Upload 7 files

config.json

@@ -0,0 +1,27 @@
+{
+  "architectures": [
+    "RetroGPTForCausalLM"
+  ],
+  "model_type": "retrogpt",
+  "vocab_size": 439,
+  "d_model": 512,
+  "n_heads": 8,
+  "n_layers": 6,
+  "num_hidden_layers": 6,
+  "hidden_size": 512,
+  "num_attention_heads": 8,
+  "max_seq_len": 256,
+  "dropout": 0.1,
+  "sep_id": 3,
+  "pad_token_id": 0,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "auto_map": {
+    "AutoConfig": "modeling_retrogpt.RetroGPTConfig",
+    "AutoModelForCausalLM": "modeling_retrogpt.RetroGPTForCausalLM",
+    "AutoTokenizer": [
+      "modeling_retrogpt.RetroGPTTokenizer",
+      null
+    ]
+  }
+}

configuration_retrogpt.py

@@ -0,0 +1,19 @@
+from transformers import PretrainedConfig
+
+class RetroGPTConfig(PretrainedConfig):
+    model_type = "retrogpt"
+    def __init__(self, d_model=512, n_heads=8, n_layers=6, max_seq_len=256, 
+                 dropout=0.1, sep_id=3, vocab_size=None, **kwargs):
+        super().__init__(**kwargs)
+        self.d_model = d_model
+        self.n_heads = n_heads
+        self.n_layers = n_layers
+        self.max_seq_len = max_seq_len
+        self.dropout = dropout
+        self.sep_id = sep_id
+        self.vocab_size = vocab_size
+        
+        # Standard aliases
+        self.num_hidden_layers = n_layers
+        self.hidden_size = d_model
+        self.num_attention_heads = n_heads

