appvoid commited on
Commit
75ecc5e
·
verified ·
1 Parent(s): 164a703

Qwen90M step 860000; complete resumable checkpoint

Browse files
Files changed (6) hide show
  1. complete.json +5 -5
  2. coverage.json +50 -50
  3. latest.json +1 -1
  4. model.safetensors +1 -1
  5. shard_scheduler.json +241 -241
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 855000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "628d1e63b12b188b6ca281f0cf6c430fc964fc9444c4b85a80a780b372b0644b",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "0b9d7502ee7a37b93921e50411dcf10cc48e5c0ea61bf54ea6e2aa3c774d65a7",
12
  "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
13
  "performance.json": "172134497605e50805257548c05018674e515323c63930c6237cd46d43786367",
14
- "shard_scheduler.json": "24de37ac4e4183ecbaeccca24a00bbf84010cf8c976979813ca172c7341fe35d",
15
- "coverage.json": "399d603729230593eb96d15af1ccf15d2fed8d8f61569fc076f477232117f990",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"
 
1
  {
2
+ "step": 860000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "76d2d488ad46202033ce64137cf12ae9086c284d7e8060eb3f80363676d0e8d0",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "892eb6396b64cb4fd594382fdc135dcfdab02c54a38e9994701368b84bd73dcd",
12
  "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
13
  "performance.json": "172134497605e50805257548c05018674e515323c63930c6237cd46d43786367",
14
+ "shard_scheduler.json": "9a29ea4bdc9cefd2e92ebd24e57d2572330e585a6e74a99d3c512c6e2de2f678",
15
+ "coverage.json": "7fba50c91c135d7273f46b84ac1af03e10b40c0850608fe8d645d5160ea5fc41",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 855000,
3
  "stats": {
4
- "climbmix/blocks": 30780000,
5
- "climbmix/targets": 51453874788,
6
- "climbmix/input_tokens": 51457135364,
7
- "ultra_qa/blocks": 11542500,
8
- "ultra_qa/targets": 18830986539,
9
- "ultra_qa/input_tokens": 18831022740,
10
- "cortex/blocks": 11542500,
11
- "cortex/targets": 438911204,
12
- "cortex/input_tokens": 18347586211,
13
- "ultra_style/blocks": 11542500,
14
- "ultra_style/targets": 18256746475,
15
- "ultra_style/input_tokens": 18257703756,
16
- "openmath/blocks": 7695000,
17
- "openmath/targets": 11890027768,
18
- "openmath/input_tokens": 11893989809,
19
- "rewrite6/blocks": 3847500,
20
- "rewrite6/targets": 6307501946,
21
- "rewrite6/input_tokens": 6314417109
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 4215,
26
- "row": 2185,
27
  "start": {
28
- "file": 204,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 2,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
- "offset": 2048,
37
- "documents": 17266825
38
  },
39
  "rewrite6": {
40
- "page": 48,
41
- "row": 1106,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 196608
47
  }
48
  },
49
- "epoch": 47,
50
- "offset": 0,
51
- "documents": 9851702
52
  },
53
  "openmath": {
54
- "page": 2634,
55
- "row": 2725,
56
  "start": {
57
  "file": 24,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 309,
61
- "row": 265
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 10791589
67
  },
68
  "ultra_qa": {
69
- "page": 1139,
70
- "row": 374,
71
  "start": {
72
- "file": 8,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
- "row_group": 7,
76
