2 is None else template.sep2.strip()
        eos_token_id = tokenizer.convert_tokens_to_ids(sep)

        generation_config['eos_token_id'] = eos_token_id
        generation_output = self.generate(
            pixel_values=pixel_values,
            input_ids=input_ids,
            attention_mask=attention_mask,
            **generation_config
        )
        responses = tokenizer.batch_decode(generation_output, skip_special_tokens=False)
        responses = [response.split(sep)[0].strip() for response in responses]
        return responses

    def chat(self, tokenizer, pixel_values, question, generation_config, history=None, return_history=False,
             num_patches_list=None, IMG_START_TOKEN='<img>', IMG_END_TOKEN='</img>', IMG_CONTEXT_TOKEN='<IMG_CONTEXT>',
             verbose=False):

        if history is None and pixel_values is not None and '<image>' not in question:
            question = '<image>\n' + question

        if num_patches_list is None:
            num_patches_list = [pixel_values.shape[0]] if pixel_values is not None else []
        assert pixel_values is None or len(pixel_values) == sum(num_patches_list)

        img_context_token_id = tokenizer.convert_tokens_to_ids(IMG_CONTEXT_TOKEN)
        self.img_context_token_id = img_context_token_id

        template = get_conv_template(self.template)
        template.system_message = self.system_message
        
        sep = template.sep.strip() if template.sep2 is None else template.sep2.strip()
        eos_token_id = tokenizer.convert_tokens_to_ids(sep)

        history = [] if history is None else history
        for (old_question, old_answer) in history:
            template.append_message(template.roles[0], old_question)
            template.append_message(template.roles[1], old_answer)
        template.append_message(template.roles[0], question)
        template.append_message(template.roles[1], None)
        query = template.get_prompt()

        if verbose and pixel_values is not None:
            image_bs = pixel_values.shape[0]
            print(f'dynamic ViT batch size: {image_bs}')

        for num_patches in num_patches_list:
            image_tokens = IMG_START_TOKEN + IMG_CONTEXT_TOKEN * self.num_image_token * num_patches + IMG_END_TOKEN
            query = query.replace('<image>', image_tokens, 1)

        model_inputs = tokenizer(query, return_tensors='pt')
        input_ids = model_inputs['input_ids'].to(self.device)
        attention_mask = model_inputs['attention_mask'].to(self.device)
        generation_config['eos_token_id'] = eos_token_id
        generation_output = self.generate(
            pixel_values=pixel_values,
            input_ids=input_ids,
            attention_mask=attention_mask,
            **generation_config
        )
        response = tokenizer.batch_decode(generation_output, skip_special_tokens=False)[0]
        response = response.split(sep)[0].strip()

        history.append((question, response))
        if return_history:
            return response, history
        else:
            query_to_print = query.replace(IMG_CONTEXT_TOKEN, '')
            query_to_print = query_to_print.replace(f'{IMG_START_TOKEN}{IMG_END_TOKEN}', '<image>')
            if verbose:
                print(query_to_print + response)
            return response

    @torch.no_grad()
    def generate(
            self,
            pixel_values: Optional[torch.FloatTensor] = None,
            input_ids: Optional[torch.FloatTensor] = None,
            attention_mask: Optional[torch.LongTensor] = None,
            visual_features: Optional[torch.FloatTensor] = None,
            generation_config: Optional[GenerationConfig] = None,
            output_hidden_states: Optional[bool] = None,
            **generate_kwargs,
    ) -> torch.LongTensor:

        assert self.img_context_token_id is not None
        if pixel_values is not None:
            if visual_features is not None:
                vit_embeds = visual_features
            else:
                vit_embeds = self.extract_feature(pixel_values)
            input_embeds = self.language_model.get_input_embeddings()(input_ids)
            B, N, C = input_embeds.shape
            input_embeds = input_embeds.reshape(B * N, C)

            input_ids = input_ids.reshape(B * N)
            selected = (input_ids == self.img_context_token_id)
            assert selected.sum() != 0
            input_embeds[selected] = vit_embeds.reshape(-1, C).to(input_embeds.device)

            input_embeds = input_embeds.reshape(B, N, C)
        else:
            input_embeds = self.language_model.get_input_embeddings()(input_ids)

        outputs = self.language_model.generate(
            inputs_embeds=input_embeds,
            attention_mask=attention_mask,
            generation_config=generation_config,
            output_hidden_states=output_hidden_states,
            use_cache=True,
            **generate_kwargs,
        )

        return outputs

    @property
    def lm_head(self):
        return self.language_model.get_output_embeddings()

    def get_output_embeddings(self):
        return self.language_model.get_output_embeddings()

    def get_input_embeddings(self):
        return self.language_model.get_input_embeddings()

    def set_input_embeddings(self, value):
        return self.language_model.set_input_embeddings(value)

    def set_output_embeddings(self, value):
        return self.language_model.set_output_embeddings(value)