modeling_retrogpt.py

@@ -0,0 +1,153 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import json
+import os
+import re
+from transformers import PreTrainedModel, PretrainedConfig, PreTrainedTokenizer
+from transformers.modeling_outputs import CausalLMOutputWithPast
+
+class RetroGPTConfig(PretrainedConfig):
+    model_type = "retrogpt"
+    def __init__(self, d_model=512, n_heads=8, n_layers=6, max_seq_len=256, 
+                 dropout=0.1, sep_id=3, vocab_size=None, 
+                 pad_token_id=0, bos_token_id=1, eos_token_id=2, **kwargs):
+        # FIX: Pass token IDs directly to the parent class
+        super().__init__(pad_token_id=pad_token_id, bos_token_id=bos_token_id, eos_token_id=eos_token_id, **kwargs)
+        self.d_model = d_model
+        self.n_heads = n_heads
+        self.n_layers = n_layers
+        self.max_seq_len = max_seq_len
+        self.dropout = dropout
+        self.sep_id = sep_id
+        self.vocab_size = vocab_size
+        
+        self.num_hidden_layers = n_layers
+        self.hidden_size = d_model
+        self.num_attention_heads = n_heads
+
+class RMSNorm(nn.Module):
+    def __init__(self, dim, eps=1e-6):
+        super().__init__()
+        self.eps = eps
+        self.weight = nn.Parameter(torch.ones(dim))
+    def forward(self, x):
+        variance = x.pow(2).mean(-1, keepdim=True)
+        return self.weight * (x * torch.rsqrt(variance + self.eps))
+
+class SwiGLU(nn.Module):
+    def __init__(self, dim):
+        super().__init__()
+        hidden_dim = int(8 * dim / 3)
+        self.w1 = nn.Linear(dim, hidden_dim, bias=False)
+        self.w2 = nn.Linear(hidden_dim, dim, bias=False)
+        self.w3 = nn.Linear(dim, hidden_dim, bias=False)
+    def forward(self, x): return self.w2(F.silu(self.w1(x)) * self.w3(x))
+
+class Attention(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.n_heads = config.n_heads
+        self.head_dim = config.d_model // config.n_heads
+        self.wq, self.wk, self.wv, self.wo = [nn.Linear(config.d_model, config.d_model, bias=False) for _ in range(4)]
+        self.dropout_p = config.dropout
+    def forward(self, x, mask=None):
+        B, T, C = x.size()
+        q = self.wq(x).view(B, T, self.n_heads, self.head_dim).transpose(1, 2)
+        k = self.wk(x).view(B, T, self.n_heads, self.head_dim).transpose(1, 2)
+        v = self.wv(x).view(B, T, self.n_heads, self.head_dim).transpose(1, 2)
+        attn_mask = (mask == 1.0) if mask is not None else None
+        y = F.scaled_dot_product_attention(q, k, v, attn_mask=attn_mask, dropout_p=self.dropout_p if self.training else 0.0)
+        return self.wo(y.transpose(1, 2).contiguous().view(B, T, C))
+
+class Block(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.norm1, self.attn, self.norm2, self.mlp = RMSNorm(config.d_model), Attention(config), RMSNorm(config.d_model), SwiGLU(config.d_model)
+    def forward(self, x, mask=None):
+        x = x + self.attn(self.norm1(x), mask)
+        x = x + self.mlp(self.norm2(x))
+        return x
+
+class RetroGPTForCausalLM(PreTrainedModel):
+    config_class = RetroGPTConfig
+    
+    def __init__(self, config):
+        super().__init__(config)
+        self.token_emb = nn.Embedding(config.vocab_size, config.d_model)
+        self.pos_emb = nn.Embedding(config.max_seq_len, config.d_model)
+        self.drop = nn.Dropout(config.dropout)
+        self.blocks = nn.ModuleList([Block(config) for _ in range(config.n_layers)])
+        self.norm = RMSNorm(config.d_model)
+        self.head = nn.Linear(config.d_model, config.vocab_size, bias=False)
+        self.post_init()
+
+    def get_input_embeddings(self): return self.token_emb
+    def set_input_embeddings(self, value): self.token_emb = value