- "row": 56622
77
  }
78
  },
79
  "epoch": 0,
80
- "offset": 2048,
81
- "documents": 4665718
82
  },
83
  "ultra_style": {
84
- "page": 1738,
85
- "row": 2213,
86
  "start": {
87
  "file": 10,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 3,
91
- "row": 14940
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
- "documents": 7121061
97
  },
98
  "cortex": {
99
- "page": 223850,
100
- "row": 23,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 28652823
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 860000,
3
  "stats": {
4
+ "climbmix/blocks": 30960000,
5
+ "climbmix/targets": 51754676559,
6
+ "climbmix/input_tokens": 51757956002,
7
+ "ultra_qa/blocks": 11610000,
8
+ "ultra_qa/targets": 18941056864,
9
+ "ultra_qa/input_tokens": 18941093290,
10
+ "cortex/blocks": 11610000,
11
+ "cortex/targets": 441483304,
12
+ "cortex/input_tokens": 18454886279,
13
+ "ultra_style/blocks": 11610000,
14
+ "ultra_style/targets": 18363767524,
15
+ "ultra_style/input_tokens": 18364730231,
16
+ "openmath/blocks": 7740000,
17
+ "openmath/targets": 11959461269,
18
+ "openmath/input_tokens": 11963446369,
19
+ "rewrite6/blocks": 3870000,
20
+ "rewrite6/targets": 6344361516,
21
+ "rewrite6/input_tokens": 6351317040
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 4240,
26
+ "row": 396,
27
  "start": {
28
+ "file": 205,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 20,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
+ "offset": 0,
37
+ "documents": 17367436
38
  },
39
  "rewrite6": {
40
+ "page": 12,
41
+ "row": 642,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 49152
47
  }
48
  },
49
+ "epoch": 48,
50
+ "offset": 2048,
51
+ "documents": 9909186
52
  },
53
  "openmath": {
54
+ "page": 2650,
55
+ "row": 161,
56
  "start": {
57
  "file": 24,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 374,
61
+ "row": 801
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 10854561
67
  },
68
  "ultra_qa": {
69
+ "page": 1145,
70
+ "row": 3108,
71
  "start": {
72
+ "file": 9,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
+ "row_group": 0,
76
+ "row": 13278
77
  }
78
  },
79
  "epoch": 0,
80
+ "offset": 4096,
81
+ "documents": 4693028
82
  },
83
  "ultra_style": {
84
+ "page": 1748,
85
+ "row": 2510,
86
  "start": {
87
  "file": 10,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 3,
91
+ "row": 55900
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
+ "documents": 7162318
97
  },
98
  "cortex": {
99
+ "page": 225159,
100
+ "row": 87,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 28820439
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 855000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "628d1e63b12b188b6ca281f0cf6c430fc964fc9444c4b85a80a780b372b0644b", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "0b9d7502ee7a37b93921e50411dcf10cc48e5c0ea61bf54ea6e2aa3c774d65a7", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "172134497605e50805257548c05018674e515323c63930c6237cd46d43786367", "shard_scheduler.json": "24de37ac4e4183ecbaeccca24a00bbf84010cf8c976979813ca172c7341fe35d", "coverage.json": "399d603729230593eb96d15af1ccf15d2fed8d8f61569fc076f477232117f990", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000855000"}
 
