Devatri commited on
Commit
c19cc10
·
verified ·
1 Parent(s): d4bff30

Upload C4/results.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. C4/results.json +358 -0
C4/results.json ADDED
@@ -0,0 +1,358 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "config": {
3
+ "name": "C4",
4
+ "pos_encoding": "sinusoidal",
5
+ "attention": "mha",
6
+ "norm": "rmsnorm",
7
+ "tokenization": "bpe",
8
+ "d_model": 256,
9
+ "n_heads": 8,
10
+ "n_kv_heads": 2,
11
+ "n_enc_layers": 4,
12
+ "n_dec_layers": 4,
13
+ "d_ff": 1024,
14
+ "dropout": 0.1,
15
+ "src_vocab_size": 4096,
16
+ "tgt_vocab_size": 8192,
17
+ "max_src_len": 768,
18
+ "max_tgt_len": 384,
19
+ "src_patch_size": 16,
20
+ "tgt_patch_size": 4,
21
+ "local_layers": 2,
22
+ "d_local": 64,
23
+ "local_heads": 4,
24
+ "blt_max_src_bytes": 10240,
25
+ "blt_max_tgt_bytes": 1288,
26
+ "batch_size": 16,
27
+ "accum_steps": 2,
28
+ "lr": 0.0003,
29
+ "warmup_steps": 500,
30
+ "epochs": 30,
31
+ "label_smoothing": 0.1,
32
+ "grad_clip": 1.0,
33
+ "seed": 42,
34
+ "amp": true,
35
+ "max_raw_cipher_chars": 10240,
36
+ "val_frac": 0.1,
37
+ "test_frac": 0.1
38
+ },
39
+ "n_params": 12606976,
40
+ "train_seconds": 695.0826947689056,
41
+ "best_val_loss": 6.000316072570575,
42
+ "peak_mem_gib": 4.5991668701171875,
43
+ "test": {
44
+ "loss": 5.996847112997218,
45
+ "ppl": 402.1588311690615,
46
+ "bit_accuracy": 0.06788805925749643,
47
+ "sequence_accuracy": 0.0,
48
+ "levenshtein": 410.08613445378154,
49
+ "levenshtein_norm": 0.7285616137385754,
50
+ "bleu": 1.330158568728205,
51
+ "rouge1": 0.24314428193372983,
52
+ "rouge2": 0.038257264493512466,
53
+ "rougeL": 0.1815693863580148,
54
+ "greedy_decode_sec": 74.43447661399841
55
+ },
56
+ "history": [
57
+ {
58
+ "epoch": 1,
59
+ "train_loss": 8.648933363926876,
60
+ "val_loss": 8.180379795472001,
61
+ "val_ppl": 3570.2103530900054,
62
+ "epoch_sec": 33.00962829589844,
63
+ "tokens_per_sec": 13895.733568650763,
64
+ "lr": 7.139999999999999e-05,
65
+ "peak_mem_gib": 4.425947189331055
66
+ },
67
+ {
68
+ "epoch": 2,
69
+ "train_loss": 7.8140762959283965,
70
+ "val_loss": 7.622608592146534,
71
+ "val_ppl": 2043.88684897409,
72
+ "epoch_sec": 22.273264408111572,
73
+ "tokens_per_sec": 20580.997540398963,
74
+ "lr": 0.00014279999999999997,
75
+ "peak_mem_gib": 4.425947189331055
76
+ },
77
+ {
78
+ "epoch": 3,
79
+ "train_loss": 7.5169678060543585,
80
+ "val_loss": 7.367284296662801,
81
+ "val_ppl": 1583.3280898205364,
82
+ "epoch_sec": 22.215421438217163,
83
+ "tokens_per_sec": 20647.008713097377,
84
+ "lr": 0.00021419999999999998,
85
+ "peak_mem_gib": 4.5991668701171875
86
+ },
87
+ {
88
+ "epoch": 4,
89
+ "train_loss": 7.21295226789657,
90
+ "val_loss": 7.0524443802046966,
91
+ "val_ppl": 1155.680214843428,
