appvoid commited on
Commit
92c6015
·
verified ·
1 Parent(s): 4cb17bd

Qwen90M step 900000; 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": 895000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "eb78dafba7ed6edbbc1920e8d24942367894be7585480ad296217c19df2e72bd",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "3cc6aff6ad2f08dece94cc67d7a302ec23b409b0da9ac246329b68a9b3231985",
12
  "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
13
  "performance.json": "49b9042fb538cca213af1e6f33ceb5a89185fc6d2bab488be3f60e812871f343",
14
- "shard_scheduler.json": "2d68de6c3624467095a90ec9bdeef51e2fc5ab4e71e38707f6c6cc942ef1b8c8",
15
- "coverage.json": "da94cd20bec68c816560b61fce43b1062241b6873d42a2e73f505b9d916b5a6b",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"
 
1
  {
2
+ "step": 900000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "ef5b5f1c6524cbec6c0355105c9149bfd6ccef1a723a9955ad891300a75022f7",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "a13d4f5c6c016e1c39c58020b4df4b5b3ced492250e72ec4c4700bb5f859df8d",
12
  "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
13
  "performance.json": "49b9042fb538cca213af1e6f33ceb5a89185fc6d2bab488be3f60e812871f343",
14
+ "shard_scheduler.json": "213f04d8224b27e9be686fc21fcb231dfccd792d5d3627d641a20236fcb17ae9",
15
+ "coverage.json": "c02d09e33c53e7e8c69ad5bbdc1296f86fb7982b915ef76d15a2ae72ffb63bd6",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 895000,
3
  "stats": {
4
- "climbmix/blocks": 32220000,
5
- "climbmix/targets": 53861724331,
6
- "climbmix/input_tokens": 53865137686,
7
- "ultra_qa/blocks": 12082500,
8
- "ultra_qa/targets": 19711743279,
9
- "ultra_qa/input_tokens": 19711781154,
10
- "cortex/blocks": 12082500,
11
- "cortex/targets": 459427225,
12
- "cortex/input_tokens": 19205987571,
13
- "ultra_style/blocks": 12082500,
14
- "ultra_style/targets": 19110329995,
15
- "ultra_style/input_tokens": 19111332267,
16
- "openmath/blocks": 8055000,
17
- "openmath/targets": 12446359341,
18
- "openmath/input_tokens": 12450506245,
19
- "rewrite6/blocks": 4027500,
20
- "rewrite6/targets": 6602556148,
21
- "rewrite6/input_tokens": 6609794679
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 4412,
26
- "row": 2552,
27
  "start": {
28
- "file": 213,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 47,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
- "offset": 4096,
37
- "documents": 18074104
38
  },
39
  "rewrite6": {
40
- "page": 10,
41
- "row": 1253,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 40960
47
  }
48
  },
49
  "epoch": 50,
50
- "offset": 4096,
51
- "documents": 10312413
52
  },
53
  "openmath": {
54
- "page": 2757,
55
- "row": 3359,
56
  "start": {
57
- "file": 25,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 376,
61
- "row": 424
62
  }
63
  },
64
  "epoch": 0,
65
- "offset": 0,
66
- "documents": 11296031
67
  },
68
  "ultra_qa": {
69
- "page": 1192,
70
- "row": 1692,
71
  "start": {
72
  "file": 9,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 3,
76
- "row": 12028
77
  }
78
  },
79
  "epoch": 0,
80
- "offset": 4096,
81
- "documents": 4884124
82
  },
