appvoid commited on
Commit
9bff059
·
verified ·
1 Parent(s): 34e40e5

Qwen90M step 250000; 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 +232 -232
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 245000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "80d6299138410b2e8858ba12acf696cdf3176dcd9f400bb585856b76acf7d993",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "727525a389ec5f45c0e2113ae9043f1f725bf96b0fc45fa3f07a6899b9d03b8b",
12
  "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
13
  "performance.json": "ac15a69eb79c2ac9ef8406edc12d9ae8abafb561b0f74b54b23d5920ee4fe3f7",
14
- "shard_scheduler.json": "53335450f9b4b65852a99c22ef387c6c7d63ed363bbf58d23d279cc75e51158e",
15
- "coverage.json": "74e3063ad472df43fbe8188bf2dafe200f98ec7afd81e2c0b072fc086e471069",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 250000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "b39daf23cbfc7a4db45dc715dda0257d7a0aceacebd6722e70e52b392a3d914f",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "e91afb32a91ba214e7473f46d6f99a0a931e405920d457559aea60d50123c7e4",
12
  "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
13
  "performance.json": "ac15a69eb79c2ac9ef8406edc12d9ae8abafb561b0f74b54b23d5920ee4fe3f7",
14
+ "shard_scheduler.json": "51e06602336cf2ff9c7fbf2702cb9edfe3a4879775d6e82cb900f0895494a34a",
15
+ "coverage.json": "fa6f7ebd74b32c3a670cb83966d87b06345ed8c181b0803237a454541a8aa878",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 245000,
3
  "stats": {
4
- "climbmix/blocks": 8820000,
5
- "climbmix/targets": 14743955462,
6
- "climbmix/input_tokens": 14744889647,
7
- "ultra_qa/blocks": 3307500,
8
- "ultra_qa/targets": 5395722257,
9
- "ultra_qa/input_tokens": 5395732630,
10
- "cortex/blocks": 3307500,
11
- "cortex/targets": 125836183,
12
- "cortex/input_tokens": 5257347187,
13
- "ultra_style/blocks": 3307500,
14
- "ultra_style/targets": 5231835740,
15
- "ultra_style/input_tokens": 5232110004,
16
- "openmath/blocks": 2205000,
17
- "openmath/targets": 3406996250,
18
- "openmath/input_tokens": 3408131868,
19
- "rewrite6/blocks": 1102500,
20
- "rewrite6/targets": 1807346506,
21
- "rewrite6/input_tokens": 1809328100
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 1207,
26
- "row": 3232,
27
  "start": {
28
- "file": 58,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 35,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
- "offset": 12288,
37
- "documents": 4947104
38
  },
39
  "rewrite6": {
40
- "page": 37,
41
- "row": 1322,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 151552
47
  }
48
  },
49
- "epoch": 13,
50
  "offset": 0,
51
- "documents": 2823126
52
  },
53
  "openmath": {
54
- "page": 755,
55
- "row": 365,
56
  "start": {
57
  "file": 7,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 35,
61
- "row": 930
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 3092845
67
  },
68
  "ultra_qa": {
69
- "page": 326,
70
- "row": 1777,
71
  "start": {
72
  "file": 2,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
- "row_group": 4,
76
- "row": 37728
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
- "documents": 1337073
82
  },
83
  "ultra_style": {
84
- "page": 498,
85
- "row": 978,
86
  "start": {
87
- "file": 2,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
- "row_group": 7,
91
- "row": 74335
92
  }
93
  },
94
  "epoch": 0,
95
- "offset": 0,
96
- "documents": 2040786
97
  },
98
  "cortex": {
99
- "page": 64142,
100
- "row": 98,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 8210274
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 250000,
3
  "stats": {
4
+ "climbmix/blocks": 9000000,
5
+ "climbmix/targets": 15044821319,
6
+ "climbmix/input_tokens": 15045774685,
7
+ "ultra_qa/blocks": 3375000,
8
+ "ultra_qa/targets": 5505900381,
9
+ "ultra_qa/input_tokens": 5505910963,
10
+ "cortex/blocks": 3375000,
11
+ "cortex/targets": 128400391,
12
+ "cortex/input_tokens": 5364742698,
13
+ "ultra_style/blocks": 3375000,
14
+ "ultra_style/targets": 5338964418,
15
+ "ultra_style/input_tokens": 5339244172,
16
+ "openmath/blocks": 2250000,
17
+ "openmath/targets": 3476637803,
18
+ "openmath/input_tokens": 3477796696,
19
+ "rewrite6/blocks": 1125000,
20
+ "rewrite6/targets": 1844270111,
21
+ "rewrite6/input_tokens": 1846292063
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 1232,
26
+ "row": 2049,
27
  "start": {
28
+ "file": 59,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 53,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
+ "offset": 2048,
37
+ "documents": 5048321
38
  },
39
  "rewrite6": {
40
+ "page": 1,
41
+ "row": 847,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 4096
47
  }
48
  },
49
+ "epoch": 14,
50
  "offset": 0,
51
+ "documents": 2880599
52
  },
53
  "openmath": {
54
+ "page": 770,
55
+ "row": 2180,
56
  "start": {
57
  "file": 7,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 97,
61
+ "row": 370
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 3156100
67
  },
68
  "ultra_qa": {
69
+ "page": 333,
70
+ "row": 364,
71
  "start": {
72
  "file": 2,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
+ "row_group": 5,
76
+ "row": 1742
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
+ "documents": 1364332
82
  },
83
  "ultra_style": {
84
+ "page": 508,
85
+ "row": 1215,
86
  "start": {
87
+ "file": 3,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
+ "row_group": 0,
91
+ "row": 25940
92
  }
93
  },
94
  "epoch": 0,
95
+ "offset": 2048,
96
+ "documents": 2081983
97
  },
98
  "cortex": {
99
+ "page": 65453,
100
+ "row": 34,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 8378018
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 245000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "80d6299138410b2e8858ba12acf696cdf3176dcd9f400bb585856b76acf7d993", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "727525a389ec5f45c0e2113ae9043f1f725bf96b0fc45fa3f07a6899b9d03b8b", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "ac15a69eb79c2ac9ef8406edc12d9ae8abafb561b0f74b54b23d5920ee4fe3f7", "shard_scheduler.json": "53335450f9b4b65852a99c22ef387c6c7d63ed363bbf58d23d279cc75e51158e", "coverage.json": "74e3063ad472df43fbe8188bf2dafe200f98ec7afd81e2c0b072fc086e471069", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000245000"}
 
1
+ {"step": 250000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "b39daf23cbfc7a4db45dc715dda0257d7a0aceacebd6722e70e52b392a3d914f", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "e91afb32a91ba214e7473f46d6f99a0a931e405920d457559aea60d50123c7e4", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "ac15a69eb79c2ac9ef8406edc12d9ae8abafb561b0f74b54b23d5920ee4fe3f7", "shard_scheduler.json": "51e06602336cf2ff9c7fbf2702cb9edfe3a4879775d6e82cb900f0895494a34a", "coverage.json": "fa6f7ebd74b32c3a670cb83966d87b06345ed8c181b0803237a454541a8aa878", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000250000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:80d6299138410b2e8858ba12acf696cdf3176dcd9f400bb585856b76acf7d993
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b39daf23cbfc7a4db45dc715dda0257d7a0aceacebd6722e70e52b392a3d914f
3
  size 360622744
shard_scheduler.json CHANGED
@@ -22,8 +22,8 @@
22
  "tokens_per_second": 1062967.6168812835,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
- "required_token_positions_per_second": 75420.9647685022,
26
- "required_pages_per_second_upper_estimate": 0.006537298073446168
27
  },
28
  {
29
  "source": "rewrite6",
@@ -36,8 +36,8 @@
36
  "tokens_per_second": 1206672.263377472,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
- "required_token_positions_per_second": 9427.620596062776,
40
- "required_pages_per_second_upper_estimate": 0.0037608188112584874
41
  },
42
  {
43
  "source": "openmath",
@@ -50,8 +50,8 @@
50
  "tokens_per_second": 410042.7581134458,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
- "required_token_positions_per_second": 18855.24119212555,
54
- "required_pages_per_second_upper_estimate": 0.0043618282088870395
55
  },
56
  {
57
  "source": "ultra_qa",
@@ -64,8 +64,8 @@
64
  "tokens_per_second": 2163486.5624068812,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
- "required_token_positions_per_second": 28282.86178818833,
68
- "required_pages_per_second_upper_estimate": 0.0017222778941929503
69
  },
70
  {
71
  "source": "ultra_style",
@@ -78,8 +78,8 @@
78
  "tokens_per_second": 1546694.2737633225,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
- "required_token_positions_per_second": 28282.86178818833,
82
- "required_pages_per_second_upper_estimate": 0.0021772929924270767
83
  },
84
  {
85
  "source": "cortex",
@@ -92,8 +92,8 @@
92
  "tokens_per_second": 1015177.8395898147,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
  "estimated_pages_for_buffer": 22,
95
- "required_token_positions_per_second": 28282.86178818833,
96
- "required_pages_per_second_upper_estimate": 0.35117848675997776
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,409 +158,409 @@
158
  "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 690670,
161
- "producer_batch_seconds": 0.5026795613593151,
162
- "producer_p95_seconds": 0.20843738300027326,
163
- "gpu_update_seconds": 0.9775531276522568,
164
- "producer_batches_per_second": 1.9893388887661585,
165
- "gpu_batches_per_second": 1.0229623042602838,
166
- "producer_headroom": 1.9446844526736233,
167
  "sustainable": true,
168
- "buffer_seconds": 60.60829391443992,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
  "upload_pages_per_commit": 78,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 284788.4705175513,
174
- "upload_bytes_per_second": 3726836.968530263,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 61.108976236771625,
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": 1.4821450189999723,
184
- "bytes": 5748205,
185
- "pages": 24,
186
- "shards": 2
187
- },
188
- {
189
- "seconds": 1.3858465779999278,
190
- "bytes": 775414,
191
- "pages": 24,
192
  "shards": 1
193
  },
194
  {
195
- "seconds": 1.4328443340000376,
196
- "bytes": 11483713,
197
- "pages": 34,
198
- "shards": 4
199
  },
200
  {
201
- "seconds": 0.909485917999973,
202
- "bytes": 1076059,
203
- "pages": 32,
204
- "shards": 1
205
  },
206
  {
207
- "seconds": 0.961313237000013,
208
- "bytes": 6243214,
209
  "pages": 35,
210
  "shards": 2
211
  },
212
  {
213
- "seconds": 1.4556228910000755,
214
- "bytes": 7655201,
215
- "pages": 35,
216
  "shards": 3
217
  },
218
  {
219
- "seconds": 1.1608069779999823,
220
- "bytes": 11166576,
221
- "pages": 32,
222
- "shards": 3
223
  },
224
  {
225
- "seconds": 1.0247551580000618,
226
- "bytes": 7489968,
227
- "pages": 35,
228
- "shards": 3
229
  },
230
  {
231
- "seconds": 0.7696208980000847,
232
- "bytes": 1062863,
233
  "pages": 33,
234
  "shards": 1
235
  },
236
  {
237
- "seconds": 1.0704197729999123,
238
- "bytes": 6206676,
239
- "pages": 34,
240
- "shards": 2
241
  },
242
  {
243
- "seconds": 2.8963105739999264,
244
- "bytes": 7237253,
245
- "pages": 32,
246
  "shards": 3
247
  },
248
  {
249
- "seconds": 1.3083305300001484,
250
- "bytes": 5580567,
251
- "pages": 35,
252
  "shards": 2
253
  },
254
  {
255
- "seconds": 1.1845111140000881,
256
- "bytes": 11699180,
257
- "pages": 29,
258
- "shards": 3
259
  },
260
  {
261
- "seconds": 0.859392975999981,
262
- "bytes": 2445063,
263
  "pages": 35,
264
  "shards": 2
265
  },
266
  {
267
- "seconds": 1.0558128229999966,
268
- "bytes": 6261983,
269
- "pages": 34,
270
  "shards": 2
271
  },
272
  {
273
- "seconds": 1.3157395409998571,
274
- "bytes": 12201352,
275
- "pages": 34,
276
  "shards": 4
277
  },
278
  {
279
- "seconds": 2.8152504160002536,
280
- "bytes": 1180754,
281
- "pages": 33,
282
- "shards": 1
283
  },
284
  {
285
- "seconds": 1.0497820410000713,
286
- "bytes": 5934890,
287
- "pages": 35,
288
  "shards": 2
289
  },
290
  {
291
- "seconds": 1.250543007000033,
292
- "bytes": 8195692,
293
- "pages": 31,
294
- "shards": 3
295
  },
296
  {
297
- "seconds": 1.6225744519997534,
298
- "bytes": 9437094,
299
- "pages": 34,
300
- "shards": 3
301
  },
302
  {
303
- "seconds": 1.1101808009998422,
304
- "bytes": 6059135,
305
- "pages": 35,
306
  "shards": 2
307
  },
308
  {
309
- "seconds": 1.0066997960002482,
310
- "bytes": 2574525,
 
 
 
 
 
 
311
  "pages": 35,
312
  "shards": 2
313
  },
314
  {
315
- "seconds": 3.0401188460000412,
316
- "bytes": 10860861,
317
  "pages": 35,
318
  "shards": 3
319
  },
320
  {
321
- "seconds": 0.9114932880002016,
322
- "bytes": 506194,
323
- "pages": 16,
324
  "shards": 1
325
  },
326
  {
327
- "seconds": 1.1319729980000375,
328
- "bytes": 7544705,
329
- "pages": 29,
330
- "shards": 3
331
  },
332
  {
333
- "seconds": 1.008947303000241,
334
- "bytes": 7015503,
335
- "pages": 38,
336
- "shards": 2
337
  },
338
  {
339
- "seconds": 1.0751419640000677,
340
- "bytes": 5874473,
341
- "pages": 27,
342
- "shards": 2
343
  },
344
  {
345
- "seconds": 1.191119417000209,
346
- "bytes": 6703440,
347
- "pages": 36,
348
- "shards": 3
349
  },
350
  {
351
- "seconds": 1.0874789810000038,
352
- "bytes": 6434937,
353
- "pages": 34,
354
- "shards": 3
355
  },
356
  {
357
- "seconds": 1.6576894219997484,
358
- "bytes": 7512513,
359
- "pages": 37,
360
- "shards": 3
361
  },
362
  {
363
- "seconds": 0.866141241999685,
364
- "bytes": 1214946,
365
- "pages": 34,
366
  "shards": 1
367
  },
368
  {
369
- "seconds": 1.2982788010003787,
370
- "bytes": 16654234,
371
- "pages": 34,
372
  "shards": 4
373
  },
374
  {
375
- "seconds": 0.9026169319995461,
376
- "bytes": 2322454,
377
- "pages": 27,
378
  "shards": 2
379
  },
380
  {
381
- "seconds": 1.0145137629997407,
382
- "bytes": 6320844,
383
- "pages": 35,
384
  "shards": 2
385
  },
386
  {
387
- "seconds": 1.26741677300015,
388
- "bytes": 7410356,
389
  "pages": 37,
390
  "shards": 3
391
  },
392
  {
393
- "seconds": 1.306584923000628,
394
- "bytes": 936913,
395
- "pages": 30,
396
- "shards": 1
397
  },
398
  {
399
- "seconds": 33.908926936999706,
400
- "bytes": 9910795,
401
- "pages": 40,
402
- "shards": 3
403
  },
404
  {
405
- "seconds": 2.0074904029997924,
406
- "bytes": 13527574,
407
- "pages": 44,
408
- "shards": 4
409
  },
410
  {
411
- "seconds": 1.034237903999383,
412
- "bytes": 890095,
413
- "pages": 25,
414
  "shards": 1
415
  },
416
  {
417
- "seconds": 1.0843116149999332,
418
- "bytes": 11777166,
419
- "pages": 39,
420
  "shards": 4
421
  },
422
  {
423
- "seconds": 1.0721420089994353,
424
- "bytes": 6211390,
425
- "pages": 35,
426
- "shards": 2
427
  },
428
  {
429
- "seconds": 0.7338976749997528,
430
- "bytes": 1218142,
431
  "pages": 34,
432
  "shards": 1
433
  },
434
  {
435
- "seconds": 1.1807161469996572,
436
- "bytes": 7525394,
437
- "pages": 28,
438
- "shards": 3
439
- },
440
- {
441
- "seconds": 1.1362227980007447,
442
- "bytes": 11027017,
443
- "pages": 35,
444
  "shards": 3
445
  },
446
  {
447
- "seconds": 0.8288988310005152,
448
- "bytes": 5801101,
449
- "pages": 35,
450
  "shards": 2
451
  },
452
  {
453
- "seconds": 1.2185628970000835,
454
- "bytes": 7809668,
455
  "pages": 37,
456
  "shards": 3
457
  },
458
  {
459
- "seconds": 1.2807725910006411,
460
- "bytes": 1082933,
461
- "pages": 33,
462
  "shards": 1
463
  },
464
  {
465
- "seconds": 1.2944532310002614,
466
- "bytes": 11460253,
 
 
 
 
 
 
467
  "pages": 33,
468
  "shards": 4
469
  },
470
  {
471
- "seconds": 1.090632098000242,
472
- "bytes": 1122627,
473
  "pages": 33,
474
  "shards": 1
475
  },
476
  {
477
- "seconds": 1.176693084999897,
478
- "bytes": 12183576,
479
- "pages": 36,
480
  "shards": 3
481
  },
482
  {
483
- "seconds": 1.0236794259999442,
484
- "bytes": 7487764,
485
- "pages": 29,
486
  "shards": 3
487
  },
488
  {
489
- "seconds": 1.1006291130006502,
490
- "bytes": 6137509,
491
- "pages": 39,
492
  "shards": 2
493
  },
494
  {
495
- "seconds": 1.0797216399996614,
496
- "bytes": 5987959,
497
- "pages": 27,
498
- "shards": 2
499
  },
500
  {
501
- "seconds": 1.322158891000072,
502
- "bytes": 2479484,
503
  "pages": 35,
504
  "shards": 2
505
  },
506
  {
507
- "seconds": 1.3229940370001714,
508
- "bytes": 6465970,
509
- "pages": 36,
510
  "shards": 2
511
  },
512
  {
513
- "seconds": 1.3629516780001723,
514
- "bytes": 11728657,
515
- "pages": 30,
516
  "shards": 4
517
  },
518
  {
519
- "seconds": 1.096408940999936,
520
- "bytes": 6091544,
 
 
 
 
 
 
521
  "pages": 35,
522
  "shards": 2
523
  },
524
  {
525
- "seconds": 1.042788374000338,
526
- "bytes": 6163555,
527
- "pages": 36,
528
- "shards": 2
529
  },
530
  {
531
- "seconds": 1.04696464799963,
532
- "bytes": 2343472,
533
- "pages": 28,
534
  "shards": 2
535
  },
536
  {
537
- "seconds": 1.3791698659997564,
538
- "bytes": 10729274,
539
- "pages": 36,
540
  "shards": 4
541
  },
542
  {
543
- "seconds": 1.2665546050002376,
544
- "bytes": 6293373,
545
- "pages": 36,
546
- "shards": 2
547
- },
548
- {
549
- "seconds": 1.4158054800000173,
550
- "bytes": 8082619,
551
- "pages": 30,
552
- "shards": 3
553
  },
554
  {
555
- "seconds": 1.100743914999839,
556
- "bytes": 6375983,
557
  "pages": 35,
558
  "shards": 2
559
  },
560
  {
561
- "seconds": 1.1518813020011294,
562
- "bytes": 6877765,
563
- "pages": 35,
564
  "shards": 3
565
  }
566
  ],
 
22
  "tokens_per_second": 1062967.6168812835,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
+ "required_token_positions_per_second": 75570.71051583829,
26
+ "required_pages_per_second_upper_estimate": 0.006550277655298134
27
  },
28
  {
29
  "source": "rewrite6",
 
36
  "tokens_per_second": 1206672.263377472,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
+ "required_token_positions_per_second": 9446.338814479786,
40
+ "required_pages_per_second_upper_estimate": 0.003768285788447338
41
  },
42
  {
43
  "source": "openmath",
 
50
  "tokens_per_second": 410042.7581134458,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
+ "required_token_positions_per_second": 18892.677628959573,
54
+ "required_pages_per_second_upper_estimate": 0.004370488469689805
55
  },
56
  {
57
  "source": "ultra_qa",
 
64
  "tokens_per_second": 2163486.5624068812,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
+ "required_token_positions_per_second": 28339.01644343936,
68
+ "required_pages_per_second_upper_estimate": 0.0017256974180770313
69
  },
70
  {
71
  "source": "ultra_style",
 
78
  "tokens_per_second": 1546694.2737633225,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
+ "required_token_positions_per_second": 28339.01644343936,
82
+ "required_pages_per_second_upper_estimate": 0.0021816159332343353
83
  },
84
  {
85
  "source": "cortex",
 
92
  "tokens_per_second": 1015177.8395898147,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
  "estimated_pages_for_buffer": 22,
95
+ "required_token_positions_per_second": 28339.01644343936,
96
+ "required_pages_per_second_upper_estimate": 0.3518757396406541
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
 
158
  "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 690670,
161
+ "producer_batch_seconds": 0.129124772499992,
162
+ "producer_p95_seconds": 0.24966627399953722,
163
+ "gpu_update_seconds": 0.9756160752855156,
164
+ "producer_batches_per_second": 7.744447332908656,
165
+ "gpu_batches_per_second": 1.0249933609461575,
166
+ "producer_headroom": 7.555607312187723,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.48819666770197,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
  "upload_pages_per_commit": 78,
172
  "max_pending_pages": 4096,
173
+ "generation_bytes_per_second": 382211.29352878686,
174
+ "upload_bytes_per_second": 4854486.8576622605,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 82.01364616773323,
179
  "selection_policy": "Smallest shard within 90% of measured local peak, enlarged to amortize 30 seconds of measured page production; bounded to 256 MiB."
180
  },
181
  "upload_measurements": [
182
  {
183
+ "seconds": 0.7338976749997528,
184
+ "bytes": 1218142,
185
+ "pages": 34,
 
 
 
 
 
 
186
  "shards": 1
187
  },
188
  {
189
+ "seconds": 1.1807161469996572,
190
+ "bytes": 7525394,
191
+ "pages": 28,
192
+ "shards": 3
193
  },
194
  {
195
+ "seconds": 1.1362227980007447,
196
+ "bytes": 11027017,
197
+ "pages": 35,
198
+ "shards": 3
199
  },
200
  {
201
+ "seconds": 0.8288988310005152,
202
+ "bytes": 5801101,
203
  "pages": 35,
204
  "shards": 2
205
  },
206
  {
207
+ "seconds": 1.2185628970000835,
208
+ "bytes": 7809668,
209
+ "pages": 37,
210
  "shards": 3
211
  },
212
  {
213
+ "seconds": 1.2807725910006411,
214
+ "bytes": 1082933,
215
+ "pages": 33,
216
+ "shards": 1
217
  },
218
  {
219
+ "seconds": 1.2944532310002614,
220
+ "bytes": 11460253,
221
+ "pages": 33,
222
+ "shards": 4
223
  },
224
  {
225
+ "seconds": 1.090632098000242,
226
+ "bytes": 1122627,
227
  "pages": 33,
228
  "shards": 1
229
  },
230
  {
231
+ "seconds": 1.176693084999897,
232
+ "bytes": 12183576,
233
+ "pages": 36,
234
+ "shards": 3
235
  },
236
  {
237
+ "seconds": 1.0236794259999442,
238
+ "bytes": 7487764,
239
+ "pages": 29,
240
  "shards": 3
241
  },
242
  {
243
+ "seconds": 1.1006291130006502,
244
+ "bytes": 6137509,
245
+ "pages": 39,
246
  "shards": 2
247
  },
248
  {
249
+ "seconds": 1.0797216399996614,
250
+ "bytes": 5987959,
251
+ "pages": 27,
252
+ "shards": 2
253
  },
254
  {
255
+ "seconds": 1.322158891000072,
256
+ "bytes": 2479484,
257
  "pages": 35,
258
  "shards": 2
259
  },
260
  {
261
+ "seconds": 1.3229940370001714,
262
+ "bytes": 6465970,
263
+ "pages": 36,
264
  "shards": 2
265
  },
266
  {
267
+ "seconds": 1.3629516780001723,
268
+ "bytes": 11728657,
269
+ "pages": 30,
270
  "shards": 4
271
  },
272
  {
273
+ "seconds": 1.096408940999936,
274
+ "bytes": 6091544,
275
+ "pages": 35,
276
+ "shards": 2
277
  },
278
  {
279
+ "seconds": 1.042788374000338,
280
+ "bytes": 6163555,
281
+ "pages": 36,
282
  "shards": 2
283
  },
284
  {
285
+ "seconds": 1.04696464799963,
286
+ "bytes": 2343472,
287
+ "pages": 28,
288
+ "shards": 2
289
  },
290
  {
291
+ "seconds": 1.3791698659997564,
292
+ "bytes": 10729274,
293
+ "pages": 36,
294
+ "shards": 4
295
  },
296
  {
297
+ "seconds": 1.2665546050002376,
298
+ "bytes": 6293373,
299
+ "pages": 36,
300
  "shards": 2
301
  },
302
  {
303
+ "seconds": 1.4158054800000173,
304
+ "bytes": 8082619,
305
+ "pages": 30,
306
+ "shards": 3
307
+ },
308
+ {
309
+ "seconds": 1.100743914999839,
310
+ "bytes": 6375983,
311
  "pages": 35,
312
  "shards": 2
313
  },
314
  {
315
+ "seconds": 1.1518813020011294,
316
+ "bytes": 6877765,
317
  "pages": 35,
318
  "shards": 3
319
  },
320
  {
321
+ "seconds": 1.007327716999498,
322
+ "bytes": 296016,
323
+ "pages": 9,
324
  "shards": 1
325
  },
326
  {
327
+ "seconds": 1.1130705590003345,
328
+ "bytes": 6042311,
329
+ "pages": 35,
330
+ "shards": 2
331
  },
332
  {
333
+ "seconds": 0.9439528090006206,
334
+ "bytes": 7900957,
335
+ "pages": 37,
336
+ "shards": 3
337
  },
338
  {
339
+ "seconds": 0.7852815879996342,
340
+ "bytes": 838334,
341
+ "pages": 26,
342
+ "shards": 1
343
  },
344
  {
345
+ "seconds": 4.030324741001095,
346
+ "bytes": 15386927,
347
+ "pages": 28,
348
+ "shards": 4
349
  },
350
  {
351
+ "seconds": 0.845626510999864,
352
+ "bytes": 2613623,
353
+ "pages": 35,
354
+ "shards": 2
355
  },
356
  {
357
+ "seconds": 1.1139741530005267,
358
+ "bytes": 5909960,
359
+ "pages": 35,
360
+ "shards": 2
361
  },
362
  {
363
+ "seconds": 1.2427591689993278,
364
+ "bytes": 910923,
365
+ "pages": 26,
366
  "shards": 1
367
  },
368
  {
369
+ "seconds": 1.2945012599993788,
370
+ "bytes": 12063808,
371
+ "pages": 38,
372
  "shards": 4
373
  },
374
  {
375
+ "seconds": 0.8872203939990868,
376
+ "bytes": 6084404,
377
+ "pages": 35,
378
  "shards": 2
379
  },
380
  {
381
+ "seconds": 1.511142709001433,
382
+ "bytes": 6655300,
383
+ "pages": 29,
384
  "shards": 2
385
  },
386
  {
387
+ "seconds": 1.0988184660000115,
388
+ "bytes": 7486365,
389
  "pages": 37,
390
  "shards": 3
391
  },
392
  {
393
+ "seconds": 1.2685245129996474,
394
+ "bytes": 10318034,
395
+ "pages": 29,
396
+ "shards": 3
397
  },
398
  {
399
+ "seconds": 1.292125334000957,
400
+ "bytes": 1081552,
401
+ "pages": 33,
402
+ "shards": 1
403
  },
404
  {
405
+ "seconds": 1.197562700999697,
406
+ "bytes": 7759288,
407
+ "pages": 35,
408
+ "shards": 3
409
  },
410
  {
411
+ "seconds": 0.8769548000000214,
412
+ "bytes": 1135557,
413
+ "pages": 34,
414
  "shards": 1
415
  },
416
  {
417
+ "seconds": 3.838755348999257,
418
+ "bytes": 11603941,
419
+ "pages": 24,
420
  "shards": 4
421
  },
422
  {
423
+ "seconds": 1.57969898599913,
424
+ "bytes": 12094782,
425
+ "pages": 45,
426
+ "shards": 3
427
  },
428
  {
429
+ "seconds": 0.8754016399998363,
430
+ "bytes": 1113722,
431
  "pages": 34,
432
  "shards": 1
433
  },
434
  {
435
+ "seconds": 4.401627224999174,
436
+ "bytes": 7622267,
437
+ "pages": 32,
 
 
 
 
 
 
438
  "shards": 3
439
  },
440
  {
441
+ "seconds": 0.9423913660011749,
442
+ "bytes": 5379592,
443
+ "pages": 34,
444
  "shards": 2
445
  },
446
  {
447
+ "seconds": 1.2426582910011348,
448
+ "bytes": 7714336,
449
  "pages": 37,
450
  "shards": 3
451
  },
452
  {
453
+ "seconds": 1.0709264350007288,
454
+ "bytes": 727110,
455
+ "pages": 23,
456
  "shards": 1
457
  },
458
  {
459
+ "seconds": 1.4975259919992823,
460
+ "bytes": 12239566,
461
+ "pages": 44,
462
+ "shards": 3
463
+ },
464
+ {
465
+ "seconds": 1.4755159030009963,
466
+ "bytes": 11647558,
467
  "pages": 33,
468
  "shards": 4
469
  },
470
  {
471
+ "seconds": 0.8895111490001,
472
+ "bytes": 1197830,
473
  "pages": 33,
474
  "shards": 1
475
  },
476
  {
477
+ "seconds": 1.169628391000515,
478
+ "bytes": 7496810,
479
+ "pages": 38,
480
  "shards": 3
481
  },
482
  {
483
+ "seconds": 1.5263510319982743,
484
+ "bytes": 5728692,
485
+ "pages": 27,
486
  "shards": 3
487
  },
488
  {
489
+ "seconds": 1.2046007990011276,
490
+ "bytes": 5154137,
491
+ "pages": 31,
492
  "shards": 2
493
  },
494
  {
495
+ "seconds": 2.8337960279986874,
496
+ "bytes": 11685666,
497
+ "pages": 29,
498
+ "shards": 3
499
  },
500
  {
501
+ "seconds": 1.282873722999284,
502
+ "bytes": 2666361,
503
  "pages": 35,
504
  "shards": 2
505
  },
506
  {
507
+ "seconds": 1.1408444070002588,
508
+ "bytes": 5962047,
509
+ "pages": 35,
510
  "shards": 2
511
  },
512
  {
513
+ "seconds": 1.150664095999673,
514
+ "bytes": 11982825,
515
+ "pages": 34,
516
  "shards": 4
517
  },
518
  {
519
+ "seconds": 0.9085387359991728,
520
+ "bytes": 1084848,
521
+ "pages": 33,
522
+ "shards": 1
523
+ },
524
+ {
525
+ "seconds": 1.2088771979997546,
526
+ "bytes": 6082441,
527
  "pages": 35,
528
  "shards": 2
529
  },
530
  {
531
+ "seconds": 1.0392706679995172,
532
+ "bytes": 8238622,
533
+ "pages": 31,
534
+ "shards": 3
535
  },
536
  {
537
+ "seconds": 1.0565556260007725,
538
+ "bytes": 5902213,
539
+ "pages": 31,
540
  "shards": 2
541
  },
542
  {
543
+ "seconds": 1.6428124459998799,
544
+ "bytes": 11197872,
545
+ "pages": 39,
546
  "shards": 4
547
  },
548
  {
549
+ "seconds": 1.1795239030016091,
550
+ "bytes": 829951,
551
+ "pages": 25,
552
+ "shards": 1
 
 
 
 
 
 
553
  },
554
  {
555
+ "seconds": 1.183389565001562,
556
+ "bytes": 6316614,
557
  "pages": 35,
558
  "shards": 2
559
  },
560
  {
561
+ "seconds": 1.805858101000922,
562
+ "bytes": 7507792,
563
+ "pages": 34,
564
  "shards": 3
565
  }
566
  ],
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:727525a389ec5f45c0e2113ae9043f1f725bf96b0fc45fa3f07a6899b9d03b8b
3
- size 721770815
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e91afb32a91ba214e7473f46d6f99a0a931e405920d457559aea60d50123c7e4
3
+ size 721772479