+    def get_output_embeddings(self): return self.head
+    def set_output_embeddings(self, new_embeddings): self.head = new_embeddings
+    def tie_weights(self, *args, **kwargs): self.head.weight = self.token_emb.weight
+
+    @property
+    def _tied_weights_keys(self): return ["head.weight"]
+        
+    # FIX: Added attention_mask parameter
+    def forward(self, input_ids, attention_mask=None, labels=None, **kwargs):
+        idx = input_ids
+        B, T = idx.size()
+        x = self.drop(self.token_emb(idx) + self.pos_emb(torch.arange(T, device=idx.device)))
+        
+        # Base causal mask
+        mask = torch.tril(torch.ones(T, T, device=idx.device)).unsqueeze(0).expand(B, T, T).clone()
+        
+        # FIX: Integrate Hugging Face's padding mask safely
+        if attention_mask is not None:
+            mask = mask * attention_mask.unsqueeze(1).expand(B, T, T)
+
+        # Apply <sep> token logic
+        if hasattr(self.config, "sep_id"):
+            sep_masks = (idx == self.config.sep_id)
+            if sep_masks.any():
+                sep_idx = sep_masks.int().argmax(dim=1)
+                has_sep = sep_masks.any(dim=1)
+                for b in range(B):
+                    if has_sep[b]: mask[b, :sep_idx[b]+1, :sep_idx[b]+1] = 1.0
+                    
+        for block in self.blocks: x = block(x, mask.unsqueeze(1))
+        logits = self.head(self.norm(x))
+        loss = F.cross_entropy(logits.view(-1, logits.size(-1)), labels.view(-1), ignore_index=-100) if labels is not None else None
+        return CausalLMOutputWithPast(loss=loss, logits=logits)
+
+    def prepare_inputs_for_generation(self, input_ids, **kwargs): 
+        # FIX: Ensure attention mask is passed during generation loop
+        attention_mask = kwargs.get("attention_mask", None)
+        return {"input_ids": input_ids, "attention_mask": attention_mask}
+
+class RetroGPTTokenizer(PreTrainedTokenizer):
+    vocab_files_names = {"vocab_file": "vocab.json"}
+    model_input_names = ["input_ids", "attention_mask"]
+    
+    def __init__(self, vocab_file=None, bos_token="<s>", eos_token="</s>", sep_token="<sep>", pad_token="<pad>", **kwargs):
+        if vocab_file is None: vocab_file = os.path.join(os.path.dirname(__file__), "vocab.json")
+        if os.path.exists(vocab_file):
+            with open(vocab_file, "r") as f: self.stoi = json.load(f)
+        else: self.stoi = {}
+        self.itos = {int(v): k for k, v in self.stoi.items()}
+        self.pattern = re.compile(r"(\[[^\]]+]|Br?|Cl?|N|O|S|P|F|I|b|c|n|o|s|p|\(|\)|\.|=|#|-|\+|\\|\/|:|~|@|\?|>|\*|\$|\%[0-9]{2}|[0-9])")
+        super().__init__(bos_token=bos_token, eos_token=eos_token, sep_token=sep_token, pad_token=pad_token, **kwargs)
+        
+        # FIX: Explicitly set special IDs so tokenizer.pad() doesn't fail
+        self.pad_token_id = self.stoi.get(pad_token, 0)
+        self.bos_token_id = self.stoi.get(bos_token, 1)
+        self.eos_token_id = self.stoi.get(eos_token, 2)
+        
+    @property
+    def vocab_size(self): return len(self.stoi)
+    def get_vocab(self): return {k: int(v) for k, v in self.stoi.items()}
+    def _tokenize(self, text): return self.pattern.findall(text)
+    def _convert_token_to_id(self, token): return self.stoi.get(token, self.pad_token_id)
+    def _convert_id_to_token(self, index): return self.itos.get(index, self.pad_token)
+    def save_vocabulary(self, save_directory, filename_prefix=None):
+        vocab_file = os.path.join(save_directory, (filename_prefix + "-" if filename_prefix else "") + "vocab.json")
+        with open(vocab_file, "w") as f: json.dump(self.stoi, f, indent=2)
+        return (vocab_file,)