1
+ {"step": 860000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "76d2d488ad46202033ce64137cf12ae9086c284d7e8060eb3f80363676d0e8d0", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "892eb6396b64cb4fd594382fdc135dcfdab02c54a38e9994701368b84bd73dcd", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "172134497605e50805257548c05018674e515323c63930c6237cd46d43786367", "shard_scheduler.json": "9a29ea4bdc9cefd2e92ebd24e57d2572330e585a6e74a99d3c512c6e2de2f678", "coverage.json": "7fba50c91c135d7273f46b84ac1af03e10b40c0850608fe8d645d5160ea5fc41", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000860000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:628d1e63b12b188b6ca281f0cf6c430fc964fc9444c4b85a80a780b372b0644b
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:76d2d488ad46202033ce64137cf12ae9086c284d7e8060eb3f80363676d0e8d0
3
  size 360622744
shard_scheduler.json CHANGED
@@ -22,8 +22,8 @@
22
  "tokens_per_second": 598315.063954628,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
- "required_token_positions_per_second": 77256.40416056558,
26
- "required_pages_per_second_upper_estimate": 0.005344777299382726
27
  },
28
  {
29
  "source": "rewrite6",
@@ -36,8 +36,8 @@
36
  "tokens_per_second": 8533.957213384612,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
- "required_token_positions_per_second": 9657.050520070698,
40
- "required_pages_per_second_upper_estimate": 0.004043205390606975
41
  },
42
  {
43
  "source": "openmath",
@@ -50,8 +50,8 @@
50
  "tokens_per_second": 380144.35761751566,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
- "required_token_positions_per_second": 19314.101040141395,
54
- "required_pages_per_second_upper_estimate": 0.004022617899564916
55
  },
56
  {
57
  "source": "ultra_qa",
@@ -64,8 +64,8 @@
64
  "tokens_per_second": 43742.16556775435,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
- "required_token_positions_per_second": 28971.151560212093,
68
- "required_pages_per_second_upper_estimate": 0.0017367762996476524
69
  },
70
  {
71
  "source": "ultra_style",
@@ -78,8 +78,8 @@
78
  "tokens_per_second": 22018.86276523721,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
- "required_token_positions_per_second": 28971.151560212093,
82
- "required_pages_per_second_upper_estimate": 0.002252379603178061
83
  },
84
  {
85
  "source": "cortex",
@@ -92,8 +92,8 @@
92
  "tokens_per_second": 910315.184299418,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
  "estimated_pages_for_buffer": 23,
95
- "required_token_positions_per_second": 28971.151560212093,
96
- "required_pages_per_second_upper_estimate": 0.37161082541542684
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,408 +158,408 @@
158
  "prefetch_batches": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
- "producer_batch_seconds": 0.07687787735935103,
162
- "producer_p95_seconds": 0.20161003999965033,
163
- "gpu_update_seconds": 0.9543286514703385,
164
- "producer_batches_per_second": 13.00764321738086,
165
- "gpu_batches_per_second": 1.0478570442785045,
166
- "producer_headroom": 12.41356661045037,
167
  "sustainable": true,
168
- "buffer_seconds": 60.122705042631324,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 82,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 446772.5461731995,
174
- "upload_bytes_per_second": 4802594.46749234,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 95.8669880761813,
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": 1.2396700680001231,
184
- "bytes": 7470065,
185
- "pages": 39,
186
- "shards": 3
187
- },
188
- {
189
- "seconds": 1.6093791190005504,
190
- "bytes": 874866,
191
- "pages": 25,
192
- "shards": 1
193
  },
194
  {
195
- "seconds": 1.1941245199996047,
196
- "bytes": 7420221,
197
- "pages": 39,
198
- "shards": 3
199
  },
200
  {
201
- "seconds": 1.6006398999998055,
202
- "bytes": 9959535,
203
- "pages": 29,
204
  "shards": 3
205
  },
206
  {
207
- "seconds": 1.2196721769996657,
208
- "bytes": 7030636,
209
- "pages": 39,
210
  "shards": 2
211
  },
212
  {
213
- "seconds": 1.0609669329996905,
214
- "bytes": 7430334,
215
- "pages": 38,
216
- "shards": 3
217
  },
218
  {
219
- "seconds": 1.0616652130001967,
220
- "bytes": 6074717,
221
  "pages": 35,
222
  "shards": 2
223
  },
224
  {
225
- "seconds": 1.3232113710000704,
226
- "bytes": 5040655,
227
- "pages": 31,
228
  "shards": 3
229
  },
230
  {
231
- "seconds": 1.5191957560000446,
232
- "bytes": 7567101,
233
- "pages": 34,
234
- "shards": 3
235
  },
236
  {
237
- "seconds": 1.065770677000728,
238
- "bytes": 897206,
239
- "pages": 26,
240
  "shards": 1
241
  },
242
  {
243
- "seconds": 1.459628982999675,
244
- "bytes": 13449017,
245
- "pages": 40,
246
- "shards": 4
247
  },
248
  {
249
- "seconds": 1.560673630000565,
250
- "bytes": 10527927,
251
- "pages": 42,
 