92
+ "epoch_sec": 22.350778341293335,
93
+ "tokens_per_sec": 20518.524813640233,
94
+ "lr": 0.00028559999999999995,
95
+ "peak_mem_gib": 4.5991668701171875
96
+ },
97
+ {
98
+ "epoch": 5,
99
+ "train_loss": 6.944078296193421,
100
+ "val_loss": 6.86633686208749,
101
+ "val_ppl": 959.4276052320746,
102
+ "epoch_sec": 22.32414698600769,
103
+ "tokens_per_sec": 20553.61847812562,
104
+ "lr": 0.00027500954910846334,
105
+ "peak_mem_gib": 4.5991668701171875
106
+ },
107
+ {
108
+ "epoch": 6,
109
+ "train_loss": 6.757989900233638,
110
+ "val_loss": 6.739306645994319,
111
+ "val_ppl": 844.9746661281619,
112
+ "epoch_sec": 22.23685121536255,
113
+ "tokens_per_sec": 20614.699246775806,
114
+ "lr": 0.0002510482226267169,
115
+ "peak_mem_gib": 4.5991668701171875
116
+ },
117
+ {
118
+ "epoch": 7,
119
+ "train_loss": 6.596686356497476,
120
+ "val_loss": 6.619671081196324,
121
+ "val_ppl": 749.6984666306424,
122
+ "epoch_sec": 22.264714002609253,
123
+ "tokens_per_sec": 20595.638459414335,
124
+ "lr": 0.00023242549051998872,
125
+ "peak_mem_gib": 4.5991668701171875
126
+ },
127
+ {
128
+ "epoch": 8,
129
+ "train_loss": 6.450608731696479,
130
+ "val_loss": 6.531970972757969,
131
+ "val_ppl": 686.7504448648347,
132
+ "epoch_sec": 22.34483504295349,
133
+ "tokens_per_sec": 20531.09808678908,
134
+ "lr": 0.00021741413836966813,
135
+ "peak_mem_gib": 4.5991668701171875
136
+ },
137
+ {
138
+ "epoch": 9,
139
+ "train_loss": 6.3212245667740445,
140
+ "val_loss": 6.446918405600694,
141
+ "val_ppl": 630.7555620281845,
142
+ "epoch_sec": 23.592225790023804,
143
+ "tokens_per_sec": 19443.311711350707,
144
+ "lr": 0.0002049800154226969,
145
+ "peak_mem_gib": 4.5991668701171875
146
+ },
147
+ {
148
+ "epoch": 10,
149
+ "train_loss": 6.206023735359433,
150
+ "val_loss": 6.382216231857776,
151
+ "val_ppl": 591.2365740779665,
152
+ "epoch_sec": 22.396533012390137,
153
+ "tokens_per_sec": 20464.863903106678,
154
+ "lr": 0.0001944611170656493,
155
+ "peak_mem_gib": 4.5991668701171875
156
+ },
157
+ {
158
+ "epoch": 11,
159
+ "train_loss": 6.099300534509986,
160
+ "val_loss": 6.324348792983374,
161
+ "val_ppl": 557.9943255048041,
162
+ "epoch_sec": 22.246440887451172,
163
+ "tokens_per_sec": 20609.81360207731,
164
+ "lr": 0.0001854114001850042,
165
+ "peak_mem_gib": 4.5991668701171875
166
+ },
167
+ {
168
+ "epoch": 12,
169
+ "train_loss": 6.000730438072846,
170
+ "val_loss": 6.2653687279543275,
171
+ "val_ppl": 526.035514202802,
172
+ "epoch_sec": 22.38110327720642,
173
+ "tokens_per_sec": 20491.483119470442,
174
+ "lr": 0.00017751790062418157,
175
+ "peak_mem_gib": 4.5991668701171875
176
+ },
177
+ {
178
+ "epoch": 13,
179
+ "train_loss": 5.910986713524583,
180
+ "val_loss": 6.226990904314323,
