appvoid commited on
Commit
8504798
·
verified ·
1 Parent(s): 13fb983

Qwen90M step 30000; complete resumable checkpoint

Browse files
Files changed (6) hide show
  1. complete.json +5 -5
  2. coverage.json +45 -45
  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": 25000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "df7303e39f8a38cb32caa015de73f0c83cb91ddb60382d6c2876d432a817c840",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "c0a2189fdee90c71c35753bc51ffdb3afecbeb4a346bd3889656ce4fe54fb477",
12
  "training_config.json": "07f331dbc4eba0ea92b45763c59c20af4874302e74a327f3562642d69645fed2",
13
  "performance.json": "b95a1befd575f8d9a02b43fd0cc41c54243f5297127b93d2656e0292ba15e77e",
14
- "shard_scheduler.json": "61528e6c5607c5d05d8a1210d80299914e613d3ffcdb126da4557c3c47fdf788",
15
- "coverage.json": "de500062e0edf2f10dd1cb2862bfcdb202710953fdc9e56a645d42f5467026c1",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 30000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "345733703ef5ba6b63da89055fd805800d7ed30d8be589277e153476b671b19c",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "91c1a40f9a6c8b01e9e50d09f40d34bfc35834f4940f699fc2705fad10d4fc7d",
12
  "training_config.json": "07f331dbc4eba0ea92b45763c59c20af4874302e74a327f3562642d69645fed2",
13
  "performance.json": "b95a1befd575f8d9a02b43fd0cc41c54243f5297127b93d2656e0292ba15e77e",
14
+ "shard_scheduler.json": "0625ffed52cc5254b512d94fd967ab92da88751d415cb467e341553610c28e5b",
15
+ "coverage.json": "37ac82d062088127476b52f93c69f4abd41ed762ba2931ba3b2597dded20b2a8",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 25000,
3
  "stats": {
4
- "climbmix/blocks": 900000,
5
- "climbmix/targets": 1503316012,
6
- "climbmix/input_tokens": 1503411635,
7
- "ultra_qa/blocks": 337500,
8
- "ultra_qa/targets": 550725921,
9
- "ultra_qa/input_tokens": 550727011,
10
- "cortex/blocks": 337500,
11
- "cortex/targets": 12839020,
12
- "cortex/input_tokens": 536531199,
13
- "ultra_style/blocks": 337500,
14
- "ultra_style/targets": 533604810,
15
- "ultra_style/input_tokens": 533633023,
16
- "openmath/blocks": 225000,
17
- "openmath/targets": 347653521,
18
- "openmath/input_tokens": 347769506,
19
- "rewrite6/blocks": 112500,
20
- "rewrite6/targets": 184371575,
21
- "rewrite6/input_tokens": 184573987
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 123,
26
- "row": 2056,
27
  "start": {
28
- "file": 5,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 76,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 0,
37
- "documents": 505864
38
  },
39
  "rewrite6": {
40
- "page": 20,
41
- "row": 960,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 81920
47
  }
48
  },
49
  "epoch": 1,
50
- "offset": 2048,
51
- "documents": 288284
52
  },
53
  "openmath": {
54
- "page": 77,
55
- "row": 320,
56
  "start": {
57
  "file": 0,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 315,
61
- "row": 392
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 315712
67
  },
68
  "ultra_qa": {
69
- "page": 33,
70
- "row": 1362,
71
  "start": {
72
  "file": 0,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 2,
76
- "row": 5892
