"...git@developer.sourcefind.cn:renzhc/diffusers_dcu.git" did not exist on "df91c44712381c021c0f4855a623b1a1c32f28b7"
Unverified Commit 2b605ab1 authored by Li Bo's avatar Li Bo Committed by GitHub
Browse files

[Feat/Fix] Refactoring Llava models into single file (#475)

parent 947bda73
...@@ -22,11 +22,7 @@ import aiohttp ...@@ -22,11 +22,7 @@ import aiohttp
import requests import requests
from llava.conversation import ( from llava.conversation import (
default_conversation,
conv_templates,
SeparatorStyle,
conv_llava_llama_3, conv_llava_llama_3,
conv_qwen,
) )
...@@ -43,7 +39,8 @@ async def test_concurrent(args): ...@@ -43,7 +39,8 @@ async def test_concurrent(args):
prompt = "<image>\nPlease generate caption towards this image." prompt = "<image>\nPlease generate caption towards this image."
conv_template = copy.deepcopy(conv_llava_llama_3) conv_template = copy.deepcopy(conv_llava_llama_3)
conv_template.append_message(role="user", message=prompt) conv_template.append_message(role=conv_template.roles[0], message=prompt)
conv_template.append_message(role=conv_template.roles[1], message=None)
prompt_with_template = conv_template.get_prompt() prompt_with_template = conv_template.get_prompt()
response = [] response = []
for i in range(1): for i in range(1):
...@@ -74,7 +71,8 @@ def test_streaming(args): ...@@ -74,7 +71,8 @@ def test_streaming(args):
url = f"{args.host}:{args.port}" url = f"{args.host}:{args.port}"
prompt = "<image>\nPlease generate caption towards this image." prompt = "<image>\nPlease generate caption towards this image."
conv_template = copy.deepcopy(conv_llava_llama_3) conv_template = copy.deepcopy(conv_llava_llama_3)
conv_template.append_message(role="user", message=prompt) conv_template.append_message(role=conv_template.roles[0], message=prompt)
conv_template.append_message(role=conv_template.roles[1], message=None)
prompt_with_template = conv_template.get_prompt() prompt_with_template = conv_template.get_prompt()
pload = { pload = {
"text": prompt_with_template, "text": prompt_with_template,
......
...@@ -22,11 +22,7 @@ import aiohttp ...@@ -22,11 +22,7 @@ import aiohttp
import requests import requests
from llava.conversation import ( from llava.conversation import (
default_conversation, conv_qwen
conv_templates,
SeparatorStyle,
conv_llava_llama_3,
conv_qwen,
) )
...@@ -43,7 +39,8 @@ async def test_concurrent(args): ...@@ -43,7 +39,8 @@ async def test_concurrent(args):
prompt = "<image>\nPlease generate caption towards this image." prompt = "<image>\nPlease generate caption towards this image."
conv_template = copy.deepcopy(conv_qwen) conv_template = copy.deepcopy(conv_qwen)
conv_template.append_message(role="user", message=prompt) conv_template.append_message(role=conv_template.roles[0], message=prompt)
conv_template.append_message(role=conv_template.roles[1], message=None)
prompt_with_template = conv_template.get_prompt() prompt_with_template = conv_template.get_prompt()
response = [] response = []
for i in range(1): for i in range(1):
...@@ -74,7 +71,8 @@ def test_streaming(args): ...@@ -74,7 +71,8 @@ def test_streaming(args):
url = f"{args.host}:{args.port}" url = f"{args.host}:{args.port}"
prompt = "<image>\nPlease generate caption towards this image." prompt = "<image>\nPlease generate caption towards this image."
