extraction 20261031_031805_P_museum_sys: 16 rows kept of 200

extractions/20261031_031805_P_museum_sys/summary.json

@@ -0,0 +1,72 @@
+{
+ "package": "0.1.14",
+ "model": "Qwen/Qwen3.5-9B",
+ "adapter": "AbstractPhil/qwen-symbolic-thinking-image-training",
+ "adapter_subfolder": "lora/Qwen3.5-9B/r32/20261031_022753_G_sys_2ep/final",
+ "system": "v1",
+ "manifest": "hub:AbstractPhil/qwen-thinking-images/manifest_v1.2.jsonl",
+ "split": "main",
+ "sources": [
+  "aic",
+  "nga",
+  "cleveland",
+  "met"
+ ],
+ "exclude": null,
+ "batch": 64,
+ "chunk": 512,
+ "max_side": 1024,
+ "max_new_tokens": 800,
+ "floors": {
+  "min_backed": 0.2,
+  "min_said": 0.1
+ },
+ "rows": 200,
+ "pictures_missing": 0,
+ "pictures_missing_by_source": {},
+ "oom": {},
+ "kept": 16,
+ "rejected": {
+  "not backed by the captions": 166,
+  "defines a word": 5,
+  "says too little of the captions": 13
+ },
+ "s_per_row": 0.385,
+ "tok_per_s": 387.4,
+ "by_source": {
+  "aic": {
+   "rows": 50,
+   "said": 0.074509843897599,
+   "backed": 0.017188810400877837
+  },
+  "cleveland": {
+   "rows": 50,
+   "said": 0.027285264248422144,
+   "backed": 0.03237416710508648
+  },
+  "met": {
+   "rows": 50,
+   "said": 0.06323773448773448,
+   "backed": 0.03629335940281257
+  },
+  "nga": {
+   "rows": 50,
+   "said": 0.0870788196328527,
+   "backed": 0.20872197961623187
+  }
+ },
+ "env": {
+  "package": "0.1.14",
+  "python": "3.13.16",
+  "torch": "2.11.0+cu130",
+  "transformers": "5.21.0",
+  "qwen3_5": true,
+  "cuda": true,
+  "gpu": "NVIDIA RTX PRO 6000 Blackwell Server Edition",
+  "gpu_gb": 95.0,
+  "hf_token": true
+ },
+ "library_notes": [
+  "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`."
+ ]
+}