Bengali model CER=0.0046
Browse files- bengali/shobdo_bengali.pth +3 -0
- bengali/shobdo_bengali.py +39 -0
- bengali/vocab.json +303 -0
bengali/shobdo_bengali.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:48c81b5727ec94e8ffc80a5b66433cf2ab4f0b25b74438f4017ae65dcafaad5c
|
| 3 |
+
size 18697093
|
bengali/shobdo_bengali.py
ADDED
|
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
|
| 2 |
+
import torch.nn as nn
|
| 3 |
+
import torch.nn.functional as F
|
| 4 |
+
|
| 5 |
+
class ConvBnRelu(nn.Module):
|
| 6 |
+
def __init__(self,i,o,k=3,s=1,p=1):
|
| 7 |
+
super().__init__()
|
| 8 |
+
self.b=nn.Sequential(nn.Conv2d(i,o,k,s,p,bias=False),nn.BatchNorm2d(o),nn.ReLU(inplace=True))
|
| 9 |
+
def forward(self,x): return self.b(x)
|
| 10 |
+
|
| 11 |
+
class LightCNN(nn.Module):
|
| 12 |
+
def __init__(self):
|
| 13 |
+
super().__init__()
|
| 14 |
+
self.b1=nn.Sequential(ConvBnRelu(1,32),ConvBnRelu(32,32),nn.MaxPool2d(2,2))
|
| 15 |
+
self.b2=nn.Sequential(ConvBnRelu(32,64),ConvBnRelu(64,64),nn.MaxPool2d(2,2))
|
| 16 |
+
self.b3=nn.Sequential(ConvBnRelu(64,128),ConvBnRelu(128,128),nn.MaxPool2d((2,1)))
|
| 17 |
+
self.b4=nn.Sequential(ConvBnRelu(128,256),ConvBnRelu(256,256),nn.MaxPool2d((2,1)))
|
| 18 |
+
self.b5=nn.Sequential(ConvBnRelu(256,256),ConvBnRelu(256,256))
|
| 19 |
+
self.pool=nn.AdaptiveAvgPool2d((1,None))
|
| 20 |
+
def forward(self,x):
|
| 21 |
+
for b in [self.b1,self.b2,self.b3,self.b4,self.b5]: x=b(x)
|
| 22 |
+
return self.pool(x).squeeze(2)
|
| 23 |
+
|
| 24 |
+
class BiLSTM(nn.Module):
|
| 25 |
+
def __init__(self,i,h,o):
|
| 26 |
+
super().__init__()
|
| 27 |
+
self.rnn=nn.LSTM(i,h,bidirectional=True,batch_first=True)
|
| 28 |
+
self.fc=nn.Linear(h*2,o)
|
| 29 |
+
def forward(self,x): o,_=self.rnn(x); return self.fc(o)
|
| 30 |
+
|
| 31 |
+
class Model(nn.Module):
|
| 32 |
+
def __init__(self,input_channel,output_channel,hidden_size,num_class):
|
| 33 |
+
super().__init__()
|
| 34 |
+
self.cnn=LightCNN()
|
| 35 |
+
self.rnn=nn.Sequential(BiLSTM(256,hidden_size,hidden_size),BiLSTM(hidden_size,hidden_size,num_class))
|
| 36 |
+
def forward(self,x):
|
| 37 |
+
f=self.cnn(x).permute(0,2,1)
|
| 38 |
+
o=self.rnn(f).permute(1,0,2)
|
| 39 |
+
return F.log_softmax(o,dim=2)
|
bengali/vocab.json
ADDED
|
