appvoid commited on
Commit
749cff3
·
verified ·
1 Parent(s): f70aa6d

Qwen90M step 40000; complete resumable checkpoint

Browse files
Files changed (6) hide show
  1. complete.json +5 -5
  2. coverage.json +49 -49
  3. latest.json +1 -1
  4. model.safetensors +1 -1
  5. shard_scheduler.json +244 -244
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 35000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "257552ddb070147bf74ac81d1972455eac1b5b94298e7820362135ec7d22bf73",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "bdb02210299e7645fb1fcfe319164a683c0fab01d39f95c7ebced4aab5172f8e",
12
  "training_config.json": "07f331dbc4eba0ea92b45763c59c20af4874302e74a327f3562642d69645fed2",
13
  "performance.json": "b95a1befd575f8d9a02b43fd0cc41c54243f5297127b93d2656e0292ba15e77e",
14
- "shard_scheduler.json": "5414324baf2ef485c06d3f1e0e9a90e7ae02cc5270937d0e6356509393fcac57",
15
- "coverage.json": "1384d2eab2585bf506f3fd13f050eebeb4ea9493ce600a26e3511d35382fe101",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 40000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "1f54edcf173e14eaa220773d6904ca3660dca9333bec4cca80621f41c11ae4ba",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "ddb86ed5b8deb7d4e788e14749c412fa48be7f3cefd33dadef82d54b10488b02",
12
  "training_config.json": "07f331dbc4eba0ea92b45763c59c20af4874302e74a327f3562642d69645fed2",
13
  "performance.json": "b95a1befd575f8d9a02b43fd0cc41c54243f5297127b93d2656e0292ba15e77e",
14
+ "shard_scheduler.json": "dd35db67300912a92de8d12b39d40858b1385a9f601def538c82db10aecb48b5",
15
+ "coverage.json": "496553c7362fb891e4a087907a1ee7bc57bd29676eca9ccbf872f730f0c086c9",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 35000,
3
  "stats": {
4
- "climbmix/blocks": 1260000,
5
- "climbmix/targets": 2105041906,
6
- "climbmix/input_tokens": 2105175390,
7
- "ultra_qa/blocks": 472500,
8
- "ultra_qa/targets": 770911552,
9
- "ultra_qa/input_tokens": 770913097,
10
- "cortex/blocks": 472500,
11
- "cortex/targets": 17983452,
12
- "cortex/input_tokens": 751155694,
13
- "ultra_style/blocks": 472500,
14
- "ultra_style/targets": 746810228,
15
- "ultra_style/input_tokens": 746849525,
16
- "openmath/blocks": 315000,
17
- "openmath/targets": 486799702,
18
- "openmath/input_tokens": 486962255,
19
- "rewrite6/blocks": 157500,
20
- "rewrite6/targets": 258227030,
21
- "rewrite6/input_tokens": 258510170
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 172,
26
- "row": 2609,
27
  "start": {
28
- "file": 8,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 25,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
- "offset": 8192,
37
- "documents": 707121
38
  },
39
  "rewrite6": {
40
- "page": 48,
41
- "row": 1325,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 196608
47
  }
48
  },
49
- "epoch": 1,
50
  "offset": 0,
51
- "documents": 403337
52
  },
53
  "openmath": {
54
- "page": 107,
55
- "row": 3947,
56
  "start": {
57
  "file": 1,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 1,
61
- "row": 622
62
  }
63
  },
64
  "epoch": 0,
65
- "offset": 0,
66
- "documents": 442219
67
  },
68
  "ultra_qa": {
69
- "page": 46,
70
- "row": 2796,
71
  "start": {
72
  "file": 0,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
- "row_group": 2,