 
 
 
 
 
252
  "shards": 3
253
  },
254
  {
255
- "seconds": 1.384273941000174,
256
- "bytes": 1166498,
257
- "pages": 35,
 
 
 
 
 
 
258
  "shards": 1
259
  },
260
  {
261
- "seconds": 1.245850141000119,
262
- "bytes": 7339507,
263
- "pages": 30,
264
  "shards": 3
265
  },
266
  {
267
- "seconds": 1.0799733499998183,
268
- "bytes": 6017108,
269
- "pages": 34,
270
  "shards": 2
271
  },
272
  {
273
- "seconds": 1.2428365799996755,
274
- "bytes": 12889604,
275
- "pages": 40,
276
- "shards": 4
277
- },
278
- {
279
- "seconds": 1.21177975899991,
280
- "bytes": 6176322,
281
- "pages": 36,
282
  "shards": 2
283
  },
284
  {
285
- "seconds": 1.4487905369996952,
286
- "bytes": 925527,
287
- "pages": 25,
288
- "shards": 1
289
  },
290
  {
291
- "seconds": 1.6813211230000888,
292
- "bytes": 7642707,
293
- "pages": 39,
294
  "shards": 3
295
  },
296
  {
297
- "seconds": 1.8502533870005209,
298
- "bytes": 10875382,
299
  "pages": 36,
300
- "shards": 3
301
  },
302
  {
303
- "seconds": 1.247339760999239,
304
- "bytes": 856828,
305
- "pages": 26,
306
- "shards": 1
307
  },
308
  {
309
- "seconds": 1.4432682090000526,
310
- "bytes": 13357057,
311
- "pages": 39,
312
- "shards": 4
313
  },
314
  {
315
- "seconds": 1.271395830000074,
316
- "bytes": 6214888,
317
- "pages": 36,
318
- "shards": 2
319
  },
320
  {
321
- "seconds": 1.4851975340006902,
322
- "bytes": 6638444,
323
- "pages": 30,
324
  "shards": 3
325
  },
326
  {
327
- "seconds": 1.5820424499997898,
328
- "bytes": 5864908,
329
- "pages": 27,
330
  "shards": 2
331
  },
332
  {
333
- "seconds": 1.0337117700000817,
334
- "bytes": 1192866,
335
- "pages": 35,
336
  "shards": 1
337
  },
338
  {
339
- "seconds": 1.3218913410000823,
340
- "bytes": 7396091,
341
  "pages": 30,
342
  "shards": 3
343
  },
344
  {
345
- "seconds": 1.1367835250002827,
346
- "bytes": 5040157,
347
- "pages": 37,
348
  "shards": 2
349
  },
350
  {
351
- "seconds": 1.2845682920014951,
352
- "bytes": 11638802,
353
- "pages": 28,
354
- "shards": 3
355
  },
356
  {
357
- "seconds": 1.3405057710006076,
358
- "bytes": 7718368,
359
  "pages": 39,
360
  "shards": 3
361
  },
362
  {
363
- "seconds": 1.021054885999547,
364
- "bytes": 1163402,
365
- "pages": 34,
366
- "shards": 1
367
  },
368
  {
369
- "seconds": 1.532974641999317,
370
- "bytes": 11431908,
371
- "pages": 35,
372
- "shards": 4
373
  },
374
  {
375
- "seconds": 1.541119751000224,
376
- "bytes": 6183503,
377
  "pages": 36,
378
  "shards": 2
379
  },
380
  {
381
- "seconds": 1.4027249899991148,
382
- "bytes": 1181995,
383
- "pages": 34,
384
  "shards": 1
385
  },
386
  {
387
- "seconds": 1.3459127579990309,
388
- "bytes": 13166079,
389
- "pages": 31,
390
- "shards": 4
391
  },
392
  {
393
- "seconds": 1.166742791001525,
394
- "bytes": 5284765,
395
- "pages": 34,
396
  "shards": 2
397
  },
398
  {
399
- "seconds": 1.747680902999491,
400
- "bytes": 7697898,
401
- "pages": 38,
402
- "shards": 3
403
  },
404
  {
405
- "seconds": 0.9798488249998627,
406
- "bytes": 6153509,
407
- "pages": 36,
408
