appvoid commited on
Commit
65772bf
·
verified ·
1 Parent(s): 3233289

Qwen90M step 720000; complete resumable checkpoint

Browse files
Files changed (6) hide show
  1. complete.json +5 -5
  2. coverage.json +48 -48
  3. latest.json +1 -1
  4. model.safetensors +1 -1
  5. shard_scheduler.json +246 -246
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 715000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "b8c254cd3e3df27a533324f081bcea4d5dc590522534c2d49abee21dbe8f10c8",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "08ba2dd523ece7c3398532396b6dd7590d07e9be7249052a8b8cf4f84043c025",
12
  "training_config.json": "78af77f846fb3dcba08002ed692d24a6e4b3900ba35e5592c606aca3466b41f8",
13
  "performance.json": "96c7d3993f53da33aeba321717c341456e5878e35efa676b6ffe17d014497e29",
14
- "shard_scheduler.json": "c6173291cb813b7c0485c738e6b5e7c240ea2f582ce47fb3f2c695916a22695f",
15
- "coverage.json": "e0b100ec4fa9dab8fcb0642981922fafa82749f90200b316989ed84ba5d792fc",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "1dee8333ad826e9fb3d197c33535515cdb9de560b94f624cae15e044b464f0ac"
 
1
  {
2
+ "step": 720000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "982ad5bd645ce1303553abc60578335d78399b9efe72845439bb9f2e39966f3b",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "8c7cadbc8ada2e114351e38f2dc65770375fbd3c5571e7a3db009d13ce7e870a",
12
  "training_config.json": "78af77f846fb3dcba08002ed692d24a6e4b3900ba35e5592c606aca3466b41f8",
13
  "performance.json": "96c7d3993f53da33aeba321717c341456e5878e35efa676b6ffe17d014497e29",
14
+ "shard_scheduler.json": "21430a5915815eb5e02a03ff691f8f7587a8b2a3d2a94f5f01d048866c5097af",
15
+ "coverage.json": "480e76c148c2db0ee587dc2579d82fe33abc92e0d0c035547dd68eda66cebf1c",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "1dee8333ad826e9fb3d197c33535515cdb9de560b94f624cae15e044b464f0ac"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 715000,
3
  "stats": {
4
- "climbmix/blocks": 25740000,
5
- "climbmix/targets": 43029519747,
6
- "climbmix/input_tokens": 43032246779,
7
- "ultra_qa/blocks": 9652500,
8
- "ultra_qa/targets": 15747139609,
9
- "ultra_qa/input_tokens": 15747170018,
10
- "cortex/blocks": 9652500,
11
- "cortex/targets": 367020487,
12
- "cortex/input_tokens": 15343129639,
13
- "ultra_style/blocks": 9652500,
14
- "ultra_style/targets": 15267128059,
15
- "ultra_style/input_tokens": 15267929164,
16
- "openmath/blocks": 6435000,
17
- "openmath/targets": 9943322469,
18
- "openmath/input_tokens": 9946635258,
19
- "rewrite6/blocks": 3217500,
20
- "rewrite6/targets": 5274681721,
21
- "rewrite6/input_tokens": 5280464589
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 3525,
26
- "row": 1183,
27
  "start": {
28
- "file": 170,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 50,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
- "offset": 38912,
37
- "documents": 14439583
38
  },
39
  "rewrite6": {
40
- "page": 5,
41
- "row": 1882,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 20480
47
  }
48
  },
49
  "epoch": 40,
50
- "offset": 2048,
51
- "documents": 8238522
52
  },
53
  "openmath": {
54
- "page": 2203,
55
- "row": 534,
56
  "start": {
57
  "file": 20,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 290,
61
- "row": 488
62
  }
63
  },
64
  "epoch": 0,
65
- "offset": 2048,
66
- "documents": 9024022
67
  },
68
  "ultra_qa": {
69
- "page": 952,
70
- "row": 2369,
71
  "start": {
72
  "file": 7,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 4,
76
- "row": 3902
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
- "documents": 3901761
82
  },
83
  "ultra_style": {
84
- "page": 1454,
85
- "row": 163,
86
  "start": {
87
  "file": 8,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
- "row_group": 5,
91
- "row": 49621
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
- "documents": 5955747
97
  },
98
  "cortex": {
99
- "page": 187198,
100
- "row": 117,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 23961461
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 720000,
3
  "stats": {
4
+ "climbmix/blocks": 25920000,
5
+ "climbmix/targets": 43330222714,
6
+ "climbmix/input_tokens": 43332969001,
7
+ "ultra_qa/blocks": 9720000,
8
+ "ultra_qa/targets": 15857175972,
9
+ "ultra_qa/input_tokens": 15857206570,
10
+ "cortex/blocks": 9720000,
11
+ "cortex/targets": 369582035,
12
+ "cortex/input_tokens": 15450455935,
13
+ "ultra_style/blocks": 9720000,
14
+ "ultra_style/targets": 15374177136,
15
+ "ultra_style/input_tokens": 15374983563,
16
+ "openmath/blocks": 6480000,
17
+ "openmath/targets": 10012814408,
18
+ "openmath/input_tokens": 10016150211,
19
+ "rewrite6/blocks": 3240000,
20
+ "rewrite6/targets": 5311509903,
21
+ "rewrite6/input_tokens": 5317333313
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 3550,
26
+ "row": 129,
27
  "start": {
28
+ "file": 171,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 68,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
+ "offset": 0,
37
+ "documents": 14540929
38
  },
39
  "rewrite6": {
40
+ "page": 19,
41
+ "row": 2309,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 77824
47
  }
48
  },
49
  "epoch": 40,
50
+ "offset": 0,
51
+ "documents": 8296293
52
  },
53
  "openmath": {
54
+ "page": 2218,
55
+ "row": 1902,
56
  "start": {
57
  "file": 20,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 351,
61
+ "row": 928
62
  }
63
  },
64
  "epoch": 0,
65
+ "offset": 0,
66
+ "documents": 9086830
67
  },
68
  "ultra_qa": {
69
+ "page": 959,
70
+ "row": 986,
71
  "start": {
72
  "file": 7,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 4,
76
+ "row": 32574
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
+ "documents": 3929050
82
  },
83
  "ultra_style": {
84
+ "page": 1464,
85
+ "row": 135,
86
  "start": {
87
  "file": 8,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
+ "row_group": 6,
91
+ "row": 6394
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
+ "documents": 5996679
97
  },
98
  "cortex": {
99
+ "page": 188509,
100
+ "row": 47,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 24129199
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 715000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "b8c254cd3e3df27a533324f081bcea4d5dc590522534c2d49abee21dbe8f10c8", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "08ba2dd523ece7c3398532396b6dd7590d07e9be7249052a8b8cf4f84043c025", "training_config.json": "78af77f846fb3dcba08002ed692d24a6e4b3900ba35e5592c606aca3466b41f8", "performance.json": "96c7d3993f53da33aeba321717c341456e5878e35efa676b6ffe17d014497e29", "shard_scheduler.json": "c6173291cb813b7c0485c738e6b5e7c240ea2f582ce47fb3f2c695916a22695f", "coverage.json": "e0b100ec4fa9dab8fcb0642981922fafa82749f90200b316989ed84ba5d792fc", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "1dee8333ad826e9fb3d197c33535515cdb9de560b94f624cae15e044b464f0ac", "folder": "step-000000715000"}
 
1
+ {"step": 720000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "982ad5bd645ce1303553abc60578335d78399b9efe72845439bb9f2e39966f3b", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "8c7cadbc8ada2e114351e38f2dc65770375fbd3c5571e7a3db009d13ce7e870a", "training_config.json": "78af77f846fb3dcba08002ed692d24a6e4b3900ba35e5592c606aca3466b41f8", "performance.json": "96c7d3993f53da33aeba321717c341456e5878e35efa676b6ffe17d014497e29", "shard_scheduler.json": "21430a5915815eb5e02a03ff691f8f7587a8b2a3d2a94f5f01d048866c5097af", "coverage.json": "480e76c148c2db0ee587dc2579d82fe33abc92e0d0c035547dd68eda66cebf1c", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "1dee8333ad826e9fb3d197c33535515cdb9de560b94f624cae15e044b464f0ac", "folder": "step-000000720000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b8c254cd3e3df27a533324f081bcea4d5dc590522534c2d49abee21dbe8f10c8
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:982ad5bd645ce1303553abc60578335d78399b9efe72845439bb9f2e39966f3b
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": 77176.89841787542,
26
- "required_pages_per_second_upper_estimate": 0.005339276907624682
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": 9647.112302234427,
40
- "required_pages_per_second_upper_estimate": 0.004039044466332516
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": 19294.224604468855,
54
- "required_pages_per_second_upper_estimate": 0.004018478162191183
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": 28941.33690670328,
68
- "required_pages_per_second_upper_estimate": 0.001734988956694139
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": 28941.33690670328,
82
- "required_pages_per_second_upper_estimate": 0.002250061645008552
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": 28941.33690670328,
96
- "required_pages_per_second_upper_estimate": 0.3712283950526966
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": 0.09608320560147376,
162
- "producer_p95_seconds": 0.27539721799985273,
163
- "gpu_update_seconds": 0.9553117773766794,
164
- "producer_batches_per_second": 10.407646099441353,
165
- "gpu_batches_per_second": 1.0467786786278674,
166
- "producer_headroom": 9.942546893564783,
167
  "sustainable": true,
168
- "buffer_seconds": 60.1846419747308,
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": 397695.1736651185,
174
- "upload_bytes_per_second": 4823252.873992993,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 85.33612639870803,
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": 1.369231013002718,
184
- "bytes": 10621268,
185
- "pages": 30,
186
- "shards": 3
187
- },
188
- {
189
- "seconds": 0.9135342390000005,
190
- "bytes": 1211398,
191
- "pages": 33,
192
  "shards": 1
193
  },
194
  {
195
- "seconds": 1.2419597409971175,
196
- "bytes": 13297063,
197
- "pages": 34,
198
- "shards": 4
199
  },
200
  {
201
- "seconds": 1.2846820170016144,
202
- "bytes": 6246650,
203
- "pages": 37,
204
- "shards": 2
205
  },
206
  {
207
- "seconds": 1.021862665998924,
208
- "bytes": 6426718,
209
- "pages": 33,
210
  "shards": 3
211
  },
212
  {
213
- "seconds": 0.9455101860003197,
214
- "bytes": 6297280,
215
- "pages": 36,
216
- "shards": 2
217
  },
218
  {
219
- "seconds": 1.347191354001552,
220
- "bytes": 5982017,
221
- "pages": 33,
222
- "shards": 2
223
  },
224
  {
225
- "seconds": 3.3193765439973504,
226
- "bytes": 2461242,
227
- "pages": 33,
228
  "shards": 2
229
  },
230
  {
231
- "seconds": 1.3424353250011336,
232
- "bytes": 15776988,
233
- "pages": 33,
234
- "shards": 4
235
  },
236
  {
237
- "seconds": 1.047338389998913,
238
- "bytes": 2628784,
239
- "pages": 38,
240
- "shards": 2
241
  },
242
  {
243
- "seconds": 1.2893916930006526,
244
- "bytes": 6001693,
245
- "pages": 33,
246
- "shards": 2
247
  },
248
  {
249
- "seconds": 1.5835260300009395,
250
- "bytes": 761598,
251
- "pages": 23,
252
  "shards": 1
253
  },
254
  {
255
- "seconds": 1.654032960999757,
256
- "bytes": 12595480,
257
- "pages": 39,
258
- "shards": 4
259
  },
260
  {
261
- "seconds": 1.5982389480013808,
262
- "bytes": 6084209,
263
- "pages": 37,
264
  "shards": 2
265
  },
266
  {
267
- "seconds": 0.98295405799945,
268
- "bytes": 6764576,
269
- "pages": 29,
270
  "shards": 2
271
  },
272
  {
273
- "seconds": 1.2006459389995143,
274
- "bytes": 7651087,
275
- "pages": 39,
276
  "shards": 3
277
  },
278
  {
279
- "seconds": 1.3339825349976309,
280
- "bytes": 10020438,
281
- "pages": 33,
282
  "shards": 3
283
  },
284
  {
285
- "seconds": 0.850007712997467,
286
- "bytes": 1050252,
287
- "pages": 30,
288
- "shards": 1
289
  },
290
  {
291
- "seconds": 1.2932650000002468,
292
- "bytes": 7701652,
293
- "pages": 38,
294
  "shards": 3
295
  },
296
  {
297
- "seconds": 1.022619493000093,
298
- "bytes": 1029637,
299
  "pages": 33,
300
- "shards": 1
301
- },
302
- {
303
- "seconds": 1.0783612599989283,
304
- "bytes": 5165514,
305
- "pages": 4,
306
- "shards": 2
307
- },
308
- {
309
- "seconds": 1.7178185550001217,
310
- "bytes": 4032890,
311
- "pages": 3,
312
  "shards": 2
313
  },
314
  {
315
- "seconds": 1.248450455001148,
316
- "bytes": 5769233,
317
- "pages": 1,
318
- "shards": 1
319
  },
320
  {
321
- "seconds": 1.0994338210002752,
322
- "bytes": 2592435,
323
  "pages": 38,
324
  "shards": 2
325
  },
326
  {
327
- "seconds": 1.061899332999019,
328
- "bytes": 6323038,
329
- "pages": 39,
330
  "shards": 2
331
  },
332
  {
333
- "seconds": 1.428008876999229,
334
- "bytes": 7759017,
335
  "pages": 37,
336
- "shards": 4
337
  },
338
  {
339
- "seconds": 0.8940000050024537,
340
- "bytes": 393500,
341
- "pages": 10,
342
- "shards": 1
343
  },
344
  {
345
- "seconds": 1.0701383749983506,
346
- "bytes": 5409485,
347
- "pages": 27,
348
  "shards": 2
349
  },
350
  {
351
- "seconds": 1.2911368530003529,
352
- "bytes": 5955841,
353
- "pages": 31,
354
  "shards": 2
355
  },
356
  {
357
- "seconds": 1.0264875259999826,
358
- "bytes": 2166910,
359
- "pages": 21,
360
- "shards": 2
361
  },
362
  {
363
- "seconds": 1.2000135250018502,
364
- "bytes": 5803444,
365
- "pages": 1,
366
  "shards": 1
367
  },
368
  {
369
- "seconds": 1.2405726069991942,
370
- "bytes": 6889170,
371
- "pages": 43,
 
 
 
 
 
 
372
  "shards": 2
373
  },
374
  {
375
- "seconds": 0.9624333549982111,
376
- "bytes": 484343,
377
- "pages": 14,
378
  "shards": 1
379
  },
380
  {
381
- "seconds": 1.4119799629988847,
382
- "bytes": 11493188,
383
- "pages": 37,
384
  "shards": 4
385
  },
386
  {
387
- "seconds": 0.9430077399993024,
388
- "bytes": 1135014,
389
  "pages": 33,
390
- "shards": 1
391
  },
392
  {
393
- "seconds": 1.1626808530018025,
394
- "bytes": 5979744,
395
- "pages": 31,
396
  "shards": 2
397
  },
398
  {
399
- "seconds": 1.2183306759980042,
400
- "bytes": 7063267,
401
- "pages": 30,
402
- "shards": 3
403
  },
404
  {
405
- "seconds": 1.0798259360017255,
406
- "bytes": 4403386,
407
- "pages": 18,
408
- "shards": 2
409
  },
410
  {
411
- "seconds": 1.2122206399981224,
412
- "bytes": 13761400,
413
- "pages": 50,
414
  "shards": 4
415
  },
416
  {
417
- "seconds": 1.1497695119978744,
418
- "bytes": 863102,
419
- "pages": 27,
420
- "shards": 1
421
  },
422
  {
423
- "seconds": 1.423836862002645,
424
- "bytes": 10998784,
425
- "pages": 40,
426
  "shards": 3
427
  },
428
  {
429
- "seconds": 1.2024595770017186,
430
- "bytes": 7444430,
431
  "pages": 30,
432
  "shards": 3
433
  },
434
  {
435
- "seconds": 1.016105377002532,
436
- "bytes": 1143206,
437
- "pages": 36,
438
  "shards": 1
439
  },
440
  {
441
- "seconds": 4.951656708999508,
442
- "bytes": 5644531,
443
- "pages": 20,
 
 
 
 
 
 
444
  "shards": 2
445
  },
446
  {
447
- "seconds": 1.038082987000962,
448
- "bytes": 8819669,
449
- "pages": 48,
450
- "shards": 3
451
  },
452
  {
453
- "seconds": 1.3574339080005302,
454
- "bytes": 10377609,
455
- "pages": 27,
456
- "shards": 3
457
  },
458
  {
459
- "seconds": 1.2949794519990974,
460
- "bytes": 7882340,
461
- "pages": 40,
462
- "shards": 3
463
  },
464
  {
465
- "seconds": 1.516641152997181,
466
- "bytes": 1120266,
467
- "pages": 34,
468
- "shards": 1
469
  },
470
  {
471
- "seconds": 1.2580970729977707,
472
- "bytes": 5952993,
473
- "pages": 28,
474
  "shards": 2
475
  },
476
  {
477
- "seconds": 1.295129406000342,
478
- "bytes": 6202154,
479
- "pages": 29,
480
  "shards": 3
481
  },
482
  {
483
- "seconds": 1.2203903480003646,
484
- "bytes": 12408575,
485
- "pages": 47,
 
 
 
 
 
 
486
  "shards": 3
487
  },
488
  {
489
- "seconds": 1.1006403920000594,
490
- "bytes": 7578962,
491
- "pages": 31,
492
  "shards": 3
493
  },
494
  {
495
- "seconds": 1.032797053001559,
496
- "bytes": 1142518,
497
- "pages": 33,
498
  "shards": 1
499
  },
500
  {
501
- "seconds": 1.317680020998523,
502
- "bytes": 10858764,
503
- "pages": 40,
504
- "shards": 3
505
  },
506
  {
507
- "seconds": 1.2862951650022296,
508
- "bytes": 2384330,
509
- "pages": 29,
510
- "shards": 2
511
  },
512
  {
513
- "seconds": 1.0045538510021288,
514
- "bytes": 5798532,
515
- "pages": 25,
 
 
 
 
 
 
516
  "shards": 2
517
  },
518
  {
519
- "seconds": 1.256682400999125,
520
- "bytes": 12134890,
521
- "pages": 46,
522
  "shards": 3
523
  },
524
  {
525
- "seconds": 1.024204646000726,
526
- "bytes": 6281125,
527
- "pages": 32,
528
  "shards": 3
529
  },
530
  {
531
- "seconds": 1.1149510969989933,
532
- "bytes": 6170756,
533
- "pages": 37,
534
  "shards": 2
535
  },
536
  {
537
- "seconds": 0.9807241070011514,
538
- "bytes": 2284775,
539
- "pages": 28,
540
  "shards": 3
541
  },
542
  {
543
- "seconds": 1.0974279740003112,
544
- "bytes": 5896433,
545
- "pages": 33,
546
  "shards": 2
547
  },
548
  {
549
- "seconds": 1.4275493559980532,
550
- "bytes": 16198066,
551
- "pages": 38,
552
- "shards": 4
553
- },
554
- {
555
- "seconds": 1.0836265130019456,
556
- "bytes": 2657150,
557
- "pages": 38,
558
  "shards": 2
559
  },
560
  {
561
- "seconds": 0.9289632090003579,
562
- "bytes": 5852982,
563
- "pages": 29,
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": 76826.23510929271,
26
+ "required_pages_per_second_upper_estimate": 0.005315017206285955
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": 9603.27938866159,
40
+ "required_pages_per_second_upper_estimate": 0.004020692540754891
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": 19206.55877732318,
54
+ "required_pages_per_second_upper_estimate": 0.004000219682300069
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": 28809.83816598477,
68
+ "required_pages_per_second_upper_estimate": 0.0017271058079750155
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": 28809.83816598477,
82
+ "required_pages_per_second_upper_estimate": 0.002239838196319521
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": 28809.83816598477,
96
+ "required_pages_per_second_upper_estimate": 0.36954167039910685
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": 0.1271348553984808,
162
+ "producer_p95_seconds": 0.3210992399981478,
163
+ "gpu_update_seconds": 0.9596721731204818,
164
+ "producer_batches_per_second": 7.865663565398208,
165
+ "gpu_batches_per_second": 1.0420225031099815,
166
+ "producer_headroom": 7.548458446840295,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.459346906590355,
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": 345923.98800616147,
174
+ "upload_bytes_per_second": 5405264.117713781,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 74.22723512781747,
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": 1.016105377002532,
184
+ "bytes": 1143206,
185
+ "pages": 36,
 
 
 
 
 
 
186
  "shards": 1
187
  },
188
  {
189
+ "seconds": 4.951656708999508,
190
+ "bytes": 5644531,
191
+ "pages": 20,
192
+ "shards": 2
193
  },
194
  {
195
+ "seconds": 1.038082987000962,
196
+ "bytes": 8819669,
197
+ "pages": 48,
198
+ "shards": 3
199
  },
200
  {
201
+ "seconds": 1.3574339080005302,
202
+ "bytes": 10377609,
203
+ "pages": 27,
204
  "shards": 3
205
  },
206
  {
207
+ "seconds": 1.2949794519990974,
208
+ "bytes": 7882340,
209
+ "pages": 40,
210
+ "shards": 3
211
  },
212
  {
213
+ "seconds": 1.516641152997181,
214
+ "bytes": 1120266,
215
+ "pages": 34,
216
+ "shards": 1
217
  },
218
  {
219
+ "seconds": 1.2580970729977707,
220
+ "bytes": 5952993,
221
+ "pages": 28,
222
  "shards": 2
223
  },
224
  {
225
+ "seconds": 1.295129406000342,
226
+ "bytes": 6202154,
227
+ "pages": 29,
228
+ "shards": 3
229
  },
230
  {
231
+ "seconds": 1.2203903480003646,
232
+ "bytes": 12408575,
233
+ "pages": 47,
234
+ "shards": 3
235
  },
236
  {
237
+ "seconds": 1.1006403920000594,
238
+ "bytes": 7578962,
239
+ "pages": 31,
240
+ "shards": 3
241
  },
242
  {
243
+ "seconds": 1.032797053001559,
244
+ "bytes": 1142518,
245
+ "pages": 33,
246
  "shards": 1
247
  },
248
  {
249
+ "seconds": 1.317680020998523,
250
+ "bytes": 10858764,
251
+ "pages": 40,
252
+ "shards": 3
253
  },
254
  {
255
+ "seconds": 1.2862951650022296,
256
+ "bytes": 2384330,
257
+ "pages": 29,
258
  "shards": 2
259
  },
260
  {
261
+ "seconds": 1.0045538510021288,
262
+ "bytes": 5798532,
263
+ "pages": 25,
264
  "shards": 2
265
  },
266
  {
267
+ "seconds": 1.256682400999125,
268
+ "bytes": 12134890,
269
+ "pages": 46,
270
  "shards": 3
271
  },
272
  {
273
+ "seconds": 1.024204646000726,
274
+ "bytes": 6281125,
275
+ "pages": 32,
276
  "shards": 3
277
  },
278
  {
279
+ "seconds": 1.1149510969989933,
280
+ "bytes": 6170756,
281
+ "pages": 37,
282
+ "shards": 2
283
  },
284
  {
285
+ "seconds": 0.9807241070011514,
286
+ "bytes": 2284775,
287
+ "pages": 28,
288
  "shards": 3
289
  },
290
  {
291
+ "seconds": 1.0974279740003112,
292
+ "bytes": 5896433,
293
  "pages": 33,
 
 
 
 
 
 
 
 
 
 
 
 
294
  "shards": 2
295
  },
296
  {
297
+ "seconds": 1.4275493559980532,
298
+ "bytes": 16198066,
299
+ "pages": 38,
300
+ "shards": 4
301
  },
302
  {
303
+ "seconds": 1.0836265130019456,
304
+ "bytes": 2657150,
305
  "pages": 38,
306
  "shards": 2
307
  },
308
  {
309
+ "seconds": 0.9289632090003579,
310
+ "bytes": 5852982,
311
+ "pages": 29,
312
  "shards": 2
313
  },
314
  {
315
+ "seconds": 1.0356132939996314,
316
+ "bytes": 7513178,
317
  "pages": 37,
318
+ "shards": 3
319
  },
320
  {
321
+ "seconds": 1.2883466929997667,
322
+ "bytes": 5484455,
323
+ "pages": 22,
324
+ "shards": 2
325
  },
326
  {
327
+ "seconds": 1.2610901230000309,
328
+ "bytes": 5822650,
329
+ "pages": 29,
330
  "shards": 2
331
  },
332
  {
333
+ "seconds": 0.9832762490004825,
334
+ "bytes": 2733834,
335
+ "pages": 37,
336
  "shards": 2
337
  },
338
  {
339
+ "seconds": 1.1357609220031009,
340
+ "bytes": 11893185,
341
+ "pages": 33,
342
+ "shards": 3
343
  },
344
  {
345
+ "seconds": 0.8795082219985488,
346
+ "bytes": 997738,
347
+ "pages": 30,
348
  "shards": 1
349
  },
350
  {
351
+ "seconds": 1.1275165429979097,
352
+ "bytes": 11698586,
353
+ "pages": 39,
354
+ "shards": 4
355
+ },
356
+ {
357
+ "seconds": 1.0235819969966542,
358
+ "bytes": 6134782,
359
+ "pages": 37,
360
  "shards": 2
361
  },
362
  {
363
+ "seconds": 1.3678917270008242,
364
+ "bytes": 1154643,
365
+ "pages": 35,
366
  "shards": 1
367
  },
368
  {
369
+ "seconds": 1.2100134370048181,
370
+ "bytes": 12201541,
371
+ "pages": 35,
372
  "shards": 4
373
  },
374
  {
375
+ "seconds": 1.149907068000175,
376
+ "bytes": 11779236,
377
  "pages": 33,
378
+ "shards": 3
379
  },
380
  {
381
+ "seconds": 1.051513154998247,
382
+ "bytes": 2589054,
383
+ "pages": 34,
384
  "shards": 2
385
  },
386
  {
387
+ "seconds": 1.0851055740058655,
388
+ "bytes": 5876732,
389
+ "pages": 32,
390
+ "shards": 2
391
  },
392
  {
393
+ "seconds": 1.0857899119946524,
394
+ "bytes": 1135012,
395
+ "pages": 33,
396
+ "shards": 1
397
  },
398
  {
399
+ "seconds": 1.3140387780003948,
400
+ "bytes": 11211496,
401
+ "pages": 39,
402
  "shards": 4
403
  },
404
  {
405
+ "seconds": 1.1181462930035195,
406
+ "bytes": 5980396,
407
+ "pages": 35,
408
+ "shards": 2
409
  },
410
  {
411
+ "seconds": 1.134683780001069,
412
+ "bytes": 8428323,
413
+ "pages": 38,
414
  "shards": 3
415
  },
416
  {
417
+ "seconds": 1.2588409719974152,
418
+ "bytes": 10473930,
419
  "pages": 30,
420
  "shards": 3
421
  },
422
  {
423
+ "seconds": 0.8643372480000835,
424
+ "bytes": 910531,
425
+ "pages": 27,
426
  "shards": 1
427
  },
428
  {
429
+ "seconds": 1.064887002001342,
430
+ "bytes": 7934391,
431
+ "pages": 43,
432
+ "shards": 3
433
+ },
434
+ {
435
+ "seconds": 1.0906057760003023,
436
+ "bytes": 6005491,
437
+ "pages": 32,
438
  "shards": 2
439
  },
440
  {
441
+ "seconds": 0.9286484479962382,
442
+ "bytes": 714826,
443
+ "pages": 21,
444
+ "shards": 1
445
  },
446
  {
447
+ "seconds": 1.5466609349969076,
448
+ "bytes": 4766956,
449
+ "pages": 1,
450
+ "shards": 1
451
  },
452
  {
453
+ "seconds": 1.2835849650000455,
454
+ "bytes": 14052046,
455
+ "pages": 50,
456
+ "shards": 4
457
  },
458
  {
459
+ "seconds": 1.258647842994833,
460
+ "bytes": 5993179,
461
+ "pages": 33,
462
+ "shards": 2
463
  },
464
  {
465
+ "seconds": 1.1118833660002565,
466
+ "bytes": 2568385,
467
+ "pages": 33,
468
  "shards": 2
469
  },
470
  {
471
+ "seconds": 0.9927107309995336,
472
+ "bytes": 6005616,
473
+ "pages": 31,
474
  "shards": 3
475
  },
476
  {
477
+ "seconds": 1.3781777679978404,
478
+ "bytes": 5044232,
479
+ "pages": 36,
480
+ "shards": 2
481
+ },
482
+ {
483
+ "seconds": 1.3652096680016257,
484
+ "bytes": 7517328,
485
+ "pages": 32,
486
  "shards": 3
487
  },
488
  {
489
+ "seconds": 1.3108274569967762,
490
+ "bytes": 11956950,
491
+ "pages": 36,
492
  "shards": 3
493
  },
494
  {
495
+ "seconds": 0.923527514001762,
496
+ "bytes": 789312,
497
+ "pages": 22,
498
  "shards": 1
499
  },
500
  {
501
+ "seconds": 1.4157909589994233,
502
+ "bytes": 12468598,
503
+ "pages": 49,
504
+ "shards": 4
505
  },
506
  {
507
+ "seconds": 0.873544275003951,
508
+ "bytes": 984851,
509
+ "pages": 27,
510
+ "shards": 1
511
  },
512
  {
513
+ "seconds": 1.161854639001831,
514
+ "bytes": 7666603,
515
+ "pages": 40,
516
+ "shards": 3
517
+ },
518
+ {
519
+ "seconds": 0.9383765300008235,
520
+ "bytes": 5995522,
521
+ "pages": 27,
522
  "shards": 2
523
  },
524
  {
525
+ "seconds": 1.2260548230042332,
526
+ "bytes": 10646786,
527
+ "pages": 36,
528
  "shards": 3
529
  },
530
  {
531
+ "seconds": 1.1512752330017975,
532
+ "bytes": 7594320,
533
+ "pages": 40,
534
  "shards": 3
535
  },
536
  {
537
+ "seconds": 1.0619875439952011,
538
+ "bytes": 5471838,
539
+ "pages": 28,
540
  "shards": 2
541
  },
542
  {
543
+ "seconds": 1.1333524080037023,
544
+ "bytes": 7101020,
545
+ "pages": 41,
546
  "shards": 3
547
  },
548
  {
549
+ "seconds": 1.235225882002851,
550
+ "bytes": 5723234,
551
+ "pages": 29,
552
  "shards": 2
553
  },
554
  {
555
+ "seconds": 0.9995385679940227,
556
+ "bytes": 6053776,
557
+ "pages": 37,
 
 
 
 
 
 
558
  "shards": 2
559
  },
560
  {
561
+ "seconds": 1.3147696169980918,
562
+ "bytes": 8376347,
563
+ "pages": 36,
564
+ "shards": 3
565
  }
566
  ],
567
  "cache_key": {
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:08ba2dd523ece7c3398532396b6dd7590d07e9be7249052a8b8cf4f84043c025
3
- size 721776063
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8c7cadbc8ada2e114351e38f2dc65770375fbd3c5571e7a3db009d13ce7e870a
3
+ size 721767167