181
+ "val_ppl": 506.22989513284784,
182
+ "epoch_sec": 22.37939167022705,
183
+ "tokens_per_sec": 20478.21525951918,
184
+ "lr": 0.00017055368216583073,
185
+ "peak_mem_gib": 4.5991668701171875
186
+ },
187
+ {
188
+ "epoch": 14,
189
+ "train_loss": 5.827012618067848,
190
+ "val_loss": 6.190348197253964,
191
+ "val_ppl": 488.0160024659686,
192
+ "epoch_sec": 22.246175050735474,
193
+ "tokens_per_sec": 20611.678140366006,
194
+ "lr": 0.00016434964046729364,
195
+ "peak_mem_gib": 4.5991668701171875
196
+ },
197
+ {
198
+ "epoch": 15,
199
+ "train_loss": 5.7479622453402675,
200
+ "val_loss": 6.158046929975107,
201
+ "val_ppl": 472.5043390069415,
202
+ "epoch_sec": 22.213014602661133,
203
+ "tokens_per_sec": 20642.943256565744,
204
+ "lr": 0.00015877683720748892,
205
+ "peak_mem_gib": 4.5991668701171875
206
+ },
207
+ {
208
+ "epoch": 16,
209
+ "train_loss": 5.677382439770528,
210
+ "val_loss": 6.134259328739738,
211
+ "val_ppl": 461.3972237645913,
212
+ "epoch_sec": 22.277552843093872,
213
+ "tokens_per_sec": 20576.990804540157,
214
+ "lr": 0.00015373501156702268,
215
+ "peak_mem_gib": 4.5991668701171875
216
+ },
217
+ {
218
+ "epoch": 17,
219
+ "train_loss": 5.609386029141837,
220
+ "val_loss": 6.10338859759755,
221
+ "val_ppl": 447.37116534670406,
222
+ "epoch_sec": 22.396621465682983,
223
+ "tokens_per_sec": 20466.926259496944,
224
+ "lr": 0.0001491448684815029,
225
+ "peak_mem_gib": 4.5991668701171875
226
+ },
227
+ {
228
+ "epoch": 18,
229
+ "train_loss": 5.5442525434562855,
230
+ "val_loss": 6.086284340994333,
231
+ "val_ppl": 439.78428309268133,
232
+ "epoch_sec": 22.174579858779907,
233
+ "tokens_per_sec": 20681.24866043045,
234
+ "lr": 0.0001449427589131121,
235
+ "peak_mem_gib": 4.5991668701171875
236
+ },
237
+ {
238
+ "epoch": 19,
239
+ "train_loss": 5.48383063302523,
240
+ "val_loss": 6.070839611150451,
241
+ "val_ppl": 433.04411765045137,
242
+ "epoch_sec": 22.02217721939087,
243
+ "tokens_per_sec": 20825.05264720984,
244
+ "lr": 0.00014107692200099125,
245
+ "peak_mem_gib": 4.5991668701171875
246
+ },
247
+ {
248
+ "epoch": 20,
249
+ "train_loss": 5.425856893251126,
250
+ "val_loss": 6.055804795285843,
251
+ "val_ppl": 426.5820785898982,
252
+ "epoch_sec": 22.400903940200806,
253
+ "tokens_per_sec": 20474.75410922585,
254
+ "lr": 0.00013750477455423167,
255
+ "peak_mem_gib": 4.5991668701171875
256
+ },
257
+ {
258
+ "epoch": 21,
259
+ "train_loss": 5.368190612471962,
260
+ "val_loss": 6.043559207690604,
261
+ "val_ppl": 421.3901841577893,
262
+ "epoch_sec": 22.224619388580322,
263
+ "tokens_per_sec": 20620.78058513266,
264
+ "lr": 0.0001341909195182463,
265
+ "peak_mem_gib": 4.5991668701171875
266
+ },
267
+ {
268
+ "epoch": 22,
269
+ "train_loss": 5.316519506229067,
270
+ "val_loss": 6.033526472574033,
271
+ "val_ppl": 417.18362499082815,
272
+ "epoch_sec": 22.36732816696167,
273
+ "tokens_per_sec": 20506.517210111,
274
+ "lr": 0.00013110565838010918,
275
+ "peak_mem_gib": 4.5991668701171875
276
+ },
277
+ {
278
+ "epoch": 23,
279
+ "train_loss": 5.265128353519914,
280
+ "val_loss": 6.031917415134938,
281
+ "val_ppl": 416.5128923439596,
282
+ "epoch_sec": 22.35172724723816,
283
+ "tokens_per_sec": 20520.830221596196,
284
+ "lr": 0.00012822386342782138,
285
+ "peak_mem_gib": 4.5991668701171875
286
+ },
287
+ {
288
+ "epoch": 24,
289
+ "train_loss": 5.2141449803611355,
290
+ "val_loss": 6.017647656819868,
291
+ "val_ppl": 410.611559470417,
292
+ "epoch_sec": 22.436333179473877,
293
+ "tokens_per_sec": 20431.85917828083,
294
+ "lr": 0.00012552411131335845,
295
+ "peak_mem_gib": 4.5991668701171875
296
+ },
297
+ {
298
+ "epoch": 25,
299
+ "train_loss": 5.167684464306953,
300
+ "val_loss": 6.012545519536085,
301
+ "val_ppl": 408.5218983258443,
302
+ "epoch_sec": 22.355741024017334,
303
+ "tokens_per_sec": 20513.969968935904,
304
+ "lr": 0.00012298800925361817,
305
+ "peak_mem_gib": 4.5991668701171875
306
+ },
307
+ {
308
+ "epoch": 26,
309
+ "train_loss": 5.119421488741977,
310
+ "val_loss": 6.013634947558981,
311
+ "val_ppl": 408.9671960457127,
312
+ "epoch_sec": 22.393715381622314,
313
+ "tokens_per_sec": 20478.200789152743,
314
+ "lr": 0.00012059966521579404,
315
+ "peak_mem_gib": 4.5991668701171875
316
+ },
317
+ {
318
+ "epoch": 27,
319
+ "train_loss": 5.075400812160661,
320
+ "val_loss": 6.018528793521887,
321
+ "val_ppl": 410.9735238323626,
322
+ "epoch_sec": 22.195430517196655,
323
+ "tokens_per_sec": 20661.36989975002,
324
+ "lr": 0.00011834526708278771,
325
+ "peak_mem_gib": 4.5991668701171875
326
+ },
327
+ {
328
+ "epoch": 28,
329
+ "train_loss": 5.029786435859967,
330
+ "val_loss": 6.011288831377613,
331
+ "val_ppl": 408.00883614084995,
332
+ "epoch_sec": 22.182212352752686,
333
+ "tokens_per_sec": 20673.636727812318,
334
+ "lr": 0.00011621274525999436,
335
+ "peak_mem_gib": 4.5991668701171875
336
+ },
337
+ {
338
+ "epoch": 29,
339
+ "train_loss": 4.9870534554046335,
340
+ "val_loss": 6.000316072570575,
341
+ "val_ppl": 403.5563264223071,
342
+ "epoch_sec": 22.22054147720337,
343
+ "tokens_per_sec": 20650.621879343696,
344
+ "lr": 0.00011419149985036504,
345
+ "peak_mem_gib": 4.5991668701171875
346
+ },
347
+ {
348
+ "epoch": 30,
349
+ "train_loss": 4.941548499995137,
350
+ "val_loss": 6.010218999569479,
351
+ "val_ppl": 407.57256871793993,
352
+ "epoch_sec": 22.192119359970093,
353
+ "tokens_per_sec": 20659.270643027845,
354
+ "lr": 0.00011227217828476796,
355
+ "peak_mem_gib": 4.5991668701171875
356
+ }
357
+ ]
358
+ }