appvoid commited on
Commit
dc2eebd
·
verified ·
1 Parent(s): 3fd6236

Qwen90M step 195000; complete resumable checkpoint

Browse files
Files changed (6) hide show
  1. complete.json +5 -5
  2. coverage.json +47 -47
  3. latest.json +1 -1
  4. model.safetensors +1 -1
  5. shard_scheduler.json +246 -246
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 190000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "5b29d7dbcb4b6dc66cbd141f4370adef39d9c5bf0d95a0cb23dbf0b3fbf77d71",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "003d5c36c105e74cd2f29e7718b5727a4a7950ad08dd430bb87b3b0378f6d788",
12
  "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf",
13
  "performance.json": "d8fc0ab4ba9b75a14bc27c39b9cdaf8c3a3b38cccaa4c038b8370fdf2cd49c08",
14
- "shard_scheduler.json": "6e7020fbb929d08abf5c36bb9ec83b6165c6ce2fb599b8042b1e42ee010831ac",
15
- "coverage.json": "379675965a16132527a2a833bac7de4bb49e2af467d73478c93a42753679aff0",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 195000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "e7124a2db2f213964f3cd249326d01e111d4002da26c586796a207bae4728dda",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "ba31b41d133fff78caf1b22d96d15106f826e2e03648c9dbff5ca60049e9ff2e",
12
  "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf",
13
  "performance.json": "d8fc0ab4ba9b75a14bc27c39b9cdaf8c3a3b38cccaa4c038b8370fdf2cd49c08",
14
+ "shard_scheduler.json": "9382ee0711b8a32284204139fa207dcc6bdbed89a4d1e96d772bcf51dbded624",
15
+ "coverage.json": "a52c19a6ab0dd4e5feeaf4d905f52d95523ae03ff4b83abba914a7c129d58cfb",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 190000,
3
  "stats": {
4
- "climbmix/blocks": 6840000,
5
- "climbmix/targets": 11431842491,
6
- "climbmix/input_tokens": 11432567043,
7
- "ultra_qa/blocks": 2565000,
8
- "ultra_qa/targets": 4184969180,
9
- "ultra_qa/input_tokens": 4184977239,
10
- "cortex/blocks": 2565000,
11
- "cortex/targets": 97573355,
12
- "cortex/input_tokens": 4076931537,
13
- "ultra_style/blocks": 2565000,
14
- "ultra_style/targets": 4057172458,
15
- "ultra_style/input_tokens": 4057385373,
16
- "openmath/blocks": 1710000,
17
- "openmath/targets": 2642162908,
18
- "openmath/input_tokens": 2643043776,
19
- "rewrite6/blocks": 855000,
20
- "rewrite6/targets": 1401575108,
21
- "rewrite6/input_tokens": 1403112022
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 937,
26
- "row": 1183,
27
  "start": {
28
- "file": 45,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 24,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
- "documents": 3839135
38
  },
39
  "rewrite6": {
40
- "page": 33,
41
- "row": 359,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 135168
47
  }
48
  },
49
  "epoch": 10,
50
  "offset": 0,
51
- "documents": 2189567
52
  },
53
  "openmath": {
54
- "page": 585,
55
- "row": 2658,
56
  "start": {
57
  "file": 5,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 212,
61
- "row": 910
62
  }
63
  },
64
  "epoch": 0,
65
- "offset": 0,
66
- "documents": 2398818
67
  },
68
  "ultra_qa": {
69
- "page": 253,
70
- "row": 592,
71
  "start": {
72
- "file": 1,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
- "row_group": 8,
76
- "row": 410