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
- "documents": 136530
82
  },
83
  "ultra_style": {
84
- "page": 50,
85
- "row": 3964,
86
  "start": {
87
  "file": 0,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 2,
91
- "row": 33851
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
- "documents": 208764
97
  },
98
  "cortex": {
99
- "page": 6544,
100
- "row": 9,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 837641
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 30000,
3
  "stats": {
4
+ "climbmix/blocks": 1080000,
5
+ "climbmix/targets": 1804038329,
6
+ "climbmix/input_tokens": 1804152840,
7
+ "ultra_qa/blocks": 405000,
8
+ "ultra_qa/targets": 660784533,
9
+ "ultra_qa/input_tokens": 660785848,
10
+ "cortex/blocks": 405000,
11
+ "cortex/targets": 15406772,
12
+ "cortex/input_tokens": 643853720,
13
+ "ultra_style/blocks": 405000,
14
+ "ultra_style/targets": 640617580,
15
+ "ultra_style/input_tokens": 640651016,
16
+ "openmath/blocks": 270000,
17
+ "openmath/targets": 417204122,
18
+ "openmath/input_tokens": 417343326,
19
+ "rewrite6/blocks": 135000,
20
+ "rewrite6/targets": 221241868,
21
+ "rewrite6/input_tokens": 221484805
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 148,
26
+ "row": 423,
27
  "start": {
28
+ "file": 7,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 11,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 0,
37
+ "documents": 606631
38
  },
39
  "rewrite6": {
40
+ "page": 34,
41
+ "row": 1394,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 139264
47
  }
48
  },
49
  "epoch": 1,
50
+ "offset": 0,
51
+ "documents": 346062
52
  },
53
  "openmath": {
54
+ "page": 92,
55
+ "row": 2045,
56
  "start": {
57
  "file": 0,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 376,
61
+ "row": 832
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 378877
67
  },
68
  "ultra_qa": {
69
+ "page": 40,
70
+ "row": 85,
71
  "start": {
72
  "file": 0,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 2,
76
+ "row": 34564
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
+ "documents": 163925
82
  },
83
  "ultra_style": {
84
+ "page": 60,
85
+ "row": 3939,
86
  "start": {
87
  "file": 0,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 2,
91
+ "row": 74811
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
+ "documents": 249699
97
  },
98
  "cortex": {
99
+ "page": 7853,
100
+ "row": 114,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 1005298
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 25000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "df7303e39f8a38cb32caa015de73f0c83cb91ddb60382d6c2876d432a817c840", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "c0a2189fdee90c71c35753bc51ffdb3afecbeb4a346bd3889656ce4fe54fb477", "training_config.json": "07f331dbc4eba0ea92b45763c59c20af4874302e74a327f3562642d69645fed2", "performance.json": "b95a1befd575f8d9a02b43fd0cc41c54243f5297127b93d2656e0292ba15e77e", "shard_scheduler.json": "61528e6c5607c5d05d8a1210d80299914e613d3ffcdb126da4557c3c47fdf788", "coverage.json": "de500062e0edf2f10dd1cb2862bfcdb202710953fdc9e56a645d42f5467026c1", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000025000"}
 