@@ -0,0 +1,303 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"charset": "অআইঈউঊঋঌএঐওঔকখগঘঙচছজঝঞটঠডঢণতথদধনপফবভমযরলশষসহ়ৎািীুূৃেৈোৌ্ংঃঁ০১২৩৪৫৬৭৮৯ ।,.?!()-–:'%/\\ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789",
|
| 3 |
+
"num_classes": 148,
|
| 4 |
+
"char2idx": {
|
| 5 |
+
"অ": 1,
|
| 6 |
+
"আ": 2,
|
| 7 |
+
"ই": 3,
|
| 8 |
+
"ঈ": 4,
|
| 9 |
+
"উ": 5,
|
| 10 |
+
"ঊ": 6,
|
| 11 |
+
"ঋ": 7,
|
| 12 |
+
"ঌ": 8,
|
| 13 |
+
"এ": 9,
|
| 14 |
+
"ঐ": 10,
|
| 15 |
+
"ও": 11,
|
| 16 |
+
"ঔ": 12,
|
| 17 |
+
"ক": 13,
|
| 18 |
+
"খ": 14,
|
| 19 |
+
"গ": 15,
|
| 20 |
+
"ঘ": 16,
|
| 21 |
+
"ঙ": 17,
|
| 22 |
+
"চ": 18,
|
| 23 |
+
"ছ": 19,
|
| 24 |
+
"জ": 20,
|
| 25 |
+
"ঝ": 21,
|
| 26 |
+
"ঞ": 22,
|
| 27 |
+
"ট": 23,
|
| 28 |
+
"ঠ": 24,
|
| 29 |
+
"ড": 25,
|
| 30 |
+
"ঢ": 26,
|
| 31 |
+
"ণ": 27,
|
| 32 |
+
"ত": 28,
|
| 33 |
+
"থ": 29,
|
| 34 |
+
"দ": 30,
|
| 35 |
+
"ধ": 31,
|
| 36 |
+
"ন": 32,
|
| 37 |
+
"প": 33,
|
| 38 |
+
"ফ": 34,
|
| 39 |
+
"ব": 35,
|
| 40 |
+
"ভ": 36,
|
| 41 |
+
"ম": 37,
|
| 42 |
+
"য": 38,
|
| 43 |
+
"র": 39,
|
| 44 |
+
"ল": 40,
|
| 45 |
+
"শ": 41,
|
| 46 |
+
"ষ": 42,
|
| 47 |
+
"স": 43,
|
| 48 |
+
"হ": 44,
|
| 49 |
+
"়": 45,
|
| 50 |
+
"ৎ": 46,
|
| 51 |
+
"া": 47,
|
| 52 |
+
"ি": 48,
|
| 53 |
+
"ী": 49,
|
| 54 |
+
"ু": 50,
|
| 55 |
+
"ূ": 51,
|
| 56 |
+
"ৃ": 52,
|
| 57 |
+
"ে": 53,
|
| 58 |
+
"ৈ": 54,
|
| 59 |
+
"ো": 55,
|
| 60 |
+
"ৌ": 56,
|
| 61 |
+
"্": 57,
|
| 62 |
+
"ং": 58,
|
| 63 |
+
"ঃ": 59,
|
| 64 |
+
"ঁ": 60,
|
| 65 |
+
"০": 61,
|
| 66 |
+
"১": 62,
|
| 67 |
+
"২": 63,
|
| 68 |
+
"৩": 64,
|
| 69 |
+
"৪": 65,
|
| 70 |
+
"৫": 66,
|
| 71 |
+
"৬": 67,
|
| 72 |
+
"৭": 68,
|
| 73 |
+
"৮": 69,
|
| 74 |
+
"৯": 70,
|
| 75 |
+
" ": 71,
|
| 76 |
+
"।": 72,
|
| 77 |
+
",": 73,
|
| 78 |
+
".": 74,
|
| 79 |
+
"?": 75,
|
| 80 |
+
"!": 76,
|
| 81 |
+
"(": 77,
|
| 82 |
+
")": 78,
|
| 83 |
+
"-": 79,
|
| 84 |
+
"–": 80,
|
| 85 |
+
":": 81,
|
| 86 |
+
"'": 82,
|
| 87 |
+
"%": 83,
|
| 88 |
+
"/": 84,
|
| 89 |
+
"\\": 85,
|
| 90 |
+
"A": 86,
|
| 91 |
+
"B": 87,
|
| 92 |
+
"C": 88,
|
| 93 |
+
"D": 89,
|
| 94 |
+
"E": 90,
|
| 95 |
+
"F": 91,
|
| 96 |
+
"G": 92,
|
| 97 |
+
"H": 93,
|
| 98 |
+
"I": 94,
|
| 99 |
+
"J": 95,
|
| 100 |
+
"K": 96,
|
| 101 |
+
"L": 97,
|
| 102 |
+
"M": 98,
|
| 103 |
+
"N": 99,
|
| 104 |
+
"O": 100,
|
| 105 |
