423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2026-05-28 02:48:54 +08:00
Code Issues Packages Projects Releases Wiki Activity

[fix] Fix MiniCPM-V-4.6 image preprocessing behavior (#10478)

Browse Source
This commit is contained in:
马境远
2026-05-12 11:35:23 +08:00
committed by GitHub
parent 53e77a9bfa
commit ca50f22c38
3 changed files with 30 additions and 135 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
src/llamafactory/chat/hf_engine.py
View File

@@ -205,9 +205,6 @@ class HuggingfaceEngine(BaseEngine):
gen_kwargs.pop("image_sizes", None)
if getattr(model.config, "model_type", None) == "minicpmv4_6":
gen_kwargs["downsample_mode"] = os.getenv("DOWNSAMPLE_MODE", "16x")
return gen_kwargs, prompt_length
@staticmethod

8
src/llamafactory/data/collator.py
View File

@@ -17,7 +17,6 @@
import copy
import inspect
import os
from dataclasses import dataclass
from typing import TYPE_CHECKING, Any, Literal, Optional
@@ -475,13 +474,6 @@ class MultiModalDataCollatorForSeq2Seq(DataCollatorForSeq2Seq):
features["position_ids"] = torch.arange(seq_length).long().repeat(bsz, 1)
return {"data": features, "input_ids": features["input_ids"], "labels": features["labels"]}
if (
self.model is not None
and getattr(self.model.config, "model_type", None) == "minicpmv4_6"
and "target_sizes" in features
): # for minicpmv4_6 with new transformers (NaViT API, no image_bound)
features["downsample_mode"] = os.getenv("DOWNSAMPLE_MODE", "16x")
return features