76
- "row": 59140
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
- "documents": 191212
82
  },
83
  "ultra_style": {
84
- "page": 71,
85
- "row": 1500,
86
  "start": {
87
  "file": 0,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 3,
91
- "row": 35792
92
  }
93
  },
94
  "epoch": 0,
95
- "offset": 2048,
96
- "documents": 292316
97
  },
98
  "cortex": {
99
- "page": 9162,
100
- "row": 44,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 1172780
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 40000,
3
  "stats": {
4
+ "climbmix/blocks": 1440000,
5
+ "climbmix/targets": 2405645996,
6
+ "climbmix/input_tokens": 2405798466,
7
+ "ultra_qa/blocks": 540000,
8
+ "ultra_qa/targets": 881103422,
9
+ "ultra_qa/input_tokens": 881105161,
10
+ "cortex/blocks": 540000,
11
+ "cortex/targets": 20545868,
12
+ "cortex/input_tokens": 858494555,
13
+ "ultra_style/blocks": 540000,
14
+ "ultra_style/targets": 853948670,
15
+ "ultra_style/input_tokens": 853993129,
16
+ "openmath/blocks": 360000,
17
+ "openmath/targets": 556374754,
18
+ "openmath/input_tokens": 556560449,
19
+ "rewrite6/blocks": 180000,
20
+ "rewrite6/targets": 295083056,
21
+ "rewrite6/input_tokens": 295406501
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 197,
26
+ "row": 1098,
27
  "start": {
28
+ "file": 9,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 43,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
+ "offset": 4096,
37
+ "documents": 808010
38
  },
39
  "rewrite6": {
40
+ "page": 12,
41
+ "row": 807,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 49152
47
  }
48
  },
49
+ "epoch": 2,
50
  "offset": 0,
51
+ "documents": 460767
52
  },
53
  "openmath": {
54
+ "page": 123,
55
+ "row": 1391,
56
  "start": {
57
  "file": 1,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 67,
61
+ "row": 158
62
  }
63
  },
64
  "epoch": 0,
65
+ "offset": 2048,
66
+ "documents": 505199
67
  },
68
  "ultra_qa": {
69
+ "page": 53,
70
+ "row": 1357,
71
  "start": {
72
  "file": 0,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
+ "row_group": 3,
76
+ "row": 23181
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
+ "documents": 218445
82
  },
83
  "ultra_style": {
84
+ "page": 81,
85
+ "row": 1172,
86
  "start": {
87
  "file": 0,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 3,
91
+ "row": 76752
92
  }
93
  },
94
  "epoch": 0,
95
+ "offset": 0,
96
+ "documents": 332948
97
  },
98
  "cortex": {
99
+ "page": 10471,
100
+ "row": 124,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 1340412
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 35000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "257552ddb070147bf74ac81d1972455eac1b5b94298e7820362135ec7d22bf73", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "bdb02210299e7645fb1fcfe319164a683c0fab01d39f95c7ebced4aab5172f8e", "training_config.json": "07f331dbc4eba0ea92b45763c59c20af4874302e74a327f3562642d69645fed2", "performance.json": "b95a1befd575f8d9a02b43fd0cc41c54243f5297127b93d2656e0292ba15e77e", "shard_scheduler.json": "5414324baf2ef485c06d3f1e0e9a90e7ae02cc5270937d0e6356509393fcac57", "coverage.json": "1384d2eab2585bf506f3fd13f050eebeb4ea9493ce600a26e3511d35382fe101", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000035000"}
 