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
- "documents": 1036880
82
  },
83
  "ultra_style": {
84
- "page": 386,
85
- "row": 1768,
86
  "start": {
87
  "file": 2,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 2,
91
- "row": 40788
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
- "documents": 1582824
97
  },
98
  "cortex": {
99
- "page": 49741,
100
- "row": 11,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 6366859
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 195000,
3
  "stats": {
4
+ "climbmix/blocks": 7020000,
5
+ "climbmix/targets": 11732742810,
6
+ "climbmix/input_tokens": 11733486473,
7
+ "ultra_qa/blocks": 2632500,
8
+ "ultra_qa/targets": 4295100300,
9
+ "ultra_qa/input_tokens": 4295108575,
10
+ "cortex/blocks": 2632500,
11
+ "cortex/targets": 100135715,
12
+ "cortex/input_tokens": 4184383493,
13
+ "ultra_style/blocks": 2632500,
14
+ "ultra_style/targets": 4163945713,
15
+ "ultra_style/input_tokens": 4164164467,
16
+ "openmath/blocks": 1755000,
17
+ "openmath/targets": 2711634953,
18
+ "openmath/input_tokens": 2712538826,
19
+ "rewrite6/blocks": 877500,
20
+ "rewrite6/targets": 1438575118,
21
+ "rewrite6/input_tokens": 1440152299
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 961,
26
+ "row": 3746,
27
  "start": {
28
+ "file": 46,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 39,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
+ "documents": 3940002
38
  },
39
  "rewrite6": {
40
+ "page": 47,
41
+ "row": 360,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 192512
47
  }
48
  },
49
  "epoch": 10,
50
  "offset": 0,
51
+ "documents": 2246912
52
  },
53
  "openmath": {
54
+ "page": 601,
55
+ "row": 98,
56
  "start": {
57
  "file": 5,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 278,
61
+ "row": 446
62
  }
63
  },
64
  "epoch": 0,
65
+ "offset": 2048,
66
+ "documents": 2461794
67
  },
68
  "ultra_qa": {
69
+ "page": 259,
70
+ "row": 3323,
71
  "start": {
72
+ "file": 2,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
+ "row_group": 0,
76
+ "row": 21613
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
+ "documents": 1064187
82
  },
83
  "ultra_style": {
84
+ "page": 396,
85
+ "row": 2960,
86
  "start": {
87
  "file": 2,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 2,
91
+ "row": 81748
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
+ "documents": 1624976
97
  },
98
  "cortex": {
99
+ "page": 51053,
100
+ "row": 14,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 6534798
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 190000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "5b29d7dbcb4b6dc66cbd141f4370adef39d9c5bf0d95a0cb23dbf0b3fbf77d71", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "003d5c36c105e74cd2f29e7718b5727a4a7950ad08dd430bb87b3b0378f6d788", "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf", "performance.json": "d8fc0ab4ba9b75a14bc27c39b9cdaf8c3a3b38cccaa4c038b8370fdf2cd49c08", "shard_scheduler.json": "6e7020fbb929d08abf5c36bb9ec83b6165c6ce2fb599b8042b1e42ee010831ac", "coverage.json": "379675965a16132527a2a833bac7de4bb49e2af467d73478c93a42753679aff0", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000190000"}
 