conv_template = copy.deepcopy(conv_qwen) conv_template = copy.deepcopy(conv_qwen)
conv_template.append_message(role="user", message=prompt) conv_template.append_message(role=conv_template.roles[0], message=prompt)
conv_template.append_message(role=conv_template.roles[1], message=None)
prompt_with_template = conv_template.get_prompt() prompt_with_template = conv_template.get_prompt()
pload = { pload = {
"text": prompt_with_template, "text": prompt_with_template,
...@@ -113,5 +111,5 @@ if __name__ == "__main__": ...@@ -113,5 +111,5 @@ if __name__ == "__main__":
parser.add_argument("--host", type=str, default="http://127.0.0.1") parser.add_argument("--host", type=str, default="http://127.0.0.1")
parser.add_argument("--port", type=int, default=30000) parser.add_argument("--port", type=int, default=30000)
args = parser.parse_args() args = parser.parse_args()
# asyncio.run(test_concurrent(args)) asyncio.run(test_concurrent(args))
test_streaming(args) test_streaming(args)
...@@ -421,7 +421,12 @@ def import_model_classes(): ...@@ -421,7 +421,12 @@ def import_model_classes():
if not ispkg: if not ispkg:
module = importlib.import_module(name) module = importlib.import_module(name)
if hasattr(module, "EntryClass"): if hasattr(module, "EntryClass"):
model_arch_name_to_cls[module.EntryClass.__name__] = module.EntryClass entry = module.EntryClass
if isinstance(entry, list): # To support multiple model classes in one module
for cls in entry:
model_arch_name_to_cls[cls.__name__] = cls
else:
model_arch_name_to_cls[entry.__name__] = entry
return model_arch_name_to_cls return model_arch_name_to_cls
......
...@@ -5,7 +5,7 @@ from typing import List, Iterable, Optional, Tuple ...@@ -5,7 +5,7 @@ from typing import List, Iterable, Optional, Tuple
import numpy as np import numpy as np
import torch import torch
from torch import nn from torch import nn
from transformers import CLIPVisionModel, LlavaConfig from transformers import CLIPVisionModel, CLIPVisionConfig, LlavaConfig, Qwen2Config, MistralConfig
from transformers.models.llava.modeling_llava import LlavaMultiModalProjector from transformers.models.llava.modeling_llava import LlavaMultiModalProjector
from vllm.model_executor.layers.quantization.base_config import QuantizationConfig from vllm.model_executor.layers.quantization.base_config import QuantizationConfig
from vllm.model_executor.model_loader.weight_utils import default_weight_loader from vllm.model_executor.model_loader.weight_utils import default_weight_loader
...@@ -18,6 +18,8 @@ from sglang.srt.mm_utils import ( ...@@ -18,6 +18,8 @@ from sglang.srt.mm_utils import (
unpad_image_shape, unpad_image_shape,
) )
from sglang.srt.models.llama2 import LlamaForCausalLM from sglang.srt.models.llama2 import LlamaForCausalLM
from sglang.srt.models.qwen2 import Qwen2ForCausalLM
from sglang.srt.models.mistral import MistralForCausalLM
class LlavaLlamaForCausalLM(nn.Module): class LlavaLlamaForCausalLM(nn.Module):
...@@ -287,8 +289,101 @@ class LlavaLlamaForCausalLM(nn.Module): ...@@ -287,8 +289,101 @@ class LlavaLlamaForCausalLM(nn.Module):
return self.image_size // self.patch_size return self.image_size // self.patch_size
first_call = True class LlavaQwenForCausalLM(LlavaLlamaForCausalLM):
def __init__(
self,
config: LlavaConfig,
quant_config: Optional[QuantizationConfig] = None,
) -> None:
super().__init__(config, quant_config=quant_config)
self.config = config
self.vision_tower = None
if getattr(self.config, "vision_config", None) is None:
self.config.vision_config = CLIPVisionConfig(self.config.mm_vision_tower)
if getattr(self.config, "text_config", None) is None:
self.config.text_config = Qwen2Config(self.config._name_or_path)
self.config.vision_config.hidden_size = config.mm_hidden_size
self.config.text_config.hidden_size = config.hidden_size
if getattr(self.config, "projector_hidden_act", None) is None:
self.config.projector_hidden_act = "gelu"
if getattr(self.config, "image_token_index", None) is None:
self.config.image_token_index = 151646
