Update tokenizer, bump hf versions

all_results.json

@@ -1,14 +1,14 @@
 {
-    "epoch": 43.0,
-    "perplexity": 1.1020061050262677,
-    "test_accuracy": 0.5884773413073098,
-    "test_loss": 0.09713225066661835,
-    "test_runtime": 9.4021,
-    "test_samples_per_second": 248.562,
-    "test_steps_per_second": 15.635,
-    "train_loss": 0.2331255912600619,
-    "train_runtime": 2746.0841,
+    "epoch": 31.0,
+    "test_accuracy": 0.5044041304365378,
+    "test_loss": 0.3887753188610077,
+    "test_perplexity": 1.4751730705300097,
+    "test_runtime": 1.9728,
+    "test_samples_per_second": 1184.637,
+    "test_steps_per_second": 74.515,
+    "train_loss": 0.7912082670869731,
+    "train_runtime": 664.7096,
     "train_samples": 7878,
-    "train_samples_per_second": 143.441,
-    "train_steps_per_second": 4.479
+    "train_samples_per_second": 592.59,
+    "train_steps_per_second": 18.504
 }

config.json

@@ -26,7 +26,7 @@
   "summary_type": "cls_index",
   "summary_use_proj": true,
   "torch_dtype": "float32",
-  "transformers_version": "4.27.1",
+  "transformers_version": "4.34.1",
   "use_cache": true,
-  "vocab_size": 52
+  "vocab_size": 952
 }

generation_config.json

@@ -1,10 +1,9 @@
 {
   "do_sample": true,
   "eos_token_id": 1,
-  "length_penalty": 0.0,
   "max_new_tokens": 896,
   "min_new_tokens": 1,
   "num_return_sequences": 100,
   "pad_token_id": 2,
-  "transformers_version": "4.27.1"
+  "transformers_version": "4.34.1"
 }

model.safetensors

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:2be9cd6ec7f88c64b3078753df5fabf780bde590630a30c1a9b13564d21bb179
-size 22231672

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:731f207586a9dda4728f79a89e66fbbe0cc2b23b3921ec47fe995fc5aeb2c63b
-size 22270333
+oid sha256:8fa4ccb5ccf594dc95a6b5c77d5389734b1a6cf87ba8a3ea5645f4fbad7d2560
+size 13147101

special_tokens_map.json

@@ -1,6 +1,30 @@
 {
-  "bos_token": "^",
-  "eos_token": "_",
-  "pad_token": " ",
-  "unk_token": "§"
+  "bos_token": {
+    "content": "^",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "_",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": " ",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "§",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
 }

test_results.json

@@ -1,9 +1,9 @@
 {
-    "epoch": 43.0,
-    "perplexity": 1.1020061050262677,
-    "test_accuracy": 0.5884773413073098,
-    "test_loss": 0.09713225066661835,
-    "test_runtime": 9.4021,
-    "test_samples_per_second": 248.562,
-    "test_steps_per_second": 15.635
+    "epoch": 31.0,
+    "test_accuracy": 0.5044041304365378,
+    "test_loss": 0.3887753188610077,
+    "test_perplexity": 1.4751730705300097,
+    "test_runtime": 1.9728,
+    "test_samples_per_second": 1184.637,
+    "test_steps_per_second": 74.515
 }