83
  "ultra_style": {
84
- "page": 1820,
85
- "row": 334,
86
  "start": {
87
  "file": 10,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 7,
91
- "row": 9554
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
- "documents": 7455054
97
  },
98
  "cortex": {
99
- "page": 234323,
100
- "row": 46,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 29993390
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 900000,
3
  "stats": {
4
+ "climbmix/blocks": 32400000,
5
+ "climbmix/targets": 54162259750,
6
+ "climbmix/input_tokens": 54165692291,
7
+ "ultra_qa/blocks": 12150000,
8
+ "ultra_qa/targets": 19821847028,
9
+ "ultra_qa/input_tokens": 19821885096,
10
+ "cortex/blocks": 12150000,
11
+ "cortex/targets": 461992778,
12
+ "cortex/input_tokens": 19313376697,
13
+ "ultra_style/blocks": 12150000,
14
+ "ultra_style/targets": 19217238700,
15
+ "ultra_style/input_tokens": 19218246348,
16
+ "openmath/blocks": 8100000,
17
+ "openmath/targets": 12515849799,
18
+ "openmath/input_tokens": 12520019747,
19
+ "rewrite6/blocks": 4050000,
20
+ "rewrite6/targets": 6639430765,
21
+ "rewrite6/input_tokens": 6646710396
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 4437,
26
+ "row": 1722,
27
  "start": {
28
+ "file": 214,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 65,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
+ "offset": 2048,
37
+ "documents": 18175674
38
  },
39
  "rewrite6": {
40
+ "page": 24,
41
+ "row": 2293,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 98304
47
  }
48
  },
49
  "epoch": 50,
50
+ "offset": 2048,
51
+ "documents": 10370797
52
  },
53
  "openmath": {
54
+ "page": 2773,
55
+ "row": 839,
56
  "start": {
57
+ "file": 26,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 5,
61
+ "row": 311
62
  }
63
  },
64
  "epoch": 0,
65
+ "offset": 2048,
66
+ "documents": 11359047
67
  },
68
  "ultra_qa": {
69
+ "page": 1199,
70
+ "row": 201,
71
  "start": {
72
  "file": 9,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 3,
76
+ "row": 40700
77
  }
78
  },
79
  "epoch": 0,
80
+ "offset": 2048,
81
+ "documents": 4911305
82
  },
83
  "ultra_style": {
84
+ "page": 1830,
85
+ "row": 482,
86
  "start": {
87
  "file": 10,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 7,
91
+ "row": 50514
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
+ "documents": 7496162
97
  },
98
  "cortex": {
99
+ "page": 235633,
100
+ "row": 81,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 30161105
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 895000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "eb78dafba7ed6edbbc1920e8d24942367894be7585480ad296217c19df2e72bd", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "3cc6aff6ad2f08dece94cc67d7a302ec23b409b0da9ac246329b68a9b3231985", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "49b9042fb538cca213af1e6f33ceb5a89185fc6d2bab488be3f60e812871f343", "shard_scheduler.json": "2d68de6c3624467095a90ec9bdeef51e2fc5ab4e71e38707f6c6cc942ef1b8c8", "coverage.json": "da94cd20bec68c816560b61fce43b1062241b6873d42a2e73f505b9d916b5a6b", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000895000"}
 