self.multi_modal_projector = LlavaMultiModalProjector(config)
self.language_model = Qwen2ForCausalLM(config, quant_config=quant_config)
if "unpad" in getattr(config, "mm_patch_merge_type", ""):
self.language_model.model.image_newline = nn.Parameter(
torch.empty(config.text_config.hidden_size, dtype=torch.float16)
)
def pad_input_ids(self, input_ids, pad_value, pt_shape=None, image_size=None):
new_image_feature_len = self.image_feature_len
# now only support spatial_unpad + anyres
if self.mm_patch_merge_type.startswith("spatial"):
height = width = self.num_patches_per_side
if pt_shape[0] > 1:
if self.image_aspect_ratio == "anyres":
num_patch_width, num_patch_height = get_anyres_image_grid_shape(
image_size,
self.image_grid_pinpoints,
self.vision_tower.config.image_size,
)
if "unpad" in self.mm_patch_merge_type:
h = num_patch_height * height
w = num_patch_width * width
new_h, new_w = unpad_image_shape(h, w, image_size)
new_image_feature_len += new_h * (new_w + 1)
pad_ids = pad_value * (
(new_image_feature_len + len(pad_value)) // len(pad_value)
)
offset = input_ids.index(self.config.image_token_index)
# old_len + pad_len - 1, because we need to remove image_token_id
new_input_ids = (
input_ids[:offset]
+ pad_ids[:new_image_feature_len]
+ input_ids[offset + 1 :]
)
return new_input_ids, offset
class LlavaMistralForCausalLM(LlavaLlamaForCausalLM):
def __init__(
self,
config: LlavaConfig,
quant_config: Optional[QuantizationConfig] = None,
) -> None:
super().__init__(config, quant_config=quant_config)
self.config = config
self.vision_tower = None
if getattr(self.config, "vision_config", None) is None:
self.config.vision_config = CLIPVisionConfig(self.config.mm_vision_tower)
if getattr(self.config, "text_config", None) is None:
self.config.text_config = MistralConfig(self.config._name_or_path)
self.config.vision_config.hidden_size = config.mm_hidden_size
self.config.text_config.hidden_size = config.hidden_size
if getattr(self.config, "projector_hidden_act", None) is None:
self.config.projector_hidden_act = "gelu"
if getattr(self.config, "image_token_index", None) is None:
self.config.image_token_index = 32000
self.multi_modal_projector = LlavaMultiModalProjector(config)
self.language_model = MistralForCausalLM(config, quant_config=quant_config)
if "unpad" in getattr(config, "mm_patch_merge_type", ""):
self.language_model.model.image_newline = nn.Parameter(
torch.empty(config.text_config.hidden_size, dtype=torch.float16)
)
first_call = True
def clip_vision_embed_forward(self, pixel_values: torch.FloatTensor) -> torch.Tensor: def clip_vision_embed_forward(self, pixel_values: torch.FloatTensor) -> torch.Tensor:
batch_size = pixel_values.shape[0] batch_size = pixel_values.shape[0]
...@@ -319,4 +414,8 @@ def monkey_path_clip_vision_embed_forward(): ...@@ -319,4 +414,8 @@ def monkey_path_clip_vision_embed_forward():
) )
EntryClass = LlavaLlamaForCausalLM EntryClass = [
LlavaLlamaForCausalLM,
LlavaQwenForCausalLM,
LlavaMistralForCausalLM
]
"""Inference-only LLaVa model compatible with HuggingFace weights."""
from typing import List, Iterable, Optional, Tuple
import numpy as np
import torch
from torch import nn
from transformers import CLIPVisionConfig, CLIPVisionModel, LlavaConfig, MistralConfig
from transformers.models.llava.modeling_llava import LlavaMultiModalProjector
from vllm.model_executor.layers.quantization.base_config import QuantizationConfig
from vllm.model_executor.model_loader.weight_utils import default_weight_loader
from sglang.srt.managers.router.infer_batch import ForwardMode
from sglang.srt.managers.router.model_runner import InputMetadata
from sglang.srt.mm_utils import (
get_anyres_image_grid_shape,
unpad_image,
unpad_image_shape,
)
from sglang.srt.models.mistral import MistralForCausalLM
class LlavaMistralForCausalLM(nn.Module):
def __init__(
self,
config: LlavaConfig,
quant_config: Optional[QuantizationConfig] = None,
) -> None:
super().__init__()
self.config = config
self.vision_tower = None
if getattr(self.config, "vision_config", None) is None:
self.config.vision_config = CLIPVisionConfig(self.config.mm_vision_tower)
if getattr(self.config, "text_config", None) is None:
self.config.text_config = MistralConfig(self.config._name_or_path)
self.config.vision_config.hidden_size = config.mm_hidden_size
self.config.text_config.hidden_size = config.hidden_size
if getattr(self.config, "projector_hidden_act", None) is None:
self.config.projector_hidden_act = "gelu"
if getattr(self.config, "image_token_index", None) is None:
self.config.image_token_index = 32000
self.multi_modal_projector = LlavaMultiModalProjector(config)
self.language_model = MistralForCausalLM(config, quant_config=quant_config)
if "unpad" in getattr(config, "mm_patch_merge_type", ""):
self.language_model.model.image_newline = nn.Parameter(
torch.empty(config.text_config.hidden_size, dtype=torch.float16)
)
def pad_input_ids(self, input_ids, pad_value, pt_shape=None, image_size=None):
new_image_feature_len = self.image_feature_len
# now only support spatial_unpad + anyres
if self.mm_patch_merge_type.startswith("spatial"):
height = width = self.num_patches_per_side
if pt_shape[0] > 1:
if self.image_aspect_ratio == "anyres":
num_patch_width, num_patch_height = get_anyres_image_grid_shape(
image_size,
self.image_grid_pinpoints,
self.vision_tower.config.image_size,
)
if "unpad" in self.mm_patch_merge_type:
h = num_patch_height * height
w = num_patch_width * width
new_h, new_w = unpad_image_shape(h, w, image_size)
new_image_feature_len += new_h * (new_w + 1)
pad_ids = pad_value * (
(new_image_feature_len + len(pad_value)) // len(pad_value)
)
offset = input_ids.index(self.config.image_token_index)
# old_len + pad_len - 1, because we need to remove image_token_id
new_input_ids = (
input_ids[:offset]
+ pad_ids[:new_image_feature_len]
+ input_ids[offset + 1 :]
)
return new_input_ids, offset
def encode_images(self, pixel_values: torch.Tensor) -> torch.Tensor:
image_outputs = self.vision_tower(pixel_values, output_hidden_states=True)
# NOTE: This is not memory efficient. (output_hidden_states=True) will save all the hidden stated.
selected_image_feature = image_outputs.hidden_states[self.vision_feature_layer]
if self.vision_feature_select_strategy in ["default", "patch"]:
selected_image_feature = selected_image_feature[:, 1:]
elif self.vision_feature_select_strategy == "full":
selected_image_feature = selected_image_feature
else:
raise ValueError(
f"Unexpected select feature strategy: {self.config.vision_feature_select_strategy}"
)
image_features = self.multi_modal_projector(selected_image_feature)
return image_features
def forward(
self,
input_ids: torch.LongTensor,
positions: torch.Tensor,
input_metadata: InputMetadata,
pixel_values: Optional[List[Optional[np.array]]] = None,
image_sizes: Optional[List[List[int]]] = None,
image_offsets: Optional[List[int]] = None,
) -> torch.Tensor:
if input_metadata.forward_mode == ForwardMode.EXTEND:
bs = input_metadata.batch_size
# Embed text input
input_embeds = self.language_model.model.embed_tokens(input_ids)
# Embed vision input
need_vision = (
(positions[input_metadata.extend_start_loc] < self.image_feature_len)
.cpu()
.numpy()
)
# FIXME: We need to substract the length of the system prompt
has_pixel = np.array([pixel_values[i] is not None for i in range(bs)])
need_vision = need_vision & has_pixel
if need_vision.any():
pixel_values = [pixel_values[i] for i in range(bs) if need_vision[i]]
image_sizes = [image_sizes[i] for i in range(bs) if need_vision[i]]
########## Encode Image ########
if pixel_values[0].ndim == 4:
# llava-hd: BS, num_patch, C=3, H=336, W=336, num_patch obtained from process_images
np.concatenate(pixel_values, axis=0)
# ndim=4
concat_images = torch.tensor(
np.concatenate(pixel_values, axis=0),
device=self.vision_tower.device,
)
image_features = self.encode_images(concat_images)
split_sizes = [image.shape[0] for image in pixel_values]
image_features = torch.split(image_features, split_sizes, dim=0)
# hd image_features: BS, num_patch, 576, 4096
else:
# normal pixel: BS, C=3, H=336, W=336
pixel_values = torch.tensor(
np.array(pixel_values), device=self.vision_tower.device
)
image_features = self.encode_images(pixel_values)
# image_features: BS, 576, 4096
if self.mm_patch_merge_type.startswith("spatial"):
new_image_features = []
for image_idx, image_feature in enumerate(image_features):
if image_feature.shape[0] > 1:
base_image_feature = image_feature[0]
image_feature = image_feature[1:]
height = width = self.num_patches_per_side
assert height * width == base_image_feature.shape[0]
if self.image_aspect_ratio == "anyres":
(
num_patch_width,
num_patch_height,
) = get_anyres_image_grid_shape(
image_sizes[image_idx],
self.image_grid_pinpoints,
self.vision_tower.config.image_size,
)
image_feature = image_feature.view(
num_patch_height, num_patch_width, height, width, -1
)
else:
raise NotImplementedError()
if "unpad" in self.mm_patch_merge_type:
image_feature = image_feature.permute(
4, 0, 2, 1, 3
).contiguous()
image_feature = image_feature.flatten(1, 2).flatten(
2, 3
)
image_feature = unpad_image(
image_feature, image_sizes[image_idx]
)
image_feature = torch.cat(
(
image_feature,
self.language_model.model.image_newline[
:, None, None
].expand(*image_feature.shape[:-1], 1),
),
dim=-1,
)
image_feature = image_feature.flatten(1, 2).transpose(
0, 1
)
else:
image_feature = image_feature.permute(
0, 2, 1, 3, 4
).contiguous()
image_feature = image_feature.flatten(0, 3)
image_feature = torch.cat(
(base_image_feature, image_feature), dim=0
)
else:
image_feature = image_feature[0]
if "unpad" in self.mm_patch_merge_type:
image_feature = torch.cat(
(
image_feature,
self.language_model.model.image_newline[None],
),
dim=0,
)
new_image_features.append(image_feature)
image_features = new_image_features
extend_start_loc_cpu = input_metadata.extend_start_loc.cpu().numpy()
pt = 0
for i in range(bs):
if not need_vision[i]:
continue
start_idx = extend_start_loc_cpu[i]
pad_len, pad_dim = image_features[pt].shape # 576, 4096
dim = input_embeds.shape[1]
assert (
pad_dim == dim
), "invalid pad_dim={}, input_embed_dim={}!".format(pad_dim, dim)
# Fill in the placeholder for the image
try:
input_embeds[
start_idx
+ image_offsets[i] : start_idx
+ image_offsets[i]
+ pad_len
] = image_features[pt]
except RuntimeError as e:
print(f"RuntimeError in llava image encoding: {e}")
print(input_embeds.shape)
print(start_idx, image_offsets[i])
pt += 1
return self.language_model(
input_ids, positions, input_metadata, input_embeds=input_embeds
)
elif input_metadata.forward_mode == ForwardMode.DECODE:
return self.language_model(input_ids, positions, input_metadata)
def load_weights(self, weights: Iterable[Tuple[str, torch.Tensor]]):
# load clip vision model by cfg['mm_vision_tower']:
# huggingface_name or path_of_clip_relative_to_llava_model_dir
vision_path = self.config.mm_vision_tower
self.vision_tower = CLIPVisionModel.from_pretrained(
vision_path, torch_dtype=torch.float16
).cuda()
self.vision_tower.eval()
self.vision_feature_layer = self.config.mm_vision_select_layer
self.vision_feature_select_strategy = self.config.mm_vision_select_feature
self.image_size = self.vision_tower.config.image_size
self.patch_size = self.vision_tower.config.patch_size
self.mm_patch_merge_type = getattr(self.config, "mm_patch_merge_type", "flat")
self.image_aspect_ratio = getattr(self.config, "image_aspect_ratio", "square")
self.image_grid_pinpoints = getattr(self.config, "image_grid_pinpoints", None)
self.image_feature_len = int((self.image_size / self.patch_size) ** 2)
if self.vision_feature_select_strategy == "patch":
pass
elif self.vision_feature_select_strategy == "cls_patch":
self.image_feature_len += 1
else:
raise ValueError(f"Unexpected select feature: {self.select_feature}")
# load mm_projector
projector_weights = {
"model.mm_projector.0": "multi_modal_projector.linear_1",
"model.mm_projector.2": "multi_modal_projector.linear_2",
"model.vision_tower.vision_tower": "vision_tower", # Update the vision tower weights if we find them in the checkpoint (it may be finetuned).