1
+ {"step": 195000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "e7124a2db2f213964f3cd249326d01e111d4002da26c586796a207bae4728dda", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "ba31b41d133fff78caf1b22d96d15106f826e2e03648c9dbff5ca60049e9ff2e", "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf", "performance.json": "d8fc0ab4ba9b75a14bc27c39b9cdaf8c3a3b38cccaa4c038b8370fdf2cd49c08", "shard_scheduler.json": "9382ee0711b8a32284204139fa207dcc6bdbed89a4d1e96d772bcf51dbded624", "coverage.json": "a52c19a6ab0dd4e5feeaf4d905f52d95523ae03ff4b83abba914a7c129d58cfb", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000195000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:5b29d7dbcb4b6dc66cbd141f4370adef39d9c5bf0d95a0cb23dbf0b3fbf77d71
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e7124a2db2f213964f3cd249326d01e111d4002da26c586796a207bae4728dda
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": 74817.5457905918,
26
- "required_pages_per_second_upper_estimate": 0.006484995245792312
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": 9352.193223823975,
40
- "required_pages_per_second_upper_estimate": 0.003730729704732717
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": 18704.38644764795,
54
- "required_pages_per_second_upper_estimate": 0.0043269306187049715
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": 28056.579671471925,
68
- "required_pages_per_second_upper_estimate": 0.0017084985004954352
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": 28056.579671471925,
82
- "required_pages_per_second_upper_estimate": 0.002159873168693258
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": 21,
95
- "required_token_positions_per_second": 28056.579671471925,
96
- "required_pages_per_second_upper_estimate": 0.3483688201878879
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,409 +158,409 @@
158
  "prefetch_batches": 61,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 688946,
161
- "producer_batch_seconds": 0.15711286308589933,
162
- "producer_p95_seconds": 0.3503623950018664,
163
- "gpu_update_seconds": 0.9854372957696134,
164
- "producer_batches_per_second": 6.364851230884027,
165
- "gpu_batches_per_second": 1.0147779105711778,
166
- "producer_headroom": 6.272161784938251,
167
  "sustainable": true,
168
- "buffer_seconds": 60.11167504194641,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 79,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 374399.8448116255,
174
- "upload_bytes_per_second": 4221736.650489915,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 80.33749111424993,
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": 1.3773436970004695,
184
- "bytes": 2708180,
185
- "pages": 37,
186
  "shards": 2
187
  },
188
  {
189
- "seconds": 1.157878524998523,
190
- "bytes": 6108198,
191
- "pages": 36,
192
- "shards": 2
193
  },
194
  {
195
- "seconds": 1.1510198380019574,
196
- "bytes": 2279083,
 