  "shards": 2
409
  },
410
  {
411
- "seconds": 1.0631558050008607,
412
- "bytes": 1018193,
413
- "pages": 28,
414
- "shards": 2
415
  },
416
  {
417
- "seconds": 1.8303398019997985,
418
- "bytes": 12284688,
419
- "pages": 40,
420
  "shards": 4
421
  },
422
  {
423
- "seconds": 1.182235641999796,
424
- "bytes": 6603515,
425
- "pages": 27,
426
- "shards": 2
427
  },
428
  {
429
- "seconds": 1.285518773000149,
430
- "bytes": 6002809,
431
  "pages": 36,
432
  "shards": 2
433
  },
434
  {
435
- "seconds": 1.7619018259993027,
436
- "bytes": 7674402,
437
- "pages": 38,
438
  "shards": 3
439
  },
440
  {
441
- "seconds": 1.6144984039983683,
442
- "bytes": 4728422,
443
- "pages": 30,
444
- "shards": 2
445
  },
446
  {
447
- "seconds": 1.6315553869990254,
448
- "bytes": 5703905,
449
- "pages": 24,
450
- "shards": 2
451
  },
452
  {
453
- "seconds": 1.11825536699871,
454
- "bytes": 2656311,
455
- "pages": 35,
456
  "shards": 2
457
  },
458
  {
459
- "seconds": 1.9071076060008636,
460
- "bytes": 12136915,
461
  "pages": 37,
462
- "shards": 3
463
  },
464
  {
465
- "seconds": 1.7720820229988021,
466
- "bytes": 11800483,
467
- "pages": 32,
468
- "shards": 4
469
  },
470
  {
471
- "seconds": 1.0689491579996684,
472
- "bytes": 1154476,
473
- "pages": 34,
474
- "shards": 1
475
  },
476
  {
477
- "seconds": 1.7027444539999124,
478
- "bytes": 6185388,
479
- "pages": 36,
480
- "shards": 2
481
  },
482
  {
483
- "seconds": 1.5859740610012523,
484
- "bytes": 2563401,
485
- "pages": 35,
486
  "shards": 2
487
  },
488
  {
489
- "seconds": 1.2778493570003775,
490
- "bytes": 10237665,
491
- "pages": 33,
492
  "shards": 3
493
  },
494
  {
495
- "seconds": 2.496101438000551,
496
- "bytes": 13297971,
497
- "pages": 33,
498
- "shards": 4
499
  },
500
  {
501
- "seconds": 1.2942262079995999,
502
- "bytes": 1101377,
503
- "pages": 34,
504
- "shards": 1
505
  },
506
  {
507
- "seconds": 1.815276637998977,
508
- "bytes": 10307802,
509
- "pages": 39,
510
  "shards": 3
511
  },
512
  {
513
- "seconds": 1.1417645319997973,
514
- "bytes": 2370286,
515
- "pages": 28,
516
- "shards": 2
517
  },
518
  {
519
- "seconds": 1.1171746460004215,
520
- "bytes": 6111968,
521
  "pages": 37,
522
  "shards": 2
523
  },
524
  {
525
- "seconds": 1.2683859379994828,
526
- "bytes": 7618531,
527
- "pages": 38,
528
- "shards": 3
529
  },
530
  {
531
- "seconds": 1.7501807870012271,
532
- "bytes": 11262011,
533
- "pages": 28,
534
  "shards": 3
535
  },
536
  {
537
- "seconds": 1.2981046400000196,
538
- "bytes": 6199716,
539
  "pages": 36,
540
  "shards": 2
541
  },
542
  {
543
- "seconds": 1.6966266469989932,
544
- "bytes": 7696834,
545
- "pages": 38,
546
- "shards": 3
547
  },
548
  {
549
- "seconds": 1.3060271180002019,
550
- "bytes": 1241981,
551
- "pages": 35,
552
- "shards": 1
553
  },
554
  {
555
- "seconds": 1.5251753980010108,
556
- "bytes": 9924170,
557
- "pages": 32,
558
- "shards": 3
559
  },
560
  {
561
- "seconds": 1.1737768090006284,
562
- "bytes": 7614180,
563
  "pages": 39,
564
  "shards": 3
565
  }
 