}
params_dict = dict(self.named_parameters())
weights = list(weights)
for name, loaded_weight in weights:
# FIXME: why projector weights read two times?
if "projector" in name or "vision_tower" in name:
for weight_name, param_name in projector_weights.items():
if weight_name in name:
name = name.replace(weight_name, param_name)
param = params_dict[name]
weight_loader = getattr(param, "weight_loader", default_weight_loader)
weight_loader(param, loaded_weight)
# load language model
self.language_model.load_weights(weights)
monkey_path_clip_vision_embed_forward()
@property
def num_patches_per_side(self):
return self.image_size // self.patch_size
first_call = True
def clip_vision_embed_forward(self, pixel_values: torch.FloatTensor) -> torch.Tensor:
batch_size = pixel_values.shape[0]
# Move this conv layer to CPU to avoid a bug in torch >= 2.1 on A10G.
global first_call
if first_call:
self.patch_embedding.cpu().float()
first_call = False
pixel_values = pixel_values.to(dtype=torch.float32, device="cpu")
patch_embeds = self.patch_embedding(pixel_values).cuda().half()
patch_embeds = patch_embeds.flatten(2).transpose(1, 2)
class_embeds = self.class_embedding.expand(batch_size, 1, -1)
embeddings = torch.cat([class_embeds, patch_embeds], dim=1)
embeddings = embeddings + self.position_embedding(self.position_ids)
return embeddings
def monkey_path_clip_vision_embed_forward():
import transformers
setattr(
transformers.models.clip.modeling_clip.CLIPVisionEmbeddings,
"forward",
clip_vision_embed_forward,
)
EntryClass = LlavaMistralForCausalLM
"""Inference-only LLaVa model compatible with HuggingFace weights."""
from typing import List, Iterable, Optional, Tuple
import numpy as np
import torch
from torch import nn
from transformers import CLIPVisionConfig, CLIPVisionModel, LlavaConfig, Qwen2Config
from transformers.models.llava.modeling_llava import LlavaMultiModalProjector
from vllm.model_executor.layers.quantization.base_config import QuantizationConfig
from vllm.model_executor.model_loader.weight_utils import default_weight_loader
from sglang.srt.managers.router.infer_batch import ForwardMode
from sglang.srt.managers.router.model_runner import InputMetadata
from sglang.srt.mm_utils import (
get_anyres_image_grid_shape,
unpad_image,
unpad_image_shape,
)
from sglang.srt.models.qwen2 import Qwen2ForCausalLM
class LlavaQwenForCausalLM(nn.Module):
def __init__(
self,
config: LlavaConfig,
quant_config: Optional[QuantizationConfig] = None,
) -> None:
super().__init__()
self.config = config
self.vision_tower = None
if getattr(self.config, "vision_config", None) is None:
self.config.vision_config = CLIPVisionConfig(self.config.mm_vision_tower)
if getattr(self.config, "text_config", None) is None:
self.config.text_config = Qwen2Config(self.config._name_or_path)
self.config.vision_config.hidden_size = config.mm_hidden_size
self.config.text_config.hidden_size = config.hidden_size
if getattr(self.config, "projector_hidden_act", None) is None:
self.config.projector_hidden_act = "gelu"
if getattr(self.config, "image_token_index", None) is None:
self.config.image_token_index = 151646
self.multi_modal_projector = LlavaMultiModalProjector(config)
self.language_model = Qwen2ForCausalLM(config, quant_config=quant_config)
if "unpad" in getattr(config, "mm_patch_merge_type", ""):
self.language_model.model.image_newline = nn.Parameter(
torch.empty(config.text_config.hidden_size, dtype=torch.float16)
)
def pad_input_ids(self, input_ids, pad_value, pt_shape=None, image_size=None):
new_image_feature_len = self.image_feature_len
# now only support spatial_unpad + anyres
if self.mm_patch_merge_type.startswith("spatial"):
height = width = self.num_patches_per_side
if pt_shape[0] > 1:
if self.image_aspect_ratio == "anyres":
num_patch_width, num_patch_height = get_anyres_image_grid_shape(
image_size,
self.image_grid_pinpoints,
self.vision_tower.config.image_size,
)
if "unpad" in self.mm_patch_merge_type:
h = num_patch_height * height
w = num_patch_width * width
new_h, new_w = unpad_image_shape(h, w, image_size)
new_image_feature_len += new_h * (new_w + 1)
pad_ids = pad_value * (
(new_image_feature_len + len(pad_value)) // len(pad_value)
)
offset = input_ids.index(self.config.image_token_index)
# old_len + pad_len - 1, because we need to remove image_token_id
new_input_ids = (
input_ids[:offset]
+ pad_ids[:new_image_feature_len]
+ input_ids[offset + 1 :]
)
return new_input_ids, offset
def encode_images(self, pixel_values: torch.Tensor) -> torch.Tensor:
image_outputs = self.vision_tower(pixel_values, output_hidden_states=True)