 
 
 
 
 
197
  "pages": 28,
198
  "shards": 2
199
  },
200
  {
201
- "seconds": 19.352378830000816,
202
- "bytes": 15868354,
203
- "pages": 40,
 
 
 
 
 
 
204
  "shards": 4
205
  },
206
  {
207
- "seconds": 1.1481948339969676,
208
- "bytes": 6679366,
209
- "pages": 36,
210
- "shards": 2
211
  },
212
  {
213
- "seconds": 1.3883708200010005,
214
- "bytes": 2398547,
215
- "pages": 33,
216
- "shards": 2
217
  },
218
  {
219
- "seconds": 1.0771264450013405,
220
- "bytes": 6009796,
221
- "pages": 32,
222
  "shards": 2
223
  },
224
  {
225
- "seconds": 0.8690516380011104,
226
- "bytes": 4841742,
227
  "pages": 29,
228
  "shards": 2
229
  },
230
  {
231
- "seconds": 1.4456373529974371,
232
- "bytes": 7578448,
233
- "pages": 35,
234
  "shards": 3
235
  },
236
  {
237
- "seconds": 1.2644371399983356,
238
- "bytes": 12142497,
239
- "pages": 38,
240
- "shards": 3
241
  },
242
  {
243
- "seconds": 0.8376236600015545,
244
- "bytes": 2410849,
245
- "pages": 28,
246
- "shards": 3
247
  },
248
  {
249
- "seconds": 1.585001501996885,
250
- "bytes": 10678100,
251
- "pages": 41,
252
  "shards": 3
253
  },
254
  {
255
- "seconds": 0.8894862470006046,
256
- "bytes": 957902,
257
- "pages": 28,
258
- "shards": 1
259
  },
260
  {
261
- "seconds": 1.1445572449993051,
262
- "bytes": 7733205,
263
- "pages": 37,
264
- "shards": 3
265
  },
266
  {
267
- "seconds": 1.1386810379990493,
268
- "bytes": 6058898,
269
- "pages": 29,
270
  "shards": 2
271
  },
272
  {
273
- "seconds": 1.2615420420006558,
274
- "bytes": 12864855,
275
- "pages": 34,
276
  "shards": 4
277
  },
278
  {
279
- "seconds": 0.9415042880027613,
280
- "bytes": 5881567,
281
- "pages": 28,
282
  "shards": 2
283
  },
284
  {
285
- "seconds": 0.8555704430000333,
286
- "bytes": 1226801,
287
- "pages": 35,
288
- "shards": 1
289
  },
290
  {
291
- "seconds": 0.896007103001466,
292
- "bytes": 7569788,
293
- "pages": 38,
294
- "shards": 3
295
  },
296
  {
297
- "seconds": 1.1828385400003754,
298
- "bytes": 9643641,
299
- "pages": 34,
300
- "shards": 3
301
  },
302
  {
303
- "seconds": 2.2667923209992296,
304
- "bytes": 913952,
305
- "pages": 26,
306
- "shards": 1
307
  },
308
  {
309
- "seconds": 1.163448106002761,
310
- "bytes": 13378237,
311
- "pages": 41,
312
- "shards": 4
313
  },
314
  {
315
- "seconds": 0.9504098300021724,
316
- "bytes": 828896,
317
- "pages": 27,
318
  "shards": 1
319
  },
320
  {
321
- "seconds": 1.1922220630003721,
322
- "bytes": 10349010,
323
- "pages": 34,
324
  "shards": 3
325
  },
326
  {
327
- "seconds": 0.9792584309980157,
328
- "bytes": 2231089,
329
- "pages": 24,
330
  "shards": 2
331
  },
332
  {
333
- "seconds": 1.3213498320001236,
334
- "bytes": 5953362,
335
- "pages": 36,
336
  "shards": 2
337
  },
338
  {
339
- "seconds": 0.9344346500001848,
340
- "bytes": 5832415,
341
- "pages": 28,
342
  "shards": 2
343
  },
344
  {
345
- "seconds": 1.198589940999227,
346
- "bytes": 12036533,
347
- "pages": 42,
 