1
+ {"step": 40000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "1f54edcf173e14eaa220773d6904ca3660dca9333bec4cca80621f41c11ae4ba", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "ddb86ed5b8deb7d4e788e14749c412fa48be7f3cefd33dadef82d54b10488b02", "training_config.json": "07f331dbc4eba0ea92b45763c59c20af4874302e74a327f3562642d69645fed2", "performance.json": "b95a1befd575f8d9a02b43fd0cc41c54243f5297127b93d2656e0292ba15e77e", "shard_scheduler.json": "dd35db67300912a92de8d12b39d40858b1385a9f601def538c82db10aecb48b5", "coverage.json": "496553c7362fb891e4a087907a1ee7bc57bd29676eca9ccbf872f730f0c086c9", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000040000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:257552ddb070147bf74ac81d1972455eac1b5b94298e7820362135ec7d22bf73
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1f54edcf173e14eaa220773d6904ca3660dca9333bec4cca80621f41c11ae4ba
3
  size 360622744
shard_scheduler.json CHANGED
@@ -22,8 +22,8 @@
22
  "tokens_per_second": 1062967.6168812835,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
- "required_token_positions_per_second": 75370.88600524074,
26
- "required_pages_per_second_upper_estimate": 0.006532957373170129
27
  },
28
  {
29
  "source": "rewrite6",
@@ -36,8 +36,8 @@
36
  "tokens_per_second": 1206672.263377472,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
- "required_token_positions_per_second": 9421.360750655092,
40
- "required_pages_per_second_upper_estimate": 0.003758321665332333
41
  },
42
  {
43
  "source": "openmath",
@@ -50,8 +50,8 @@
50
  "tokens_per_second": 410042.7581134458,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
- "required_token_positions_per_second": 18842.721501310185,
54
- "required_pages_per_second_upper_estimate": 0.004358931998755937
55
  },
56
  {
57
  "source": "ultra_qa",
@@ -64,8 +64,8 @@
64
  "tokens_per_second": 2163486.5624068812,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
- "required_token_positions_per_second": 28264.082251965276,
68
- "required_pages_per_second_upper_estimate": 0.0017211343189655782
69
  },
70
  {
71
  "source": "ultra_style",
@@ -78,8 +78,8 @@
78
  "tokens_per_second": 1546694.2737633225,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
- "required_token_positions_per_second": 28264.082251965276,
82
- "required_pages_per_second_upper_estimate": 0.0021758472917435424
83
  },
84
  {
85
  "source": "cortex",
@@ -92,8 +92,8 @@
92
  "tokens_per_second": 1015177.8395898147,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
  "estimated_pages_for_buffer": 22,
95
- "required_token_positions_per_second": 28264.082251965276,
96
- "required_pages_per_second_upper_estimate": 0.35094530777115207
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,410 +158,410 @@
158
  "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 688946,
161
- "producer_batch_seconds": 0.1257512091639228,
162
- "producer_p95_seconds": 0.3557953109993832,
163
- "gpu_update_seconds": 0.9782026443854395,
164
- "producer_batches_per_second": 7.952209816896882,
165
- "gpu_batches_per_second": 1.0222830675624015,
166
- "producer_headroom": 7.778872671596381,
167
  "sustainable": true,
168
- "buffer_seconds": 60.648563951897245,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 84,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 387927.91475817043,
174
- "upload_bytes_per_second": 4221895.326348876,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 83.24030000742755,
179
  "selection_policy": "Smallest shard within 90% of measured local peak, enlarged to amortize 30 seconds of measured page production; bounded to 256 MiB."
180
  },
181
  "upload_measurements": [
182
  {
183
- "seconds": 0.8321514729977935,
184
- "bytes": 6044525,
185
- "pages": 36,
186
- "shards": 2
187
- },
188
- {
189
- "seconds": 1.4008962299994892,
190
- "bytes": 7557731,
191
- "pages": 39,
192
- "shards": 3
193
- },
194
- {
195
- "seconds": 0.8883114669988572,
196
- "bytes": 823542,
197
  "pages": 25,
198
- "shards": 1
199
- },
200
- {
201
- "seconds": 1.2165515029992093,
202
- "bytes": 11904873,
203
- "pages": 42,
204
- "shards": 4
205
  },
206
  {
207
- "seconds": 0.9436809830003767,
208
- "bytes": 889392,
209
- "pages": 26,
210
- "shards": 1
211
  },
212
  {
213
- "seconds": 1.1113080279974383,
214
- "bytes": 11835067,
215
  "pages": 35,
216
- "shards": 3
217
  },
218
  {
219
- "seconds": 0.9863674269981857,
220
- "bytes": 7525026,
221
- "pages": 38,
222
- "shards": 3
223
  },
224
  {
225
- "seconds": 0.7792704010025773,
226
- "bytes": 5012728,
227
- "pages": 27,
228
  "shards": 2
229
  },
230
  {
231
- "seconds": 1.0204439990011451,
232
- "bytes": 7601581,
233
- "pages": 38,
234
- "shards": 3
235
  },
236
  {
237
- "seconds": 0.844439466996846,
238
- "bytes": 1090867,
239
- "pages": 34,
240
  "shards": 1
241
  },
242
  {
243
- "seconds": 0.8992618750016845,
244
- "bytes": 5707571,
245
- "pages": 29,
246
- "shards": 2
247
  },
248
  {
249
- "seconds": 1.1450656299966795,
250
- "bytes": 17577071,
251
  "pages": 37,
252
- "shards": 5
253
  },
254
  {
255
- "seconds": 1.0943805140013865,
256
- "bytes": 891607,
257
- "pages": 26,
258
  "shards": 1
259
  },
260
  {
261
- "seconds": 0.9035562749995734,
262
- "bytes": 6454400,
263
  "pages": 35,
264
- "shards": 2
265
  },
266
  {
267
- "seconds": 1.2266940429981332,
268
- "bytes": 2571680,
269
- "pages": 38,
270
- "shards": 2
271
  },
272
  {
273
- "seconds": 1.0315104050023365,
274
- "bytes": 10269533,
275
- "pages": 32,
276
  "shards": 3
277
  },
278
  {
279
- "seconds": 0.8940106660011224,
280
- "bytes": 5798800,
281
- "pages": 31,
282
- "shards": 2
283
- },
284
- {
285
- "seconds": 1.0886239710016525,
286
- "bytes": 8189805,
287
- "pages": 33,
288
  "shards": 3
289
  },
290
  {
291
- "seconds": 0.7768403989975923,
292
- "bytes": 6370933,
293
- "pages": 36,
294
  "shards": 2
295
  },
296
  {
297
- "seconds": 1.1446712159995514,
298
- "bytes": 7008906,
299
- "pages": 33,
300
  "shards": 3
301
  },
302
  {
303
- "seconds": 0.9077771720003511,
304
- "bytes": 6012024,
305
- "pages": 36,
306
  "shards": 2
307
  },
308
  {
309
- "seconds": 1.4875493239997013,
310
- "bytes": 6032602,
311
- "pages": 34,
 
