Skip to content

Allow fallback to loading from Auto"SubProcessor".from_pretrained when model_type can't be inferred from config #42402

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
yonigozlan merged 3 commits into from
Dec 2, 2025
Merged

Allow fallback to loading from Auto"SubProcessor".from_pretrained when model_type can't be inferred from config #42402

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
afc5313
fix raise error early
yonigozlan Nov 25, 2025
9379edf
add back feature extractor saving logic
yonigozlan Dec 2, 2025
295d136
Merge remote-tracking branch 'upstream/main' into fix-auto-processor-…
yonigozlan Dec 2, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
27 changes: 17 additions & 10 deletions src/transformers/models/auto/processing_auto.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -317,7 +317,6 @@ def from_pretrained(cls, pretrained_model_name_or_path, **kwargs):
processor_class = config_dict.get("processor_class", None)
if "AutoProcessor" in config_dict.get("auto_map", {}):
processor_auto_map = config_dict["auto_map"]["AutoProcessor"]

# Saved as feature extractor
if preprocessor_config_file is None:
preprocessor_config_file = cached_file(
Expand Down Expand Up @@ -345,16 +344,24 @@ def from_pretrained(cls, pretrained_model_name_or_path, **kwargs):
processor_auto_map = config_dict["auto_map"]["AutoProcessor"]

if processor_class is None:
# Otherwise, load config, if it can be loaded.
if not isinstance(config, PreTrainedConfig):
config = AutoConfig.from_pretrained(
pretrained_model_name_or_path, trust_remote_code=trust_remote_code, **kwargs
)
# Last resort: try loading the model config to get processor_class.
# This handles cases where processor info is only in config.json (not in any
# preprocessor/tokenizer config files). AutoConfig.from_pretrained may raise
# ValueError if the model_type is unrecognized or the config is invalid -
# we catch and ignore this to allow fallback to AutoTokenizer/AutoImageProcessor.
try:
if not isinstance(config, PreTrainedConfig):
config = AutoConfig.from_pretrained(
pretrained_model_name_or_path, trust_remote_code=trust_remote_code, **kwargs
)

# And check if the config contains the processor class.
processor_class = getattr(config, "processor_class", None)
if hasattr(config, "auto_map") and "AutoProcessor" in config.auto_map:
processor_auto_map = config.auto_map["AutoProcessor"]
processor_class = getattr(config, "processor_class", None)
if hasattr(config, "auto_map") and "AutoProcessor" in config.auto_map:
processor_auto_map = config.auto_map["AutoProcessor"]
except ValueError:
# Config loading failed (unrecognized model_type, invalid config, etc.)
# Continue to fallback logic below (AutoTokenizer, AutoImageProcessor, etc.)
pass

if processor_class is not None:
processor_class = processor_class_from_name(processor_class)
Expand Down