Add Gemma-based image recognition scaffold
This commit is contained in:
@@ -69,6 +69,23 @@ def loadOcrConfig():
|
||||
}
|
||||
|
||||
|
||||
def loadVisionConfig():
|
||||
config = loadTomlFile(CONFIG_FILE)
|
||||
vision = config.get("vision", {})
|
||||
|
||||
return {
|
||||
"model_name": str(vision.get("model_name", "google/gemma-3-4b-it")).strip(),
|
||||
"prompt": str(
|
||||
vision.get(
|
||||
"prompt",
|
||||
"Describe the image briefly and extract any readable text.",
|
||||
)
|
||||
).strip(),
|
||||
"max_new_tokens": int(vision.get("max_new_tokens", 256)),
|
||||
"temperature": float(vision.get("temperature", 0.2)),
|
||||
}
|
||||
|
||||
|
||||
def loadVrcLogConfig():
|
||||
config = loadTomlFile(CONFIG_FILE)
|
||||
vrc_log = config.get("vrc_log", {})
|
||||
@@ -108,6 +125,14 @@ def loadVrcLogConfig():
|
||||
notice_section = event_config.get("notice", {})
|
||||
missing_count = notice_section.get("missing_count", 0)
|
||||
|
||||
log_patterns = (
|
||||
vrc_log.get("patterns")
|
||||
or vrc_log.get("log_patterns")
|
||||
or event_config.get("patterns")
|
||||
or event_config.get("log_patterns")
|
||||
or []
|
||||
)
|
||||
|
||||
if not missing_count:
|
||||
missing_count = event_config.get("missing_count", 0)
|
||||
|
||||
@@ -127,4 +152,5 @@ def loadVrcLogConfig():
|
||||
"staff_names": [str(name).strip() for name in staff_names if str(name).strip()],
|
||||
"missing_count": int(missing_count),
|
||||
"self_name": str(self_name).strip(),
|
||||
"log_patterns": [str(pattern).strip() for pattern in log_patterns if str(pattern).strip()],
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user