+
"P": 101,
|
| 106 |
+
"Q": 102,
|
| 107 |
+
"R": 103,
|
| 108 |
+
"S": 104,
|
| 109 |
+
"T": 105,
|
| 110 |
+
"U": 106,
|
| 111 |
+
"V": 107,
|
| 112 |
+
"W": 108,
|
| 113 |
+
"X": 109,
|
| 114 |
+
"Y": 110,
|
| 115 |
+
"Z": 111,
|
| 116 |
+
"a": 112,
|
| 117 |
+
"b": 113,
|
| 118 |
+
"c": 114,
|
| 119 |
+
"d": 115,
|
| 120 |
+
"e": 116,
|
| 121 |
+
"f": 117,
|
| 122 |
+
"g": 118,
|
| 123 |
+
"h": 119,
|
| 124 |
+
"i": 120,
|
| 125 |
+
"j": 121,
|
| 126 |
+
"k": 122,
|
| 127 |
+
"l": 123,
|
| 128 |
+
"m": 124,
|
| 129 |
+
"n": 125,
|
| 130 |
+
"o": 126,
|
| 131 |
+
"p": 127,
|
| 132 |
+
"q": 128,
|
| 133 |
+
"r": 129,
|
| 134 |
+
"s": 130,
|
| 135 |
+
"t": 131,
|
| 136 |
+
"u": 132,
|
| 137 |
+
"v": 133,
|
| 138 |
+
"w": 134,
|
| 139 |
+
"x": 135,
|
| 140 |
+
"y": 136,
|
| 141 |
+
"z": 137,
|
| 142 |
+
"0": 138,
|
| 143 |
+
"1": 139,
|
| 144 |
+
"2": 140,
|
| 145 |
+
"3": 141,
|
| 146 |
+
"4": 142,
|
| 147 |
+
"5": 143,
|
| 148 |
+
"6": 144,
|
| 149 |
+
"7": 145,
|
| 150 |
+
"8": 146,
|
| 151 |
+
"9": 147
|
| 152 |
+
},
|
| 153 |
+
"idx2char": {
|
| 154 |
+
"1": "অ",
|
| 155 |
+
"2": "আ",
|
| 156 |
+
"3": "ই",
|
| 157 |
+
"4": "ঈ",
|
| 158 |
+
"5": "উ",
|
| 159 |
+
"6": "ঊ",
|
| 160 |
+
"7": "ঋ",
|
| 161 |
+
"8": "ঌ",
|
| 162 |
+
"9": "এ",
|
| 163 |
+
"10": "ঐ",
|
| 164 |
+
"11": "ও",
|
| 165 |
+
"12": "ঔ",
|
| 166 |
+
"13": "ক",
|
| 167 |
+
"14": "খ",
|
| 168 |
+
"15": "গ",
|
| 169 |
+
"16": "ঘ",
|
| 170 |
+
"17": "ঙ",
|
| 171 |
+
"18": "চ",
|
| 172 |
+
"19": "ছ",
|
| 173 |
+
"20": "জ",
|
| 174 |
+
"21": "ঝ",
|
| 175 |
+
"22": "ঞ",
|
| 176 |
+
"23": "ট",
|
| 177 |
+
"24": "ঠ",
|
| 178 |
+
"25": "ড",
|
| 179 |
+
"26": "ঢ",
|
| 180 |
+
"27": "ণ",
|
| 181 |
+
"28": "ত",
|
| 182 |
+
"29": "থ",
|
| 183 |
+
"30": "দ",
|
| 184 |
+
"31": "ধ",
|
| 185 |
+
"32": "ন",
|
| 186 |
+
"33": "প",
|
| 187 |
+
"34": "ফ",
|
| 188 |
+
"35": "ব",
|
| 189 |
+
"36": "ভ",
|
| 190 |
+
"37": "ম",
|
| 191 |
+
"38": "য",
|
| 192 |
+
"39": "র",
|
| 193 |
+
"40": "ল",
|
| 194 |
+
"41": "শ",
|
| 195 |
+
"42": "ষ",
|
| 196 |
+
"43": "স",
|
| 197 |
+
"44": "হ",
|
| 198 |
+
"45": "়",
|
| 199 |
+
"46": "ৎ",
|
| 200 |
+
"47": "া",
|
| 201 |
+
"48": "ি",
|
| 202 |
+
"49": "ী",
|
| 203 |
+
"50": "ু",
|
| 204 |
+
"51": "ূ",
|
| 205 |
+
"52": "ৃ",
|
| 206 |
+
