ChatUniVi/model/builder.py

import os
import shutil
from transformers import AutoTokenizer, AutoModelForCausalLM, AutoConfig, BitsAndBytesConfig
import torch
from ChatUniVi.model import *
from ChatUniVi.constants import DEFAULT_IMAGE_PATCH_TOKEN, DEFAULT_IM_START_TOKEN, DEFAULT_IM_END_TOKEN
from accelerate import init_empty_weights, load_checkpoint_and_dispatch
from transformers import AutoConfig, AutoModelForCausalLM


def load_pretrained_model(model_path, model_base, model_name, load_8bit=False, load_4bit=False, device_map="auto"):
    kwargs = {"device_map": device_map}

    if load_8bit:
        kwargs['load_in_8bit'] = True
    elif load_4bit:
        kwargs['load_in_4bit'] = True
        kwargs['quantization_config'] = BitsAndBytesConfig(
            load_in_4bit=True,
            bnb_4bit_compute_dtype=torch.float16,
            bnb_4bit_use_double_quant=True,
            bnb_4bit_quant_type='nf4'
        )
    else:
        kwargs['torch_dtype'] = torch.float16

    if 'chatunivi' in model_name.lower():
        # Load ChatUniVi model
        if 'lora' in model_name.lower() and model_base is not None:
            lora_cfg_pretrained = AutoConfig.from_pretrained(model_path)
            tokenizer = AutoTokenizer.from_pretrained(model_base, use_fast=False)
            print('Loading ChatUniVi from base model...')
            model = ChatUniViLlamaForCausalLM.from_pretrained(model_base, low_cpu_mem_usage=True, config=lora_cfg_pretrained, **kwargs)
            token_num, tokem_dim = model.lm_head.out_features, model.lm_head.in_features
            if model.lm_head.weight.shape[0] != token_num:
                model.lm_head.weight = torch.nn.Parameter(torch.empty(token_num, tokem_dim, device=model.device, dtype=model.dtype))
                model.model.embed_tokens.weight = torch.nn.Parameter(torch.empty(token_num, tokem_dim, device=model.device, dtype=model.dtype))

            print('Loading additional ChatUniVi weights...')
            if os.path.exists(os.path.join(model_path, 'non_lora_trainables.bin')):
                non_lora_trainables = torch.load(os.path.join(model_path, 'non_lora_trainables.bin'), map_location='cpu')
            else:
                # this is probably from HF Hub
                from huggingface_hub import hf_hub_download
                def load_from_hf(repo_id, filename, subfolder=None):
                    cache_file = hf_hub_download(
                        repo_id=repo_id,
                        filename=filename,
                        subfolder=subfolder)
                    return torch.load(cache_file, map_location='cpu')
                non_lora_trainables = load_from_hf(model_path, 'non_lora_trainables.bin')
            non_lora_trainables = {(k[11:] if k.startswith('base_model.') else k): v for k, v in non_lora_trainables.items()}
            if any(k.startswith('model.model.') for k in non_lora_trainables):
                non_lora_trainables = {(k[6:] if k.startswith('model.') else k): v for k, v in non_lora_trainables.items()}
            model.load_state_dict(non_lora_trainables, strict=False)

            from peft import PeftModel
            print('Loading LoRA weights...')
            model = PeftModel.from_pretrained(model, model_path)
            print('Merging LoRA weights...')
            model = model.merge_and_unload()
            print('Model is loaded...')
        elif model_base is not None:
            # this may be mm projector only
            print('Loading ChatUniVi from base model...')
            tokenizer = AutoTokenizer.from_pretrained(model_base, use_fast=False)
            cfg_pretrained = AutoConfig.from_pretrained(model_path)
            model = ChatUniViLlamaForCausalLM.from_pretrained(model_base, low_cpu_mem_usage=True, config=cfg_pretrained, **kwargs)

            mm_projector_weights = torch.load(os.path.join(model_path, 'mm_projector.bin'), map_location='cpu')
            mm_projector_weights = {k: v.to(torch.float16) for k, v in mm_projector_weights.items()}
            model.load_state_dict(mm_projector_weights, strict=False)
        else:
            tokenizer = AutoTokenizer.from_pretrained(model_path, use_fast=False)
            model = AutoModelForCausalLM.from_pretrained(model_path, low_cpu_mem_usage=True, **kwargs)
    else:
        # Load language model
        if model_base is not None:
            # PEFT model
            from peft import PeftModel
            tokenizer = AutoTokenizer.from_pretrained(model_base, use_fast=False)
            model = AutoModelForCausalLM.from_pretrained(model_base, torch_dtype=torch.float16, low_cpu_mem_usage=True, device_map="auto")
            print(f"Loading LoRA weights from {model_path}")
            model = PeftModel.from_pretrained(model, model_path)
            print(f"Merging weights")
            model = model.merge_and_unload()
            print('Convert to FP16...')
            model.to(torch.float16)
        else:
            use_fast = False
            tokenizer = AutoTokenizer.from_pretrained(model_path, use_fast=False)
            model = AutoModelForCausalLM.from_pretrained(model_path, low_cpu_mem_usage=True, **kwargs)

    image_processor = None

    if 'chatunivi' in model_name.lower():
        mm_use_im_start_end = getattr(model.config, "mm_use_im_start_end", False)
        mm_use_im_patch_token = getattr(model.config, "mm_use_im_patch_token", True)
        if mm_use_im_patch_token:
            tokenizer.add_tokens([DEFAULT_IMAGE_PATCH_TOKEN], special_tokens=True)
        if mm_use_im_start_end:
            tokenizer.add_tokens([DEFAULT_IM_START_TOKEN, DEFAULT_IM_END_TOKEN], special_tokens=True)
        model.resize_token_embeddings(len(tokenizer))

        vision_tower = model.get_vision_tower()
        if not vision_tower.is_loaded:
            vision_tower.load_model()
            vision_tower.to(device='cuda', dtype=torch.float16)

        image_processor = vision_tower.image_eval_processor

    if hasattr(model.config, "max_sequence_length"):
        context_len = model.config.max_sequence_length
    else:
        context_len = 2048

    return tokenizer, model, image_processor, context_len