1
+ {"step": 900000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "ef5b5f1c6524cbec6c0355105c9149bfd6ccef1a723a9955ad891300a75022f7", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "a13d4f5c6c016e1c39c58020b4df4b5b3ced492250e72ec4c4700bb5f859df8d", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "49b9042fb538cca213af1e6f33ceb5a89185fc6d2bab488be3f60e812871f343", "shard_scheduler.json": "213f04d8224b27e9be686fc21fcb231dfccd792d5d3627d641a20236fcb17ae9", "coverage.json": "c02d09e33c53e7e8c69ad5bbdc1296f86fb7982b915ef76d15a2ae72ffb63bd6", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000900000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:eb78dafba7ed6edbbc1920e8d24942367894be7585480ad296217c19df2e72bd
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ef5b5f1c6524cbec6c0355105c9149bfd6ccef1a723a9955ad891300a75022f7
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": 76601.26916859495,
26
- "required_pages_per_second_upper_estimate": 0.005299453540515585
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": 9575.15864607437,
40
- "required_pages_per_second_upper_estimate": 0.004008918973061503
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": 19150.31729214874,
54
- "required_pages_per_second_upper_estimate": 0.003988506064125937
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": 28725.47593822311,
68
- "required_pages_per_second_upper_estimate": 0.0017220484212343672
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": 28725.47593822311,
82
- "required_pages_per_second_upper_estimate": 0.0022332794041812756
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": 28725.47593822311,
96
- "required_pages_per_second_upper_estimate": 0.36845956232248317
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,410 +158,410 @@
158
  "prefetch_batches": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
- "producer_batch_seconds": 0.09046511610937102,
162
- "producer_p95_seconds": 0.19785067599877948,
163
- "gpu_update_seconds": 0.9624905801198795,
164
- "producer_batches_per_second": 11.053984596571063,
165
- "gpu_batches_per_second": 1.0389712072563335,
166
- "producer_headroom": 10.639356046989894,
167
  "sustainable": true,
168
- "buffer_seconds": 60.636906547552414,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 80,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 539194.1073240534,
174
- "upload_bytes_per_second": 5396381.570381515,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 115.6985036353226,
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": 1.0580149020001954,
184
- "bytes": 6239919,
185
  "pages": 36,
186
  "shards": 2
187
  },
188
  {
189
- "seconds": 0.9808759570000802,
190
- "bytes": 5815079,
191
- "pages": 28,
192
  "shards": 2
193
  },
194
  {
195
- "seconds": 1.6121151610000197,
196
- "bytes": 12779121,
197
- "pages": 35,
198
- "shards": 4
199
  },
200
  {
201
- "seconds": 1.091388643999835,
202
- "bytes": 6015354,
203
- "pages": 36,
204
  "shards": 2
205
  },
206
  {
207
- "seconds": 1.609872815000017,
208
- "bytes": 7812095,
209
- "pages": 39,
210
- "shards": 3
211
- },
212
- {
213
- "seconds": 0.9189778770000885,
214
- "bytes": 1227062,
215
- "pages": 33,
216
  "shards": 1
217
  },
218
  {
219
- "seconds": 1.2558460740001465,
220
- "bytes": 9834347,
221
- "pages": 36,
222
- "shards": 3
223
  },
224
  {
225
- "seconds": 1.1021917439998106,
226
- "bytes": 2521958,
227
- "pages": 32,
228
  "shards": 2
229
  },
230
  {
231
- "seconds": 1.2651067010001498,
232
- "bytes": 11919197,
233
- "pages": 34,
234
- "shards": 3
235
  },
236
  {
237
- "seconds": 1.3387522560001344,
238
- "bytes": 10892645,
239
- "pages": 35,
240
- "shards": 3
241
  },
242
  {
243
- "seconds": 1.3189305679998142,
244
- "bytes": 2654594,
245
- "pages": 37,
246
- "shards": 2
247
  },
248
  {
249
- "seconds": 1.483484325000063,
250
- "bytes": 6080107,
251
- "pages": 36,
252
- "shards": 2
253
  },
254
  {
255
- "seconds": 1.1274267459998555,
256
- "bytes": 2324297,
257
- "pages": 28,
258
- "shards": 2
259
  },
260
  {
261
- "seconds": 1.315684739000062,
262
- "bytes": 5833727,
263
- "pages": 30,
264
- "shards": 2
265
  },
266
  {
267
- "seconds": 1.434490552999705,
268
- "bytes": 16176474,
269
- "pages": 43,
270
  "shards": 4
271
  },
272
  {
273
- "seconds": 1.33469606700055,
274
- "bytes": 2416191,
275
- "pages": 31,
276
  "shards": 2
277
  },
278
  {
279
- "seconds": 1.1527810859997771,
280
- "bytes": 6214716,
281
- "pages": 33,
 
 
 
 
 
 
282
  "shards": 2
283
  },
284
  {
285
- "seconds": 1.56234888400013,
286
- "bytes": 11134116,
287
- "pages": 40,
288
  "shards": 4
289
  },
290
  {
291
- "seconds": 1.07002194100005,
292
- "bytes": 1145768,
293
- "pages": 34,
294
- "shards": 1
295
  },
296
  {
297
- "seconds": 1.216849121999985,
298
- "bytes": 5739383,
299
  "pages": 27,
300
  "shards": 2
301
  },
302
  {
303
- "seconds": 1.4537456819998624,
304
- "bytes": 13399952,
305
- "pages": 41,
306
- "shards": 4
307
- },
308
- {
309
- "seconds": 1.1661972729998524,
310
- "bytes": 5666265,
311
- "pages": 35,
312
  "shards": 2
313
  },
314
  {
315
- "seconds": 1.290868875000342,
316
- "bytes": 7144629,
317
- "pages": 30,
318
  "shards": 3
319
  },
320
  {
321
- "seconds": 1.0513503590000255,
322
- "bytes": 1139099,
323
- "pages": 34,
324
  "shards": 2
325
  },
326
  {
327
- "seconds": 0.996597190999637,
328
- "bytes": 6046442,
329
- "pages": 28,
 
 
 
 
 
 
330
  "shards": 2
331
  },
332
  {
333
- "seconds": 1.1579751299996133,
334
- "bytes": 6933490,
335
  "pages": 30,
336
- "shards": 3
337
  },
338
  {
339
- "seconds": 1.4397287989995675,
340
- "bytes": 11874214,
341
- "pages": 32,
342
  "shards": 3
343
  },
344
  {
345
- "seconds": 1.0771185140001762,
346
- "bytes": 6044727,
347
- "pages": 37,
348
  "shards": 2
349
  },
350
  {
351
- "seconds": 1.0941710249999232,
352
- "bytes": 2736868,
353
- "pages": 37,
354
  "shards": 2
355
  },
356
  {
357
- "seconds": 1.062443419000374,
358
- "bytes": 5796244,
359
- "pages": 20,
360
- "shards": 2
361
  },
362
  {
363
- "seconds": 1.0528361209999275,
364
- "bytes": 5176693,
365
- "pages": 44,
366
- "shards": 2
367
  },
368
  {
369
- "seconds": 1.123353902999952,
370
- "bytes": 7475982,
371
- "pages": 39,
372
  "shards": 3
373
  },
374
  {
375
- "seconds": 1.3653379679999489,
376
- "bytes": 11617938,
377
- "pages": 31,
378
- "shards": 3
379
  },
380
  {
381
- "seconds": 1.2881032879995473,
382
- "bytes": 6772198,
383
- "pages": 31,
384
  "shards": 3
385
  },
386
  {
387
- "seconds": 1.1199771049996343,
388
- "bytes": 6074153,
389
  "pages": 36,
390
  "shards": 2
391
  },
392
  {
393
- "seconds": 1.1042831229997319,
394
- "bytes": 6145766,
395
  "pages": 36,
396
- "shards": 2
397
  },
398
  {
399
- "seconds": 1.2581367550001232,
400
- "bytes": 2542671,
401
- "pages": 33,
402
  "shards": 2
403
  },
404
  {
405
- "seconds": 1.7602326240003094,
406
- "bytes": 10170709,
407
- "pages": 35,
408
  "shards": 3
409
  },
410
  {
411
- "seconds": 1.23988030000055,
412
- "bytes": 8293432,
413
  "pages": 33,
414
- "shards": 3
415
  },
416
  {
417
- "seconds": 1.1932894399997167,
418
- "bytes": 6531618,
419
- "pages": 36,
420
  "shards": 2
421
  },
422
  {
423
- "seconds": 1.1864704220006388,
424
- "bytes": 6136727,
425
- "pages": 35,
426
  "shards": 2
427
  },
428
  {
429
- "seconds": 1.2697396670000671,
430
- "bytes": 6182466,
431
- "pages": 33,
432
- "shards": 3
433
  },
434
  {
435
- "seconds": 1.2363136280000617,
436
- "bytes": 6022734,
437
  "pages": 37,
438
  "shards": 2
439
  },
440
  {
441
- "seconds": 1.2937560119999034,
442
- "bytes": 947666,
443
- "pages": 30,
 
 
 
 
 
 
444
  "shards": 1
445
  },
446
  {
447
- "seconds": 1.5761228700002903,
448
- "bytes": 18281686,
449
- "pages": 39,
450
- "shards": 5
451
  },
452
  {
453
- "seconds": 1.1403170100002171,
454
- "bytes": 5965661,
455
- "pages": 36,
456
- "shards": 2
457
  },
458
  {
459
- "seconds": 1.088612870000361,
460
- "bytes": 2479743,
461
- "pages": 29,
462
- "shards": 2
463
  },
464
  {
465
- "seconds": 1.3296366939994186,
466
- "bytes": 6147791,
467
- "pages": 36,
468
  "shards": 2
469
  },
470
  {
471
- "seconds": 1.1267749469989212,
472
- "bytes": 894977,
473
- "pages": 27,
474
- "shards": 1
475
  },
476
  {
477
- "seconds": 1.4399002789996302,
478
- "bytes": 10842331,
479
- "pages": 40,
480
- "shards": 4
481
  },
482
  {
483
- "seconds": 1.632049243999063,
484
- "bytes": 11967346,
485
- "pages": 40,
486
  "shards": 3
487
  },
488
  {
489
- "seconds": 0.9512228559997311,
490
- "bytes": 901267,
491
- "pages": 26,
492
  "shards": 1
493
  },
494
  {
495
- "seconds": 1.6983987419989717,
496
- "bytes": 12559131,
497
- "pages": 36,
498
  "shards": 4
499
  },
500
  {
501
- "seconds": 1.1707839329992566,
502
- "bytes": 6060902,
503
- "pages": 32,
504
- "shards": 2
505
- },
506
- {
507
- "seconds": 1.1801391750013863,
508
- "bytes": 2749619,
509
- "pages": 37,
510
  "shards": 3
511
  },
512
  {
513
- "seconds": 1.1686728980002954,
514
- "bytes": 6121581,
515
- "pages": 36,
516
- "shards": 2
517
  },
518
  {
519
- "seconds": 1.3895534080002108,
520
- "bytes": 15955272,
521
- "pages": 35,
522
  "shards": 4
523
  },
524
  {
525
- "seconds": 0.9986490080009389,
526
- "bytes": 2653371,
527
- "pages": 37,
528
- "shards": 2
529
  },
530
  {
531
- "seconds": 1.1648389490001136,
532
- "bytes": 5951393,
533
- "pages": 27,
534
  "shards": 2
535
  },
536
  {
537
- "seconds": 1.1927761070000997,
538
- "bytes": 2669572,
539
  "pages": 36,
540
- "shards": 2
541
  },
542
  {
543
- "seconds": 1.1629383399995277,
544
- "bytes": 10339749,
545
- "pages": 40,
546
- "shards": 3
547
  },
548
  {
549
- "seconds": 1.6129309269999794,
550
- "bytes": 5876330,
551
- "pages": 28,
552
- "shards": 2
553
  },
554
  {
555
- "seconds": 1.7068213439997635,
556
- "bytes": 8291481,
557
- "pages": 35,
558
- "shards": 3
559
  },
560
  {
561
- "seconds": 1.2685745549988496,
562
- "bytes": 6220542,
563
- "pages": 34,
564
- "shards": 2
565
  }
566
  ],
567
  "cache_key": {
 
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": 77365.33532253442,
26
+ "required_pages_per_second_upper_estimate": 0.005352313409922849
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": 9670.666915316802,
40
+ "required_pages_per_second_upper_estimate": 0.004048906290953853
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": 19341.333830633605,
54
+ "required_pages_per_second_upper_estimate": 0.004028289771647462
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": 29012.00074595041,
68
+ "required_pages_per_second_upper_estimate": 0.0017392251459595694
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": 29012.00074595041,
82
+ "required_pages_per_second_upper_estimate": 0.002255555447692636
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": 29012.00074595041,
96
+ "required_pages_per_second_upper_estimate": 0.3721347949096395
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": 0.10793024863289702,
162
+ "producer_p95_seconds": 0.2442270199990162,
163
+ "gpu_update_seconds": 0.9529849472328861,
164
+ "producer_batches_per_second": 9.265243179428767,
165
+ "gpu_batches_per_second": 1.0493345177210072,
166
+ "producer_headroom": 8.82963728244778,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.038051675671824,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 83,
172
  "max_pending_pages": 4096,
173
+ "generation_bytes_per_second": 398086.69224948075,
174
+ "upload_bytes_per_second": 5232468.799184691,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 85.42013717282597,
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": 1.1932894399997167,
184
+ "bytes": 6531618,
185
  "pages": 36,
186
  "shards": 2
187
  },
188
  {
189
+ "seconds": 1.1864704220006388,
190
+ "bytes": 6136727,
191
+ "pages": 35,
192
  "shards": 2
193
  },
194
  {
195
+ "seconds": 1.2697396670000671,
196
+ "bytes": 6182466,
197
+ "pages": 33,
198
+ "shards": 3
199
  },
200
  {
201
+ "seconds": 1.2363136280000617,
202
+ "bytes": 6022734,
203
+ "pages": 37,
204
  "shards": 2
205
  },
206
  {
207
+ "seconds": 1.2937560119999034,
208
+ "bytes": 947666,
209
+ "pages": 30,
 
 
 
 
 
 
210
  "shards": 1
211
  },
212
  {
213
+ "seconds": 1.5761228700002903,
214
+ "bytes": 18281686,
215
+ "pages": 39,
216
+ "shards": 5
217
  },
218
  {
219
+ "seconds": 1.1403170100002171,
220
+ "bytes": 5965661,
221
+ "pages": 36,
222
  "shards": 2
223
  },
224
  {
225
+ "seconds": 1.088612870000361,
226
+ "bytes": 2479743,
227
+ "pages": 29,
228
+ "shards": 2
229
  },
230
  {
231
+ "seconds": 1.3296366939994186,
232
+ "bytes": 6147791,
233
+ "pages": 36,
234
+ "shards": 2
235
  },
236
  {
237
+ "seconds": 1.1267749469989212,
238
+ "bytes": 894977,
239
+ "pages": 27,
240
+ "shards": 1
241
  },
242
  {
243
+ "seconds": 1.4399002789996302,
244
+ "bytes": 10842331,
245
+ "pages": 40,
246
+ "shards": 4
247
  },
248
  {
249
+ "seconds": 1.632049243999063,
250
+ "bytes": 11967346,
251
+ "pages": 40,
252
+ "shards": 3
253
  },
254
  {
255
+ "seconds": 0.9512228559997311,
256
+ "bytes": 901267,
257
+ "pages": 26,
258
+ "shards": 1
259
  },
260
  {
261
+ "seconds": 1.6983987419989717,
262
+ "bytes": 12559131,
263
+ "pages": 36,
264
  "shards": 4
265
  },
266
  {
267
+ "seconds": 1.1707839329992566,
268
+ "bytes": 6060902,
269
+ "pages": 32,
270
  "shards": 2
271
  },
272
  {
273
+ "seconds": 1.1801391750013863,
274
+ "bytes": 2749619,
275
+ "pages": 37,
276
+ "shards": 3
277
+ },
278
+ {
279
+ "seconds": 1.1686728980002954,
280
+ "bytes": 6121581,
281
+ "pages": 36,
282
  "shards": 2
283
  },
284
  {
285
+ "seconds": 1.3895534080002108,
286
+ "bytes": 15955272,
287
+ "pages": 35,
288
  "shards": 4
289
  },
290
  {
291
+ "seconds": 0.9986490080009389,
292
+ "bytes": 2653371,
293
+ "pages": 37,
294
+ "shards": 2
295
  },
296
  {
297
+ "seconds": 1.1648389490001136,
298
+ "bytes": 5951393,
299
  "pages": 27,
300
  "shards": 2
301
  },
302
  {
303
+ "seconds": 1.1927761070000997,
304
+ "bytes": 2669572,
305
+ "pages": 36,
 
 
 
 
 
 
306
  "shards": 2
307
  },
308
  {
309
+ "seconds": 1.1629383399995277,
310
+ "bytes": 10339749,
311
+ "pages": 40,
312
  "shards": 3
313
  },
314
  {
315
+ "seconds": 1.6129309269999794,
316
+ "bytes": 5876330,
317
+ "pages": 28,
318
  "shards": 2
319
  },
320
  {
321
+ "seconds": 1.7068213439997635,
322
+ "bytes": 8291481,
323
+ "pages": 35,
324
+ "shards": 3
325
+ },
326
+ {
327
+ "seconds": 1.2685745549988496,
328
+ "bytes": 6220542,
329
+ "pages": 34,
330
  "shards": 2
331
  },
332
  {
333
+ "seconds": 1.144102094000118,
334
+ "bytes": 5891615,
335
  "pages": 30,
336
+ "shards": 2
337
  },
338
  {
339
+ "seconds": 1.2515704560009908,
340
+ "bytes": 7512167,
341
+ "pages": 38,
342
  "shards": 3
343
  },
344
  {
345
+ "seconds": 1.4014940459983336,
346
+ "bytes": 5931776,
347
+ "pages": 29,
348
  "shards": 2
349
  },
350
  {
351
+ "seconds": 1.1321277869992628,
352
+ "bytes": 2736110,
353
+ "pages": 36,
354
  "shards": 2
355
  },
356
  {
357
+ "seconds": 1.411381372001415,
358
+ "bytes": 15991084,
359
+ "pages": 40,
360
+ "shards": 4
361
  },
362
  {
363
+ "seconds": 0.8898140989986132,
364
+ "bytes": 1156490,
365
+ "pages": 32,
366
+ "shards": 1
367
  },
368
  {
369
+ "seconds": 1.3950513529998716,
370
+ "bytes": 7312918,
371
+ "pages": 32,
372
  "shards": 3
373
  },
374
  {
375
+ "seconds": 1.2647196729994903,
376
+ "bytes": 6433157,
377
+ "pages": 36,
378
+ "shards": 2
379
  },
380
  {
381
+ "seconds": 1.2063595479994547,
382
+ "bytes": 6907543,
383
+ "pages": 33,
384
  "shards": 3
385
  },
386
  {
387
+ "seconds": 1.290804393000144,
388
+ "bytes": 6092692,
389
  "pages": 36,
390
  "shards": 2
391
  },
392
  {
393
+ "seconds": 1.4008768420007982,
394
+ "bytes": 11876531,
395
  "pages": 36,
396
+ "shards": 3
397
  },
398
  {
399
+ "seconds": 0.9798814419991686,
400
+ "bytes": 2581068,
401
+ "pages": 36,
402
  "shards": 2
403
  },
404
  {
405
+ "seconds": 1.9553985610000382,
406
+ "bytes": 10094743,
407
+ "pages": 34,
408
  "shards": 3
409
  },
410
  {
411
+ "seconds": 1.0605444640004862,
412
+ "bytes": 2480385,
413
  "pages": 33,
414
+ "shards": 2
415
  },
416
  {
417
+ "seconds": 1.4249819679989741,
418
+ "bytes": 6270908,
419
+ "pages": 32,
420
  "shards": 2
421
  },
422
  {
423
+ "seconds": 1.2829982630009908,
424
+ "bytes": 5992379,
425
+ "pages": 37,
426
  "shards": 2
427
  },
428
  {
429
+ "seconds": 1.3964128229999915,
430
+ "bytes": 12688722,
431
+ "pages": 35,
432
+ "shards": 4
433
  },
434
  {
435
+ "seconds": 1.4358708510008,
436
+ "bytes": 6164418,
437
  "pages": 37,
438
  "shards": 2
439
  },
440
  {
441
+ "seconds": 1.5897931969993806,
442
+ "bytes": 7707081,
443
+ "pages": 36,
444
+ "shards": 3
445
+ },
446
+ {
447
+ "seconds": 1.0018968719996337,
448
+ "bytes": 927002,
449
+ "pages": 29,
450
  "shards": 1
451
  },
452
  {
453
+ "seconds": 1.3852189129993349,
454
+ "bytes": 10288556,
455
+ "pages": 35,
456
+ "shards": 3
457
  },
458
  {
459
+ "seconds": 1.2743636189989047,
460
+ "bytes": 2631065,
461
+ "pages": 37,
462
+ "shards": 3
463
  },
464
  {
465
+ "seconds": 1.369053440999778,
466
+ "bytes": 11976832,
467
+ "pages": 37,
468
+ "shards": 3
469
  },
470
  {
471
+ "seconds": 1.3680221860013262,
472
+ "bytes": 5930978,
473
+ "pages": 33,
474
  "shards": 2
475
  },
476
  {
477
+ "seconds": 1.3103218420001213,
478
+ "bytes": 6746870,
479
+ "pages": 31,
480
+ "shards": 3
481
  },
482
  {
483
+ "seconds": 1.4409255560003658,
484
+ "bytes": 6062763,
485
+ "pages": 36,
486
+ "shards": 2
487
  },
488
  {
489
+ "seconds": 1.1668520560015168,
490
+ "bytes": 7696896,
491
+ "pages": 38,
492
  "shards": 3
493
  },
494
  {
495
+ "seconds": 1.1897505889992317,
496
+ "bytes": 1090330,
497
+ "pages": 30,
498
  "shards": 1
499
  },
500
  {
501
+ "seconds": 1.3477415189991007,
502
+ "bytes": 15713674,
503
+ "pages": 35,
504
  "shards": 4
505
  },
506
  {
507
+ "seconds": 1.2074203040010616,
508
+ "bytes": 7463350,
509
+ "pages": 39,
 
 
 
 
 
 
510
  "shards": 3
511
  },
512
  {
513
+ "seconds": 1.190478869000799,
514
+ "bytes": 1226986,
515
+ "pages": 34,
516
+ "shards": 1
517
  },
518
  {
519
+ "seconds": 1.5480925499996374,
520
+ "bytes": 11721285,
521
+ "pages": 32,
522
  "shards": 4
523
  },
524
  {
525
+ "seconds": 1.0147747470000468,
526
+ "bytes": 1162064,
527
+ "pages": 33,
528
+ "shards": 1
529
  },
530
  {
531
+ "seconds": 1.132629542000359,
532
+ "bytes": 6067377,
533
+ "pages": 33,
534
  "shards": 2
535
  },
536
  {
537
+ "seconds": 1.3739411900005507,
538
+ "bytes": 13347857,
539
  "pages": 36,
540
+ "shards": 4
541
  },
542
  {
543
+ "seconds": 1.5033581300012884,
544
+ "bytes": 979335,
545
+ "pages": 32,
546
+ "shards": 1
547
  },
548
  {
549
+ "seconds": 1.363758608000353,
550
+ "bytes": 11974676,
551
+ "pages": 44,
552
+ "shards": 4
553
  },
554
  {
555
+ "seconds": 1.1600040840003203,
556
+ "bytes": 6119662,
557
+ "pages": 36,
558
+ "shards": 2
559
  },
560
  {
561
+ "seconds": 0.8502458040002239,
562
+ "bytes": 891680,
563
+ "pages": 27,
564
+ "shards": 1
565
  }
566
  ],
567
  "cache_key": {
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:3cc6aff6ad2f08dece94cc67d7a302ec23b409b0da9ac246329b68a9b3231985
3
- size 721774207
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a13d4f5c6c016e1c39c58020b4df4b5b3ced492250e72ec4c4700bb5f859df8d
3
+ size 721766271