1
+ {"step": 30000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "345733703ef5ba6b63da89055fd805800d7ed30d8be589277e153476b671b19c", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "91c1a40f9a6c8b01e9e50d09f40d34bfc35834f4940f699fc2705fad10d4fc7d", "training_config.json": "07f331dbc4eba0ea92b45763c59c20af4874302e74a327f3562642d69645fed2", "performance.json": "b95a1befd575f8d9a02b43fd0cc41c54243f5297127b93d2656e0292ba15e77e", "shard_scheduler.json": "0625ffed52cc5254b512d94fd967ab92da88751d415cb467e341553610c28e5b", "coverage.json": "37ac82d062088127476b52f93c69f4abd41ed762ba2931ba3b2597dded20b2a8", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000030000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:df7303e39f8a38cb32caa015de73f0c83cb91ddb60382d6c2876d432a817c840
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:345733703ef5ba6b63da89055fd805800d7ed30d8be589277e153476b671b19c
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": 75999.12812604562,
26
- "required_pages_per_second_upper_estimate": 0.006587411807936398
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": 9499.891015755702,
40
- "required_pages_per_second_upper_estimate": 0.0037896485622130613
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": 18999.782031511404,
54
- "required_pages_per_second_upper_estimate": 0.004395265188247066
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": 28499.673047267104,
68
- "required_pages_per_second_upper_estimate": 0.0017354805623500836
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": 28499.673047267104,
82
- "required_pages_per_second_upper_estimate": 0.0021939837233229385
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": 28499.673047267104,
96
- "required_pages_per_second_upper_estimate": 0.3538705569771298
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": 688946,
161
- "producer_batch_seconds": 0.12395837693742351,
162
- "producer_p95_seconds": 0.32236783699772786,
163
- "gpu_update_seconds": 0.970116392358095,
164
- "producer_batches_per_second": 8.067224052996584,
165
- "gpu_batches_per_second": 1.0308041466748807,
166
- "producer_headroom": 7.826146294637495,
167
  "sustainable": true,
168
- "buffer_seconds": 60.14721632620189,
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": 466257.7871084776,
174
- "upload_bytes_per_second": 5728821.4302538335,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 100.04806718769785,
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": 1.0224713220013655,
184
- "bytes": 6909652,
185
- "pages": 35,
186
- "shards": 2
187
  },
188
  {
189
- "seconds": 1.018912650000857,
190
- "bytes": 5819747,
191
- "pages": 22,
192
  "shards": 2
193
  },
194
  {
195
- "seconds": 0.8826371689974621,
196
- "bytes": 4172226,
197
- "pages": 46,
198
  "shards": 3
199
  },
200
  {
201
- "seconds": 2.508591282999987,
202
- "bytes": 5665007,
203
- "pages": 21,
204
- "shards": 2
205
  },
206
  {
207
- "seconds": 1.1267430049992981,
208
- "bytes": 4735418,
209
- "pages": 32,
210
- "shards": 2
211
  },
212
  {
213
- "seconds": 1.0588101130015275,
214
- "bytes": 8209294,
215
- "pages": 47,
216
- "shards": 4
217
  },
218
  {
219
- "seconds": 1.0802435129990045,
220
- "bytes": 11583568,
221
- "pages": 29,
222
  "shards": 3
223
  },
224
  {
225
- "seconds": 1.2433872929977952,
226
- "bytes": 7696542,
227
- "pages": 38,
228
- "shards": 3
229
  },
230
  {
231
- "seconds": 0.9942282960000739,
232
- "bytes": 5730133,
233
- "pages": 28,
234
- "shards": 2
235
  },
236
  {
237
- "seconds": 0.9958830260002287,
238
- "bytes": 6097937,
239
- "pages": 35,
240
- "shards": 2
241
  },
242
  {
243
- "seconds": 1.0204584569983126,
244
- "bytes": 2643615,
245
  "pages": 36,
246
  "shards": 2
247
  },
248
  {
249
- "seconds": 1.1489640459985822,
250
- "bytes": 6010015,
251
- "pages": 32,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
252
  "shards": 2
253
  },
254
  {
255
- "seconds": 1.3807983080005215,
256
- "bytes": 15814135,
257
- "pages": 39,
258
  "shards": 4
259
  },
260
  {
261
- "seconds": 1.1238995200001227,
262
- "bytes": 2339467,
263
  "pages": 27,
264
  "shards": 2
265
  },
266
  {
267
- "seconds": 1.365642591001233,
268
- "bytes": 6320146,
269
  "pages": 36,
270
  "shards": 2
271
  },
272
  {
273
- "seconds": 1.0016697349965398,
274
- "bytes": 761758,
275
- "pages": 22,
276
  "shards": 2
277
  },
278
  {
279
- "seconds": 1.1483951949994662,
280
- "bytes": 11735955,
281
- "pages": 47,
282
- "shards": 4
283
  },
284
  {
285
- "seconds": 1.0657177410030272,
286
- "bytes": 1129396,
287
- "pages": 33,
288
- "shards": 1
289
  },
290
  {
291
- "seconds": 1.0460155590008071,
292
- "bytes": 6970566,
293
- "pages": 24,
294
  "shards": 3
295
  },
296
  {
297
- "seconds": 2.029252282998641,
298
- "bytes": 16530913,
299
- "pages": 39,
300
- "shards": 4
301
  },
302
  {
303
- "seconds": 0.8704905230006261,
304
- "bytes": 1186552,
305
- "pages": 33,
306
- "shards": 1
307
  },
308
  {
309
- "seconds": 3.2236978209984954,
310
- "bytes": 7889017,
311
- "pages": 38,
312
  "shards": 3
313
  },
314
  {
315
- "seconds": 0.9988030870008515,
316
- "bytes": 1104863,
317
- "pages": 33,
318
- "shards": 1
319
  },
320
  {
321
- "seconds": 1.0168222699976468,
322
- "bytes": 4937209,
323
- "pages": 2,
324
  "shards": 2
325
  },
326
  {
327
- "seconds": 1.0983402380006737,
328
- "bytes": 4472235,
329
- "pages": 20,
330
- "shards": 2
331
  },
332
  {
333
- "seconds": 1.3725059809985396,
334
- "bytes": 7833557,
335
- "pages": 41,
336
- "shards": 3
337
  },
338
  {
339
- "seconds": 1.04416848999972,
340
- "bytes": 7007035,
341
- "pages": 42,
342
  "shards": 2
343
  },
344
  {
345
- "seconds": 1.018632732997503,
346
- "bytes": 7180638,
347
- "pages": 29,
 
 
 
 
 
 
348
  "shards": 3
349
  },
350
  {
351
- "seconds": 1.1994012059985835,
352
- "bytes": 10151059,
353
- "pages": 38,
354
  "shards": 3
355
  },
356
  {
357
- "seconds": 0.9821173639975314,
358
- "bytes": 2386369,
359
- "pages": 28,
360
- "shards": 2
361
  },
362
  {
363
- "seconds": 1.2013670170017576,
364
- "bytes": 6446691,
365
- "pages": 35,
366
  "shards": 2
367
  },
368
  {
369
- "seconds": 1.0164053049993527,
370
- "bytes": 1100982,
371
- "pages": 30,
372
  "shards": 1
373
  },
374
  {
375
- "seconds": 1.2673739959973318,
376
- "bytes": 17927773,
377
- "pages": 41,
378
- "shards": 5
379
  },
380
  {
381
- "seconds": 0.8297740179987159,
382
- "bytes": 928094,
383
  "pages": 28,
384
- "shards": 1
385
  },
386
  {
387
- "seconds": 0.8634128480007348,
388
- "bytes": 6113242,
389
- "pages": 34,
390
- "shards": 2
391
  },
392
  {
393
- "seconds": 0.979562092001288,
394
- "bytes": 7577775,
395
- "pages": 38,
396
  "shards": 3
397
  },
398
  {
399
- "seconds": 0.9862535339998431,
400
- "bytes": 4746157,
401
- "pages": 29,
402
- "shards": 2
403
  },
404
  {
405
- "seconds": 1.0129832909988181,
406
- "bytes": 7553178,
407
  "pages": 36,
408
- "shards": 3
409
  },
410
  {
411
- "seconds": 1.2109974509985477,
412
- "bytes": 11754963,
413
- "pages": 35,
414
  "shards": 3
415
  },
416
  {
417
- "seconds": 0.878410489000089,
418
- "bytes": 971318,
419
- "pages": 31,
420
  "shards": 1
421
  },
422
  {
423
- "seconds": 1.079362858999957,
424
- "bytes": 12108906,
425
- "pages": 37,
426
  "shards": 4
427
  },
428
  {
429
- "seconds": 0.808165985999949,
430
- "bytes": 1135299,
431
- "pages": 34,
432
  "shards": 1
433
  },
434
  {
435
- "seconds": 1.9975283729982038,
436
- "bytes": 6018894,
437
- "pages": 27,
438
- "shards": 2
439
  },
440
  {
441
- "seconds": 1.1784233340003993,
442
- "bytes": 7683890,
443
  "pages": 38,
444
  "shards": 3
445
  },
446
  {
447
- "seconds": 1.1247381559987844,
448
- "bytes": 11305256,
449
- "pages": 34,
450
- "shards": 3
451
  },
452
  {
453
- "seconds": 0.900414362000447,
454
- "bytes": 7293854,
455
- "pages": 29,
456
  "shards": 3
457
  },
458
  {
459
- "seconds": 1.0943440850023762,
460
- "bytes": 1102979,
461
  "pages": 34,
462
  "shards": 1
463
  },
464
  {
465
- "seconds": 1.0738140280009247,
466
- "bytes": 6022869,
467
- "pages": 36,
468
- "shards": 3
469
  },
470
  {
471
- "seconds": 1.1826581230016018,
472
- "bytes": 10993511,
473
- "pages": 35,
474
- "shards": 4
475
  },
476
  {
477
- "seconds": 1.020132988000114,
478
- "bytes": 882161,
479
  "pages": 26,
480
  "shards": 1
481
  },
482
  {
483
- "seconds": 1.030573132000427,
484
- "bytes": 11789498,
485
- "pages": 32,
486
- "shards": 3
487
- },
488
- {
489
- "seconds": 0.8314693719985371,
490
- "bytes": 2720719,
491
- "pages": 36,
492
  "shards": 2
493
  },
494
  {
495
- "seconds": 0.9525506049976684,
496
- "bytes": 10860039,
497
- "pages": 34,
498
- "shards": 3
499
- },
500
- {
501
- "seconds": 1.2853242279998085,
502
- "bytes": 7445426,
503
  "pages": 38,
504
- "shards": 3
505
- },
506
- {
507
- "seconds": 0.7597367620001023,
508
- "bytes": 831447,
509
- "pages": 25,
510
- "shards": 1
511
- },
512
- {
513
- "seconds": 1.0118765539991728,
514
- "bytes": 6408285,
515
- "pages": 37,
516
  "shards": 2
517
  },
518
  {
519
- "seconds": 1.0685828210007458,
520
- "bytes": 12919611,
521
- "pages": 37,
522
- "shards": 4
523
  },
524
  {
525
- "seconds": 1.0399747580013354,
526
- "bytes": 5973768,
527
- "pages": 27,
528
  "shards": 2
529
  },
530
  {
531
- "seconds": 1.0144561509987398,
532
- "bytes": 6361193,
533
- "pages": 36,
534
- "shards": 2
535
  },
536
  {
537
- "seconds": 1.0571298460017715,
538
- "bytes": 2527314,
539
  "pages": 36,
540
  "shards": 2
541
  },
542
  {
543
- "seconds": 1.1346876719981083,
544
- "bytes": 10084333,
545
- "pages": 34,
546
  "shards": 3
547
  },
548
  {
549
- "seconds": 0.9850222569984908,
550
- "bytes": 2356425,
551
- "pages": 28,
552
  "shards": 2
553
  },
554
  {
555
- "seconds": 1.0524133049984812,
556
- "bytes": 12178249,
557
- "pages": 40,
558
- "shards": 3
559
  },
560
  {
561
- "seconds": 1.109467931000836,
562
- "bytes": 5783909,
563
- "pages": 28,
564
  "shards": 2
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": 75157.17042322725,
26
+ "required_pages_per_second_upper_estimate": 0.00651443304817839
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": 9394.646302903406,
40
+ "required_pages_per_second_upper_estimate": 0.0037476648727075975
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": 18789.29260580681,
54
+ "required_pages_per_second_upper_estimate": 0.004346572164097676
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": 28183.938908710217,
68
+ "required_pages_per_second_upper_estimate": 0.0017162540098409707
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": 28183.938908710217,
82
+ "required_pages_per_second_upper_estimate": 0.002169677635328795
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": 28183.938908710217,
96
+ "required_pages_per_second_upper_estimate": 0.3499501956704399
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
 
158
  "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 688946,
161
+ "producer_batch_seconds": 0.3413639609687493,
162
+ "producer_p95_seconds": 0.3689845970002352,
163
+ "gpu_update_seconds": 0.9809842438827957,
164
+ "producer_batches_per_second": 2.929424644482452,
165
+ "gpu_batches_per_second": 1.0193843644643452,
166
+ "producer_headroom": 2.8737194198792455,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.82102312073334,
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": 341581.6420395341,
174
+ "upload_bytes_per_second": 6088898.701746191,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 73.29546876801984,
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": 0.808165985999949,
184
+ "bytes": 1135299,
185
+ "pages": 34,
186
+ "shards": 1
187
  },
188
  {
189
+ "seconds": 1.9975283729982038,
190
+ "bytes": 6018894,
191
+ "pages": 27,
192
  "shards": 2
193
  },
194
  {
195
+ "seconds": 1.1784233340003993,
196
+ "bytes": 7683890,
197
+ "pages": 38,
198
  "shards": 3
199
  },
200
  {
201
+ "seconds": 1.1247381559987844,
202
+ "bytes": 11305256,
203
+ "pages": 34,
204
+ "shards": 3
205
  },
206
  {
207
+ "seconds": 0.900414362000447,
208
+ "bytes": 7293854,
209
+ "pages": 29,
210
+ "shards": 3
211
  },
212
  {
213
+ "seconds": 1.0943440850023762,
214
+ "bytes": 1102979,
215
+ "pages": 34,
216
+ "shards": 1
217
  },
218
  {
219
+ "seconds": 1.0738140280009247,
220
+ "bytes": 6022869,
221
+ "pages": 36,
222
  "shards": 3
223
  },
224
  {
225
+ "seconds": 1.1826581230016018,
226
+ "bytes": 10993511,
227
+ "pages": 35,
228
+ "shards": 4
229
  },
230
  {
231
+ "seconds": 1.020132988000114,
232
+ "bytes": 882161,
233
+ "pages": 26,
234
+ "shards": 1
235
  },
236
  {
237
+ "seconds": 1.030573132000427,
238
+ "bytes": 11789498,
239
+ "pages": 32,
240
+ "shards": 3
241
  },
242
  {
243
+ "seconds": 0.8314693719985371,
244
+ "bytes": 2720719,
245
  "pages": 36,
246
  "shards": 2
247
  },
248
  {
249
+ "seconds": 0.9525506049976684,
250
+ "bytes": 10860039,
251
+ "pages": 34,
252
+ "shards": 3
253
+ },
254
+ {
255
+ "seconds": 1.2853242279998085,
256
+ "bytes": 7445426,
257
+ "pages": 38,
258
+ "shards": 3
259
+ },
260
+ {
261
+ "seconds": 0.7597367620001023,
262
+ "bytes": 831447,
263
+ "pages": 25,
264
+ "shards": 1
265
+ },
266
+ {
267
+ "seconds": 1.0118765539991728,
268
+ "bytes": 6408285,
269
+ "pages": 37,
270
  "shards": 2
271
  },
272
  {
273
+ "seconds": 1.0685828210007458,
274
+ "bytes": 12919611,
275
+ "pages": 37,
276
  "shards": 4
277
  },
278
  {
279
+ "seconds": 1.0399747580013354,
280
+ "bytes": 5973768,
281
  "pages": 27,
282
  "shards": 2
283
  },
284
  {
285
+ "seconds": 1.0144561509987398,
286
+ "bytes": 6361193,
287
  "pages": 36,
288
  "shards": 2
289
  },
290
  {
291
+ "seconds": 1.0571298460017715,
292
+ "bytes": 2527314,
293
+ "pages": 36,
294
  "shards": 2
295
  },
296
  {
297
+ "seconds": 1.1346876719981083,
298
+ "bytes": 10084333,
299
+ "pages": 34,
300
+ "shards": 3
301
  },
302
  {
303
+ "seconds": 0.9850222569984908,
304
+ "bytes": 2356425,
305
+ "pages": 28,
306
+ "shards": 2
307
  },
308
  {
309
+ "seconds": 1.0524133049984812,
310
+ "bytes": 12178249,
311
+ "pages": 40,
312
  "shards": 3
313
  },
314
  {
315
+ "seconds": 1.109467931000836,
316
+ "bytes": 5783909,
317
+ "pages": 28,
318
+ "shards": 2
319
  },
320
  {
321
+ "seconds": 0.8641120300017064,
322
+ "bytes": 2329685,
323
+ "pages": 28,
324
+ "shards": 2
325
  },
326
  {
327
+ "seconds": 1.4142594379991351,
328
+ "bytes": 10485303,
329
+ "pages": 41,
330
  "shards": 3
331
  },
332
  {
333
+ "seconds": 0.9442589440004667,
334
+ "bytes": 6417253,
335
+ "pages": 36,
336
+ "shards": 2
337
  },
338
  {
339
+ "seconds": 0.9173370479984442,
340
+ "bytes": 2397470,
341
+ "pages": 28,
342
  "shards": 2
343
  },
344
  {
345
+ "seconds": 1.938626930997998,
346
+ "bytes": 16310135,
347
+ "pages": 36,
348
+ "shards": 4
349
  },
350
  {
351
+ "seconds": 1.1073444919966278,
352
+ "bytes": 2450943,
353
+ "pages": 33,
354
+ "shards": 2
355
  },
356
  {
357
+ "seconds": 0.8343555079991347,
358
+ "bytes": 5999582,
359
+ "pages": 37,
360
  "shards": 2
361
  },
362
  {
363
+ "seconds": 0.9876751570009219,
364
+ "bytes": 1005909,
365
+ "pages": 31,
366
+ "shards": 1
367
+ },
368
+ {
369
+ "seconds": 1.0930767579993699,
370
+ "bytes": 7383476,
371
+ "pages": 31,
372
  "shards": 3
373
  },
374
  {
375
+ "seconds": 1.2710551140007738,
376
+ "bytes": 11134870,
377
+ "pages": 40,
378
  "shards": 3
379
  },
380
  {
381
+ "seconds": 1.053878970000369,
382
+ "bytes": 8139088,
383
+ "pages": 33,
384
+ "shards": 3
385
  },
386
  {
387
+ "seconds": 1.03730981300032,
388
+ "bytes": 6215262,
389
+ "pages": 28,
390
  "shards": 2
391
  },
392
  {
393
+ "seconds": 0.9270265450031729,
394
+ "bytes": 1100519,
395
+ "pages": 35,
396
  "shards": 1
397
  },
398
  {
399
+ "seconds": 1.1968161780023365,
400
+ "bytes": 11473828,
401
+ "pages": 39,
402
+ "shards": 4
403
  },
404
  {
405
+ "seconds": 0.9470581830028095,
406
+ "bytes": 5899775,
407
  "pages": 28,
408
+ "shards": 3
409
  },
410
  {
411
+ "seconds": 1.0518160799983889,
412
+ "bytes": 1140036,
413
+ "pages": 35,
414
+ "shards": 1
415
  },
416
  {
417
+ "seconds": 1.1662064560005092,
418
+ "bytes": 7349596,
419
+ "pages": 29,
420
  "shards": 3
421
  },
422
  {
423
+ "seconds": 1.122332668001036,
424
+ "bytes": 11201873,
425
+ "pages": 34,
426
+ "shards": 3
427
  },
428
  {
429
+ "seconds": 0.8321514729977935,
430
+ "bytes": 6044525,
431
  "pages": 36,
432
+ "shards": 2
433
  },
434
  {
435
+ "seconds": 1.4008962299994892,
436
+ "bytes": 7557731,
437
+ "pages": 39,
438
  "shards": 3
439
  },
440
  {
441
+ "seconds": 0.8883114669988572,
442
+ "bytes": 823542,
443
+ "pages": 25,
444
  "shards": 1
445
  },
446
  {
447
+ "seconds": 1.2165515029992093,
448
+ "bytes": 11904873,
449
+ "pages": 42,
450
  "shards": 4
451
  },
452
  {
453
+ "seconds": 0.9436809830003767,
454
+ "bytes": 889392,
455
+ "pages": 26,
456
  "shards": 1
457
  },
458
  {
459
+ "seconds": 1.1113080279974383,
460
+ "bytes": 11835067,
461
+ "pages": 35,
462
+ "shards": 3
463
  },
464
  {
465
+ "seconds": 0.9863674269981857,
466
+ "bytes": 7525026,
467
  "pages": 38,
468
  "shards": 3
469
  },
470
  {
471
+ "seconds": 0.7792704010025773,
472
+ "bytes": 5012728,
473
+ "pages": 27,
474
+ "shards": 2
475
  },
476
  {
477
+ "seconds": 1.0204439990011451,
478
+ "bytes": 7601581,
479
+ "pages": 38,
480
  "shards": 3
481
  },
482
  {
483
+ "seconds": 0.844439466996846,
484
+ "bytes": 1090867,
485
  "pages": 34,
486
  "shards": 1
487
  },
488
  {
489
+ "seconds": 0.8992618750016845,
490
+ "bytes": 5707571,
491
+ "pages": 29,
492
+ "shards": 2
493
  },
494
  {
495
+ "seconds": 1.1450656299966795,
496
+ "bytes": 17577071,
497
+ "pages": 37,
498
+ "shards": 5
499
  },
500
  {
501
+ "seconds": 1.0943805140013865,
502
+ "bytes": 891607,
503
  "pages": 26,
504
  "shards": 1
505
  },
506
  {
507
+ "seconds": 0.9035562749995734,
508
+ "bytes": 6454400,
509
+ "pages": 35,
 
 
 
 
 
 
510
  "shards": 2
511
  },
512
  {
513
+ "seconds": 1.2266940429981332,
514
+ "bytes": 2571680,
 
 
 
 
 
 
515
  "pages": 38,
 
 
 
 
 
 
 
 
 
 
 
 
516
  "shards": 2
517
  },
518
  {
519
+ "seconds": 1.0315104050023365,
520
+ "bytes": 10269533,
521
+ "pages": 32,
522
+ "shards": 3
523
  },
524
  {
525
+ "seconds": 0.8940106660011224,
526
+ "bytes": 5798800,
527
+ "pages": 31,
528
  "shards": 2
529
  },
530
  {
531
+ "seconds": 1.0886239710016525,
532
+ "bytes": 8189805,
533
+ "pages": 33,
534
+ "shards": 3
535
  },
536
  {
537
+ "seconds": 0.7768403989975923,
538
+ "bytes": 6370933,
539
  "pages": 36,
540
  "shards": 2
541
  },
542
  {
543
+ "seconds": 1.1446712159995514,
544
+ "bytes": 7008906,
545
+ "pages": 33,
546
  "shards": 3
547
  },
548
  {
549
+ "seconds": 0.9077771720003511,
550
+ "bytes": 6012024,
551
+ "pages": 36,
552
  "shards": 2
553
  },
554
  {
555
+ "seconds": 1.4875493239997013,
556
+ "bytes": 6032602,
557
+ "pages": 34,
558
+ "shards": 2
559
  },
560
  {
561
+ "seconds": 0.8995478449978691,
562
+ "bytes": 2436827,
563
+ "pages": 30,
564
  "shards": 2
565
  }
566
  ],
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:c0a2189fdee90c71c35753bc51ffdb3afecbeb4a346bd3889656ce4fe54fb477
3
- size 721767167
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:91c1a40f9a6c8b01e9e50d09f40d34bfc35834f4940f699fc2705fad10d4fc7d
3
+ size 721767935