22
  "tokens_per_second": 598315.063954628,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
+ "required_token_positions_per_second": 76235.28940132579,
26
+ "required_pages_per_second_upper_estimate": 0.0052741342110258485
27
  },
28
  {
29
  "source": "rewrite6",
 
36
  "tokens_per_second": 8533.957213384612,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
+ "required_token_positions_per_second": 9529.411175165724,
40
+ "required_pages_per_second_upper_estimate": 0.003989765462307878
41
  },
42
  {
43
  "source": "openmath",
 
50
  "tokens_per_second": 380144.35761751566,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
+ "required_token_positions_per_second": 19058.822350331448,
54
+ "required_pages_per_second_upper_estimate": 0.003969450080629271
55
  },
56
  {
57
  "source": "ultra_qa",
 
64
  "tokens_per_second": 43742.16556775435,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
+ "required_token_positions_per_second": 28588.23352549717,
68
+ "required_pages_per_second_upper_estimate": 0.0017138209481484775
69
  },
70
  {
71
  "source": "ultra_style",
 
78
  "tokens_per_second": 22018.86276523721,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
+ "required_token_positions_per_second": 28588.23352549717,
82
+ "required_pages_per_second_upper_estimate": 0.00222260941025741
83
  },
84
  {
85
  "source": "cortex",
 
92
  "tokens_per_second": 910315.184299418,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
  "estimated_pages_for_buffer": 23,
95
+ "required_token_positions_per_second": 28588.23352549717,
96
+ "required_pages_per_second_upper_estimate": 0.36669916401145664
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
 
158
  "prefetch_batches": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
+ "producer_batch_seconds": 0.06858718393758068,
162
+ "producer_p95_seconds": 0.22669146000043838,
163
+ "gpu_update_seconds": 0.9671111709417584,
164
+ "producer_batches_per_second": 14.579983352430284,
165
+ "gpu_batches_per_second": 1.0340072889719751,
166
+ "producer_headroom": 14.100464772280196,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.92800376933078,
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": 389341.2469068264,
174
+ "upload_bytes_per_second": 3447999.215230209,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 83.54356818584054,
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": 1.182235641999796,
184
+ "bytes": 6603515,
185
+ "pages": 27,
186
+ "shards": 2
 
 
 
 
 
 
187
  },
188
  {
189
+ "seconds": 1.285518773000149,
190
+ "bytes": 6002809,
191
+ "pages": 36,
192
+ "shards": 2
193
  },
194
  {
195
+ "seconds": 1.7619018259993027,
196
+ "bytes": 7674402,
197
+ "pages": 38,
198
  "shards": 3
199
  },
200
  {
201
+ "seconds": 1.6144984039983683,
202
+ "bytes": 4728422,
203
+ "pages": 30,
204
  "shards": 2
205
  },
206
  {
207
+ "seconds": 1.6315553869990254,
208
+ "bytes": 5703905,
209
+ "pages": 24,
210
+ "shards": 2
211
  },
212
  {
213
+ "seconds": 1.11825536699871,
214
+ "bytes": 2656311,
215
  "pages": 35,
216
  "shards": 2
217
  },
218
  {
219
+ "seconds": 1.9071076060008636,
220
+ "bytes": 12136915,
221
+ "pages": 37,
222
  "shards": 3
223
  },
224
  {
225
+ "seconds": 1.7720820229988021,
226
+ "bytes": 11800483,
227
+ "pages": 32,
228
+ "shards": 4
229
  },
230
  {
231
+ "seconds": 1.0689491579996684,
232
+ "bytes": 1154476,
233
+ "pages": 34,
234
  "shards": 1
235
  },
236
  {
237
+ "seconds": 1.7027444539999124,
238
+ "bytes": 6185388,
239
+ "pages": 36,
240
+ "shards": 2
241
  },
242
  {
243
+ "seconds": 1.5859740610012523,
244
+ "bytes": 2563401,
245
+ "pages": 35,
246
+ "shards": 2
247
+ },
248
+ {
249
+ "seconds": 1.2778493570003775,
250
+ "bytes": 10237665,
251
+ "pages": 33,
252
  "shards": 3
253
  },
254
  {
255
+ "seconds": 2.496101438000551,
256
+ "bytes": 13297971,
257
+ "pages": 33,
258
+ "shards": 4
259
+ },
260
+ {
261
+ "seconds": 1.2942262079995999,
262
+ "bytes": 1101377,
263
+ "pages": 34,
264
  "shards": 1
265
  },
266
  {
267
+ "seconds": 1.815276637998977,
268
+ "bytes": 10307802,
269
+ "pages": 39,
270
  "shards": 3
271
  },
272
  {
273
+ "seconds": 1.1417645319997973,
274
+ "bytes": 2370286,
275
+ "pages": 28,
276
  "shards": 2
277
  },
278
  {
279
+ "seconds": 1.1171746460004215,
280
+ "bytes": 6111968,
281
+ "pages": 37,
 
