|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
import os |
|
import shutil |
|
import warnings |
|
|
|
import torch |
|
from longvu.constants import ( |
|
DEFAULT_IM_END_TOKEN, |
|
DEFAULT_IM_START_TOKEN, |
|
DEFAULT_IMAGE_PATCH_TOKEN, |
|
) |
|
|
|
from longvu.language_model.cambrian_llama import CambrianLlamaForCausalLM |
|
from longvu.language_model.cambrian_qwen import CambrianQwenForCausalLM |
|
|
|
from transformers import ( |
|
AutoConfig, |
|
AutoModelForCausalLM, |
|
AutoTokenizer, |
|
BitsAndBytesConfig, |
|
) |
|
|
|
|
|
def load_pretrained_model( |
|
model_path, |
|
model_base, |
|
model_name, |
|
load_8bit=False, |
|
load_4bit=False, |
|
device_map="auto", |
|
device="cuda", |
|
use_flash_attn=False, |
|
model_args=None, |
|
**kwargs, |
|
): |
|
kwargs = {"device_map": device_map, **kwargs} |
|
|
|
if device != "cuda": |
|
kwargs["device_map"] = {"": device} |
|
|
|
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 use_flash_attn: |
|
kwargs["attn_implementation"] = "flash_attention_2" |
|
|
|
if "cambrian" in model_name.lower(): |
|
|
|
if "lora" in model_name.lower() and model_base is None: |
|
warnings.warn( |
|
"There is `lora` in model name but no `model_base` is provided. If you are loading a LoRA model, please provide the `model_base` argument. Detailed instruction: https://github.com/haotian-liu/LLaVA#launch-a-model-worker-lora-weights-unmerged." |
|
) |
|
if "lora" in model_name.lower() and model_base is not None: |
|
|
|
|
|
from core_ai.llava.language_model.cambrian_llama import CambrianConfig |
|
|
|
lora_cfg_pretrained = CambrianConfig.from_pretrained(model_path) |
|
tokenizer = AutoTokenizer.from_pretrained(model_base, use_fast=False) |
|
print("Loading Cambrian from base model...") |
|
model = CambrianLlamaForCausalLM.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 Cambrian 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: |
|
|
|
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: |
|
|
|
print(f"Loading Cambrian-1 from base model... {model_base}") |
|
tokenizer = AutoTokenizer.from_pretrained(model_base, use_fast=False) |
|
cfg_pretrained = AutoConfig.from_pretrained(model_path) |
|
model = CambrianLlamaForCausalLM.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: |
|
if "qwen" in model_name.lower(): |
|
tokenizer = AutoTokenizer.from_pretrained(model_path) |
|
model = CambrianQwenForCausalLM.from_pretrained( |
|
model_path, low_cpu_mem_usage=True, **kwargs |
|
) |
|
else: |
|
print(f"Loading Cambrian from {model_path}") |
|
tokenizer = AutoTokenizer.from_pretrained(model_path, use_fast=False) |
|
model = CambrianLlamaForCausalLM.from_pretrained( |
|
model_path, low_cpu_mem_usage=True, **kwargs |
|
) |
|
else: |
|
|
|
if model_base is not None: |
|
|
|
from peft import PeftModel |
|
|
|
tokenizer = AutoTokenizer.from_pretrained(model_base, use_fast=False) |
|
model = AutoModelForCausalLM.from_pretrained( |
|
model_base, low_cpu_mem_usage=True, **kwargs |
|
) |
|
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 |
|
if "mpt" in model_name.lower(): |
|
tokenizer = AutoTokenizer.from_pretrained(model_path, use_fast=True) |
|
model = AutoModelForCausalLM.from_pretrained( |
|
model_path, low_cpu_mem_usage=True, trust_remote_code=True, **kwargs |
|
) |
|
else: |
|
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 "llava" 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: |
|
try: |
|
vision_tower.load_model(device_map=device_map) |
|
except ValueError: |
|
|
|
vision_tower.load_model() |
|
vision_tower.to(device="cuda", dtype=torch.float16) |
|
if device_map != "auto": |
|
vision_tower.to(device=device_map, dtype=torch.float16) |
|
image_processor = vision_tower.image_processor |
|
elif "cambrian" 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_aux_list = model.get_vision_tower_aux_list() |
|
|
|
for vision_tower_aux in vision_tower_aux_list: |
|
if not vision_tower_aux.is_loaded: |
|
vision_tower_aux.load_model(device_map=device_map) |
|
vision_tower_aux.to(device=device, dtype=torch.float16) |
|
|
|
image_processor = [ |
|
vision_tower_aux.image_processor |
|
for vision_tower_aux in vision_tower_aux_list |
|
] |
|
|
|
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 |
|
|