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7869b691cd478f54bf63a451031d6d17a04c901ccad9cb9788bd9854250e1738
+size 76953330

tokenization_retrogpt.py

@@ -0,0 +1,54 @@
+import json
+import os
+import re
+from transformers import PreTrainedTokenizer
+
+class RetroGPTTokenizer(PreTrainedTokenizer):
+    vocab_files_names = {"vocab_file": "vocab.json"}
+    model_input_names = ["input_ids", "attention_mask"]
+    
+    def __init__(self, vocab_file=None, bos_token="<s>", eos_token="</s>", sep_token="<sep>", pad_token="<pad>", **kwargs):
+        if vocab_file is None:
+             # Look for vocab.json in the same directory as the script
+             vocab_file = os.path.join(os.path.dirname(__file__), "vocab.json")
+             
+        if os.path.exists(vocab_file):
+            with open(vocab_file, "r") as f:
+                self.stoi = json.load(f)
+        else:
+            self.stoi = {}
+            
+        self.itos = {int(v): k for k, v in self.stoi.items()}
+        
+        # Exact pattern from app.py
+        self.pattern = re.compile(r"(\[[^\]]+]|Br?|Cl?|N|O|S|P|F|I|b|c|n|o|s|p|\(|\)|\.|=|#|-|\+|\\|\/|:|~|@|\?|>|\*|\$|\%[0-9]{2}|[0-9])")
+        
+        super().__init__(
+            bos_token=bos_token,
+            eos_token=eos_token,
+            sep_token=sep_token,
+            pad_token=pad_token,
+            **kwargs
+        )
+
+    @property
+    def vocab_size(self):
+        return len(self.stoi)
+
+    def get_vocab(self):
+        return {k: int(v) for k, v in self.stoi.items()}
+
+    def _tokenize(self, text):
+        return self.pattern.findall(text)
+
+    def _convert_token_to_id(self, token):
+        return self.stoi.get(token, self.stoi.get(self.pad_token))
+
+    def _convert_id_to_token(self, index):
+        return self.itos.get(index, self.pad_token)
+
+    def save_vocabulary(self, save_directory, filename_prefix=None):
+        vocab_file = os.path.join(save_directory, (filename_prefix + "-" if filename_prefix else "") + "vocab.json")
+        with open(vocab_file, "w") as f:
+            json.dump(self.stoi, f, indent=2)
+        return (vocab_file,)

tokenizer_config.json

@@ -0,0 +1,9 @@
+{
+  "bos_token": "<s>",
+  "eos_token": "</s>",
+  "unk_token": "<pad>",
+  "pad_token": "<pad>",
+  "sep_token": "<sep>",
+  "model_max_length": 256,
+  "tokenizer_class": "RetroGPTTokenizer"
+}

vocab.json

@@ -0,0 +1,441 @@
+{
+  "<pad>": 0,
+  "<s>": 1,
+  "</s>": 2,
+  "<sep>": 3,
+  "#": 4,
+  "%10": 5,
+  "%11": 6,
+  "%12": 7,
+  "%13": 8,
+  "%14": 9,
+  "%15": 10,
+  "%16": 11,
+  "%17": 12,
+  "%18": 13,
+  "%19": 14,
+  "%20": 15,
+  "%21": 16,
+  "%22": 17,
+  "%23": 18,
+  "(": 19,
+  ")": 20,
+  "-": 21,
+  ".": 22,
+  "/": 23,
+  "1": 24,
+  "2": 25,
+  "3": 26,
+  "4": 27,
+  "5": 28,
+  "6": 29,
+  "7": 30,
+  "8": 31,
+  "9": 32,
+  "=": 33,
+  "B": 34,
+  "Br": 35,
+  "C": 36,
+  "Cl": 37,
+  "F": 38,
+  "I": 39,
+  "N": 40,
+  "O": 41,
+  "P": 42,
+  "S": 43,
+  "[11CH3]": 44,
+  "[11CH4]": 45,
+  "[123I-]": 46,
+  "[125I]": 47,
+  "[131I]": 48,
+  "[13CH4]": 49,
+  "[13NH3]": 50,
+  "[13c]": 51,
+  "[14C]": 52,
+  "[15OH2]": 53,
+  "[18FH]": 54,
+  "[18F]": 55,
+  "[1HH]": 56,
+  "[2H-]": 57,
+  "[2H]": 58,
+  "[32P]": 59,
+  "[3H]": 60,
+  "[99Tc+4]": 61,
+  "[99Tc]": 62,
+  "[Ag+2]": 63,
+  "[Ag+3]": 64,
+  "[Ag+]": 65,
+  "[Ag-]": 66,
+  "[Ag]": 67,
+  "[Al+2]": 68,
+  "[Al+3]": 69,
+  "[Al+]": 70,
+  "[Al-3]": 71,
+  "[Al-]": 72,
+  "[AlH-]": 73,
+  "[AlH2+]": 74,
+  "[AlH2-]": 75,
+  "[AlH3]": 76,
+  "[AlH4-]": 77,
+  "[AlH]": 78,
+  "[Al]": 79,
+  "[Ar]": 80,
+  "[As+3]": 81,
+  "[As+]": 82,
+  "[As-]": 83,
+  "[AsH+]": 84,
+  "[AsH2]": 85,
+  "[AsH3]": 86,
+  "[AsH4+]": 87,
+  "[AsH]": 88,
+  "[As]": 89,
+  "[Au+3]": 90,
+  "[Au-]": 91,
+  "[Au]": 92,
+  "[B+2]": 93,
+  "[B+3]": 94,
+  "[B+]": 95,
+  "[B-]": 96,
+  "[BH-]": 97,
+  "[BH2-]": 98,
+  "[BH3-]": 99,
+  "[BH4-]": 100,
+  "[BH]": 101,
+  "[B]": 102,
+  "[Ba+2]": 103,
+  "[Ba]": 104,
+  "[Be+2]": 105,
+  "[Bi+2]": 106,
+  "[Bi+3]": 107,
+  "[BiH3]": 108,
+  "[Bi]": 109,
+  "[Br+2]": 110,
+  "[Br+]": 111,
+  "[Br-]": 112,
+  "[BrH+]": 113,
+  "[Br]": 114,
+  "[C+4]": 115,
+  "[C+]": 116,
+  "[C-]": 117,
+  "[C@@H]": 118,
+  "[C@@]": 119,
+  "[C@H]": 120,
+  "[C@]": 121,
+  "[CH+]": 122,
+  "[CH-]": 123,
+  "[CH2+]": 124,
+  "[CH2-]": 125,
+  "[CH3+3]": 126,
+  "[CH3+]": 127,
+  "[CH3-]": 128,
+  "[CH]": 129,
+  "[C]": 130,
+  "[Ca+2]": 131,
+  "[Ca]": 132,
+  "[Cd+2]": 133,
+  "[Cd+]": 134,
+  "[Cd]": 135,
+  "[Ce+2]": 136,
+  "[Ce+3]": 137,
+  "[Ce+4]": 138,
+  "[Ce]": 139,
+  "[Cl+2]": 140,
+  "[Cl+3]": 141,
+  "[Cl+]": 142,
+  "[Cl-]": 143,
+  "[ClH+]": 144,
+  "[Cl]": 145,
+  "[Co+2]": 146,
+  "[Co+3]": 147,
+  "[Co]": 148,
+  "[Cr+2]": 149,
+  "[Cr+3]": 150,
+  "[Cr+4]": 151,
+  "[Cr+6]": 152,
+  "[Cr-]": 153,
+  "[Cr]": 154,
+  "[Cs+]": 155,
+  "[Cs]": 156,
+  "[Cu+2]": 157,
+  "[Cu+3]": 158,
+  "[Cu+4]": 159,
+  "[Cu+]": 160,
+  "[Cu-2]": 161,
+  "[Cu-]": 162,
+  "[Cu]": 163,
+  "[Dy+3]": 164,
+  "[Dy]": 165,
+  "[Er+3]": 166,
+  "[Eu+3]": 167,
+  "[Eu]": 168,
+  "[F-]": 169,
+  "[F]": 170,
+  "[Fe+2]": 171,
+  "[Fe+3]": 172,
+  "[Fe+4]": 173,
+  "[Fe+6]": 174,
+  "[Fe+]": 175,
+  "[Fe-3]": 176,
+  "[Fe-4]": 177,
+  "[Fe]": 178,
+  "[Ga+2]": 179,
+  "[Ga+3]": 180,
+  "[Ga+]": 181,
+  "[Ga]": 182,
+  "[Gd+3]": 183,
+  "[Gd]": 184,
+  "[Ge+2]": 185,
+  "[Ge+3]": 186,
+  "[GeH2]": 187,
+  "[GeH3]": 188,
+  "[GeH]": 189,
+  "[Ge]": 190,
+  "[H+]": 191,
+  "[H-]": 192,
+  "[H]": 193,
+  "[HeH]": 194,
+  "[He]": 195,
+  "[Hf+2]": 196,
+  "[Hf+3]": 197,
+  "[Hf+4]": 198,
+  "[Hf]": 199,
+  "[Hg+2]": 200,
+  "[Hg+]": 201,
+  "[Hg]": 202,
+  "[I+2]": 203,
+  "[I+3]": 204,
+  "[I+]": 205,
+  "[I-]": 206,
+  "[IH+]": 207,
+  "[IH2+]": 208,
+  "[IH]": 209,
+  "[I]": 210,
+  "[In+3]": 211,
+  "[In+]": 212,
+  "[In]": 213,
+  "[Ir+3]": 214,
+  "[Ir]": 215,
+  "[K+]": 216,
+  "[K]": 217,
+  "[La+3]": 218,
+  "[La]": 219,
+  "[Li+]": 220,
+  "[Li]": 221,
+  "[Mg+2]": 222,
+  "[Mg+]": 223,
+  "[Mg]": 224,
+  "[Mn+2]": 225,
+  "[Mn+3]": 226,
+  "[Mn+4]": 227,
+  "[Mn]": 228,
+  "[Mo+2]": 229,
+  "[Mo+4]": 230,
+  "[Mo]": 231,
+  "[N+3]": 232,
+  "[N+]": 233,
+  "[N-]": 234,
+  "[N@+]": 235,
+  "[N@@+]": 236,
+  "[N@@H+]": 237,
+  "[N@@]": 238,
+  "[N@]": 239,
+  "[NH+]": 240,
+  "[NH-]": 241,
+  "[NH2+]": 242,
+  "[NH2-]": 243,
+  "[NH3+]": 244,
+  "[NH4+]": 245,
+  "[N]": 246,
+  "[Na+]": 247,
+  "[Na]": 248,
+  "[Nb+3]": 249,
+  "[Nb+4]": 250,
+  "[Nb+5]": 251,
+  "[Nb]": 252,
+  "[Nd+3]": 253,
+  "[Nd+]": 254,
+  "[Nd]": 255,
+  "[Ni+2]": 256,
+  "[Ni+4]": 257,
+  "[Ni]": 258,
+  "[O+]": 259,
+  "[O-2]": 260,
+  "[O-]": 261,
+  "[OH+]": 262,
+  "[OH-]": 263,
+  "[OH2+]": 264,
+  "[OH3+]": 265,
+  "[OH]": 266,
+  "[O]": 267,
+  "[Os]": 268,
+  "[P+2]": 269,
+  "[P+3]": 270,
+  "[P+5]": 271,
+  "[P+]": 272,
+  "[P-3]": 273,
+  "[P-]": 274,
+  "[P@@]": 275,
+  "[P@]": 276,
+  "[PH+]": 277,
+  "[PH2+]": 278,
+  "[PH2-]": 279,
+  "[PH2]": 280,
+  "[PH3+]": 281,
+  "[PH3]": 282,
+  "[PH4+]": 283,
+  "[PH4]": 284,
+  "[PH5]": 285,
+  "[PH]": 286,
+  "[P]": 287,
+  "[Pb+2]": 288,
+  "[Pb+3]": 289,
+  "[Pb+4]": 290,
+  "[Pb]": 291,
+  "[Pd+2]": 292,
+  "[Pd-2]": 293,
+  "[Pd]": 294,
+  "[Pr+3]": 295,
+  "[Pr]": 296,
+  "[Pt+2]": 297,
+  "[Pt+]": 298,
+  "[Pt-2]": 299,
+  "[Pt-]": 300,
+  "[Pt]": 301,
+  "[Rb+]": 302,
+  "[Re+5]": 303,
+  "[Re+]": 304,
+  "[Re]": 305,
+  "[Rh+2]": 306,
+  "[Rh+3]": 307,
+  "[Rh]": 308,
+  "[Ru+2]": 309,
+  "[Ru+3]": 310,
+  "[Ru+]": 311,
+  "[Ru-]": 312,
+  "[Ru]": 313,
+  "[S+]": 314,
+  "[S-2]": 315,
+  "[S-]": 316,
+  "[S@@]": 317,
+  "[S@]": 318,
+  "[SH+]": 319,
+  "[SH-]": 320,
+  "[SH2+]": 321,
+  "[SH2]": 322,
+  "[SH3+]": 323,
+  "[SH]": 324,
+  "[S]": 325,
+  "[Sb+2]": 326,
+  "[Sb+3]": 327,
+  "[Sb+5]": 328,
+  "[Sb+]": 329,
+  "[Sb-]": 330,
+  "[Sb]": 331,
+  "[Sc+3]": 332,
+  "[Sc]": 333,
+  "[Se+]": 334,
+  "[Se-2]": 335,
+  "[Se-]": 336,
+  "[SeH+]": 337,
+  "[SeH-]": 338,
+  "[SeH2]": 339,
+  "[SeH]": 340,
+  "[Se]": 341,
+  "[Si+2]": 342,
+  "[Si+4]": 343,
+  "[Si+]": 344,
+  "[Si-]": 345,
+  "[Si@@H]": 346,
+  "[Si@@]": 347,
+  "[Si@H]": 348,
+  "[Si@]": 349,
+  "[SiH-]": 350,
+  "[SiH2]": 351,
+  "[SiH3]": 352,
+  "[SiH4]": 353,
+  "[SiH]": 354,
+  "[Si]": 355,
+  "[Sm+2]": 356,
+  "[Sm+3]": 357,
+  "[Sm]": 358,
+  "[Sn+2]": 359,
+  "[Sn+3]": 360,
+  "[Sn+4]": 361,
+  "[Sn+6]": 362,
+  "[Sn+]": 363,
+  "[SnH2]": 364,
+  "[SnH3]": 365,
+  "[SnH4]": 366,
+  "[SnH]": 367,
+  "[Sn]": 368,
+  "[Sr+2]": 369,
+  "[Sr]": 370,
+  "[Ta+2]": 371,
+  "[Ta+5]": 372,
+  "[TaH3]": 373,
+  "[Ta]": 374,
+  "[Tb+3]": 375,
+  "[Tb]": 376,
+  "[Tc+6]": 377,
+  "[Tc]": 378,
+  "[TeH2]": 379,
+  "[TeH]": 380,
+  "[Te]": 381,
+  "[Ti+2]": 382,
+  "[Ti+3]": 383,
+  "[Ti+4]": 384,
+  "[Ti+]": 385,
+  "[Ti]": 386,
+  "[Tl+2]": 387,
+  "[Tl+3]": 388,
+  "[Tl+]": 389,
+  "[Tl]": 390,
+  "[U+2]": 391,
+  "[U+4]": 392,
+  "[U+6]": 393,
+  "[U]": 394,
+  "[V+2]": 395,
+  "[V+3]": 396,
+  "[V+4]": 397,
+  "[V+5]": 398,
+  "[V+]": 399,
+  "[V]": 400,
+  "[W+6]": 401,
+  "[W]": 402,
+  "[Xe]": 403,
+  "[Y+3]": 404,
+  "[Y]": 405,
+  "[Yb+2]": 406,
+  "[Yb+3]": 407,
+  "[Yb]": 408,
+  "[Zn+2]": 409,
+  "[Zn+]": 410,
+  "[Zn]": 411,
+  "[Zr+2]": 412,
+  "[Zr+3]": 413,
+  "[Zr+4]": 414,
+  "[Zr+]": 415,
+  "[Zr]": 416,
+  "[c+]": 417,
+  "[c-]": 418,
+  "[cH+]": 419,
+  "[cH-]": 420,
+  "[n+]": 421,
+  "[n-]": 422,
+  "[nH+]": 423,
+  "[nH]": 424,
+  "[o+]": 425,
+  "[pH]": 426,
+  "[s+]": 427,
+  "[sH+]": 428,
+  "[se+]": 429,
+  "[se]": 430,
+  "[siH]": 431,
+  "[te]": 432,
+  "b": 433,
+  "c": 434,
+  "n": 435,
+  "o": 436,
+  "p": 437,
+  "s": 438
+}