tokenizer.json

@@ -1,7 +1,14 @@
 {
   "version": "1.0",
   "truncation": null,
-  "padding": null,
+  "padding": {
+    "strategy": "BatchLongest",
+    "direction": "Right",
+    "pad_to_multiple_of": null,
+    "pad_id": 2,
+    "pad_type_id": 0,
+    "pad_token": " "
+  },
   "added_tokens": [
     {
       "id": 0,
@@ -53,7 +60,7 @@
   "pre_tokenizer": {
     "type": "Split",
     "pattern": {
-      "Regex": ""
+      "Regex": "\\[[^\\]]+]|A[cglmrstu]?|B[aeihkr]?|C[adeflmorsu]?|D[bsy]?|E[rsu]?|F[emr]?|G[ade]?|H[efgos]?|I[nr]?|K[r]?|L[aru]?|M[dgnot]?|N[abdeiop]?|Os?|P[abdmortu]?|R[aefghnu]?|S[bcdegimnr]?|T[abcehilm]?|U|V|W|Xe|Yb?|Z[nr]?|as|b|c|n|o|p|se?|\\(|\\)|@@?|\\+\\d+|\\++|-\\d+|-+|\\.|-|=|#\\d{1}|\\$\\(?|:[0-9]*|~|@|\\/\\??|\\\\\\??|\\%[0-9]{2}|[0-9]|\\*|>>?|D\\d{1}|H\\d{1}|h\\d{1}|R\\d{1}|r\\d{1}|v\\d{1}|X\\d{1}|x\\d{1}|#\\d{1}|!|&|,|;"
     },
     "behavior": "Isolated",
     "invert": false
@@ -124,53 +131,953 @@
       " ": 2,
       "§": 3,
       "°": 4,
-      ":": 5,
-      ";": 6,
-      "[": 7,
-      "]": 8,
-      "0": 9,
-      "-": 10,
-      "1": 11,
-      "H": 12,
-      "D": 13,
-      "C": 14,
-      "2": 15,
-      "c": 16,
-      "3": 17,
-      "+": 18,
-      "7": 19,
-      "(": 20,
-      ")": 21,
-      "#": 22,
-      "O": 23,
-      "4": 24,
-      "a": 25,
-      ">": 26,
-      "=": 27,
-      "6": 28,
-      "5": 29,
-      "8": 30,
-      "N": 31,
-      ".": 32,
-      "9": 33,
-      "B": 34,
-      "l": 35,
-      "n": 36,
-      "r": 37,
-      "S": 38,
-      "F": 39,
-      "/": 40,
-      "@": 41,
-      "I": 42,
-      "\\": 43,
-      "P": 44,
-      "M": 45,
-      "g": 46,
-      "s": 47,
-      "Z": 48,
-      "i": 49,
-      "o": 50,
-      "u": 51
+      "-": 5,
+      ":": 6,
+      "(": 7,
+      ")": 8,
+      "=": 9,
+      ">>": 10,
+      ".": 11,
+      "C": 12,
+      "[c:3]": 13,
+      "O": 14,
+      "[c:2]": 15,
+      "[c;H0;D3;+0:1]": 16,
+      "[C:2]": 17,
+      "1": 18,
+      "[c:5]": 19,
+      "[C;H0;D3;+0:1]": 20,
+      "[#7;a:4]": 21,
+      "[c:6]": 22,
+      ":1": 23,
+      "[C:4]": 24,
+      "[CH2;D2;+0:1]": 25,
+      "c": 26,
+      "[c:4]": 27,
+      "[C:3]": 28,
+      "[C:5]": 29,
+      "[C:6]": 30,
+      "[c:7]": 31,
+      "[#7;a:5]": 32,
+      "[O;D1;H0:2]": 33,
+      "Cl": 34,
+      "[C:7]": 35,
+      "[c:8]": 36,
+      "[#7;a:3]": 37,
+      "[O;D1;H0:3]": 38,
+      "Br": 39,
+      "[#7;a:2]": 40,
+      "[#7;a:6]": 41,
+      "[C:8]": 42,
+      "[c:9]": 43,
+      "[#7;a:7]": 44,
+      "[O;D1;H0:5]": 45,
+      "[O;D1;H0:4]": 46,
+      "/": 47,
+      "[O;D1;H0:6]": 48,
+      "-1": 49,
+      "[C;D1;H3:2]": 50,
+      "[C:9]": 51,
+      "#": 52,
+      "[#7;a:8]": 53,
+      "[CH;D2;+0:1]": 54,
+      "[O;D1;H0:7]": 55,
+      "[CH;D3;+0:1]": 56,
+      "B": 57,
+      "[c;H0;D3;+0:4]": 58,
+      "[O;D1;H0:8]": 59,
+      "[c:10]": 60,
+      "[NH;D2;+0:7]": 61,
+      "I": 62,
+      "[NH;D2;+0:6]": 63,
+      "[#7:4]": 64,
+      "[#8:3]": 65,
+      "[C;D1;H3:6]": 66,
+      "[#7;a:9]": 67,
+      "[#8:4]": 68,
+      "[c;H0;D3;+0:6]": 69,
+      "F": 70,
+      "[NH;D2;+0:5]": 71,
+      "[C;D1;H3:5]": 72,
+      "[NH;D2;+0:8]": 73,
+      "[O;D1;H0:9]": 74,
+      "[C:10]": 75,
+      "[#16;a:4]": 76,
+      "[C;D1;H3:3]": 77,
+      "[C:1]": 78,
+      "[#7;a:1]": 79,
+      "[c;H0;D3;+0:5]": 80,
+      "[#7:5]": 81,
+      "[NH2;D1;+0:7]": 82,
+      "[C;D1;H3:4]": 83,
+      "[C;D1;H3:7]": 84,
+      "[c:11]": 85,
+      "[#7:3]": 86,
+      "[S;H0;D4;+0:1]": 87,
+      "[#8:5]": 88,
+      "[c;H0;D3;+0:7]": 89,
+      "[NH2;D1;+0:6]": 90,
+      "[C;H0;D3;+0:4]": 91,
+      "[#8:2]": 92,
+      "[#7;a:10]": 93,
+      "[N;H0;D3;+0:6]": 94,
+      "[N;H0;D3;+0:7]": 95,
+      "[NH2;D1;+0:8]": 96,
+      "[#7:2]": 97,
+      "[#16;a:2]": 98,
+      "[#8;a:4]": 99,
+      "[NH2;D1;+0:5]": 100,
+      "[#16;a:5]": 101,
+      "[NH;D2;+0:9]": 102,
+      "[N;H0;D3;+0:5]": 103,
+      "[c;H0;D3;+0:3]": 104,
+      "[CH3;D1;+0:1]": 105,
+      "[NH;D2;+0:4]": 106,
+      "[#7:1]": 107,
+      "[#8:1]": 108,
+      "[C;H0;D3;+0:5]": 109,
+      "[#8:7]": 110,
+      "[CH2;D2;+0:4]": 111,
+      "[CH;D2;+0:4]": 112,
+      "2": 113,
+      "[OH;D1;+0:5]": 114,
+      "[C;D1;H3:8]": 115,
+      "[#8:6]": 116,
+      "[O;D1;H0:10]": 117,
+      "[OH;D1;+0:6]": 118,
+      "[#7:6]": 119,
+      "[N;H0;D3;+0:8]": 120,
+      "[OH;D1;+0:7]": 121,
+      "\\": 122,
+      "[CH;D3;+0:4]": 123,
+      "[C;H0;D3;+0:3]": 124,
+      "[NH2;D1;+0:4]": 125,
+      "[#16;a:3]": 126,
+      "[#8;a:5]": 127,
+      "[CH2;D2;+0:5]": 128,
+      ":2": 129,
+      "[NH2;D1;+0:9]": 130,
+      "S": 131,
+      "[O;H0;D2;+0:7]": 132,
+      "[#8:8]": 133,
+      "[c:12]": 134,
+      "[C:11]": 135,
+      "[C;H0;D3;+0:2]": 136,
+      "[c;H0;D3;+0:8]": 137,
+      "[c;H0;D3;+0:2]": 138,
+      "[CH;D2;+0:3]": 139,
+      "[#7:7]": 140,
+      "[C;H0;D3;+0:6]": 141,
+      "[CH2;D2;+0:3]": 142,
+      "[O;H0;D2;+0:6]": 143,
+      "[Sn]": 144,
+      "[C;H0;D4;+0:1]": 145,
+      "[NH;D2;+0:10]": 146,
+      "[#16;a:1]": 147,
+      "[CH;D2;+0:5]": 148,
+      "[CH;D3;+0:3]": 149,
+      "[OH;D1;+0:1]": 150,
+      "[C;D1;H3:9]": 151,
+      "[NH;D2;+0:1]": 152,
+      "[OH;D1;+0:4]": 153,
+      "[#7;a:11]": 154,
+      "N": 155,
+      "[#8;a:3]": 156,
+      "[n;H0;D3;+0:7]": 157,
+      "[CH2;D2;+0:6]": 158,
+      "[CH;D3;+0:5]": 159,
+      "[nH;D2;+0:7]": 160,
+      "[O;D1;H0:11]": 161,
+      "[Mg+]": 162,
+      "[#16:4]": 163,
+      "[CH;D3;+0:2]": 164,
+      "[NH2;D1;+0:10]": 165,
+      "[O;H0;D2;+0:5]": 166,
+      "[CH2;D2;+0:2]": 167,
+      "[CH2;D2;+0:7]": 168,
+      "[O;D1;H1:6]": 169,
+      "[#8;a:6]": 170,
+      "[F;D1;H0:5]": 171,
+      "[O;H0;D1;+0:5]": 172,
+      "[N;H0;D3;+0:4]": 173,
+      "[Br;H0;D1;+0:1]": 174,
+      "[c;H0;D3;+0:9]": 175,
+      "[C;D1;H3:1]": 176,
+      "[F;D1;H0:6]": 177,
+      "[O;H0;D2;+0:8]": 178,
+      "[#16;a:6]": 179,
+      "[nH;D2;+0:6]": 180,
+      "[C;H0;D3;+0:7]": 181,
+      "[CH;D2;+0:6]": 182,
+      "[#16;a:7]": 183,
+      "[OH;D1;+0:8]": 184,
+      "[C;D1;H3:10]": 185,
+      "[N;H0;D3;+0:9]": 186,
+      "[C;D1;H2:3]": 187,
+      "[O;H0;D2;+0:4]": 188,
+      "[O;H0;D2;+0:1]": 189,
+      "[n;H0;D3;+0:6]": 190,
+      "[#7:8]": 191,
+      "[cH;D2;+0:1]": 192,
+      "[nH;D2;+0:5]": 193,
+      "[C;D1;H2:4]": 194,
+      "[C@@H;D3;+0:1]": 195,
+      "[NH2;D1;+0:1]": 196,
+      "[C@H;D3;+0:1]": 197,
+      "[C;D1;H3:11]": 198,
+      "[F;D1;H0:4]": 199,
+      "[N;H0;D2;+0:5]": 200,
+      "[n;H0;D3;+0:5]": 201,
+      "[NH2;D1;+0:11]": 202,
+      "[NH;D2;+0:11]": 203,
+      "[CH;D2;+0:2]": 204,
+      "[N;H0;D2;+0:6]": 205,
+      "[n;H0;D3;+0:8]": 206,
+      "[NH;D2;+0:3]": 207,
+      "[O;H0;D1;+0:6]": 208,
+      "[#8:9]": 209,
+      "[C:12]": 210,
+      "[CH;D3;+0:6]": 211,
+      "[c:13]": 212,
+      "[nH;D2;+0:8]": 213,
+      "[N;H0;D3;+0:1]": 214,
+      "[cH;D2;+0:3]": 215,
+      "[O;H0;D1;+0:4]": 216,
+      "[C;H0;D2;+0:5]": 217,
+      "[#16:3]": 218,
+      "[#8;a:7]": 219,
+      "[C;H0;D2;+0:4]": 220,
+      "[O;D1;H0:12]": 221,
+      "[NH;D2;+0:2]": 222,
+      "[N;H0;D2;+0:4]": 223,
+      "[O;D1;H1:5]": 224,
+      "[CH;D2;+0:7]": 225,
+      "[P+]": 226,
+      "[OH;D1;+0:3]": 227,
+      "[C;H0;D2;+0:7]": 228,
+      "[O;D1;H1:9]": 229,
+      "P": 230,
+      "[c;H0;D3;+0:10]": 231,
+      "[CH2;D1;+0:1]": 232,
+      "[C;D1;H3:12]": 233,
+      "[C;H0;D3;+0:8]": 234,
+      "[CH;D2;+0:8]": 235,
+      "[cH;D2;+0:4]": 236,
+      "[cH;D2;+0:6]": 237,
+      "[#16:5]": 238,
+      "[#7:9]": 239,
+      "[#7;a:12]": 240,
+      "[C;H0;D2;+0:6]": 241,
+      "[O;H0;D1;+0:2]": 242,
+      "[C;H0;D2;+0:1]": 243,
+      "[CH2;D2;+0:8]": 244,
+      "[N;H0;D2;+0:7]": 245,
+      "[O;H0;D1;+0:1]": 246,
+      "[nH;D2;+0:1]": 247,
+      "[#8;a:2]": 248,
+      "[CH;D3;+0:7]": 249,
+      "[N;D1;H0:3]": 250,
+      "[N;H0;D3;+0:10]": 251,
+      "[O;D1;H0:1]": 252,
+      "[O;H0;D2;+0:9]": 253,
+      "[OH;D1;+0:2]": 254,
+      "[OH;D1;+0:9]": 255,
+      "[cH;D2;+0:5]": 256,
+      "[O-]": 257,
+      "[F;D1;H0:3]": 258,
+      "[O;D1;H1:8]": 259,
+      "[cH;D2;+0:7]": 260,
+      "[#16:2]": 261,
+      "[#8:10]": 262,
+      "[O;D1;H1:7]": 263,
+      "[S;H0;D2;+0:5]": 264,
+      "[S;H0;D2;+0:6]": 265,
+      "[C;H0;D2;+0:3]": 266,
+      "[N;D1;H2:4]": 267,
+      "[N;D1;H2:6]": 268,
+      "[O;H0;D1;+0:7]": 269,
+      "[n;H0;D2;+0:7]": 270,
+      "[C:13]": 271,
+      "[F;D1;H0:7]": 272,
+      "[nH;D2;+0:9]": 273,
+      "[#8;a:8]": 274,
+      "[C;H0;D2;+0:2]": 275,
+      "[c:1]": 276,
+      "[O;H0;D1;+0:3]": 277,
+      "[n;H0;D3;+0:9]": 278,
+      "[#16:1]": 279,
+      "[C;D1;H2:2]": 280,
+      "[C;D1;H2:5]": 281,
+      "[C;H0;D2;+0:8]": 282,
+      "[NH2;D1;+0:3]": 283,
+      "[O;H0;D2;+0:3]": 284,
+      "[#8;a:1]": 285,
+      "[C;D1;H1:3]": 286,
+      "[N;D1;H2:3]": 287,
+      "[C;H0;D3;+0:9]": 288,
+      "[C@@H;D3;+0:3]": 289,
+      "[CH3;D1;+0:5]": 290,
+      "[N;D1;H2:5]": 291,
+      "[S;H0;D2;+0:7]": 292,
+      "[c;H0;D3;+0:11]": 293,
+      "[n;H0;D2;+0:8]": 294,
+      "[n;H0;D2;+0:6]": 295,
+      "n": 296,
+      "[#7:10]": 297,
+      "[N;D1;H0:5]": 298,
+      "[N;D1;H2:7]": 299,
+      "[SH;D1;+0:7]": 300,
+      "[Si]": 301,
+      "[C@H;D3;+0:2]": 302,
+      "[C@H;D3;+0:3]": 303,
+      "[N;H0;D2;+0:8]": 304,
+      "[N;H0;D3;+0:3]": 305,
+      "[Zn+]": 306,
+      "[#8:11]": 307,
+      "[#8;a:9]": 308,
+      "[C@@H;D3;+0:2]": 309,
+      "[Cl;D1;H0:2]": 310,
+      "[N;D1;H0:4]": 311,
+      "[N;D1;H0:6]": 312,
+      "[NH2;D1;+0:12]": 313,
+      "[O;D1;H0:13]": 314,
+      "[SH;D1;+0:6]": 315,
+      "[N;H0;D2;+0:2]": 316,
+      "[n;H0;D2;+0:9]": 317,
+      "-2": 318,
+      "[#16;a:9]": 319,
+      "[#7;+:5]": 320,
+      "[C;D1;H2:7]": 321,
+      "[CH3;D1;+0:4]": 322,
+      "[S;H0;D2;+0:4]": 323,
+      "[c:14]": 324,
+      "[cH;D2;+0:2]": 325,
+      "[nH;D2;+0:4]": 326,
+      "[C;H0;D4;+0:2]": 327,
+      "[C;H0;D4;+0:5]": 328,
+      "[n;H0;D3;+0:10]": 329,
+      "[#16:6]": 330,
+      "[C;H0;D3;+0:10]": 331,
+      "[F;D1;H0:2]": 332,
+      "[NH2;D1;+0:2]": 333,
+      "[O;D1;H1:2]": 334,
+      "[O;D1;H1:4]": 335,
+      "[SH;D1;+0:5]": 336,
+      "[c:15]": 337,
+      "[C@@H;D3;+0:4]": 338,
+      "[CH;D3;+0:8]": 339,
+      "[n;H0;D2;+0:5]": 340,
+      "[n;H0;D3;+0:1]": 341,
+      "[C;D1;H1:5]": 342,
+      "[C;D1;H2:8]": 343,
+      "[C@@H;D3;+0:5]": 344,
+      "[N;-;D1;H0:6]": 345,
+      "[N;D1;H2:8]": 346,
+      "[N;H0;D3;+0:11]": 347,
+      "[S;D1;H0:6]": 348,
+      "[S;H0;D2;+0:8]": 349,
+      "[cH;D2;+0:8]": 350,
+      "[C;H0;D4;+0:4]": 351,
+      "[CH2;D1;+0:5]": 352,
+      "[CH3;D1;+0:6]": 353,
+      "[n;H0;D2;+0:11]": 354,
+      "[#16;a:8]": 355,
+      "[#7:11]": 356,
+      "[CH2;D2;+0:9]": 357,
+      "[CH;D2;+0:9]": 358,
+      "[N;D1;H0:7]": 359,
+      "[N;H0;D2;+0:3]": 360,
+      "[n;H0;D2;+0:3]": 361,
+      "[n;H0;D2;+0:4]": 362,
+      "[n;H0;D3;+0:4]": 363,
+      "[n;H0;D2;+0:10]": 364,
+      "[C;D1;H3:13]": 365,
+      "[C;H0;D4;+0:3]": 366,
+      "[C@H;D3;+0:4]": 367,
+      "[F;D1;H0:8]": 368,
+      "[N;D1;H0:8]": 369,
+      "[O;D1;H1:3]": 370,
+      "[SH;D1;+0:4]": 371,
+      "[CH2;D2;+0:10]": 372,
+      "[#7;+:7]": 373,
+      "[C;D1;H1:4]": 374,
+      "[C@H;D3;+0:5]": 375,
+      "[CH3;D1;+0:7]": 376,
+      "[CH;D1;+0:7]": 377,
+      "[CH;D3;+0:9]": 378,
+      "[N;-;D1;H0:8]": 379,
+      "[NH;D2;+0:12]": 380,
+      "[O;H0;D2;+0:10]": 381,
+      "[SH;D1;+0:8]": 382,
+      "[OH;D1;+0:10]": 383,
+      "[#7;a:13]": 384,
+      "[#8:12]": 385,
+      "[C;D1;H2:6]": 386,
+      "[C;H0;D4;+0:6]": 387,
+      "[N;H0;D3;+0:2]": 388,
+      "[O;D1;H1:10]": 389,
+      "[O;H0;D1;+0:8]": 390,
+      "[S;H0;D2;+0:3]": 391,
+      "[Sn;H0;D4;+0:1]": 392,
+      "[nH;D2;+0:10]": 393,
+      "[C@@H;D3;+0:6]": 394,
+      "[c;H0;D3;+0:12]": 395,
+      "[C;H0;D2;+0:9]": 396,
+      "[CH;D2;+0:10]": 397,
+      "[F;D1;H0:9]": 398,
+      "[N-;H0;D1:6]": 399,
+      "[S;D1;H0:4]": 400,
+      "[n;H0;D2;+0:2]": 401,
+      "[n;H0;D3;+0:11]": 402,
+      "[nH;D2;+0:11]": 403,
+      "[nH;D2;+0:3]": 404,
+      "[CH2;D1;+0:4]": 405,
+      "[Cu]": 406,
+      "[N-;H0;D1:4]": 407,
+      "[NH3;D0;+0:6]": 408,
+      "[O-;H0;D1:6]": 409,
+      "[O;H0;D2;+0:2]": 410,
+      "[S;H0;D1;+0:9]": 411,
+      "[cH;D2;+0:9]": 412,
+      "[s;H0;D2;+0:9]": 413,
+      "[C:15]": 414,
+      "[C;D1;H1:1]": 415,
+      "[C;D1;H2:1]": 416,
+      "[C@H;D3;+0:6]": 417,
+      "[NH3;D0;+0:5]": 418,
+      "[NH3;D0;+0:7]": 419,
+      "[O;D1;H0:14]": 420,
+      "[S;D1;H0:3]": 421,
+      "[O-;H0;D1:7]": 422,
+      "[S;H0;D3;+0:3]": 423,
+      "[#16;a:10]": 424,
+      "[#7:12]": 425,
+      "[#7;+:6]": 426,
+      "[#7;a:14]": 427,
+      "[#8;a:10]": 428,
+      "[CH;D1;+0:5]": 429,
+      "[CH;D1;+0:8]": 430,
+      "[Cl;H0;D1;+0:1]": 431,
+      "[N;-;D1;H0:7]": 432,
+      "[N;D1;H0:2]": 433,
+      "[N;D1;H2:9]": 434,
+      "[NH2;D1;+0:13]": 435,
+      "[C@H;D3;+0:7]": 436,
+      "[N;H0;D2;+0:1]": 437,
+      "[O-;H0;D1:5]": 438,
+      "[c;H0;D3;+0:13]": 439,
+      "[n;H0;D2;+0:12]": 440,
+      "[#15:2]": 441,
+      "[#16:8]": 442,
+      "[C:14]": 443,
+      "[C;D1;H1:7]": 444,
+      "[C;D1;H3:14]": 445,
+      "[N;H0;D1;+0:2]": 446,
+      "[O-;H0;D1:8]": 447,
+      "[S;D1;H0:8]": 448,
+      "[C;H0;D2;+0:10]": 449,
+      "[CH2;D1;+0:6]": 450,
+      "[CH2;D1;+0:8]": 451,
+      "[CH3;D1;+0:3]": 452,
+      "[I;H0;D1;+0:1]": 453,
+      "[NH3;D0;+0:4]": 454,
+      "[S;H0;D1;+0:10]": 455,
+      "[S;H0;D1;+0:12]": 456,
+      "[S;H0;D3;+0:5]": 457,
+      "[cH;D2;+0:10]": 458,
+      "[n;H0;D3;+0:3]": 459,
+      "[s;H0;D2;+0:10]": 460,
+      "[s;H0;D2;+0:12]": 461,
+      "3": 462,
+      "[#7;+:1]": 463,
+      "[#7;a;+:4]": 464,
+      "[Br;D1;H0:2]": 465,
+      "[C-;H0;D1:4]": 466,
+      "[C:16]": 467,
+      "[C;D1;H3:15]": 468,
+      "[C;D1;H3:16]": 469,
+      "[C@@H;D3;+0:7]": 470,
+      "[N+;H0;D2:2]": 471,
+      "[N+;H0;D3:1]": 472,
+      "[N+]": 473,
+      "[N-;H0;D1:3]": 474,
+      "[N;D1;H0:9]": 475,
+      "[N;D1;H1:4]": 476,
+      "[N;D1;H2:1]": 477,
+      "[N;H0;D1;+0:4]": 478,
+      "[N;H0;D3;+0:12]": 479,
+      "[NH;D2;+0:13]": 480,
+      "[O;D1;H1:11]": 481,
+      "[O;H0;D2;+0:11]": 482,
+      "[S;D1;H0:7]": 483,
+      "[S;H0;D2;+0:9]": 484,
+      "[SH;D1;+0:9]": 485,
+      "[cH;D2;+0:11]": 486,
+      "[n;H0;D3;+0:2]": 487,
+      "[N+;H0;D4:4]": 488,
+      "[NH;D2;+0:14]": 489,
+      "[O;H0;D1;+0:9]": 490,
+      "[P;H0;D4;+0:2]": 491,
+      "[S;H0;D2;+0:10]": 492,
+      "[SH;D1;+0:10]": 493,
+      "[#16:7]": 494,
+      "[#16:9]": 495,
+      "[#7;a;+:2]": 496,
+      "[#8:13]": 497,
+      "[Br;D1;H0:3]": 498,
+      "[C;D1;H3:17]": 499,
+      "[C@H;D3;+0:8]": 500,
+      "[C@H;D3;+0:9]": 501,
+      "[CH2;D1;+0:3]": 502,
+      "[CH2;D1;+0:7]": 503,
+      "[CH;D1;+0:2]": 504,
+      "[CH;D2;+0:11]": 505,
+      "[N-;H0;D1:5]": 506,
+      "[N-]": 507,
+      "[N;D1;H2:2]": 508,
+      "[N;H0;D2;+0:9]": 509,
+      "[O-;H0;D1:1]": 510,
+      "[OH;D1;+0:11]": 511,
+      "[P;H0;D3;+0:2]": 512,
+      "[S+]": 513,
+      "[S-;H0;D1:6]": 514,
+      "[S;D1;H0:2]": 515,
+      "[S;H0;D1;+0:11]": 516,
+      "[S;H0;D3;+0:4]": 517,
+      "[c;H0;D3;+0:14]": 518,
+      "[s;H0;D2;+0:11]": 519,
+      "[C-;H0;D1:6]": 520,
+      "[C;H0;D4;+0:7]": 521,
+      "[NH4+;D0:4]": 522,
+      "[NH4+;D0:5]": 523,
+      "[S-;H0;D1:5]": 524,
+      "[Sn;H0;D4;+0:5]": 525,
+      "[n;H0;D3;+0:12]": 526,
+      "[#14:2]": 527,
+      "[#16;a:11]": 528,
+      "[#5:2]": 529,
+      "[#7:13]": 530,
+      "[#7;+:8]": 531,
+      "[B;H0;D3;+0:1]": 532,
+      "[B;H0;D3;+0:6]": 533,
+      "[C;D1;H1:6]": 534,
+      "[C;D1;H2:10]": 535,
+      "[CH;D1;+0:9]": 536,
+      "[F;D1;H0:10]": 537,
+      "[F;D1;H0:1]": 538,
+      "[N-;H0;D1:7]": 539,
+      "[N;-;D1;H0:9]": 540,
+      "[N;D1;H2:12]": 541,
+      "[NH2;D1;+0:15]": 542,
+      "[O-;H0;D1:4]": 543,
+      "[O;D1;H0:15]": 544,
+      "[S;D1;H0:9]": 545,
+      "[S;H0;D3;+0:6]": 546,
+      "[o;H0;D2;+0:6]": 547,
+      "[s;H0;D2;+0:8]": 548,
+      ":3": 549,
+      "[C;H0;D3;+0:11]": 550,
+      "[C@@H;D3;+0:8]": 551,
+      "[CH2;D2;+0:16]": 552,
+      "[CH;D1;+0:1]": 553,
+      "[CH;D1;+0:6]": 554,
+      "[N-;H0;D1:8]": 555,
+      "[NH2;D1;+0:14]": 556,
+      "[NH4+;D0:6]": 557,
+      "[S;H0;D1;+0:8]": 558,
+      "[SH;D1;+0:3]": 559,
+      "[n;H0;D2;+0:1]": 560,
+      "[o;H0;D2;+0:5]": 561,
+      "[#14:7]": 562,
+      "[#7;+:9]": 563,
+      "[#7;a;+:5]": 564,
+      "[#8:15]": 565,
+      "[C;D1;H2:9]": 566,
+      "[C;D1;H3:18]": 567,
+      "[C;H0;D3;+0:13]": 568,
+      "[CH2;D2;+0:11]": 569,
+      "[CH2;D2;+0:12]": 570,
+      "[CH2;D2;+0:18]": 571,
+      "[CH2;D2;+0:19]": 572,
+      "[CH;D3;+0:10]": 573,
+      "[Cl;D1;H0:3]": 574,
+      "[N+;H0;D2:7]": 575,
+      "[N;-;D1;H0:10]": 576,
+      "[N;D1;H0:10]": 577,
+      "[N;H0;D1;+0:5]": 578,
+      "[N;H0;D2;+0:10]": 579,
+      "[N;H0;D2;+0:12]": 580,
+      "[NH3+;D1:5]": 581,
+      "[NH4+;D0:9]": 582,
+      "[NH;D2;+0:16]": 583,
+      "[O;D1;H1:12]": 584,
+      "[O;D1;H1:13]": 585,
+      "[S;D1;H0:10]": 586,
+      "[S;D1;H0:5]": 587,
+      "[n;H0;D2;+0:15]": 588,
+      "[nH;D2;+0:12]": 589,
+      "[nH;D2;+0:13]": 590,
+      "[s;H0;D2;+0:6]": 591,
+      "-3": 592,
+      "[BrH;D0;+0:1]": 593,
+      "[C-;H0;D1:3]": 594,
+      "[CH2;D1;+0:11]": 595,
+      "[CH2;D1;+0:9]": 596,
+      "[CH3;D1;+0:8]": 597,
+      "[CH;D1;+0:3]": 598,
+      "[CH;D2;+0:12]": 599,
+      "[N+;H0;D2:5]": 600,
+      "[N+;H0;D4:3]": 601,
+      "[N+;H0;D4:6]": 602,
+      "[N;H0;D2;+0:11]": 603,
+      "[NH2;D1;+0:16]": 604,
+      "[NH2;D1;+0:17]": 605,
+      "[NH3+;D1:1]": 606,
+      "[NH;D1;+0:5]": 607,
+      "[NH;D1;+0:7]": 608,
+      "[OH2;D0;+0:5]": 609,
+      "[S-;H0;D1:8]": 610,
+      "[S;H0;D1;+0:13]": 611,
+      "[S;H0;D1;+0:6]": 612,
+      "[S;H0;D1;+0:7]": 613,
+      "[Sn;H0;D4;+0:6]": 614,
+      "[Zn]": 615,
+      "[n;H0;D3;+0:14]": 616,
+      "[s;H0;D2;+0:13]": 617,
+      "[#14:8]": 618,
+      "[#16:13]": 619,
+      "[#16;a:14]": 620,
+      "[#34;a:2]": 621,
+      "[#7:17]": 622,
+      "[#7;a:16]": 623,
+      "[#7;a;+:3]": 624,
+      "[#8:14]": 625,
+      "[C-;H0;D1:7]": 626,
+      "[C:17]": 627,
+      "[C:18]": 628,
+      "[C:19]": 629,
+      "[C;D1;H1:2]": 630,
+      "[C;D1;H1:8]": 631,
+      "[C;D1;H1:9]": 632,
+      "[C;D1;H2:14]": 633,
+      "[C;D1;H3:21]": 634,
+      "[C@@H;D3;+0:9]": 635,
+      "[C@@H]": 636,
+      "[CH2;D1;+0:10]": 637,
+      "[CH2;D2;+0:15]": 638,
+      "[CH2;D2;+0:22]": 639,
+      "[CH3;D1;+0:14]": 640,
+      "[CH3;D1;+0:2]": 641,
+      "[CH;D1;+0:10]": 642,
+      "[CH;D2;+0:13]": 643,
+      "[CH;D3;+0:15]": 644,
+      "[Cl;D1;H0:6]": 645,
+      "[N;D1;H0:12]": 646,
+      "[N;D1;H1:3]": 647,
+      "[N;D1;H1:6]": 648,
+      "[N;D1;H1:7]": 649,
+      "[N;D1;H1:8]": 650,
+      "[N;D1;H2:10]": 651,
+      "[N;H0;D1;+0:6]": 652,
+      "[N;H0;D3;+0:13]": 653,
+      "[NH3+;D1:8]": 654,
+      "[NH3;D0;+0:8]": 655,
+      "[NH;D1;+0:6]": 656,
+      "[NH;D2;+0:15]": 657,
+      "[O-;H0;D1:3]": 658,
+      "[O;D1;H1:14]": 659,
+      "[O;H0;D1;+0:10]": 660,
+      "[O;H0;D2;+0:12]": 661,
+      "[OH;D1;+0:12]": 662,
+      "[S;H0;D3;+0:7]": 663,
+      "[S;H0;D4;+0:4]": 664,
+      "[S;H0;D4;+0:5]": 665,
+      "[Sn;H0;D4;+0:4]": 666,
+      "[Sn;H0;D4;+0:7]": 667,
+      "[n+;H0;D3:5]": 668,
+      "[n;H0;D2;+0:14]": 669,
+      "[n;H0;D3;+0:13]": 670,
+      "[n;H0;D3;+0:15]": 671,
+      "[n;H0;D3;+0:16]": 672,
+      "[nH;D2;+0:2]": 673,
+      "[o;H0;D2;+0:3]": 674,
+      "[s;H0;D2;+0:7]": 675,
+      "[C-;H0;D1:5]": 676,
+      "[C;H0;D2;+0:11]": 677,
+      "[C;H0;D3;+0:14]": 678,
+      "[C;H0;D4;+0:8]": 679,
+      "[C@;H0;D4;+0:2]": 680,
+      "[C@;H0;D4;+0:3]": 681,
+      "[C@@;H0;D4;+0:5]": 682,
+      "[C@@H;D3;+0:10]": 683,
+      "[C@@H;D3;+0:13]": 684,
+      "[CH2;D1;+0:2]": 685,
+      "[CH2;D2;+0:13]": 686,
+      "[CH2;D2;+0:14]": 687,
+      "[CH;D1;+0:4]": 688,
+      "[CH;D3;+0:13]": 689,
+      "[Cl;H0;D1;+0:6]": 690,
+      "[N+;H0;D2:6]": 691,
+      "[N+;H0;D2:8]": 692,
+      "[N-;H0;D1:9]": 693,
+      "[N;H0;D3;+0:14]": 694,
+      "[N;H0;D3;+0:17]": 695,
+      "[NH2+;D2:5]": 696,
+      "[NH3+;D1:9]": 697,
+      "[NH;D1;+0:9]": 698,
+      "[NH;D2;+0:17]": 699,
+      "[O-;H0;D1:10]": 700,
+      "[OH2;D0;+0:6]": 701,
+      "[OH;D1;+0:13]": 702,
+      "[S-;H0;D1:7]": 703,
+      "[S;H0;D1;+0:15]": 704,
+      "[S;H0;D3;+0:1]": 705,
+      "[S;H0;D4;+0:7]": 706,
+      "[SH;D1;+0:1]": 707,
+      "[cH;D2;+0:12]": 708,
+      "[n;H0;D2;+0:13]": 709,
+      "[n;H0;D2;+0:16]": 710,
+      "[n;H0;D2;+0:17]": 711,
+      "[o;H0;D2;+0:2]": 712,
+      "[o;H0;D2;+0:7]": 713,
+      "[s;H0;D2;+0:15]": 714,
+      "-4": 715,
+      "4": 716,
+      "[#14:11]": 717,
+      "[#14:1]": 718,
+      "[#14:3]": 719,
+      "[#14:9]": 720,
+      "[#15:1]": 721,
+      "[#15:3]": 722,
+      "[#15;a:5]": 723,
+      "[#16:11]": 724,
+      "[#16;a:13]": 725,
+      "[#34:3]": 726,
+      "[#34;a:4]": 727,
+      "[#50:2]": 728,
+      "[#50:4]": 729,
+      "[#5:4]": 730,
+      "[#5:8]": 731,
+      "[#7:15]": 732,
+      "[#7;+:10]": 733,
+      "[#7;+:4]": 734,
+      "[#8;a:11]": 735,
+      "[#8;a:12]": 736,
+      "[#8;a:13]": 737,
+      "[B;H0;D3;+0:5]": 738,
+      "[B;H0;D3;+0:8]": 739,
+      "[Br;D1;H0:1]": 740,
+      "[Br;D1;H0:4]": 741,
+      "[Br;D1;H0:6]": 742,
+      "[C;D1;H1:12]": 743,
+      "[C;D1;H2:11]": 744,
+      "[C;D1;H2:20]": 745,
+      "[C;D1;H3:19]": 746,
+      "[C;D1;H3:20]": 747,
+      "[C;D1;H3:23]": 748,
+      "[C;D1;H3:29]": 749,
+      "[C;D1;H3:30]": 750,
+      "[C;H0;D2;+0:12]": 751,
+      "[C;H0;D3;+0:12]": 752,
+      "[C;H0;D3;+0:24]": 753,
+      "[C;H0;D3;+0:26]": 754,
+      "[C;H0;D4;+0:28]": 755,
+      "[C@;H0;D4;+0:14]": 756,
+      "[C@@;H0;D4;+0:1]": 757,
+      "[C@@;H0;D4;+0:2]": 758,
+      "[C@@;H0;D4;+0:4]": 759,
+      "[C@@H;D3;+0:12]": 760,
+      "[C@H;D3;+0:10]": 761,
+      "[C@H;D3;+0:11]": 762,
+      "[C@H;D3;+0:12]": 763,
+      "[C@H;D3;+0:14]": 764,
+      "[C@H;D3;+0:17]": 765,
+      "[C@H]": 766,
+      "[CH2;D2;+0:20]": 767,
+      "[CH2;D2;+0:21]": 768,
+      "[CH2;D2;+0:23]": 769,
+      "[CH2;D2;+0:24]": 770,
+      "[CH3;D1;+0:16]": 771,
+      "[CH3;D1;+0:17]": 772,
+      "[CH3;D1;+0:31]": 773,
+      "[CH3;D1;+0:9]": 774,
+      "[CH;D1;+0:11]": 775,
+      "[CH;D1;+0:13]": 776,
+      "[CH;D2;+0:17]": 777,
+      "[CH;D3;+0:12]": 778,
+      "[CH;D3;+0:14]": 779,
+      "[CH;D3;+0:19]": 780,
+      "[Cl;H0;D1;+0:7]": 781,
+      "[F;D1;H0:11]": 782,
+      "[F;D1;H0:12]": 783,
+      "[Mg]": 784,
+      "[N+;H0;D2:11]": 785,
+      "[N+;H0;D2:1]": 786,
+      "[N+;H0;D4:7]": 787,
+      "[N-;H0;D1:12]": 788,
+      "[N-;H0;D1:1]": 789,
+      "[N-;H0;D1:2]": 790,
+      "[N;-;D1;H0:11]": 791,
+      "[N;-;D1;H0:5]": 792,
+      "[N;D1;H0:11]": 793,
+      "[N;D1;H0:1]": 794,
+      "[N;D1;H1:12]": 795,
+      "[N;D1;H1:1]": 796,
+      "[N;D1;H1:5]": 797,
+      "[N;D1;H1:9]": 798,
+      "[N;H0;D1;+0:10]": 799,
+      "[N;H0;D1;+0:3]": 800,
+      "[N;H0;D2;+0:13]": 801,
+      "[N;H0;D3;+0:21]": 802,
+      "[N;H0;D3;+0:22]": 803,
+      "[NH2+;D2:7]": 804,
+      "[NH3+;D1:6]": 805,
+      "[NH3;D0;+0:9]": 806,
+      "[NH;D1;+0:10]": 807,
+      "[NH;D1;+0:11]": 808,
+      "[O-;H0;D1:11]": 809,
+      "[O-;H0;D1:12]": 810,
+      "[O-;H0;D1:9]": 811,
+      "[O;-;D1;H0:8]": 812,
+      "[O;D1;H0:16]": 813,
+      "[O;D1;H0:18]": 814,
+      "[O;D1;H0:25]": 815,
+      "[O;D1;H0:27]": 816,
+      "[O;D1;H1:1]": 817,
+      "[O;H0;D1;+0:12]": 818,
+      "[O;H0;D2;+0:13]": 819,
+      "[O;H0;D2;+0:14]": 820,
+      "[OH2;D0;+0:1]": 821,
+      "[OH;D1;+0:14]": 822,
+      "[P;H0;D3;+0:4]": 823,
+      "[P;H0;D3;+0:5]": 824,
+      "[P;H0;D4;+0:4]": 825,
+      "[P;H0;D4;+0:5]": 826,
+      "[S-;H0;D1:4]": 827,
+      "[S;D1;H0:11]": 828,
+      "[S;H0;D2;+0:1]": 829,
+      "[S;H0;D4;+0:3]": 830,
+      "[S;H0;D4;+0:6]": 831,
+      "[Si;H0;D4;+0:1]": 832,
+      "[SiH;D3;+0:1]": 833,
+      "[c:16]": 834,
+      "[n+;H0;D3:4]": 835,
+      "[o;H0;D2;+0:4]": 836,
+      "[o;H0;D2;+0:8]": 837,
+      "[o;H0;D2;+0:9]": 838,
+      "[s;H0;D2;+0:5]": 839,
+      "[C-;H0;D1:1]": 840,
+      "[C-;H0;D1:8]": 841,
+      "[C;H0;D2;+0:13]": 842,
+      "[C;H0;D2;+0:16]": 843,
+      "[C@;H0;D4;+0:1]": 844,
+      "[C@;H0;D4;+0:4]": 845,
+      "[C@;H0;D4;+0:5]": 846,
+      "[C@;H0;D4;+0:7]": 847,
+      "[C@@;H0;D4;+0:3]": 848,
+      "[C@@;H0;D4;+0:6]": 849,
+      "[C@@;H0;D4;+0:9]": 850,
+      "[C@@H;D3;+0:15]": 851,
+      "[C@@H;D3;+0:18]": 852,
+      "[C@@H;D3;+0:20]": 853,
+      "[C@@H;D3;+0:26]": 854,
+      "[C@@]": 855,
+      "[C@H;D3;+0:13]": 856,
+      "[C@H;D3;+0:16]": 857,
+      "[C@H;D3;+0:18]": 858,
+      "[C@H;D3;+0:20]": 859,
+      "[C@H;D3;+0:24]": 860,
+      "[C@H;D3;+0:26]": 861,
+      "[CH2;D1;+0:12]": 862,
+      "[CH2;D1;+0:13]": 863,
+      "[CH2;D1;+0:16]": 864,
+      "[CH;D1;+0:17]": 865,
+      "[CH;D2;+0:15]": 866,
+      "[CH;D3;+0:11]": 867,
+      "[CH;D3;+0:16]": 868,
+      "[CH;D3;+0:24]": 869,
+      "[Cl;H0;D1;+0:5]": 870,
+      "[Cl;H0;D1;+0:9]": 871,
+      "[ClH;D0;+0:5]": 872,
+      "[ClH;D0;+0:6]": 873,
+      "[ClH;D0;+0:9]": 874,
+      "[IH;D0;+0:1]": 875,
+      "[N+;H0;D2:10]": 876,
+      "[N+;H0;D2:3]": 877,
+      "[N+;H0;D2:9]": 878,
+      "[N+;H0;D3:3]": 879,
+      "[N+;H0;D3:4]": 880,
+      "[N+;H0;D4:11]": 881,
+      "[N+;H0;D4:5]": 882,
+      "[N+;H0;D4:9]": 883,
+      "[N-;H0;D1:10]": 884,
+      "[N-;H0;D1:11]": 885,
+      "[N;H0;D1;+0:11]": 886,
+      "[N;H0;D2;+0:16]": 887,
+      "[N;H0;D3;+0:15]": 888,
+      "[N;H0;D3;+0:16]": 889,
+      "[N;H0;D3;+0:18]": 890,
+      "[N;H0;D3;+0:23]": 891,
+      "[N@@+;H0;D4:3]": 892,
+      "[NH2+;D2:1]": 893,
+      "[NH2+;D2:8]": 894,
+      "[NH2;D1;+0:22]": 895,
+      "[NH2;D1;+0:25]": 896,
+      "[NH2;D1;+0:27]": 897,
+      "[NH3+;D1:10]": 898,
+      "[NH3+;D1:12]": 899,
+      "[NH3+;D1:4]": 900,
+      "[NH3+;D1:7]": 901,
+      "[NH3;D0;+0:10]": 902,
+      "[NH3;D0;+0:3]": 903,
+      "[NH4+;D0:11]": 904,
+      "[NH4+;D0:13]": 905,
+      "[NH4+;D0:3]": 906,
+      "[NH4+;D0:8]": 907,
+      "[NH;D1;+0:12]": 908,
+      "[NH;D1;+0:3]": 909,
+      "[NH;D1;+0:4]": 910,
+      "[NH;D1;+0:8]": 911,
+      "[NH;D2;+0:18]": 912,
+      "[NH;D2;+0:22]": 913,
+      "[NH;D2;+0:23]": 914,
+      "[NH;D2;+0:25]": 915,
+      "[NH;D2;+0:27]": 916,
+      "[O;H0;D1;+0:13]": 917,
+      "[OH2;D0;+0:4]": 918,
+      "[OH2;D0;+0:7]": 919,
+      "[OH2;D0;+0:9]": 920,
+      "[P+;H0;D4:2]": 921,
+      "[P;H0;D3;+0:1]": 922,
+      "[P;H0;D3;+0:6]": 923,
+      "[P;H0;D3;+0:8]": 924,
+      "[P;H0;D4;+0:6]": 925,
+      "[P;H0;D4;+0:8]": 926,
+      "[PH;D2;+0:1]": 927,
+      "[S;H0;D1;+0:14]": 928,
+      "[S;H0;D1;+0:16]": 929,
+      "[S;H0;D1;+0:5]": 930,
+      "[S;H0;D2;+0:12]": 931,
+      "[S;H0;D3;+0:10]": 932,
+      "[S;H0;D3;+0:11]": 933,
+      "[S;H0;D3;+0:8]": 934,
+      "[S;H0;D4;+0:10]": 935,
+      "[S;H0;D4;+0:11]": 936,
+      "[S;H0;D4;+0:8]": 937,
+      "[S@@;H0;D3;+0:7]": 938,
+      "[SH;D1;+0:12]": 939,
+      "[SnH;D3;+0:5]": 940,
+      "[SnH;D3;+0:6]": 941,
+      "[c;H0;D3;+0:16]": 942,
+      "[cH;D2;+0:13]": 943,
+      "[cH;D2;+0:17]": 944,
+      "[n+;H0;D3:3]": 945,
+      "[n+;H0;D3:7]": 946,
+      "[n-;H0;D2:6]": 947,
+      "[nH;D2;+0:15]": 948,
+      "[s;H0;D2;+0:14]": 949,
+      "[s;H0;D2;+0:16]": 950,
+      "s": 951
     },
     "unk_token": "§"
   }

tokenizer_config.json

@@ -1,8 +1,54 @@
 {
+  "added_tokens_decoder": {
+    "0": {
+      "content": "^",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "_",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": " ",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "§",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "4": {
+      "content": "°",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
   "bos_token": "^",
+  "clean_up_tokenization_spaces": true,
   "eos_token": "_",
-  "model_max_length": 896,
+  "max_length": null,
+  "model_max_length": 1024,
+  "pad_to_multiple_of": null,
   "pad_token": " ",
+  "pad_token_type_id": 0,
   "padding_side": "right",
   "tokenizer_class": "PreTrainedTokenizerFast",
   "truncation_side": "left",

train_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 43.0,
-    "train_loss": 0.2331255912600619,
-    "train_runtime": 2746.0841,
+    "epoch": 31.0,
+    "train_loss": 0.7912082670869731,
+    "train_runtime": 664.7096,
     "train_samples": 7878,
-    "train_samples_per_second": 143.441,
-    "train_steps_per_second": 4.479
+    "train_samples_per_second": 592.59,
+    "train_steps_per_second": 18.504
 }

trainer_state.json

@@ -1,8 +1,9 @@
 {
-  "best_metric": 0.0970187559723854,
-  "best_model_checkpoint": "/home/stephan/code/molreactgen/checkpoints/2023-05-12_13-06-58_experiment/checkpoint-9850",
-  "epoch": 42.99695431472081,
-  "global_step": 10588,
+  "best_metric": 0.39055171608924866,
+  "best_model_checkpoint": "/home/stephan/code/molreactgen/checkpoints/2023-11-07_18-29-28_experiment/checkpoint-6895",
+  "epoch": 30.996954314720814,
+  "eval_steps": 500,
+  "global_step": 7633,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -10,1033 +11,753 @@
     {
       "epoch": 0.41,
       "learning_rate": 4.065040650406504e-05,
-      "loss": 3.0981,
+      "loss": 6.0993,
       "step": 100
     },
     {
       "epoch": 0.81,
       "learning_rate": 8.130081300813008e-05,
-      "loss": 2.1418,
+      "loss": 5.0338,
       "step": 200
     },
     {
       "epoch": 1.0,
-      "eval_accuracy": 0.2829720403871191,
-      "eval_loss": 1.5341426134109497,
-      "eval_runtime": 6.1779,
-      "eval_samples_per_second": 390.745,
-      "eval_steps_per_second": 24.442,
+      "eval_accuracy": 0.2072072072072072,
+      "eval_loss": 3.800293445587158,
+      "eval_runtime": 1.7789,
+      "eval_samples_per_second": 1357.054,
+      "eval_steps_per_second": 84.886,
       "step": 246
     },
     {
       "epoch": 1.22,
       "learning_rate": 0.00012195121951219512,
-      "loss": 1.588,
+      "loss": 3.8428,
       "step": 300
     },
     {
       "epoch": 1.62,
       "learning_rate": 0.00016260162601626016,
-      "loss": 1.1354,
+      "loss": 2.8076,
       "step": 400
     },
     {
       "epoch": 2.0,
-      "eval_accuracy": 0.32898021613845974,
-      "eval_loss": 0.7312331795692444,
-      "eval_runtime": 6.0715,
-      "eval_samples_per_second": 397.598,
-      "eval_steps_per_second": 24.87,
+      "eval_accuracy": 0.25823462236055444,
+      "eval_loss": 1.976719856262207,
+      "eval_runtime": 1.774,
+      "eval_samples_per_second": 1360.786,
+      "eval_steps_per_second": 85.12,
       "step": 492
     },
     {
       "epoch": 2.03,
       "learning_rate": 0.0002032520325203252,
-      "loss": 0.8911,
+      "loss": 2.2493,
       "step": 500
     },
     {
       "epoch": 2.44,
       "learning_rate": 0.00024390243902439024,
-      "loss": 0.743,
+      "loss": 1.9599,
       "step": 600
     },
     {
       "epoch": 2.84,
       "learning_rate": 0.0002845528455284553,
-      "loss": 0.664,
+      "loss": 1.7151,
       "step": 700
     },
     {
       "epoch": 3.0,
-      "eval_accuracy": 0.3475759534583064,
-      "eval_loss": 0.5133360028266907,
-      "eval_runtime": 6.1012,
-      "eval_samples_per_second": 395.662,
-      "eval_steps_per_second": 24.749,
+      "eval_accuracy": 0.2894563992327041,
+      "eval_loss": 1.4048570394515991,
+      "eval_runtime": 1.775,
+      "eval_samples_per_second": 1360.028,
+      "eval_steps_per_second": 85.072,
       "step": 738
     },
     {
       "epoch": 3.25,
       "learning_rate": 0.0003252032520325203,
-      "loss": 0.6077,
+      "loss": 1.5375,
       "step": 800
     },
     {
       "epoch": 3.65,
       "learning_rate": 0.00036585365853658537,
-      "loss": 0.5517,
+      "loss": 1.3954,
       "step": 900
     },
     {
       "epoch": 4.0,
-      "eval_accuracy": 0.35866693372907127,
-      "eval_loss": 0.40151524543762207,
-      "eval_runtime": 6.0531,
-      "eval_samples_per_second": 398.803,
-      "eval_steps_per_second": 24.946,
+      "eval_accuracy": 0.312210304753801,
+      "eval_loss": 1.0894988775253296,
+      "eval_runtime": 1.8067,
+      "eval_samples_per_second": 1336.115,
+      "eval_steps_per_second": 83.576,
       "step": 985
     },
     {
       "epoch": 4.06,
       "learning_rate": 0.0004065040650406504,
-      "loss": 0.5002,
+      "loss": 1.2718,
       "step": 1000
     },
     {
       "epoch": 4.47,
       "learning_rate": 0.00044715447154471545,
-      "loss": 0.4461,
+      "loss": 1.1709,
       "step": 1100
     },
     {
       "epoch": 4.87,
       "learning_rate": 0.0004878048780487805,
-      "loss": 0.3986,
+      "loss": 1.0895,
       "step": 1200
     },
     {
       "epoch": 5.0,
-      "eval_accuracy": 0.36889936906506915,
-      "eval_loss": 0.30596020817756653,
-      "eval_runtime": 6.0649,
-      "eval_samples_per_second": 398.03,
-      "eval_steps_per_second": 24.897,
+      "eval_accuracy": 0.3280339162692104,
+      "eval_loss": 0.8805840015411377,
+      "eval_runtime": 1.7957,
+      "eval_samples_per_second": 1344.309,
+      "eval_steps_per_second": 84.089,
       "step": 1231
     },
     {
       "epoch": 5.28,
       "learning_rate": 0.0004999506716812021,
-      "loss": 0.3583,
+      "loss": 0.9914,
       "step": 1300
     },
     {
       "epoch": 5.69,
       "learning_rate": 0.0004997091104496882,
-      "loss": 0.3313,
+      "loss": 0.9375,
       "step": 1400
     },
     {
       "epoch": 6.0,
-      "eval_accuracy": 0.3759261451059297,
-      "eval_loss": 0.241128608584404,
-      "eval_runtime": 6.1261,
-      "eval_samples_per_second": 394.054,
-      "eval_steps_per_second": 24.649,
+      "eval_accuracy": 0.3402263075005411,
+      "eval_loss": 0.7313582301139832,
+      "eval_runtime": 1.7974,
+      "eval_samples_per_second": 1343.047,
+      "eval_steps_per_second": 84.01,
       "step": 1477
     },
     {
       "epoch": 6.09,
       "learning_rate": 0.0004992664502959351,
-      "loss": 0.3005,
+      "loss": 0.8598,
       "step": 1500
     },
     {
       "epoch": 6.5,
       "learning_rate": 0.0004986230477086575,
-      "loss": 0.2758,
+      "loss": 0.8097,
       "step": 1600
     },
     {
       "epoch": 6.9,
       "learning_rate": 0.0004977794208410241,
-      "loss": 0.259,
+      "loss": 0.7668,
       "step": 1700
     },
     {
       "epoch": 7.0,
-      "eval_accuracy": 0.3807942678696614,
-      "eval_loss": 0.19934649765491486,
-      "eval_runtime": 6.0599,
-      "eval_samples_per_second": 398.358,
-      "eval_steps_per_second": 24.918,
+      "eval_accuracy": 0.3481269173067019,
+      "eval_loss": 0.6367093324661255,
+      "eval_runtime": 1.8149,
+      "eval_samples_per_second": 1330.094,
+      "eval_steps_per_second": 83.2,
       "step": 1723
     },
     {
       "epoch": 7.31,
       "learning_rate": 0.0004967362490933723,
-      "loss": 0.2409,
+      "loss": 0.716,
       "step": 1800
     },
     {
       "epoch": 7.72,
       "learning_rate": 0.0004954943725660643,
-      "loss": 0.2288,
+      "loss": 0.6978,
       "step": 1900
     },
     {
       "epoch": 8.0,
-      "eval_accuracy": 0.38369857151961545,
-      "eval_loss": 0.17374461889266968,
-      "eval_runtime": 6.1335,
-      "eval_samples_per_second": 393.573,
-      "eval_steps_per_second": 24.619,
+      "eval_accuracy": 0.35578121617889635,
+      "eval_loss": 0.5603917241096497,
+      "eval_runtime": 1.8199,
+      "eval_samples_per_second": 1326.454,
+      "eval_steps_per_second": 82.972,
       "step": 1970
     },
     {
       "epoch": 8.12,
       "learning_rate": 0.0004940547913829275,
-      "loss": 0.2201,
+      "loss": 0.6556,
       "step": 2000
     },
     {
       "epoch": 8.53,
       "learning_rate": 0.0004924186648858207,
-      "loss": 0.2064,
+      "loss": 0.627,
       "step": 2100
     },
     {
       "epoch": 8.93,
       "learning_rate": 0.0004905873107009799,
-      "loss": 0.1992,
+      "loss": 0.6133,
       "step": 2200
     },
     {
       "epoch": 9.0,
-      "eval_accuracy": 0.385633256552892,
-      "eval_loss": 0.15620127320289612,
-      "eval_runtime": 6.1272,
-      "eval_samples_per_second": 393.981,
-      "eval_steps_per_second": 24.644,
+      "eval_accuracy": 0.3603864842472962,
+      "eval_loss": 0.5122300386428833,
+      "eval_runtime": 1.8088,
+      "eval_samples_per_second": 1334.561,
+      "eval_steps_per_second": 83.479,
       "step": 2216
     },
     {
       "epoch": 9.34,
       "learning_rate": 0.0004885622036778897,
-      "loss": 0.1894,
+      "loss": 0.5846,
       "step": 2300
     },
     {
       "epoch": 9.75,
       "learning_rate": 0.0004863449747015384,
-      "loss": 0.182,
+      "loss": 0.5681,
       "step": 2400
     },
     {
       "epoch": 10.0,
-      "eval_accuracy": 0.38733486894215974,
-      "eval_loss": 0.1415354460477829,
-      "eval_runtime": 6.1286,
-      "eval_samples_per_second": 393.894,
-      "eval_steps_per_second": 24.639,
+      "eval_accuracy": 0.36240548750905005,
+      "eval_loss": 0.48733416199684143,
+      "eval_runtime": 1.781,
+      "eval_samples_per_second": 1355.443,
+      "eval_steps_per_second": 84.785,
       "step": 2462
     },
     {
       "epoch": 10.15,
       "learning_rate": 0.0004839374093790139,
-      "loss": 0.181,
+      "loss": 0.5537,
       "step": 2500
     },
     {
       "epoch": 10.56,
       "learning_rate": 0.00048134144660149535,
-      "loss": 0.1695,
+      "loss": 0.5314,
       "step": 2600
     },
     {
       "epoch": 10.96,
       "learning_rate": 0.0004785591769828005,
-      "loss": 0.1687,
+      "loss": 0.536,
       "step": 2700
     },
     {
       "epoch": 11.0,
-      "eval_accuracy": 0.3880878028351102,
-      "eval_loss": 0.13479964435100555,
-      "eval_runtime": 6.1178,
-      "eval_samples_per_second": 394.588,
-      "eval_steps_per_second": 24.682,
+      "eval_accuracy": 0.36355120655037804,
+      "eval_loss": 0.47042036056518555,
+      "eval_runtime": 1.8486,
+      "eval_samples_per_second": 1305.878,
+      "eval_steps_per_second": 81.685,
       "step": 2708
     },
     {
       "epoch": 11.37,
       "learning_rate": 0.00047559284117574613,
-      "loss": 0.1629,
+      "loss": 0.5126,
       "step": 2800
     },
     {
       "epoch": 11.78,
       "learning_rate": 0.0004724448280676768,
-      "loss": 0.1592,
+      "loss": 0.511,
       "step": 2900
     },
     {
       "epoch": 12.0,
-      "eval_accuracy": 0.38904467529156844,
-      "eval_loss": 0.12684421241283417,
-      "eval_runtime": 6.118,
-      "eval_samples_per_second": 394.571,
-      "eval_steps_per_second": 24.681,
+      "eval_accuracy": 0.3647827612202094,
+      "eval_loss": 0.4570145606994629,
+      "eval_runtime": 1.8132,
+      "eval_samples_per_second": 1331.315,
+      "eval_steps_per_second": 83.276,
       "step": 2955
     },
     {
       "epoch": 12.18,
       "learning_rate": 0.00046911767285661587,
-      "loss": 0.1527,
+      "loss": 0.4918,
       "step": 3000
     },
     {
       "epoch": 12.59,
       "learning_rate": 0.0004656140550095876,
-      "loss": 0.1516,
+      "loss": 0.4883,
       "step": 3100
     },
     {
       "epoch": 12.99,
       "learning_rate": 0.00046193679610475414,
-      "loss": 0.1476,
+      "loss": 0.4929,
       "step": 3200
     },
     {
       "epoch": 13.0,
-      "eval_accuracy": 0.3893742545772372,
-      "eval_loss": 0.12305640429258347,
-      "eval_runtime": 6.1457,
-      "eval_samples_per_second": 392.794,
-      "eval_steps_per_second": 24.57,
+      "eval_accuracy": 0.3655067660867164,
+      "eval_loss": 0.4465464651584625,
+      "eval_runtime": 1.8353,
+      "eval_samples_per_second": 1315.328,
+      "eval_steps_per_second": 82.276,
       "step": 3201
     },
     {
       "epoch": 13.4,
       "learning_rate": 0.0004580888575591068,
-      "loss": 0.1447,
+      "loss": 0.4634,
       "step": 3300
     },
     {
       "epoch": 13.81,
       "learning_rate": 0.00045407333824353966,
-      "loss": 0.1417,
+      "loss": 0.4757,
       "step": 3400
     },
     {
       "epoch": 14.0,
-      "eval_accuracy": 0.3899678614674472,
-      "eval_loss": 0.11744081974029541,
-      "eval_runtime": 6.1575,
-      "eval_samples_per_second": 392.044,
-      "eval_steps_per_second": 24.523,
+      "eval_accuracy": 0.36605909969621653,
+      "eval_loss": 0.43762096762657166,
+      "eval_runtime": 1.8426,
+      "eval_samples_per_second": 1310.127,
+      "eval_steps_per_second": 81.951,
       "step": 3447
     },
     {
       "epoch": 14.21,
       "learning_rate": 0.00044989347198722777,
-      "loss": 0.1412,
+      "loss": 0.4605,
       "step": 3500
     },
     {
       "epoch": 14.62,
       "learning_rate": 0.00044555262497331783,
-      "loss": 0.1403,
+      "loss": 0.4507,
       "step": 3600
     },
     {
       "epoch": 15.0,
-      "eval_accuracy": 0.3901754417910176,
-      "eval_loss": 0.11636122316122055,
-      "eval_runtime": 6.1213,
-      "eval_samples_per_second": 394.36,
-      "eval_steps_per_second": 24.668,
+      "eval_accuracy": 0.3666450211603484,
+      "eval_loss": 0.42967188358306885,
+      "eval_runtime": 1.7693,
+      "eval_samples_per_second": 1364.379,
+      "eval_steps_per_second": 85.344,
       "step": 3693
     },
     {
       "epoch": 15.03,
       "learning_rate": 0.0004410542930280316,
-      "loss": 0.1354,
+      "loss": 0.4591,
       "step": 3700
     },
     {
       "epoch": 15.43,
       "learning_rate": 0.0004364020988053623,
-      "loss": 0.1314,
+      "loss": 0.4366,
       "step": 3800
     },
     {
       "epoch": 15.84,
       "learning_rate": 0.00043159978886963223,
-      "loss": 0.1342,
+      "loss": 0.4449,
       "step": 3900
     },
     {
       "epoch": 16.0,
-      "eval_accuracy": 0.3905068419567177,
-      "eval_loss": 0.11423930525779724,
-      "eval_runtime": 6.1098,
-      "eval_samples_per_second": 395.1,
-      "eval_steps_per_second": 24.714,
+      "eval_accuracy": 0.3675183053807743,
+      "eval_loss": 0.4223393499851227,
+      "eval_runtime": 1.82,
+      "eval_samples_per_second": 1326.352,
+      "eval_steps_per_second": 82.966,
       "step": 3940
     },
     {
       "epoch": 16.24,
       "learning_rate": 0.0004266512306782628,
-      "loss": 0.1307,
+      "loss": 0.4323,
       "step": 4000
     },
     {
       "epoch": 16.65,
       "learning_rate": 0.00042156040946718344,
-      "loss": 0.1273,
+      "loss": 0.4312,
       "step": 4100
     },
     {
       "epoch": 17.0,
-      "eval_accuracy": 0.39085189872265264,
-      "eval_loss": 0.1110914945602417,
-      "eval_runtime": 6.0921,
-      "eval_samples_per_second": 396.254,
-      "eval_steps_per_second": 24.786,
+      "eval_accuracy": 0.36820125842495355,
+      "eval_loss": 0.4195675849914551,
+      "eval_runtime": 1.8599,
+      "eval_samples_per_second": 1297.945,
+      "eval_steps_per_second": 81.189,
       "step": 4186
     },
     {
       "epoch": 17.06,
       "learning_rate": 0.00041633142504139133,
-      "loss": 0.128,
+      "loss": 0.4315,
       "step": 4200
     },
     {
       "epoch": 17.46,
       "learning_rate": 0.00041096848847324417,
-      "loss": 0.1247,
+      "loss": 0.4158,
       "step": 4300
     },
     {
       "epoch": 17.87,
       "learning_rate": 0.0004054759187111451,
-      "loss": 0.1243,
+      "loss": 0.4252,
       "step": 4400
     },
     {
       "epoch": 18.0,
-      "eval_accuracy": 0.39107313564645796,
-      "eval_loss": 0.10974010080099106,
-      "eval_runtime": 6.1271,
-      "eval_samples_per_second": 393.988,
-      "eval_steps_per_second": 24.645,
+      "eval_accuracy": 0.36844383737507186,
+      "eval_loss": 0.41086554527282715,
+      "eval_runtime": 1.8281,
+      "eval_samples_per_second": 1320.494,
+      "eval_steps_per_second": 82.599,
       "step": 4432
     },
     {
       "epoch": 18.27,
       "learning_rate": 0.00039985813910135305,
-      "loss": 0.1231,
+      "loss": 0.4129,
       "step": 4500
     },
     {
       "epoch": 18.68,
       "learning_rate": 0.00039411967382571643,
-      "loss": 0.1205,
+      "loss": 0.4102,
       "step": 4600
     },
     {
       "epoch": 19.0,
-      "eval_accuracy": 0.39124429836940194,
-      "eval_loss": 0.10775511711835861,
-      "eval_runtime": 6.146,
-      "eval_samples_per_second": 392.775,
-      "eval_steps_per_second": 24.569,
+      "eval_accuracy": 0.3685520649066631,
+      "eval_loss": 0.40959808230400085,
+      "eval_runtime": 1.835,
+      "eval_samples_per_second": 1315.545,
+      "eval_steps_per_second": 82.29,
       "step": 4678
     },
     {
       "epoch": 19.09,
       "learning_rate": 0.0003882651442582019,
-      "loss": 0.1193,
+      "loss": 0.4104,
       "step": 4700
     },
     {
       "epoch": 19.49,
       "learning_rate": 0.00038229926524315015,
-      "loss": 0.1175,
+      "loss": 0.3982,
       "step": 4800
     },
     {
       "epoch": 19.9,
       "learning_rate": 0.0003762268412982577,
-      "loss": 0.1202,
+      "loss": 0.4092,
       "step": 4900
     },
     {
       "epoch": 20.0,
-      "eval_accuracy": 0.3915374600544443,
-      "eval_loss": 0.10535065084695816,
-      "eval_runtime": 6.1087,
-      "eval_samples_per_second": 395.174,
-      "eval_steps_per_second": 24.719,
+      "eval_accuracy": 0.3690036349522679,
+      "eval_loss": 0.4085357189178467,
+      "eval_runtime": 1.7636,
+      "eval_samples_per_second": 1368.753,
+      "eval_steps_per_second": 85.618,
       "step": 4925
     },
     {
       "epoch": 20.3,
       "learning_rate": 0.00037005276274534144,
-      "loss": 0.1151,
+      "loss": 0.3863,
       "step": 5000
     },
     {
       "epoch": 20.71,
       "learning_rate": 0.0003637820017720022,
-      "loss": 0.1146,
+      "loss": 0.3941,
       "step": 5100
     },
     {
       "epoch": 21.0,
-      "eval_accuracy": 0.39146917705326983,
-      "eval_loss": 0.10494475811719894,
-      "eval_runtime": 6.135,
-      "eval_samples_per_second": 393.48,
-      "eval_steps_per_second": 24.613,
+      "eval_accuracy": 0.36924248191853826,
+      "eval_loss": 0.4053109884262085,
+      "eval_runtime": 1.7869,
+      "eval_samples_per_second": 1350.961,
+      "eval_steps_per_second": 84.505,
       "step": 5171
     },
     {
       "epoch": 21.12,
       "learning_rate": 0.00035741960842735953,
-      "loss": 0.1152,
+      "loss": 0.3992,
       "step": 5200
     },
     {
       "epoch": 21.52,
       "learning_rate": 0.0003509707065550817,
-      "loss": 0.1133,
+      "loss": 0.3846,
       "step": 5300
     },
     {
       "epoch": 21.93,
       "learning_rate": 0.00034444048966698643,
-      "loss": 0.1119,
+      "loss": 0.3882,
       "step": 5400
     },
     {
       "epoch": 22.0,
-      "eval_accuracy": 0.39147099793330115,
-      "eval_loss": 0.10487605631351471,
-      "eval_runtime": 6.1054,
-      "eval_samples_per_second": 395.389,
-      "eval_steps_per_second": 24.732,
+      "eval_accuracy": 0.3694365450786329,
+      "eval_loss": 0.40214401483535767,
+      "eval_runtime": 1.8091,
+      "eval_samples_per_second": 1334.335,
+      "eval_steps_per_second": 83.465,
       "step": 5417
     },
     {
       "epoch": 22.34,
       "learning_rate": 0.0003378342167605362,
-      "loss": 0.11,
+      "loss": 0.3787,
       "step": 5500
     },
     {
       "epoch": 22.74,
       "learning_rate": 0.00033115720808359495,
-      "loss": 0.1107,
+      "loss": 0.3821,
       "step": 5600
     },
     {
       "epoch": 23.0,
-      "eval_accuracy": 0.39174139861795204,
-      "eval_loss": 0.10271785408258438,
-      "eval_runtime": 6.1004,
-      "eval_samples_per_second": 395.715,
-      "eval_steps_per_second": 24.753,
+      "eval_accuracy": 0.3694477410301768,
+      "eval_loss": 0.4013039767742157,
+      "eval_runtime": 1.8106,
+      "eval_samples_per_second": 1333.273,
+      "eval_steps_per_second": 83.399,
       "step": 5663
     },
     {
       "epoch": 23.15,
       "learning_rate": 0.0003244148408498587,
-      "loss": 0.1095,
+      "loss": 0.3724,
       "step": 5700
     },
     {
       "epoch": 23.55,
       "learning_rate": 0.000317612544908409,
-      "loss": 0.1067,
+      "loss": 0.372,
       "step": 5800
     },
     {
       "epoch": 23.96,
       "learning_rate": 0.000310755798370878,
-      "loss": 0.1085,
+      "loss": 0.3769,
       "step": 5900
     },
     {
       "epoch": 24.0,
-      "eval_accuracy": 0.391701339257263,
-      "eval_loss": 0.10227391123771667,
-      "eval_runtime": 6.1722,
-      "eval_samples_per_second": 391.109,
-      "eval_steps_per_second": 24.465,
+      "eval_accuracy": 0.36991797099502155,
+      "eval_loss": 0.399305135011673,
+      "eval_runtime": 1.7971,
+      "eval_samples_per_second": 1343.28,
+      "eval_steps_per_second": 84.025,
       "step": 5910
     },
     {
       "epoch": 24.37,
       "learning_rate": 0.00030385012319974537,
-      "loss": 0.1055,
+      "loss": 0.3671,
       "step": 6000
     },
     {
       "epoch": 24.77,
       "learning_rate": 0.00029690108076132154,
-      "loss": 0.1068,
+      "loss": 0.3696,
       "step": 6100
     },
     {
       "epoch": 25.0,
-      "eval_accuracy": 0.3917787266585941,
-      "eval_loss": 0.10184060782194138,
-      "eval_runtime": 6.1443,
-      "eval_samples_per_second": 392.883,
-      "eval_steps_per_second": 24.576,
+      "eval_accuracy": 0.3692051620800585,
+      "eval_loss": 0.3980158865451813,
+      "eval_runtime": 1.8534,
+      "eval_samples_per_second": 1302.488,
+      "eval_steps_per_second": 81.473,
       "step": 6156
     },
     {
       "epoch": 25.18,
       "learning_rate": 0.0002899142673469971,
-      "loss": 0.1049,
+      "loss": 0.3577,
       "step": 6200
     },
     {
       "epoch": 25.58,
       "learning_rate": 0.00028289530966636625,
-      "loss": 0.1038,
+      "loss": 0.3604,
       "step": 6300
     },
     {
       "epoch": 25.99,
       "learning_rate": 0.000275849860315853,
-      "loss": 0.1045,
+      "loss": 0.3628,
       "step": 6400
     },
     {
       "epoch": 26.0,
-      "eval_accuracy": 0.3920181723827126,
-      "eval_loss": 0.1005973368883133,
-      "eval_runtime": 6.0791,
-      "eval_samples_per_second": 397.098,
-      "eval_steps_per_second": 24.839,
+      "eval_accuracy": 0.36994782686580535,
+      "eval_loss": 0.39616090059280396,
+      "eval_runtime": 1.7695,
+      "eval_samples_per_second": 1364.248,
+      "eval_steps_per_second": 85.336,
       "step": 6402
     },
     {
       "epoch": 26.4,
       "learning_rate": 0.0002687835932264908,
-      "loss": 0.1008,
+      "loss": 0.3498,
       "step": 6500
     },
     {
       "epoch": 26.8,
       "learning_rate": 0.0002617021990945197,
-      "loss": 0.1022,
+      "loss": 0.3587,
       "step": 6600
     },
     {
       "epoch": 27.0,
-      "eval_accuracy": 0.39193259102124056,
-      "eval_loss": 0.10028840601444244,
-      "eval_runtime": 6.0776,
-      "eval_samples_per_second": 397.197,
-      "eval_steps_per_second": 24.845,
+      "eval_accuracy": 0.37061958395844063,
+      "eval_loss": 0.3926030397415161,
+      "eval_runtime": 1.8394,
+      "eval_samples_per_second": 1312.376,
+      "eval_steps_per_second": 82.091,
       "step": 6648
     },
     {
       "epoch": 27.21,
       "learning_rate": 0.0002546113807984821,
-      "loss": 0.1011,
+      "loss": 0.3534,
       "step": 6700
     },
     {
       "epoch": 27.61,
       "learning_rate": 0.00024751684880650884,
-      "loss": 0.0995,
+      "loss": 0.3492,
       "step": 6800
     },
     {
       "epoch": 28.0,
-      "eval_accuracy": 0.39223394666642386,
-      "eval_loss": 0.09900100529193878,
-      "eval_runtime": 6.1395,
-      "eval_samples_per_second": 393.191,
-      "eval_steps_per_second": 24.595,
+      "eval_accuracy": 0.37057106816841695,
+      "eval_loss": 0.39055171608924866,
+      "eval_runtime": 1.8305,
+      "eval_samples_per_second": 1318.747,
+      "eval_steps_per_second": 82.49,
       "step": 6895
     },
     {
       "epoch": 28.02,
       "learning_rate": 0.00024042431657749118,
-      "loss": 0.1009,
+      "loss": 0.3534,
       "step": 6900
     },
     {
       "epoch": 28.43,
       "learning_rate": 0.0002333394959598461,
-      "loss": 0.0977,
+      "loss": 0.3418,
       "step": 7000
     },
     {
       "epoch": 28.83,
       "learning_rate": 0.00022626809259157726,
-      "loss": 0.0989,
+      "loss": 0.3461,
       "step": 7100
     },
     {
       "epoch": 29.0,
-      "eval_accuracy": 0.39221664830612635,
-      "eval_loss": 0.09937664866447449,
-      "eval_runtime": 6.1179,
-      "eval_samples_per_second": 394.581,
-      "eval_steps_per_second": 24.682,
+      "eval_accuracy": 0.3706457078453764,
+      "eval_loss": 0.3932913541793823,
+      "eval_runtime": 1.7675,
+      "eval_samples_per_second": 1365.756,
+      "eval_steps_per_second": 85.43,
       "step": 7141
     },
     {
       "epoch": 29.24,
       "learning_rate": 0.00021921580130533828,
-      "loss": 0.0977,
+      "loss": 0.3412,
       "step": 7200
     },
     {
       "epoch": 29.64,
       "learning_rate": 0.0002121883015421973,
-      "loss": 0.0975,
+      "loss": 0.3363,
       "step": 7300
     },
     {
       "epoch": 30.0,
-      "eval_accuracy": 0.3923249906679898,
-      "eval_loss": 0.09780476242303848,
-      "eval_runtime": 6.0619,
-      "eval_samples_per_second": 398.224,
-      "eval_steps_per_second": 24.91,
+      "eval_accuracy": 0.3706942236354001,
+      "eval_loss": 0.39353010058403015,
+      "eval_runtime": 1.8169,
+      "eval_samples_per_second": 1328.629,
+      "eval_steps_per_second": 83.108,
       "step": 7387
     },
     {
       "epoch": 30.05,
       "learning_rate": 0.00020519125277779733,
-      "loss": 0.0964,
+      "loss": 0.3422,
       "step": 7400
     },
     {
       "epoch": 30.46,
       "learning_rate": 0.00019823028996459485,
-      "loss": 0.0941,
+      "loss": 0.3356,
       "step": 7500
     },
     {
       "epoch": 30.86,
       "learning_rate": 0.00019131101899384867,
-      "loss": 0.0956,
+      "loss": 0.3337,
       "step": 7600
     },
     {
       "epoch": 31.0,
-      "eval_accuracy": 0.39216930542531203,
-      "eval_loss": 0.0990176796913147,
-      "eval_runtime": 6.1212,
-      "eval_samples_per_second": 394.37,
-      "eval_steps_per_second": 24.669,
+      "eval_accuracy": 0.3702277256544034,
+      "eval_loss": 0.3950214684009552,
+      "eval_runtime": 1.7977,
+      "eval_samples_per_second": 1342.849,
+      "eval_steps_per_second": 83.998,
       "step": 7633
     },
     {
-      "epoch": 31.27,
-      "learning_rate": 0.00018443901218101152,
-      "loss": 0.0942,
-      "step": 7700
-    },
-    {
-      "epoch": 31.68,
-      "learning_rate": 0.00017761980377816285,
-      "loss": 0.0931,
-      "step": 7800
-    },
-    {
-      "epoch": 32.0,
-      "eval_accuracy": 0.39236231870863186,
-      "eval_loss": 0.09768786281347275,
-      "eval_runtime": 6.0866,
-      "eval_samples_per_second": 396.608,
-      "eval_steps_per_second": 24.809,
-      "step": 7880
-    },
-    {
-      "epoch": 32.08,
-      "learning_rate": 0.00017085888551709338,
-      "loss": 0.0945,
-      "step": 7900
-    },
-    {
-      "epoch": 32.49,
-      "learning_rate": 0.00016416170218663446,
-      "loss": 0.0916,
-      "step": 8000
-    },
-    {
-      "epoch": 32.89,
-      "learning_rate": 0.0001575336472477909,
-      "loss": 0.0926,
-      "step": 8100
-    },
-    {
-      "epoch": 33.0,
-      "eval_accuracy": 0.3922312153463769,
-      "eval_loss": 0.09789934009313583,
-      "eval_runtime": 6.1446,
-      "eval_samples_per_second": 392.866,
-      "eval_steps_per_second": 24.574,
-      "step": 8126
-    },
-    {
-      "epoch": 33.3,
-      "learning_rate": 0.0001509800584902108,
-      "loss": 0.0912,
-      "step": 8200
-    },
-    {
-      "epoch": 33.71,
-      "learning_rate": 0.00014450621373348819,
-      "loss": 0.0908,
-      "step": 8300
-    },
-    {
-      "epoch": 34.0,
-      "eval_accuracy": 0.3923732439888198,
-      "eval_loss": 0.09784437716007233,
-      "eval_runtime": 6.1319,
-      "eval_samples_per_second": 393.678,
-      "eval_steps_per_second": 24.625,
-      "step": 8372
-    },
-    {
-      "epoch": 34.11,
-      "learning_rate": 0.0001381173265767623,
-      "loss": 0.0905,
-      "step": 8400
-    },
-    {
-      "epoch": 34.52,
-      "learning_rate": 0.00013181854220003567,
-      "loss": 0.0895,
-      "step": 8500
-    },
-    {
-      "epoch": 34.92,
-      "learning_rate": 0.00012561493322059202,
-      "loss": 0.0896,
-      "step": 8600
-    },
-    {
-      "epoch": 35.0,
-      "eval_accuracy": 0.3924324225898377,
-      "eval_loss": 0.09740225225687027,
-      "eval_runtime": 6.1438,
-      "eval_samples_per_second": 392.916,
-      "eval_steps_per_second": 24.578,
-      "step": 8618
-    },
-    {
-      "epoch": 35.33,
-      "learning_rate": 0.00011951149560785166,
-      "loss": 0.0892,
-      "step": 8700
-    },
-    {
-      "epoch": 35.74,
-      "learning_rate": 0.0001135131446599535,
-      "loss": 0.0883,
-      "step": 8800
-    },
-    {
-      "epoch": 36.0,
-      "eval_accuracy": 0.3924469896300882,
-      "eval_loss": 0.09739205241203308,
-      "eval_runtime": 6.1304,
-      "eval_samples_per_second": 393.776,
-      "eval_steps_per_second": 24.631,
-      "step": 8865
-    },
-    {
-      "epoch": 36.14,
-      "learning_rate": 0.00010762471104530472,
-      "loss": 0.0877,
-      "step": 8900
-    },
-    {
-      "epoch": 36.55,
-      "learning_rate": 0.00010185093691228534,
-      "loss": 0.0866,
-      "step": 9000
-    },
-    {
-      "epoch": 36.95,
-      "learning_rate": 9.619647207024071e-05,
-      "loss": 0.0878,
-      "step": 9100
-    },
-    {
-      "epoch": 37.0,
-      "eval_accuracy": 0.3924242286296967,
-      "eval_loss": 0.0978410392999649,
-      "eval_runtime": 6.1135,
-      "eval_samples_per_second": 394.865,
-      "eval_steps_per_second": 24.699,
-      "step": 9111
-    },
-    {
-      "epoch": 37.36,
-      "learning_rate": 9.0665870244838e-05,
-      "loss": 0.0859,
-      "step": 9200
-    },
-    {
-      "epoch": 37.77,
-      "learning_rate": 8.526358541080173e-05,
-      "loss": 0.0865,
-      "step": 9300
-    },
-    {
-      "epoch": 38.0,
-      "eval_accuracy": 0.3925216457113723,
-      "eval_loss": 0.09711020439863205,
-      "eval_runtime": 6.1232,
-      "eval_samples_per_second": 394.237,
-      "eval_steps_per_second": 24.66,
-      "step": 9357
-    },
-    {
-      "epoch": 38.17,
-      "learning_rate": 7.999396820498208e-05,
-      "loss": 0.0857,
-      "step": 9400
-    },
-    {
-      "epoch": 38.58,
-      "learning_rate": 7.486126242264468e-05,
-      "loss": 0.085,
-      "step": 9500
-    },
-    {
-      "epoch": 38.98,
-      "learning_rate": 6.986960159980326e-05,
-      "loss": 0.0855,
-      "step": 9600
-    },
-    {
-      "epoch": 39.0,
-      "eval_accuracy": 0.3925689885921866,
-      "eval_loss": 0.09748318791389465,
-      "eval_runtime": 6.1096,
-      "eval_samples_per_second": 395.116,
-      "eval_steps_per_second": 24.715,
-      "step": 9603
-    },
-    {
-      "epoch": 39.39,
-      "learning_rate": 6.502300568434777e-05,
-      "loss": 0.0838,
-      "step": 9700
-    },
-    {
-      "epoch": 39.8,
-      "learning_rate": 6.0325377798648745e-05,
-      "loss": 0.0847,
-      "step": 9800
-    },
-    {
-      "epoch": 40.0,
-      "eval_accuracy": 0.3925434962717481,
-      "eval_loss": 0.0970187559723854,
-      "eval_runtime": 6.2108,
-      "eval_samples_per_second": 388.675,
-      "eval_steps_per_second": 24.312,
-      "step": 9850
-    },
-    {
-      "epoch": 40.2,
-      "learning_rate": 5.578050109624511e-05,
-      "loss": 0.0827,
-      "step": 9900
-    },
-    {
-      "epoch": 40.61,
-      "learning_rate": 5.139203571514672e-05,
-      "loss": 0.0835,
-      "step": 10000
-    },
-    {
-      "epoch": 41.0,
-      "eval_accuracy": 0.3925671677121553,
-      "eval_loss": 0.09726004302501678,
-      "eval_runtime": 6.0996,
-      "eval_samples_per_second": 395.763,
-      "eval_steps_per_second": 24.756,
-      "step": 10096
-    },
-    {
-      "epoch": 41.02,
-      "learning_rate": 4.716351583020542e-05,
-      "loss": 0.0831,
-      "step": 10100
-    },
-    {
-      "epoch": 41.42,
-      "learning_rate": 4.3098346806928315e-05,
-      "loss": 0.0821,
-      "step": 10200
-    },
-    {
-      "epoch": 41.83,
-      "learning_rate": 3.919980245902524e-05,
-      "loss": 0.0825,
-      "step": 10300
-    },
-    {
-      "epoch": 42.0,
-      "eval_accuracy": 0.392557152871983,
-      "eval_loss": 0.09756684303283691,
-      "eval_runtime": 6.1355,
-      "eval_samples_per_second": 393.449,
-      "eval_steps_per_second": 24.611,
-      "step": 10342
-    },
-    {
-      "epoch": 42.23,
-      "learning_rate": 3.5471022411899925e-05,
-      "loss": 0.0825,
-      "step": 10400
-    },
-    {
-      "epoch": 42.64,
-      "learning_rate": 3.1915009574206264e-05,
-      "loss": 0.0814,
-      "step": 10500
-    },
-    {
-      "epoch": 43.0,
-      "eval_accuracy": 0.3925817347524058,
-      "eval_loss": 0.09748771786689758,
-      "eval_runtime": 6.1103,
-      "eval_samples_per_second": 395.068,
-      "eval_steps_per_second": 24.712,
-      "step": 10588
-    },
-    {
-      "epoch": 43.0,
-      "step": 10588,
-      "total_flos": 3084220612961280.0,
-      "train_loss": 0.2331255912600619,
-      "train_runtime": 2746.0841,
-      "train_samples_per_second": 143.441,
-      "train_steps_per_second": 4.479
+      "epoch": 31.0,
+      "step": 7633,
+      "total_flos": 564540063409152.0,
+      "train_loss": 0.7912082670869731,
+      "train_runtime": 664.7096,
+      "train_samples_per_second": 592.59,
+      "train_steps_per_second": 18.504
     }
   ],
+  "logging_steps": 100,
   "max_steps": 12300,
   "num_train_epochs": 50,
-  "total_flos": 3084220612961280.0,
+  "save_steps": 500,
+  "total_flos": 564540063409152.0,
   "trial_name": null,
   "trial_params": null
 }

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c0db7b0d2599b04668617c424a3337e9f602ee02316377c982902ccf0e0c3346
-size 3643
+oid sha256:7436c8435d06f3e361bd8cf0f410e726279c9cc132dc8d75ad123af026548abd
+size 4155