154
src/llamafactory/data/mm_plugin.py
View File

@@ -1209,23 +1209,6 @@ class LlavaNextVideoPlugin(BasePlugin):
@dataclass
class MiniCPMVPlugin(BasePlugin):
def _resolve_token_id(self, tokenizer: Any, attr_name: str, token_text: str | None = None) -> int | None:
token_id = getattr(tokenizer, attr_name, None)
if isinstance(token_id, int) and token_id >= 0:
return token_id
if token_text is None or not hasattr(tokenizer, "convert_tokens_to_ids"):
return None
converted_id = tokenizer.convert_tokens_to_ids(token_text)
if isinstance(converted_id, list):
converted_id = converted_id[0] if len(converted_id) else None
if isinstance(converted_id, int) and converted_id >= 0:
return converted_id
return None
@override
def _get_mm_inputs(
self,
@@ -1237,8 +1220,6 @@ class MiniCPMVPlugin(BasePlugin):
) -> dict[str, "torch.Tensor"]:
image_processor: BaseImageProcessor = getattr(processor, "image_processor")
mm_inputs = {}
preprocess_params = inspect.signature(image_processor.preprocess).parameters
downsample_mode = os.getenv("DOWNSAMPLE_MODE", "16x") if "downsample_mode" in preprocess_params else None
if len(images) != 0:
images = self._regularize_images(
images,
@@ -1255,15 +1236,9 @@ class MiniCPMVPlugin(BasePlugin):
images = new_images
image_processor_kwargs = {
"do_pad": True,
"max_slice_nums": image_processor.max_slice_nums,
"return_tensors": "pt",
}
if downsample_mode is not None:
image_processor_kwargs["downsample_mode"] = downsample_mode
image_inputs = image_processor(images, **image_processor_kwargs)
image_inputs = image_processor(
images, do_pad=True, max_slice_nums=image_processor.max_slice_nums, return_tensors="pt"
)
mm_inputs.update(image_inputs)
if len(videos) != 0:
@@ -1274,15 +1249,7 @@ class MiniCPMVPlugin(BasePlugin):
video_fps=getattr(processor, "video_fps", 2.0),
video_maxlen=getattr(processor, "video_maxlen", 128),
)["videos"]
video_processor_kwargs = {
"do_pad": True,
"max_slice_nums": 2,
"return_tensors": "pt",
}
if downsample_mode is not None:
video_processor_kwargs["downsample_mode"] = downsample_mode
video_inputs = image_processor(videos, **video_processor_kwargs)
video_inputs = image_processor(videos, do_pad=True, max_slice_nums=2, return_tensors="pt")
mm_inputs.update(video_inputs)
if len(audios) != 0:
@@ -1367,8 +1334,7 @@ class MiniCPMVPlugin(BasePlugin):
if self.expand_mm_tokens and mm_inputs:
pattern = "(<image>./</image>)"
image_sizes = mm_inputs.get("image_sizes")
image_grids = mm_inputs.get("grids")
image_sizes = mm_inputs["image_sizes"]
idx = 0
for index, message in enumerate(messages):
text = message["content"]
@@ -1376,21 +1342,13 @@ class MiniCPMVPlugin(BasePlugin):
text_chunks = text.split(pattern)
final_text = ""
for i in range(len(image_tags)):
grid = image_grids[0][idx] if image_grids and len(image_grids[0]) > idx else [1, 1]
image_size = image_sizes[0][idx] if image_sizes and len(image_sizes[0]) > idx else None
placeholder_fn = image_processor.get_slice_image_placeholder
if image_size is not None:
image_placeholder = placeholder_fn(
image_size,
image_idx=idx,
max_slice_nums=max_slice_nums,
use_image_id=use_image_id,
final_text = (
final_text
+ text_chunks[i]
+ image_processor.get_slice_image_placeholder(
image_sizes[0][idx], idx, max_slice_nums, use_image_id
)
else:
image_placeholder = placeholder_fn(grid)
final_text = final_text + text_chunks[i] + image_placeholder
)
idx += 1
final_text += text_chunks[-1]
@@ -1427,25 +1385,15 @@ class MiniCPMVPlugin(BasePlugin):
processor: Optional["MMProcessor"],
) -> dict[str, Union[list[int], "torch.Tensor"]]:
self._validate_input(processor, images, videos, audios)
tokenizer = processor.tokenizer
im_start_id = self._resolve_token_id(tokenizer, "im_start_id", "<image>")
slice_start_id = self._resolve_token_id(tokenizer, "slice_start_id", "<slice>")
im_end_id = self._resolve_token_id(tokenizer, "im_end_id", "</image>")
slice_end_id = self._resolve_token_id(tokenizer, "slice_end_id", "</slice>")
if None in (im_start_id, slice_start_id, im_end_id, slice_end_id):
raise AttributeError(
"Cannot resolve MiniCPM image boundary token ids from tokenizer. "
"Expected attributes (im_start_id/slice_start_id/im_end_id/slice_end_id) "
"or corresponding special tokens (<image>, <slice>, </image>, </slice>)."
)
# image bound
image_bounds_list = []
valid_image_nums_ls = []
for i, input_ids in enumerate(batch_ids):
input_ids_ = torch.tensor(input_ids)
start_cond = (input_ids_ == im_start_id) | (input_ids_ == slice_start_id)
end_cond = (input_ids_ == im_end_id) | (input_ids_ == slice_end_id)
start_cond = (input_ids_ == processor.tokenizer.im_start_id) | (
input_ids_ == processor.tokenizer.slice_start_id
)
end_cond = (input_ids_ == processor.tokenizer.im_end_id) | (input_ids_ == processor.tokenizer.slice_end_id)
image_start_tokens = torch.where(start_cond)[0]
image_start_tokens += 1
image_end_tokens = torch.where(end_cond)[0]
@@ -1466,16 +1414,6 @@ class MiniCPMVPlugin(BasePlugin):
mm_inputs.update({"image_bound": image_bounds_list})
if len(audios) > 0:
audio_start_id = self._resolve_token_id(tokenizer, "audio_start_id", "<audio>")
audio_end_id = self._resolve_token_id(tokenizer, "audio_end_id", "</audio>")
spk_start_id = self._resolve_token_id(tokenizer, "spk_start_id", "<spk>")
spk_end_id = self._resolve_token_id(tokenizer, "spk_end_id", "</spk>")
if None in (audio_start_id, audio_end_id, spk_start_id, spk_end_id):
raise AttributeError(
"Cannot resolve MiniCPM audio/speaker boundary token ids from tokenizer. "
"Expected *_id attributes or corresponding special tokens."
)
# audio bound
audio_bounds_ls = []
spk_bounds_ls = []
@@ -1483,15 +1421,15 @@ class MiniCPMVPlugin(BasePlugin):
for input_ids, audiolen in zip(batch_ids, audlens):
input_ids_ = torch.tensor(input_ids)
audio_start_idx = torch.where(input_ids_ == audio_start_id)[0]
audio_end_idx = torch.where(input_ids_ == audio_end_id)[0]
audio_start_idx = torch.where(input_ids_ == processor.tokenizer.audio_start_id)[0]
audio_end_idx = torch.where(input_ids_ == processor.tokenizer.audio_end_id)[0]
assert len(audio_start_idx) == len(audio_end_idx)
audio_bounds = torch.hstack([(audio_start_idx + 1).unsqueeze(-1), audio_end_idx.unsqueeze(-1)])
audio_bounds_ls.append(audio_bounds)
valid_audio_nums_ls.append(audiolen)
spk_start_idx = torch.where(input_ids_ == spk_start_id)[0]
spk_end_idx = torch.where(input_ids_ == spk_end_id)[0]
spk_start_idx = torch.where(input_ids_ == processor.tokenizer.spk_start_id)[0]
spk_end_idx = torch.where(input_ids_ == processor.tokenizer.spk_end_id)[0]
assert len(spk_start_idx) == len(spk_end_idx)
spk_bounds = torch.hstack([(spk_start_idx + 1).unsqueeze(-1), spk_end_idx.unsqueeze(-1)])
spk_bounds_ls.append(spk_bounds)
@@ -1500,8 +1438,6 @@ class MiniCPMVPlugin(BasePlugin):
mm_inputs.update(audio_inputs)
mm_inputs.update({"audio_bounds": audio_bounds_ls, "spk_bounds": spk_bounds_ls})
return mm_inputs
@dataclass
class MiniCPMV4_6Plugin(BasePlugin):
@@ -1518,59 +1454,23 @@ class MiniCPMV4_6Plugin(BasePlugin):
image_processor = getattr(processor, "image_processor")
video_processor = getattr(processor, "video_processor", None)
mm_inputs = {}
preprocess_params = inspect.signature(image_processor.preprocess).parameters
downsample_mode = os.getenv("DOWNSAMPLE_MODE", "16x") if "downsample_mode" in preprocess_params else None
if len(images) != 0:
images = self._regularize_images(
images,
image_max_pixels=getattr(processor, "image_max_pixels", 768 * 768),
image_min_pixels=getattr(processor, "image_min_pixels", 32 * 32),
)["images"]
image_processor_kwargs = {
"max_slice_nums": getattr(image_processor, "max_slice_nums", 9),
"return_tensors": "pt",
}
if downsample_mode is not None:
image_processor_kwargs["downsample_mode"] = downsample_mode
image_inputs = image_processor(images, **image_processor_kwargs)
mm_inputs.update(image_inputs)
# The image_processor ignores downsample_mode; target_sizes are always based on patch_size.
# downsample_mode only affects the token divisor in _build_v4_6_placeholder and model forward.
mm_inputs.update(image_processor(images, return_tensors="pt"))
if len(videos) != 0:
videos = self._regularize_videos(
videos,
image_max_pixels=getattr(processor, "video_max_pixels", 256 * 256),
image_min_pixels=getattr(processor, "video_min_pixels", 16 * 16),
video_fps=getattr(processor, "video_fps", 2.0),
video_maxlen=getattr(processor, "video_maxlen", 128),
)["videos"]
if video_processor is not None:
video_processor_kwargs = {
"max_slice_nums": 2,
"return_tensors": "pt",
}
if downsample_mode is not None:
video_processor_kwargs["downsample_mode"] = downsample_mode
video_inputs = video_processor(videos, **video_processor_kwargs)
video_inputs = video_processor(videos, return_tensors="pt")
mm_inputs["pixel_values_videos"] = video_inputs["pixel_values_videos"]
mm_inputs["target_sizes_videos"] = video_inputs["target_sizes_videos"]
else:
# Fallback to image processor for video
video_processor_kwargs = {
"max_slice_nums": 2,
"return_tensors": "pt",
}
if downsample_mode is not None:
video_processor_kwargs["downsample_mode"] = downsample_mode
video_inputs = image_processor(videos, **video_processor_kwargs)
video_inputs = image_processor(videos, return_tensors="pt")
mm_inputs["pixel_values_videos"] = video_inputs["pixel_values"]
mm_inputs["target_sizes_videos"] = video_inputs["target_sizes"]
if len(audios) != 0:
audios = self._regularize_audios(
audios,
sampling_rate=getattr(processor, "audio_sampling_rate", 16000),
)["audios"]
audio_features, audio_feature_lens, audio_phs = processor.audio_feature_extract(
[audios],
chunk_input=True,
@@ -1745,6 +1645,12 @@ class MiniCPMV4_6Plugin(BasePlugin):
if "pixel_values" not in mm_inputs:
mm_inputs["pixel_values"] = torch.empty(1, 3, 14, 0)
# Pass downsample_mode to model forward so it matches the placeholder divisor
_ds = os.getenv("DOWNSAMPLE_MODE")
if _ds is None:
_ds = getattr(getattr(processor, "image_processor", None), "downsample_mode", "16x")
mm_inputs["downsample_mode"] = _ds
if len(audios) > 0:
audio_inputs = self._get_mm_inputs([], [], audios, processor)
mm_inputs.update(audio_inputs)