# NOTE: This is not memory efficient. (output_hidden_states=True) will save all the hidden stated.
selected_image_feature = image_outputs.hidden_states[self.vision_feature_layer]
if self.vision_feature_select_strategy in ["default", "patch"]:
selected_image_feature = selected_image_feature[:, 1:]
elif self.vision_feature_select_strategy == "full":
selected_image_feature = selected_image_feature
else:
raise ValueError(
f"Unexpected select feature strategy: {self.config.vision_feature_select_strategy}"
)
image_features = self.multi_modal_projector(selected_image_feature)
return image_features
def forward(
self,
input_ids: torch.LongTensor,
positions: torch.Tensor,
input_metadata: InputMetadata,
pixel_values: Optional[List[Optional[np.array]]] = None,
image_sizes: Optional[List[List[int]]] = None,
image_offsets: Optional[List[int]] = None,
) -> torch.Tensor:
if input_metadata.forward_mode == ForwardMode.EXTEND:
bs = input_metadata.batch_size
# Embed text input
input_embeds = self.language_model.model.embed_tokens(input_ids)
# Embed vision input
need_vision = (
(positions[input_metadata.extend_start_loc] < self.image_feature_len)
.cpu()
.numpy()
)
# FIXME: We need to substract the length of the system prompt
has_pixel = np.array([pixel_values[i] is not None for i in range(bs)])
need_vision = need_vision & has_pixel
if need_vision.any():
pixel_values = [pixel_values[i] for i in range(bs) if need_vision[i]]
image_sizes = [image_sizes[i] for i in range(bs) if need_vision[i]]
########## Encode Image ########
if pixel_values[0].ndim == 4:
# llava-hd: BS, num_patch, C=3, H=336, W=336, num_patch obtained from process_images
np.concatenate(pixel_values, axis=0)
# ndim=4
concat_images = torch.tensor(
np.concatenate(pixel_values, axis=0),
device=self.vision_tower.device,
)
image_features = self.encode_images(concat_images)
split_sizes = [image.shape[0] for image in pixel_values]
image_features = torch.split(image_features, split_sizes, dim=0)
# hd image_features: BS, num_patch, 576, 4096
else:
# normal pixel: BS, C=3, H=336, W=336
pixel_values = torch.tensor(
np.array(pixel_values), device=self.vision_tower.device
)
image_features = self.encode_images(pixel_values)
# image_features: BS, 576, 4096
if self.mm_patch_merge_type.startswith("spatial"):
new_image_features = []
for image_idx, image_feature in enumerate(image_features):
if image_feature.shape[0] > 1:
base_image_feature = image_feature[0]
image_feature = image_feature[1:]
height = width = self.num_patches_per_side
assert height * width == base_image_feature.shape[0]
if self.image_aspect_ratio == "anyres":
(
num_patch_width,
num_patch_height,
) = get_anyres_image_grid_shape(
image_sizes[image_idx],
self.image_grid_pinpoints,
self.vision_tower.config.image_size,
)
image_feature = image_feature.view(
num_patch_height, num_patch_width, height, width, -1
)
else:
raise NotImplementedError()
if "unpad" in self.mm_patch_merge_type:
image_feature = image_feature.permute(
4, 0, 2, 1, 3
).contiguous()
image_feature = image_feature.flatten(1, 2).flatten(
2, 3
)
image_feature = unpad_image(
image_feature, image_sizes[image_idx]
)
image_feature = torch.cat(
(
image_feature,
self.language_model.model.image_newline[
:, None, None
].expand(*image_feature.shape[:-1], 1),
),
dim=-1,
)
image_feature = image_feature.flatten(1, 2).transpose(
0, 1
)
else:
image_feature = image_feature.permute(
0, 2, 1, 3, 4
).contiguous()
image_feature = image_feature.flatten(0, 3)
image_feature = torch.cat(
(base_image_feature, image_feature), dim=0
)
else:
image_feature = image_feature[0]
if "unpad" in self.mm_patch_merge_type:
image_feature = torch.cat(
(
image_feature,
self.language_model.model.image_newline[None],
),
dim=0,
)
new_image_features.append(image_feature)
image_features = new_image_features
extend_start_loc_cpu = input_metadata.extend_start_loc.cpu().numpy()