 
 
 
 
 
312
  "shards": 2
313
  },
314
  {
315
- "seconds": 0.8995478449978691,
316
- "bytes": 2436827,
317
- "pages": 30,
318
  "shards": 2
319
  },
320
  {
321
- "seconds": 1.0387542439966637,
322
- "bytes": 11801776,
323
- "pages": 24,
324
  "shards": 3
325
  },
326
  {
327
- "seconds": 1.245932874000573,
328
- "bytes": 7158906,
329
- "pages": 42,
330
  "shards": 3
331
  },
332
  {
333
- "seconds": 0.8130094469997857,
334
- "bytes": 5715102,
335
- "pages": 28,
336
  "shards": 2
337
  },
338
  {
339
- "seconds": 0.8107168300011836,
340
- "bytes": 6329429,
341
- "pages": 36,
 
 
 
 
 
 
342
  "shards": 2
343
  },
344
  {
345
- "seconds": 1.009647611001128,
346
- "bytes": 2308548,
347
- "pages": 27,
348
  "shards": 2
349
  },
350
  {
351
- "seconds": 0.98430517000088,
352
- "bytes": 10007399,
353
- "pages": 44,
354
  "shards": 4
355
  },
356
  {
357
- "seconds": 0.9780213199992431,
358
- "bytes": 6711146,
359
- "pages": 29,
360
  "shards": 2
361
  },
362
  {
363
- "seconds": 0.9533051110011002,
364
- "bytes": 7632407,
365
- "pages": 38,
366
- "shards": 3
367
  },
368
  {
369
- "seconds": 1.273418650001986,
370
- "bytes": 11058609,
371
  "pages": 31,
 
 
 
 
 
 
372
  "shards": 3
373
  },
374
  {
375
- "seconds": 0.9772542809987499,
376
- "bytes": 2642702,
377
- "pages": 36,
378
- "shards": 2
379
  },
380
  {
381
- "seconds": 0.8478547829981835,
382
- "bytes": 5863488,
383
- "pages": 28,
384
  "shards": 2
385
  },
386
  {
387
- "seconds": 0.6997049530000368,
388
- "bytes": 1148192,
389
- "pages": 35,
390
  "shards": 1
391
  },
392
  {
393
- "seconds": 1.378931018000003,
394
- "bytes": 13331524,
395
- "pages": 36,
396
  "shards": 4
397
  },
398
  {
399
- "seconds": 1.0631361519990605,
400
- "bytes": 9774660,
401
  "pages": 36,
402
- "shards": 3
403
  },
404
  {
405
- "seconds": 1.0518995679994987,
406
- "bytes": 1242140,
407
- "pages": 33,
408
- "shards": 1
409
  },
410
  {
411
- "seconds": 1.0469969259975187,
412
- "bytes": 7461003,
413
  "pages": 31,
414
  "shards": 3
415
  },
416
  {
417
- "seconds": 1.1135563850002654,
418
- "bytes": 5487788,
419
- "pages": 31,
420
  "shards": 2
421
  },
422
  {
423
- "seconds": 1.0679467740010296,
424
- "bytes": 7651397,
425
- "pages": 39,
426
  "shards": 3
427
  },
428
  {
429
- "seconds": 0.9655926410050597,
430
- "bytes": 5648372,
431
- "pages": 25,
432
  "shards": 2
433
  },
434
  {
435
- "seconds": 2.317049978002615,
436
- "bytes": 7094864,
437
- "pages": 41,
438
  "shards": 2
439
  },
440
  {
441
- "seconds": 1.0156349210010376,
442
- "bytes": 12015881,
443
- "pages": 35,
444
- "shards": 4
445
  },
446
  {
447
- "seconds": 1.0830463239981327,
448
- "bytes": 6195752,
449
- "pages": 36,
450
- "shards": 2
451
  },
452
  {
453
- "seconds": 0.7641328610043274,
454
- "bytes": 2288843,
455
- "pages": 28,
456
- "shards": 2
457
  },
458
  {
459
- "seconds": 1.014886224998918,
460
- "bytes": 6118927,
461
- "pages": 37,
462
  "shards": 2
463
  },
464
  {
465
- "seconds": 0.9393350060054217,
466
- "bytes": 807847,
467
- "pages": 26,
468
  "shards": 1
469
  },
470
  {
471
- "seconds": 1.4682032510027057,
472
- "bytes": 16949786,
473
- "pages": 39,
474
- "shards": 5
475
  },
476
  {
477
- "seconds": 0.9623997290036641,
478
- "bytes": 6092433,
479
  "pages": 37,
480
  "shards": 2
481
  },
482
  {
483
- "seconds": 0.7187483379966579,
484
- "bytes": 1132933,
485
- "pages": 34,
486
- "shards": 1
487
  },
488
  {
489
- "seconds": 1.4302320659990073,
490
- "bytes": 11751486,
491
- "pages": 35,
492
  "shards": 4
493
  },
494
  {
495
- "seconds": 0.8936330840006121,
496
- "bytes": 843076,
497
- "pages": 27,
498
- "shards": 1
499
  },
500
  {
501
- "seconds": 1.0585707080026623,
502
- "bytes": 7490877,
503
- "pages": 34,
504
  "shards": 3
505
  },
506
  {
507
- "seconds": 33.798858237001696,
508
- "bytes": 12238534,
509
- "pages": 38,
510
- "shards": 3
511
  },
512
  {
513
- "seconds": 0.911944024002878,
514
- "bytes": 5136667,
515
- "pages": 41,
516
- "shards": 2
517
  },
518
  {
519
- "seconds": 1.2792566019998048,
520
- "bytes": 7551543,
521
- "pages": 38,
522
- "shards": 3
523
  },
524
  {
525
- "seconds": 0.9091981460005627,
526
- "bytes": 5814772,
527
- "pages": 27,
528
  "shards": 2
529
  },
530
  {
531
- "seconds": 1.0604712630010908,
532
- "bytes": 6743320,
533
- "pages": 36,
534
- "shards": 4
535
  },
536
  {
537
- "seconds": 1.0128854309950839,
538
- "bytes": 6059900,
539
- "pages": 26,
540
- "shards": 2
541
  },
542
  {
543
- "seconds": 1.0482589550010744,
544
- "bytes": 7056720,
545
- "pages": 42,
546
  "shards": 2
547
  },
548
  {
549
- "seconds": 1.9086563699966064,
550
- "bytes": 7678172,
551
- "pages": 33,
552
  "shards": 3
553
  },
554
  {
555
- "seconds": 1.1078665649984032,
556
- "bytes": 9761910,
557
- "pages": 33,
 
 
 
 
 
 
558
  "shards": 3
559
  },
560
  {
561
- "seconds": 0.90664235100121,
562
- "bytes": 2695308,
563
- "pages": 37,
564
- "shards": 2
565
  }
566
  ],
567
  "cache_key": {
 
22
  "tokens_per_second": 1062967.6168812835,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
+ "required_token_positions_per_second": 75414.82490767069,
26
+ "required_pages_per_second_upper_estimate": 0.006536765885870627
27
  },
28
  {
29
  "source": "rewrite6",
 
36
  "tokens_per_second": 1206672.263377472,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
+ "required_token_positions_per_second": 9426.853113458836,
40
+ "required_pages_per_second_upper_estimate": 0.003760512650972888
41
  },
42
  {
43
  "source": "openmath",
 
50
  "tokens_per_second": 410042.7581134458,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
+ "required_token_positions_per_second": 18853.70622691767,
54
+ "required_pages_per_second_upper_estimate": 0.004361473121700661
55
  },
56
  {
57
  "source": "ultra_qa",
 
64
  "tokens_per_second": 2163486.5624068812,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
+ "required_token_positions_per_second": 28280.55934037651,
68
+ "required_pages_per_second_upper_estimate": 0.0017221376872012205
69
  },
70
  {
71
  "source": "ultra_style",
 
78
  "tokens_per_second": 1546694.2737633225,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
+ "required_token_positions_per_second": 28280.55934037651,
82
+ "required_pages_per_second_upper_estimate": 0.002177115743620939
83
  },
84
  {
85
  "source": "cortex",
 
92
  "tokens_per_second": 1015177.8395898147,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
  "estimated_pages_for_buffer": 22,
95
+ "required_token_positions_per_second": 28280.55934037651,
96
+ "required_pages_per_second_upper_estimate": 0.3511498980639521
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
 
158
  "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 688946,
161
+ "producer_batch_seconds": 0.1152253453594767,
162
+ "producer_p95_seconds": 0.4034826810020604,
163
+ "gpu_update_seconds": 0.9776327146587446,
164
+ "producer_batches_per_second": 8.678646150985522,
165
+ "gpu_batches_per_second": 1.0228790270680161,
166
+ "producer_headroom": 8.48452839615064,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.61322830884216,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 79,
172
  "max_pending_pages": 4096,
173
+ "generation_bytes_per_second": 346780.9633111328,
174
+ "upload_bytes_per_second": 3718135.8829697864,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 74.41112207890022,
179
  "selection_policy": "Smallest shard within 90% of measured local peak, enlarged to amortize 30 seconds of measured page production; bounded to 256 MiB."
180
  },
181
  "upload_measurements": [
182
  {
183
+ "seconds": 0.9655926410050597,
184
+ "bytes": 5648372,
 
 
 
 
 
 
 
 
 
 
 
 
185
  "pages": 25,
186
+ "shards": 2
 
 
 
 
 
 
187
  },
188
  {
189
+ "seconds": 2.317049978002615,
190
+ "bytes": 7094864,
191
+ "pages": 41,
192
+ "shards": 2
193
  },
194
  {
195
+ "seconds": 1.0156349210010376,
196
+ "bytes": 12015881,
197
  "pages": 35,
198
+ "shards": 4
199
  },
200
  {
201
+ "seconds": 1.0830463239981327,
202
+ "bytes": 6195752,
203
+ "pages": 36,
204
+ "shards": 2
205
  },
206
  {
207
+ "seconds": 0.7641328610043274,
208
+ "bytes": 2288843,
209
+ "pages": 28,
210
  "shards": 2
211
  },
212
  {
213
+ "seconds": 1.014886224998918,
214
+ "bytes": 6118927,
215
+ "pages": 37,
216
+ "shards": 2
217
  },
218
  {
219
+ "seconds": 0.9393350060054217,
220
+ "bytes": 807847,
221
+ "pages": 26,
222
  "shards": 1
223
  },
224
  {
225
+ "seconds": 1.4682032510027057,
226
+ "bytes": 16949786,
227
+ "pages": 39,
228
+ "shards": 5
229
  },
230
  {
231
+ "seconds": 0.9623997290036641,
232
+ "bytes": 6092433,
233
  "pages": 37,
234
+ "shards": 2
235
  },
236
  {
237
+ "seconds": 0.7187483379966579,
238
+ "bytes": 1132933,
239
+ "pages": 34,
240
  "shards": 1
241
  },
242
  {
243
+ "seconds": 1.4302320659990073,
244
+ "bytes": 11751486,
245
  "pages": 35,
246
+ "shards": 4
247
  },
248
  {
249
+ "seconds": 0.8936330840006121,
250
+ "bytes": 843076,
251
+ "pages": 27,
252
+ "shards": 1
253
  },
254
  {
255
+ "seconds": 1.0585707080026623,
256
+ "bytes": 7490877,
257
+ "pages": 34,
258
  "shards": 3
259
  },
260
  {
261
+ "seconds": 33.798858237001696,
262
+ "bytes": 12238534,
263
+ "pages": 38,
 
 
 
 
 
 
264
  "shards": 3
265
  },
266
  {
267
+ "seconds": 0.911944024002878,
268
+ "bytes": 5136667,
269
+ "pages": 41,
270
  "shards": 2
271
  },
272
  {
273
+ "seconds": 1.2792566019998048,
274
+ "bytes": 7551543,
275
+ "pages": 38,
276
  "shards": 3
277
  },
278
  {
279
+ "seconds": 0.9091981460005627,
280
+ "bytes": 5814772,
281
+ "pages": 27,
282
  "shards": 2
283
  },
284
  {
285
+ "seconds": 1.0604712630010908,
286
+ "bytes": 6743320,
287
+ "pages": 36,
288
+ "shards": 4
289
+ },
290
+ {
291
+ "seconds": 1.0128854309950839,
292
+ "bytes": 6059900,
293
+ "pages": 26,
294
  "shards": 2
295
  },
296
  {
297
+ "seconds": 1.0482589550010744,
298
+ "bytes": 7056720,
299
+ "pages": 42,
300
  "shards": 2
301
  },
302
  {
303
+ "seconds": 1.9086563699966064,
304
+ "bytes": 7678172,
305
+ "pages": 33,
306
  "shards": 3
307
  },
308
  {
309
+ "seconds": 1.1078665649984032,
310
+ "bytes": 9761910,
311
+ "pages": 33,
312
  "shards": 3
313
  },
314
  {
315
+ "seconds": 0.90664235100121,
316
+ "bytes": 2695308,
317
+ "pages": 37,
318
  "shards": 2
319
  },
320
  {
321
+ "seconds": 0.7943546639944543,
322
+ "bytes": 258770,
323
+ "pages": 8,
324
+ "shards": 1
325
+ },
326
+ {
327
+ "seconds": 1.1189483790003578,
328
+ "bytes": 5921688,
329
+ "pages": 28,
330
  "shards": 2
331
  },
332
  {
333
+ "seconds": 1.2605290909996256,
334
+ "bytes": 6824656,
335
+ "pages": 33,
336
  "shards": 2
337
  },
338
  {
339
+ "seconds": 11.391802897000161,
340
+ "bytes": 12565360,
341
+ "pages": 37,
342
  "shards": 4
343
  },
344
  {
345
+ "seconds": 0.941881225000543,
346
+ "bytes": 6023516,
347
+ "pages": 37,
348
  "shards": 2
349
  },
350
  {
351
+ "seconds": 0.9670609070017235,
352
+ "bytes": 2689085,
353
+ "pages": 35,
354
+ "shards": 2
355
  },
356
  {
357
+ "seconds": 1.1650014270053362,
358
+ "bytes": 5886860,
359
  "pages": 31,
360
+ "shards": 2
361
+ },
362
+ {
363
+ "seconds": 1.1598776869941503,
364
+ "bytes": 10363400,
365
+ "pages": 37,
366
  "shards": 3
367
  },
368
  {
369
+ "seconds": 1.1336681269967812,
370
+ "bytes": 8324153,
371
+ "pages": 31,
372
+ "shards": 3
373
  },
374
  {
375
+ "seconds": 1.0718341040046653,
376
+ "bytes": 6483305,
377
+ "pages": 36,
378
  "shards": 2
379
  },
380
  {
381
+ "seconds": 0.9573828110005707,
382
+ "bytes": 873310,
383
+ "pages": 27,
384
  "shards": 1
385
  },
386
  {
387
+ "seconds": 1.2830225369980326,
388
+ "bytes": 12398676,
389
+ "pages": 40,
390
  "shards": 4
391
  },
392
  {
393
+ "seconds": 1.2898951360039064,
394
+ "bytes": 6161152,
395
  "pages": 36,
396
+ "shards": 2
397
  },
398
  {
399
+ "seconds": 0.9979428110018489,
400
+ "bytes": 2470155,
401
+ "pages": 30,
402
+ "shards": 2
403
  },
404
  {
405
+ "seconds": 1.098158920998685,
406
+ "bytes": 11654949,
407
  "pages": 31,
408
  "shards": 3
409
  },
410
  {
411
+ "seconds": 1.0713990720032598,
412
+ "bytes": 5874732,
413
+ "pages": 36,
414
  "shards": 2
415
  },
416
  {
417
+ "seconds": 0.9352636539988453,
418
+ "bytes": 6076485,
419
+ "pages": 31,
420
  "shards": 3
421
  },
422
  {
423
+ "seconds": 1.2692730029957602,
424
+ "bytes": 6086693,
425
+ "pages": 37,
426
  "shards": 2
427
  },
428
  {
429
+ "seconds": 1.036868928000331,
430
+ "bytes": 2682212,
431
+ "pages": 38,
432
  "shards": 2
433
  },
434
  {
435
+ "seconds": 1.1587374009977793,
436
+ "bytes": 10973458,
437
+ "pages": 30,
438
+ "shards": 3
439
  },
440
  {
441
+ "seconds": 0.8646178900016821,
442
+ "bytes": 947338,
443
+ "pages": 27,
444
+ "shards": 1
445
  },
446
  {
447
+ "seconds": 1.0228583920033998,
448
+ "bytes": 13452885,
449
+ "pages": 43,
450
+ "shards": 4
451
  },
452
  {
453
+ "seconds": 1.0817572929954622,
454
+ "bytes": 6130820,
455
+ "pages": 28,
456
  "shards": 2
457
  },
458
  {
459
+ "seconds": 0.9214774369975203,
460
+ "bytes": 1243052,
461
+ "pages": 35,
462
  "shards": 1
463
  },
464
  {
465
+ "seconds": 1.3106369909946807,
466
+ "bytes": 11452061,
467
+ "pages": 33,
468
+ "shards": 4
469
  },
470
  {
471
+ "seconds": 1.0961634519990184,
472
+ "bytes": 6066738,
473
  "pages": 37,
474
  "shards": 2
475
  },
476
  {
477
+ "seconds": 0.9469924939985503,
478
+ "bytes": 890694,
479
+ "pages": 27,
480
+ "shards": 2
481
  },
482
  {
483
+ "seconds": 1.2476149669964798,
484
+ "bytes": 13021486,
485
+ "pages": 30,
486
  "shards": 4
487
  },
488
  {
489
+ "seconds": 0.8867707609970239,
490
+ "bytes": 6143048,
491
+ "pages": 43,
492
+ "shards": 2
493
  },
494
  {
495
+ "seconds": 1.05023945299763,
496
+ "bytes": 7400872,
497
+ "pages": 31,
498
  "shards": 3
499
  },
500
  {
501
+ "seconds": 1.1047442850031075,
502
+ "bytes": 6203278,
503
+ "pages": 37,
504
+ "shards": 2
505
  },
506
  {
507
+ "seconds": 1.0125872680000612,
508
+ "bytes": 1072664,
509
+ "pages": 31,
510
+ "shards": 1
511
  },
512
  {
513
+ "seconds": 1.132626735001395,
514
+ "bytes": 11459123,
515
+ "pages": 36,
516
+ "shards": 4
517
  },
518
  {
519
+ "seconds": 1.060930259001907,
520
+ "bytes": 6808205,
521
+ "pages": 34,
522
  "shards": 2
523
  },
524
  {
525
+ "seconds": 3.3303450339953997,
526
+ "bytes": 5952538,
527
+ "pages": 29,
528
+ "shards": 2
529
  },
530
  {
531
+ "seconds": 0.9529956519982079,
532
+ "bytes": 7196338,
533
+ "pages": 30,
534
+ "shards": 3
535
  },
536
  {
537
+ "seconds": 1.1464728400023887,
538
+ "bytes": 6003121,
539
+ "pages": 37,
540
  "shards": 2
541
  },
542
  {
543
+ "seconds": 0.9798986519963364,
544
+ "bytes": 7781027,
545
+ "pages": 39,
546
  "shards": 3
547
  },
548
  {
549
+ "seconds": 0.9489700769991032,
550
+ "bytes": 911408,
551
+ "pages": 26,
552
+ "shards": 1
553
+ },
554
+ {
555
+ "seconds": 1.3251626939963899,
556
+ "bytes": 11870118,
557
+ "pages": 36,
558
  "shards": 3
559
  },
560
  {
561
+ "seconds": 1.1823667450007633,
562
+ "bytes": 10876587,
563
+ "pages": 32,
564
+ "shards": 4
565
  }
566
  ],
567
  "cache_key": {
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:bdb02210299e7645fb1fcfe319164a683c0fab01d39f95c7ebced4aab5172f8e
3
- size 721769407
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ddb86ed5b8deb7d4e788e14749c412fa48be7f3cefd33dadef82d54b10488b02
3
+ size 721776959