 
 
 
 
 
282
  "shards": 2
283
  },
284
  {
285
+ "seconds": 1.2683859379994828,
286
+ "bytes": 7618531,
287
+ "pages": 38,
288
+ "shards": 3
289
  },
290
  {
291
+ "seconds": 1.7501807870012271,
292
+ "bytes": 11262011,
293
+ "pages": 28,
294
  "shards": 3
295
  },
296
  {
297
+ "seconds": 1.2981046400000196,
298
+ "bytes": 6199716,
299
  "pages": 36,
300
+ "shards": 2
301
  },
302
  {
303
+ "seconds": 1.6966266469989932,
304
+ "bytes": 7696834,
305
+ "pages": 38,
306
+ "shards": 3
307
  },
308
  {
309
+ "seconds": 1.3060271180002019,
310
+ "bytes": 1241981,
311
+ "pages": 35,
312
+ "shards": 1
313
  },
314
  {
315
+ "seconds": 1.5251753980010108,
316
+ "bytes": 9924170,
317
+ "pages": 32,
318
+ "shards": 3
319
  },
320
  {
321
+ "seconds": 1.1737768090006284,
322
+ "bytes": 7614180,
323
+ "pages": 39,
324
  "shards": 3
325
  },
326
  {
327
+ "seconds": 1.2979090160006308,
328
+ "bytes": 6818365,
329
+ "pages": 30,
330
  "shards": 2
331
  },
332
  {
333
+ "seconds": 0.9478092970002763,
334
+ "bytes": 253929,
335
+ "pages": 8,
336
  "shards": 1
337
  },
338
  {
339
+ "seconds": 1.6955216110000038,
340
+ "bytes": 7314599,
341
  "pages": 30,
342
  "shards": 3
343
  },
344
  {
345
+ "seconds": 1.2749757900000986,
346
+ "bytes": 5237846,
347
+ "pages": 31,
348
  "shards": 2
349
  },
350
  {
351
+ "seconds": 1.5798416210000141,
352
+ "bytes": 6237294,
353
+ "pages": 36,
354
+ "shards": 2
355
  },
356
  {
357
+ "seconds": 1.5424033700001019,
358
+ "bytes": 7705972,
359
  "pages": 39,
360
  "shards": 3
361
  },
362
  {
363
+ "seconds": 0.9795733470000414,
364
+ "bytes": 923741,
365
+ "pages": 28,
366
+ "shards": 2
367
  },
368
  {
369
+ "seconds": 1.5211821389984834,
370
+ "bytes": 18268501,
371
+ "pages": 39,
372
+ "shards": 5
373
  },
374
  {
375
+ "seconds": 1.3536671069996373,
376
+ "bytes": 6151717,
377
  "pages": 36,
378
  "shards": 2
379
  },
380
  {
381
+ "seconds": 1.1601749330002349,
382
+ "bytes": 883373,
383
+ "pages": 25,
384
  "shards": 1
385
  },
386
  {
387
+ "seconds": 1.1801181450009608,
388
+ "bytes": 7664644,
389
+ "pages": 38,
390
+ "shards": 3
391
  },
392
  {
393
+ "seconds": 1.0472080479994474,
394
+ "bytes": 5473896,
395
+ "pages": 33,
396
  "shards": 2
397
  },
398
  {
399
+ "seconds": 1.548700262001148,
400
+ "bytes": 13538856,
401
+ "pages": 41,
402
+ "shards": 4
403
  },
404
  {
405
+ "seconds": 1.3191421729989088,
406
+ "bytes": 5997715,
407
+ "pages": 32,
408
  "shards": 2
409
  },
410
  {
411
+ "seconds": 1.765130539000893,
412
+ "bytes": 937778,
413
+ "pages": 29,
414
+ "shards": 1
415
  },
416
  {
417
+ "seconds": 2.7900099470007262,
418
+ "bytes": 11587181,
419
+ "pages": 37,
420
  "shards": 4
421
  },
422
  {
423
+ "seconds": 0.987754063000466,
424
+ "bytes": 1220161,
425
+ "pages": 35,
426
+ "shards": 1
427
  },
428
  {
429
+ "seconds": 1.716705255999841,
430
+ "bytes": 6235687,