pt = 0
for i in range(bs):
if not need_vision[i]:
continue
start_idx = extend_start_loc_cpu[i]
pad_len, pad_dim = image_features[pt].shape # 576, 4096
dim = input_embeds.shape[1]
assert (
pad_dim == dim
), "invalid pad_dim={}, input_embed_dim={}!".format(pad_dim, dim)
# Fill in the placeholder for the image
try:
input_embeds[
start_idx
+ image_offsets[i] : start_idx
+ image_offsets[i]
+ pad_len
] = image_features[pt]
except RuntimeError as e:
print(f"RuntimeError in llava image encoding: {e}")
print(input_embeds.shape)
print(start_idx, image_offsets[i])
pt += 1
return self.language_model(
input_ids, positions, input_metadata, input_embeds=input_embeds
)
elif input_metadata.forward_mode == ForwardMode.DECODE:
return self.language_model(input_ids, positions, input_metadata)
def load_weights(self, weights: Iterable[Tuple[str, torch.Tensor]]):
# load clip vision model by cfg['mm_vision_tower']:
# huggingface_name or path_of_clip_relative_to_llava_model_dir
vision_path = self.config.mm_vision_tower
self.vision_tower = CLIPVisionModel.from_pretrained(
vision_path, torch_dtype=torch.float16
).cuda()
self.vision_tower.eval()
self.vision_feature_layer = self.config.mm_vision_select_layer
self.vision_feature_select_strategy = self.config.mm_vision_select_feature
self.image_size = self.vision_tower.config.image_size
self.patch_size = self.vision_tower.config.patch_size
self.mm_patch_merge_type = getattr(self.config, "mm_patch_merge_type", "flat")
self.image_aspect_ratio = getattr(self.config, "image_aspect_ratio", "square")
self.image_grid_pinpoints = getattr(self.config, "image_grid_pinpoints", None)
self.image_feature_len = int((self.image_size / self.patch_size) ** 2)
if self.vision_feature_select_strategy == "patch":
pass
elif self.vision_feature_select_strategy == "cls_patch":
self.image_feature_len += 1
else:
raise ValueError(f"Unexpected select feature: {self.select_feature}")
# load mm_projector
projector_weights = {
"model.mm_projector.0": "multi_modal_projector.linear_1",
"model.mm_projector.2": "multi_modal_projector.linear_2",
"model.vision_tower.vision_tower": "vision_tower", # Update the vision tower weights if we find them in the checkpoint (it may be finetuned).
}
params_dict = dict(self.named_parameters())
weights = list(weights)
for name, loaded_weight in weights:
# FIXME: why projector weights read two times?
if "projector" in name or "vision_tower" in name:
for weight_name, param_name in projector_weights.items():
if weight_name in name:
name = name.replace(weight_name, param_name)
param = params_dict[name]
weight_loader = getattr(param, "weight_loader", default_weight_loader)
weight_loader(param, loaded_weight)
# load language model
self.language_model.load_weights(weights)
monkey_path_clip_vision_embed_forward()
@property
def num_patches_per_side(self):
return self.image_size // self.patch_size
first_call = True
def clip_vision_embed_forward(self, pixel_values: torch.FloatTensor) -> torch.Tensor:
batch_size = pixel_values.shape[0]
# Move this conv layer to CPU to avoid a bug in torch >= 2.1 on A10G.
global first_call
if first_call:
self.patch_embedding.cpu().float()
first_call = False
pixel_values = pixel_values.to(dtype=torch.float32, device="cpu")
patch_embeds = self.patch_embedding(pixel_values).cuda().half()
patch_embeds = patch_embeds.flatten(2).transpose(1, 2)
class_embeds = self.class_embedding.expand(batch_size, 1, -1)
embeddings = torch.cat([class_embeds, patch_embeds], dim=1)
embeddings = embeddings + self.position_embedding(self.position_ids)
return embeddings
def monkey_path_clip_vision_embed_forward():
import transformers
setattr(
transformers.models.clip.modeling_clip.CLIPVisionEmbeddings,
"forward",
clip_vision_embed_forward,
)
EntryClass = LlavaQwenForCausalLM
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment