appvoid commited on
Commit
5d38250
·
verified ·
1 Parent(s): 1102e43

Qwen90M step 525000; complete resumable checkpoint

Browse files
Files changed (6) hide show
  1. complete.json +5 -5
  2. coverage.json +46 -46
  3. latest.json +1 -1
  4. model.safetensors +1 -1
  5. shard_scheduler.json +242 -242
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 520000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "94689d39f31e52e9199eeddbbd7aaa83322651ef0705b3f99ac0f5afe885fe7f",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "84af1488eb650b70be0d37c145b2c99c81009be6ddae8e4bb3426d935b6fde18",
12
  "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
13
  "performance.json": "234744063e094dfd7b525da4e181ac493fac400c830f871c7e8c59127ed10dc9",
14
- "shard_scheduler.json": "ca45b34a5ce482493ede09362d54c536a11ccac97331918b9089c1ba2346d3c8",
15
- "coverage.json": "9839d8d5634e978124cc5b5644cf2998b990933ad09e03d5f004fa841278743f",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 525000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "555a0b375b5d8b6df17309058cb3fac983e079e022e3453c5bb40694dc0dfd3c",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "550740c374917e84e0c390d9c1282f579154291db3c66058d82c3775128b44da",
12
  "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
13
  "performance.json": "234744063e094dfd7b525da4e181ac493fac400c830f871c7e8c59127ed10dc9",
14
+ "shard_scheduler.json": "ded67c5cf32a54e2366cba7242e168a5494026aca2d41bea5b359a7947e97553",
15
+ "coverage.json": "8e20b86d8845516d961760222cf2e262fe8d2ed820d4ef25ec6904b4dc6e9d4b",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 520000,
3
  "stats": {
4
- "climbmix/blocks": 18720000,
5
- "climbmix/targets": 31291918409,
6
- "climbmix/input_tokens": 31293901346,
7
- "ultra_qa/blocks": 7020000,
8
- "ultra_qa/targets": 11452247156,
9
- "ultra_qa/input_tokens": 11452269248,
10
- "cortex/blocks": 7020000,
11
- "cortex/targets": 266976897,
12
- "cortex/input_tokens": 11159205229,
13
- "ultra_style/blocks": 7020000,
14
- "ultra_style/targets": 11102583967,
15
- "ultra_style/input_tokens": 11103166058,
16
- "openmath/blocks": 4680000,
17
- "openmath/targets": 7231495445,
18
- "openmath/input_tokens": 7233904958,
19
- "rewrite6/blocks": 2340000,
20
- "rewrite6/targets": 3836125690,
21
- "rewrite6/input_tokens": 3840331312
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 2564,
26
- "row": 700,
27
  "start": {
28
- "file": 124,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 6,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
- "documents": 10502844
38
  },
39
  "rewrite6": {
40
- "page": 8,
41
- "row": 2036,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 32768
47
  }
48
  },
49
  "epoch": 29,
50
  "offset": 0,
51
- "documents": 5991520
52
  },
53
  "openmath": {
54
- "page": 1602,
55
- "row": 1247,
56
  "start": {
57
  "file": 15,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 12,
61
- "row": 42
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 6563039
67
  },
68
  "ultra_qa": {
69
- "page": 692,
70
- "row": 3292,
71
  "start": {
72
  "file": 5,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
- "row_group": 3,
76
- "row": 42729
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
- "documents": 2837724
82
  },
83
  "ultra_style": {
84
- "page": 1057,
85
- "row": 1722,
86
  "start": {
87
  "file": 6,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
- "row_group": 2,
91
- "row": 49196
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
- "documents": 4331194
97
  },
98
  "cortex": {
99
- "page": 136152,
100
- "row": 51,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 17427507
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 525000,
3
  "stats": {
4
+ "climbmix/blocks": 18900000,
5
+ "climbmix/targets": 31592797042,
6
+ "climbmix/input_tokens": 31594799016,
7
+ "ultra_qa/blocks": 7087500,
8
+ "ultra_qa/targets": 11562286610,
9
+ "ultra_qa/input_tokens": 11562308928,
10
+ "cortex/blocks": 7087500,
11
+ "cortex/targets": 269544792,
12
+ "cortex/input_tokens": 11266328102,
13
+ "ultra_style/blocks": 7087500,
14
+ "ultra_style/targets": 11209247634,
15
+ "ultra_style/input_tokens": 11209834992,
16
+ "openmath/blocks": 4725000,
17
+ "openmath/targets": 7301172470,
18
+ "openmath/input_tokens": 7303605185,
19
+ "rewrite6/blocks": 2362500,
20
+ "rewrite6/targets": 3872985034,
21
+ "rewrite6/input_tokens": 3877231666
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 2588,
26
+ "row": 3394,
27
  "start": {
28
+ "file": 125,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 20,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
+ "documents": 10603842
38
  },
39
  "rewrite6": {
40
+ "page": 22,
41
+ "row": 2978,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 90112
47
  }
48
  },
49
  "epoch": 29,
50
  "offset": 0,
51
+ "documents": 6049806
52
  },
53
  "openmath": {
54
+ "page": 1617,
55
+ "row": 2871,
56
  "start": {
57
  "file": 15,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 73,
61
+ "row": 482
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 6626103
67
  },
68
  "ultra_qa": {
69
+ "page": 699,
70
+ "row": 1991,
71
  "start": {
72
  "file": 5,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
+ "row_group": 4,
76
+ "row": 6884
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
+ "documents": 2865095
82
  },
83
  "ultra_style": {
84
+ "page": 1067,
85
+ "row": 1860,
86
  "start": {
87
  "file": 6,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
+ "row_group": 3,
91
+ "row": 4718
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
+ "documents": 4372292
97
  },
98
  "cortex": {
99
+ "page": 137458,
100
+ "row": 115,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 17594739
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 520000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "94689d39f31e52e9199eeddbbd7aaa83322651ef0705b3f99ac0f5afe885fe7f", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "84af1488eb650b70be0d37c145b2c99c81009be6ddae8e4bb3426d935b6fde18", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "234744063e094dfd7b525da4e181ac493fac400c830f871c7e8c59127ed10dc9", "shard_scheduler.json": "ca45b34a5ce482493ede09362d54c536a11ccac97331918b9089c1ba2346d3c8", "coverage.json": "9839d8d5634e978124cc5b5644cf2998b990933ad09e03d5f004fa841278743f", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000520000"}
 
1
+ {"step": 525000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "555a0b375b5d8b6df17309058cb3fac983e079e022e3453c5bb40694dc0dfd3c", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "550740c374917e84e0c390d9c1282f579154291db3c66058d82c3775128b44da", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "234744063e094dfd7b525da4e181ac493fac400c830f871c7e8c59127ed10dc9", "shard_scheduler.json": "ded67c5cf32a54e2366cba7242e168a5494026aca2d41bea5b359a7947e97553", "coverage.json": "8e20b86d8845516d961760222cf2e262fe8d2ed820d4ef25ec6904b4dc6e9d4b", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000525000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:94689d39f31e52e9199eeddbbd7aaa83322651ef0705b3f99ac0f5afe885fe7f
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:555a0b375b5d8b6df17309058cb3fac983e079e022e3453c5bb40694dc0dfd3c
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": 76689.73514938472,
26
- "required_pages_per_second_upper_estimate": 0.005305573822335977
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": 9586.21689367309,
40
- "required_pages_per_second_upper_estimate": 0.004013548830408618
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": 19172.43378734618,
54
- "required_pages_per_second_upper_estimate": 0.003993112346824364
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": 28758.65068101927,
68
- "required_pages_per_second_upper_estimate": 0.0017240371964100977
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": 28758.65068101927,
82
- "required_pages_per_second_upper_estimate": 0.002235858594513405
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": 28758.65068101927,
96
- "required_pages_per_second_upper_estimate": 0.3688850923028087
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": 704566,
161
- "producer_batch_seconds": 0.12374689964809704,
162
- "producer_p95_seconds": 0.2969493060008972,
163
- "gpu_update_seconds": 0.9613802923739986,
164
- "producer_batches_per_second": 8.081010537183005,
165
- "gpu_batches_per_second": 1.0401711039141808,
166
- "producer_headroom": 7.76892427291436,
167
  "sustainable": true,
168
- "buffer_seconds": 60.566958419561914,
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": 353411.1862394783,
174
- "upload_bytes_per_second": 5844464.712245397,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 75.83381357563269,
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": 1.3237921879990608,
184
- "bytes": 13505764,
185
  "pages": 42,
186
- "shards": 5
187
- },
188
- {
189
- "seconds": 1.3861956760010798,
190
- "bytes": 10089467,
191
- "pages": 32,
192
  "shards": 3
193
  },
194
  {
195
- "seconds": 1.1900551170001563,
196
- "bytes": 2519802,
197
- "pages": 34,
198
  "shards": 2
199
  },
200
  {
201
- "seconds": 1.0123512749996735,
202
- "bytes": 5925157,
203
- "pages": 33,
204
- "shards": 2
205
  },
206
  {
207
- "seconds": 0.84063605999836,
208
- "bytes": 1225280,
209
- "pages": 36,
210
- "shards": 1
211
  },
212
  {
213
- "seconds": 1.2510575380001683,
214
- "bytes": 12042377,
215
- "pages": 34,
216
- "shards": 4
217
  },
218
  {
219
- "seconds": 1.23156422700049,
220
- "bytes": 11911377,
221
- "pages": 32,
222
- "shards": 3
223
  },
224
  {
225
- "seconds": 0.8838252950008609,
226
- "bytes": 2597092,
227
- "pages": 37,
228
  "shards": 2
229
  },
230
  {
231
- "seconds": 1.0776322439996875,
232
- "bytes": 6282866,
233
- "pages": 37,
234
  "shards": 2
235
  },
236
  {
237
- "seconds": 1.0567811459986842,
238
- "bytes": 4557138,
239
- "pages": 29,
240
  "shards": 2
241
  },
242
  {
243
- "seconds": 1.3797404860015376,
244
- "bytes": 7238899,
245
- "pages": 30,
246
- "shards": 3
247
  },
248
  {
249
- "seconds": 1.0933739200008858,
250
- "bytes": 6120770,
251
- "pages": 37,
252
  "shards": 2
253
  },
254
  {
255
- "seconds": 0.9898015409999061,
256
- "bytes": 867435,
257
- "pages": 29,
258
- "shards": 1
259
  },
260
  {
261
- "seconds": 1.4056065479999234,
262
- "bytes": 17665783,
263
  "pages": 34,
264
- "shards": 5
265
  },
266
  {
267
- "seconds": 1.0921460530007607,
268
- "bytes": 6276750,
269
  "pages": 37,
270
  "shards": 2
271
  },
272
  {
273
- "seconds": 1.0008894900020096,
274
- "bytes": 2319506,
275
- "pages": 29,
276
  "shards": 2
277
  },
278
  {
279
- "seconds": 1.162798567002028,
280
- "bytes": 6350229,
281
- "pages": 37,
282
- "shards": 2
283
  },
284
  {
285
- "seconds": 1.0111840479985403,
286
- "bytes": 5239532,
287
- "pages": 38,
288
- "shards": 2
289
  },
290
  {
291
- "seconds": 1.2082140299971798,
292
- "bytes": 13368136,
293
- "pages": 36,
294
- "shards": 4
295
  },
296
  {
297
- "seconds": 1.2696354079998855,
298
- "bytes": 6070118,
299
- "pages": 28,
300
  "shards": 2
301
  },
302
  {
303
- "seconds": 0.9898074049997376,
304
- "bytes": 2654933,
305
- "pages": 38,
306
  "shards": 2
307
  },
308
  {
309
- "seconds": 1.3211888400001044,
310
- "bytes": 9995304,
311
- "pages": 31,
312
- "shards": 3
313
  },
314
  {
315
- "seconds": 0.8846473650010012,
316
- "bytes": 1315669,
317
- "pages": 36,
318
- "shards": 1
319
  },
320
  {
321
- "seconds": 1.6196549160013092,
322
- "bytes": 7467278,
323
- "pages": 37,
324
- "shards": 3
325
  },
326
  {
327
- "seconds": 1.2983564580026723,
328
- "bytes": 12027088,
329
- "pages": 34,
330
  "shards": 3
331
  },
332
  {
333
- "seconds": 1.2329098789996351,
334
- "bytes": 6852339,
335
- "pages": 35,
336
- "shards": 3
337
  },
338
  {
339
- "seconds": 1.050476451000577,
340
- "bytes": 6070361,
341
- "pages": 29,
342
  "shards": 2
343
  },
344
  {
345
- "seconds": 1.0647863270023663,
346
- "bytes": 1121901,
347
- "pages": 35,
348
  "shards": 1
349
  },
350
  {
351
- "seconds": 1.296304608000355,
352
- "bytes": 7362354,
353
- "pages": 34,
 
 
 
 
 
 
354
  "shards": 3
355
  },
356
  {
357
- "seconds": 1.2885511929998756,
358
- "bytes": 10359776,
 
 
 
 
 
 
359
  "pages": 39,
360
  "shards": 3
361
  },
362
  {
363
- "seconds": 1.3631092460018408,
364
- "bytes": 8446837,
365
- "pages": 36,
366
- "shards": 4
367
  },
368
  {
369
- "seconds": 1.1454860439989716,
370
- "bytes": 6061904,
371
  "pages": 28,
372
  "shards": 2
373
  },
374
  {
375
- "seconds": 1.4194609969999874,
376
- "bytes": 9675770,
377
- "pages": 36,
378
  "shards": 3
379
  },
380
  {
381
- "seconds": 1.3136005369960913,
382
- "bytes": 2685859,
383
- "pages": 36,
384
  "shards": 2
385
  },
386
  {
387
- "seconds": 1.3078916630038293,
388
- "bytes": 5962936,
389
- "pages": 38,
390
  "shards": 2
391
  },
392
  {
393
- "seconds": 1.0175234589987667,
394
- "bytes": 928482,
395
- "pages": 27,
396
- "shards": 1
397
- },
398
- {
399
- "seconds": 1.5923971939992043,
400
- "bytes": 18139508,
401
  "pages": 43,
402
- "shards": 5
403
  },
404
  {
405
- "seconds": 0.9787014109970187,
406
- "bytes": 5956252,
407
- "pages": 28,
408
  "shards": 2
409
  },
410
  {
411
- "seconds": 0.8562311600035173,
412
- "bytes": 2647079,
413
- "pages": 38,
414
  "shards": 2
415
  },
416
  {
417
- "seconds": 1.054913633997785,
418
- "bytes": 5917684,
419
- "pages": 31,
420
- "shards": 2
421
  },
422
  {
423
- "seconds": 1.521548144002736,
424
- "bytes": 10001036,
425
- "pages": 42,
426
- "shards": 3
427
  },
428
  {
429
- "seconds": 1.037408388001495,
430
- "bytes": 2352962,
431
- "pages": 29,
432
  "shards": 2
433
  },
434
  {
435
- "seconds": 1.4666764929934288,
436
- "bytes": 11848768,
437
- "pages": 38,
438
- "shards": 3
439
  },
440
  {
441
- "seconds": 1.1809732929978054,
442
- "bytes": 12035253,
443
- "pages": 37,
444
  "shards": 4
445
  },
446
  {
447
- "seconds": 0.976799342999584,
448
- "bytes": 895629,
449
- "pages": 27,
450
- "shards": 1
451
  },
452
  {
453
- "seconds": 1.1010828580037924,
454
- "bytes": 6038977,
 
 
 
 
 
 
455
  "pages": 38,
456
  "shards": 2
457
  },
458
  {
459
- "seconds": 1.0995818070005043,
460
- "bytes": 2583346,
461
  "pages": 34,
462
- "shards": 2
463
  },
464
  {
465
- "seconds": 0.9759389660030138,
466
- "bytes": 6031085,
467
- "pages": 33,
468
  "shards": 2
469
  },
470
  {
471
- "seconds": 1.012582071001816,
472
- "bytes": 5087118,
473
- "pages": 25,
474
  "shards": 2
475
  },
476
  {
477
- "seconds": 1.1673960980042466,
478
- "bytes": 13547167,
479
- "pages": 41,
480
- "shards": 4
481
  },
482
  {
483
- "seconds": 1.3716537860018434,
484
- "bytes": 6334955,
485
- "pages": 38,
486
- "shards": 2
487
  },
488
  {
489
- "seconds": 1.127826156996889,
490
- "bytes": 2574706,
491
- "pages": 33,
492
  "shards": 2
493
  },
494
  {
495
- "seconds": 1.1363400170012028,
496
- "bytes": 10297243,
497
- "pages": 34,
498
- "shards": 3
499
- },
500
- {
501
- "seconds": 1.2236347540019779,
502
- "bytes": 6431007,
503
  "pages": 37,
504
  "shards": 2
505
  },
506
  {
507
- "seconds": 0.9796522880060365,
508
- "bytes": 2218113,
509
- "pages": 23,
510
  "shards": 2
511
  },
512
  {
513
- "seconds": 1.3331365960038966,
514
- "bytes": 12026403,
515
- "pages": 40,
516
  "shards": 3
517
  },
518
  {
519
- "seconds": 1.336267955994117,
520
- "bytes": 7188821,
521
- "pages": 41,
522
  "shards": 3
523
  },
524
  {
525
- "seconds": 1.3482005179976113,
526
- "bytes": 6310167,
527
- "pages": 32,
528
  "shards": 2
529
  },
530
  {
531
- "seconds": 1.184783762000734,
532
- "bytes": 5973673,
533
- "pages": 38,
534
- "shards": 2
535
  },
536
  {
537
- "seconds": 0.9036034050004673,
538
- "bytes": 2269487,
539
- "pages": 25,
540
  "shards": 2
541
  },
542
  {
543
- "seconds": 1.6477199560031295,
544
- "bytes": 16059670,
545
- "pages": 45,
546
- "shards": 5
547
  },
548
  {
549
- "seconds": 1.0709953250043327,
550
- "bytes": 6125208,
551
- "pages": 31,
552
- "shards": 2
553
  },
554
  {
555
- "seconds": 0.8623744039941812,
556
- "bytes": 2649650,
557
  "pages": 39,
558
- "shards": 2
559
  },
560
  {
561
- "seconds": 1.4692237859999295,
562
- "bytes": 10546134,
563
- "pages": 31,
564
- "shards": 3
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": 77213.6919826638,
26
+ "required_pages_per_second_upper_estimate": 0.005341822371809575
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": 9651.711497832976,
40
+ "required_pages_per_second_upper_estimate": 0.004040970053487503
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": 19303.42299566595,
54
+ "required_pages_per_second_upper_estimate": 0.00402039394449965
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": 28955.13449349893,
68
+ "required_pages_per_second_upper_estimate": 0.001735816100954843
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": 28955.13449349893,
82
+ "required_pages_per_second_upper_estimate": 0.002251134346692743
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": 28955.13449349893,
96
+ "required_pages_per_second_upper_estimate": 0.37140537568141674
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": 704566,
161
+ "producer_batch_seconds": 0.11904607814852852,
162
+ "producer_p95_seconds": 0.3377903690052335,
163
+ "gpu_update_seconds": 0.9548565559661825,
164
+ "producer_batches_per_second": 8.400108727247144,
165
+ "gpu_batches_per_second": 1.0472777232891683,
166
+ "producer_headroom": 8.02089888904068,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.1559630258695,
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": 384588.48339903663,
174
+ "upload_bytes_per_second": 3837530.023601385,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 82.52373577574086,
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": 1.521548144002736,
184
+ "bytes": 10001036,
185
  "pages": 42,
 
 
 
 
 
 
186
  "shards": 3
187
  },
188
  {
189
+ "seconds": 1.037408388001495,
190
+ "bytes": 2352962,
191
+ "pages": 29,
192
  "shards": 2
193
  },
194
  {
195
+ "seconds": 1.4666764929934288,
196
+ "bytes": 11848768,
197
+ "pages": 38,
198
+ "shards": 3
199
  },
200
  {
201
+ "seconds": 1.1809732929978054,
202
+ "bytes": 12035253,
203
+ "pages": 37,
204
+ "shards": 4
205
  },
206
  {
207
+ "seconds": 0.976799342999584,
208
+ "bytes": 895629,
209
+ "pages": 27,
210
+ "shards": 1
211
  },
212
  {
213
+ "seconds": 1.1010828580037924,
214
+ "bytes": 6038977,
215
+ "pages": 38,
216
+ "shards": 2
217
  },
218
  {
219
+ "seconds": 1.0995818070005043,
220
+ "bytes": 2583346,
221
+ "pages": 34,
222
  "shards": 2
223
  },
224
  {
225
+ "seconds": 0.9759389660030138,
226
+ "bytes": 6031085,
227
+ "pages": 33,
228
  "shards": 2
229
  },
230
  {
231
+ "seconds": 1.012582071001816,
232
+ "bytes": 5087118,
233
+ "pages": 25,
234
  "shards": 2
235
  },
236
  {
237
+ "seconds": 1.1673960980042466,
238
+ "bytes": 13547167,
239
+ "pages": 41,
240
+ "shards": 4
241
  },
242
  {
243
+ "seconds": 1.3716537860018434,
244
+ "bytes": 6334955,
245
+ "pages": 38,
246
  "shards": 2
247
  },
248
  {
249
+ "seconds": 1.127826156996889,
250
+ "bytes": 2574706,
251
+ "pages": 33,
252
+ "shards": 2
253
  },
254
  {
255
+ "seconds": 1.1363400170012028,
256
+ "bytes": 10297243,
257
  "pages": 34,
258
+ "shards": 3
259
  },
260
  {
261
+ "seconds": 1.2236347540019779,
262
+ "bytes": 6431007,
263
  "pages": 37,
264
  "shards": 2
265
  },
266
  {
267
+ "seconds": 0.9796522880060365,
268
+ "bytes": 2218113,
269
+ "pages": 23,
270
  "shards": 2
271
  },
272
  {
273
+ "seconds": 1.3331365960038966,
274
+ "bytes": 12026403,
275
+ "pages": 40,
276
+ "shards": 3
277
  },
278
  {
279
+ "seconds": 1.336267955994117,
280
+ "bytes": 7188821,
281
+ "pages": 41,
282
+ "shards": 3
283
  },
284
  {
285
+ "seconds": 1.3482005179976113,
286
+ "bytes": 6310167,
287
+ "pages": 32,
288
+ "shards": 2
289
  },
290
  {
291
+ "seconds": 1.184783762000734,
292
+ "bytes": 5973673,
293
+ "pages": 38,
294
  "shards": 2
295
  },
296
  {
297
+ "seconds": 0.9036034050004673,
298
+ "bytes": 2269487,
299
+ "pages": 25,
300
  "shards": 2
301
  },
302
  {
303
+ "seconds": 1.6477199560031295,
304
+ "bytes": 16059670,
305
+ "pages": 45,
306
+ "shards": 5
307
  },
308
  {
309
+ "seconds": 1.0709953250043327,
310
+ "bytes": 6125208,
311
+ "pages": 31,
312
+ "shards": 2
313
  },
314
  {
315
+ "seconds": 0.8623744039941812,
316
+ "bytes": 2649650,
317
+ "pages": 39,
318
+ "shards": 2
319
  },
320
  {
321
+ "seconds": 1.4692237859999295,
322
+ "bytes": 10546134,
323
+ "pages": 31,
324
  "shards": 3
325
  },
326
  {
327
+ "seconds": 1.1281011890023365,
328
+ "bytes": 2729862,
329
+ "pages": 38,
330
+ "shards": 2
331
  },
332
  {
333
+ "seconds": 2.671323365997523,
334
+ "bytes": 5596441,
335
+ "pages": 20,
336
  "shards": 2
337
  },
338
  {
339
+ "seconds": 1.0342092059945571,
340
+ "bytes": 930671,
341
+ "pages": 29,
342
  "shards": 1
343
  },
344
  {
345
+ "seconds": 1.692445526998199,
346
+ "bytes": 13078461,
347
+ "pages": 32,
348
+ "shards": 4
349
+ },
350
+ {
351
+ "seconds": 1.3595879119966412,
352
+ "bytes": 10348219,
353
+ "pages": 44,
354
  "shards": 3
355
  },
356
  {
357
+ "seconds": 1.0515126419995795,
358
+ "bytes": 1003171,
359
+ "pages": 30,
360
+ "shards": 1
361
+ },
362
+ {
363
+ "seconds": 1.000422951001383,
364
+ "bytes": 7649877,
365
  "pages": 39,
366
  "shards": 3
367
  },
368
  {
369
+ "seconds": 1.2835066039988305,
370
+ "bytes": 10437972,
371
+ "pages": 31,
372
+ "shards": 3
373
  },
374
  {
375
+ "seconds": 0.8616337460043724,
376
+ "bytes": 2430044,
377
  "pages": 28,
378
  "shards": 2
379
  },
380
  {
381
+ "seconds": 1.3565980959974695,
382
+ "bytes": 12199549,
383
+ "pages": 43,
384
  "shards": 3
385
  },
386
  {
387
+ "seconds": 1.012810416999855,
388
+ "bytes": 6201884,
389
+ "pages": 37,
390
  "shards": 2
391
  },
392
  {
393
+ "seconds": 0.8832714039963321,
394
+ "bytes": 2126873,
395
+ "pages": 22,
396
  "shards": 2
397
  },
398
  {
399
+ "seconds": 1.4999150360017666,
400
+ "bytes": 10166456,
 
 
 
 
 
 
401
  "pages": 43,
402
+ "shards": 3
403
  },
404
  {
405
+ "seconds": 0.9999620900052832,
406
+ "bytes": 2535870,
407
+ "pages": 34,
408
  "shards": 2
409
  },
410
  {
411
+ "seconds": 1.0642618760030018,
412
+ "bytes": 5883611,
413
+ "pages": 24,
414
  "shards": 2
415
  },
416
  {
417
+ "seconds": 2.4567707870010054,
418
+ "bytes": 16775004,
419
+ "pages": 44,
420
+ "shards": 4
421
  },
422
  {
423
+ "seconds": 1.3269825160023174,
424
+ "bytes": 2735395,
425
+ "pages": 38,
426
+ "shards": 2
427
  },
428
  {
429
+ "seconds": 2.3559383290048572,
430
+ "bytes": 6286319,
431
+ "pages": 38,
432
  "shards": 2
433
  },
434
  {
435
+ "seconds": 0.9432465139980195,
436
+ "bytes": 858963,
437
+ "pages": 27,
438
+ "shards": 1
439
  },
440
  {
441
+ "seconds": 1.6026845509986742,
442
+ "bytes": 11607409,
443
+ "pages": 29,
444
  "shards": 4
445
  },
446
  {
447
+ "seconds": 1.04873497600056,
448
+ "bytes": 6766779,
449
+ "pages": 32,
450
+ "shards": 2
451
  },
452
  {
453
+ "seconds": 1.4542191169966827,
454
+ "bytes": 7891732,
455
+ "pages": 46,
456
+ "shards": 3
457
+ },
458
+ {
459
+ "seconds": 1.0971116009968682,
460
+ "bytes": 6168592,
461
  "pages": 38,
462
  "shards": 2
463
  },
464
  {
465
+ "seconds": 1.321997156002908,
466
+ "bytes": 9700555,
467
  "pages": 34,
468
+ "shards": 3
469
  },
470
  {
471
+ "seconds": 1.1995469100002083,
472
+ "bytes": 2400407,
473
+ "pages": 29,
474
  "shards": 2
475
  },
476
  {
477
+ "seconds": 1.5711508849999518,
478
+ "bytes": 6147230,
479
+ "pages": 38,
480
  "shards": 2
481
  },
482
  {
483
+ "seconds": 1.4379745080004795,
484
+ "bytes": 8148657,
485
+ "pages": 31,
486
+ "shards": 3
487
  },
488
  {
489
+ "seconds": 1.2602131539970287,
490
+ "bytes": 10952650,
491
+ "pages": 39,
492
+ "shards": 4
493
  },
494
  {
495
+ "seconds": 1.072258756001247,
496
+ "bytes": 6000592,
497
+ "pages": 29,
498
  "shards": 2
499
  },
500
  {
501
+ "seconds": 0.8994364020036301,
502
+ "bytes": 2733894,
 
 
 
 
 
 
503
  "pages": 37,
504
  "shards": 2
505
  },
506
  {
507
+ "seconds": 3.8195672939982614,
508
+ "bytes": 6121969,
509
+ "pages": 34,
510
  "shards": 2
511
  },
512
  {
513
+ "seconds": 1.2331268449997879,
514
+ "bytes": 7295036,
515
+ "pages": 32,
516
  "shards": 3
517
  },
518
  {
519
+ "seconds": 31.867704994001542,
520
+ "bytes": 12094685,
521
+ "pages": 36,
522
  "shards": 3
523
  },
524
  {
525
+ "seconds": 1.475563729996793,
526
+ "bytes": 6268574,
527
+ "pages": 37,
528
  "shards": 2
529
  },
530
  {
531
+ "seconds": 1.4652358960011043,
532
+ "bytes": 7414584,
533
+ "pages": 40,
534
+ "shards": 3
535
  },
536
  {
537
+ "seconds": 2.4988657550056814,
538
+ "bytes": 5294252,
539
+ "pages": 30,
540
  "shards": 2
541
  },
542
  {
543
+ "seconds": 1.1802646279975306,
544
+ "bytes": 7575689,
545
+ "pages": 41,
546
+ "shards": 3
547
  },
548
  {
549
+ "seconds": 0.8871877959973062,
550
+ "bytes": 955502,
551
+ "pages": 27,
552
+ "shards": 1
553
  },
554
  {
555
+ "seconds": 1.3968777099944418,
556
+ "bytes": 12101140,
557
  "pages": 39,
558
+ "shards": 3
559
  },
560
  {
561
+ "seconds": 1.1570743730044342,
562
+ "bytes": 12058095,
563
+ "pages": 40,
564
+ "shards": 4
565
  }
566
  ],
567
  "cache_key": {
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:84af1488eb650b70be0d37c145b2c99c81009be6ddae8e4bb3426d935b6fde18
3
- size 721776895
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:550740c374917e84e0c390d9c1282f579154291db3c66058d82c3775128b44da
3
+ size 721783935