431
  "pages": 36,
432
  "shards": 2
433
  },
434
  {
435
+ "seconds": 33.88602663399979,
436
+ "bytes": 7355362,
437
+ "pages": 31,
438
  "shards": 3
439
  },
440
  {
441
+ "seconds": 1.5573600560001069,
442
+ "bytes": 11648033,
443
+ "pages": 46,
444
+ "shards": 3
445
  },
446
  {
447
+ "seconds": 1.2928704839996499,
448
+ "bytes": 7315946,
449
+ "pages": 39,
450
+ "shards": 3
451
  },
452
  {
453
+ "seconds": 1.4081739580014982,
454
+ "bytes": 5891309,
455
+ "pages": 28,
456
  "shards": 2
457
  },
458
  {
459
+ "seconds": 1.286470338000072,
460
+ "bytes": 2778002,
461
  "pages": 37,
462
+ "shards": 2
463
  },
464
  {
465
+ "seconds": 1.1679330120005034,
466
+ "bytes": 10820666,
467
+ "pages": 37,
468
+ "shards": 3
469
  },
470
  {
471
+ "seconds": 1.4827527400011604,
472
+ "bytes": 6647922,
473
+ "pages": 30,
474
+ "shards": 2
475
  },
476
  {
477
+ "seconds": 1.248109399999521,
478
+ "bytes": 7422482,
479
+ "pages": 37,
480
+ "shards": 3
481
  },
482
  {
483
+ "seconds": 1.451942487999986,
484
+ "bytes": 6077412,
485
+ "pages": 30,
486
  "shards": 2
487
  },
488
  {
489
+ "seconds": 1.4631003799986502,
490
+ "bytes": 7075109,
491
+ "pages": 43,
492
  "shards": 3
493
  },
494
  {
495
+ "seconds": 2.129564337999909,
496
+ "bytes": 5884538,
497
+ "pages": 31,
498
+ "shards": 2
499
  },
500
  {
501
+ "seconds": 1.1728028199995606,
502
+ "bytes": 6147166,
503
+ "pages": 33,
504
+ "shards": 2
505
  },
506
  {
507
+ "seconds": 1.2317090710002958,
508
+ "bytes": 8202394,
509
+ "pages": 32,
510
  "shards": 3
511
  },
512
  {
513
+ "seconds": 1.6838970150001842,
514
+ "bytes": 9992623,
515
+ "pages": 38,
516
+ "shards": 3
517
  },
518
  {
519
+ "seconds": 1.2029927740004496,
520
+ "bytes": 6068424,
521
  "pages": 37,
522
  "shards": 2
523
  },
524
  {
525
+ "seconds": 1.1870869580016006,
526
+ "bytes": 2666964,
527
+ "pages": 36,
528
+ "shards": 2
529
  },
530
  {
531
+ "seconds": 1.4462591800001974,
532
+ "bytes": 10549433,
533
+ "pages": 31,
534
  "shards": 3
535
  },
536
  {
537
+ "seconds": 1.0944702749984572,
538
+ "bytes": 2560030,
539
  "pages": 36,
540
  "shards": 2
541
  },
542
  {
543
+ "seconds": 1.3531785280028998,
544
+ "bytes": 11787584,
545
+ "pages": 32,
546
+ "shards": 4
547
  },
548
  {
549
+ "seconds": 1.0967427190007584,
550
+ "bytes": 6096101,
551
+ "pages": 36,
552
+ "shards": 2
553
  },
554
  {
555
+ "seconds": 1.3933399300003657,
556
+ "bytes": 2445888,
557
+ "pages": 28,
558
+ "shards": 2
559
  },
560
  {
561
+ "seconds": 1.9226194889997714,
562
+ "bytes": 9744014,
563
  "pages": 39,
564
  "shards": 3
565
  }
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:0b9d7502ee7a37b93921e50411dcf10cc48e5c0ea61bf54ea6e2aa3c774d65a7
3
- size 721794431
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:892eb6396b64cb4fd594382fdc135dcfdab02c54a38e9994701368b84bd73dcd
3
+ size 721770367