Initial Commit

config.toml.example

@@ -0,0 +1,36 @@
+# Configuration for Florence-2 caption pipeline
+
+[model]
+# HuggingFace model id (replace with your model)
+id = "your-org/your-model"
+# Device to run model on (cuda/cpu)
+device = "cuda"
+# Prompt token used for captioning (replace with your token)
+prompt_token = "<PROMPT_TOKEN>"
+
+[preprocessing]
+# Max side for resizing (longest dimension)
+image_max_side = 768
+
+[pipeline]
+# Batch sizes and queue sizing
+gpu_batch_size = 8
+download_concurrency = 16
+image_queue_max_size = 64
+result_queue_max_size = 128
+db_write_batch_size = 64
+
+[database]
+# Async DB connection string
+dsn = ""
+# Table and column names
+table = "images"
+id_column = "id"
+url_column = "url"
+caption_column = "caption"
+# Default WHERE clause used to fetch rows to caption
+query_where = "caption IS NULL"
+
+[debug]
+# Dry run disables writing to DB for quick local tests
+dry_run = true