"53": "ে",
|
| 207 |
+
"54": "ৈ",
|
| 208 |
+
"55": "ো",
|
| 209 |
+
"56": "ৌ",
|
| 210 |
+
"57": "্",
|
| 211 |
+
"58": "ং",
|
| 212 |
+
"59": "ঃ",
|
| 213 |
+
"60": "ঁ",
|
| 214 |
+
"61": "০",
|
| 215 |
+
"62": "১",
|
| 216 |
+
"63": "২",
|
| 217 |
+
"64": "৩",
|
| 218 |
+
"65": "৪",
|
| 219 |
+
"66": "৫",
|
| 220 |
+
"67": "৬",
|
| 221 |
+
"68": "৭",
|
| 222 |
+
"69": "৮",
|
| 223 |
+
"70": "৯",
|
| 224 |
+
"71": " ",
|
| 225 |
+
"72": "।",
|
| 226 |
+
"73": ",",
|
| 227 |
+
"74": ".",
|
| 228 |
+
"75": "?",
|
| 229 |
+
"76": "!",
|
| 230 |
+
"77": "(",
|
| 231 |
+
"78": ")",
|
| 232 |
+
"79": "-",
|
| 233 |
+
"80": "–",
|
| 234 |
+
"81": ":",
|
| 235 |
+
"82": "'",
|
| 236 |
+
"83": "%",
|
| 237 |
+
"84": "/",
|
| 238 |
+
"85": "\\",
|
| 239 |
+
"86": "A",
|
| 240 |
+
"87": "B",
|
| 241 |
+
"88": "C",
|
| 242 |
+
"89": "D",
|
| 243 |
+
"90": "E",
|
| 244 |
+
"91": "F",
|
| 245 |
+
"92": "G",
|
| 246 |
+
"93": "H",
|
| 247 |
+
"94": "I",
|
| 248 |
+
"95": "J",
|
| 249 |
+
"96": "K",
|
| 250 |
+
"97": "L",
|
| 251 |
+
"98": "M",
|
| 252 |
+
"99": "N",
|
| 253 |
+
"100": "O",
|
| 254 |
+
"101": "P",
|
| 255 |
+
"102": "Q",
|
| 256 |
+
"103": "R",
|
| 257 |
+
"104": "S",
|
| 258 |
+
"105": "T",
|
| 259 |
+
"106": "U",
|
| 260 |
+
"107": "V",
|
| 261 |
+
"108": "W",
|
| 262 |
+
"109": "X",
|
| 263 |
+
"110": "Y",
|
| 264 |
+
"111": "Z",
|
| 265 |
+
"112": "a",
|
| 266 |
+
"113": "b",
|
| 267 |
+
"114": "c",
|
| 268 |
+
"115": "d",
|
| 269 |
+
"116": "e",
|
| 270 |
+
"117": "f",
|
| 271 |
+
"118": "g",
|
| 272 |
+
"119": "h",
|
| 273 |
+
"120": "i",
|
| 274 |
+
"121": "j",
|
| 275 |
+
"122": "k",
|
| 276 |
+
"123": "l",
|
| 277 |
+
"124": "m",
|
| 278 |
+
"125": "n",
|
| 279 |
+
"126": "o",
|
| 280 |
+
"127": "p",
|
| 281 |
+
"128": "q",
|
| 282 |
+
"129": "r",
|
| 283 |
+
"130": "s",
|
| 284 |
+
"131": "t",
|
| 285 |
+
"132": "u",
|
| 286 |
+
"133": "v",
|
| 287 |
+
"134": "w",
|
| 288 |
+
"135": "x",
|
| 289 |
+
"136": "y",
|
| 290 |
+
"137": "z",
|
| 291 |
+
"138": "0",
|
| 292 |
+
"139": "1",
|
| 293 |
+
"140": "2",
|
| 294 |
+
"141": "3",
|
| 295 |
+
"142": "4",
|
| 296 |
+
"143": "5",
|
| 297 |
+
"144": "6",
|
| 298 |
+
"145": "7",
|
| 299 |
+
"146": "8",
|
| 300 |
+
"147": "9",
|
| 301 |
+
"0": ""
|
| 302 |
+
}
|
| 303 |
+
}
|