 
 
 
 
 
348
  "shards": 4
349
  },
350
  {
351
- "seconds": 0.8354435640030715,
352
- "bytes": 5772648,
353
- "pages": 28,
354
- "shards": 2
355
  },
356
  {
357
- "seconds": 1.2891920199981541,
358
- "bytes": 7891753,
359
- "pages": 40,
360
  "shards": 3
361
  },
362
  {
363
- "seconds": 0.9012639180000406,
364
- "bytes": 5714275,
365
- "pages": 30,
366
  "shards": 2
367
  },
368
  {
369
- "seconds": 1.098721825001121,
370
- "bytes": 5875494,
371
- "pages": 37,
372
- "shards": 2
373
  },
374
  {
375
- "seconds": 1.1078113340008713,
376
- "bytes": 2283092,
377
  "pages": 28,
378
  "shards": 2
379
  },
380
  {
381
- "seconds": 0.9954401819995837,
382
- "bytes": 6010708,
 
 
 
 
 
 
383
  "pages": 29,
384
  "shards": 2
385
  },
386
  {
387
- "seconds": 1.1992983269992692,
388
- "bytes": 13571269,
389
- "pages": 40,
390
  "shards": 4
391
  },
392
  {
393
- "seconds": 1.1989007740012312,
394
- "bytes": 5658429,
395
- "pages": 39,
396
- "shards": 2
397
- },
398
- {
399
- "seconds": 1.222567180000624,
400
- "bytes": 5708293,
401
  "pages": 27,
402
- "shards": 2
403
  },
404
  {
405
- "seconds": 1.0141177349978534,
406
- "bytes": 2660846,
407
  "pages": 38,
408
  "shards": 2
409
  },
410
  {
411
- "seconds": 1.5187353149995033,
412
- "bytes": 5879969,
413
- "pages": 35,
414
- "shards": 2
415
  },
416
  {
417
- "seconds": 1.1908320559996355,
418
- "bytes": 10104452,
419
- "pages": 27,
420
  "shards": 3
421
  },
422
  {
423
- "seconds": 1.954254078998929,
424
- "bytes": 8371700,
425
- "pages": 38,
426
  "shards": 3
427
  },
428
  {
429
- "seconds": 1.207307631000731,
430
- "bytes": 6036390,
431
- "pages": 36,
432
- "shards": 2
433
  },
434
  {
435
- "seconds": 1.257034783000563,
436
- "bytes": 2415905,
437
- "pages": 31,
438
- "shards": 3
439
  },
440
  {
441
- "seconds": 1.2379260390007403,
442
- "bytes": 10087351,
443
- "pages": 39,
444
  "shards": 3
445
  },
446
  {
447
- "seconds": 1.0402059740008553,
448
- "bytes": 6293747,
449
- "pages": 28,
450
- "shards": 2
451
- },
452
- {
453
- "seconds": 1.1847501439988264,
454
- "bytes": 2572744,
455
- "pages": 33,
456
  "shards": 2
457
  },
458
  {
459
- "seconds": 1.268939194000268,
460
- "bytes": 16767391,
461
- "pages": 42,
462
- "shards": 4
463
  },
464
  {
465
- "seconds": 1.2854500899993582,
466
- "bytes": 7334524,
467
- "pages": 31,
468
- "shards": 3
469
  },
470
  {
471
- "seconds": 1.0401804770008312,
472
- "bytes": 1254870,
473
  "pages": 34,
474
- "shards": 1
475
  },
476
  {
477
- "seconds": 1.428139841002121,
478
- "bytes": 6304931,
479
- "pages": 37,
480
  "shards": 2
481
  },
482
  {
483
- "seconds": 0.987766936996195,
484
- "bytes": 2378061,
485
- "pages": 29,
486
  "shards": 2
487
  },
488
  {
489
- "seconds": 1.0273188179999124,
490
- "bytes": 9379394,
491
- "pages": 33,
492
- "shards": 3
493
  },
494
  {
495
- "seconds": 1.4798621720037772,
496
- "bytes": 13649971,
497
  "pages": 37,
498
  "shards": 4
499
  },
500
  {
501
- "seconds": 0.8202600639997399,
502
- "bytes": 1256174,
503
- "pages": 35,
 
 
 
 
 
 
504
  "shards": 1
505
  },
506
  {
507
- "seconds": 1.0791088629994192,
508
- "bytes": 10296860,
509
- "pages": 35,
510
  "shards": 3
511
  },
512
  {
513
- "seconds": 1.2926067090011202,
514
- "bytes": 2391669,
515
- "pages": 32,
516
- "shards": 2
517
  },
518
  {
519
- "seconds": 1.033323986994219,
520
- "bytes": 6001182,
521
  "pages": 33,
522
  "shards": 2
523
  },
524
  {
525
- "seconds": 3.9172589509980753,
526
- "bytes": 5890320,
527
- "pages": 32,
528
- "shards": 2
529
- },
530
- {
531
- "seconds": 1.3426188599987654,
532
- "bytes": 12956908,
533
- "pages": 37,
534
- "shards": 4
535
  },
536
  {
537
- "seconds": 1.2346246180022717,
538
- "bytes": 6315095,
539
- "pages": 36,
540
- "shards": 2
541
  },
542
  {
543
- "seconds": 3.684220068003924,
544
- "bytes": 2396656,
545
- "pages": 29,
546
- "shards": 2
547
  },
548
  {
549
- "seconds": 1.1850955770059954,
550
- "bytes": 6553717,
551
  "pages": 37,
552
- "shards": 2
553
- },
554
- {
555
- "seconds": 1.714512132006348,
556
- "bytes": 6167274,
557
- "pages": 33,
558
- "shards": 2
559
- },
560
- {
561
- "seconds": 1.2202410780009814,
562
- "bytes": 6552255,
563
- "pages": 33,
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": 74454.29039832999,
26
+ "required_pages_per_second_upper_estimate": 0.006453509189053433
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": 9306.786299791249,
40
+ "required_pages_per_second_upper_estimate": 0.0037126162038420493
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": 18613.572599582498,
54
+ "required_pages_per_second_upper_estimate": 0.004305922433224167
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": 27920.35889937375,
68
+ "required_pages_per_second_upper_estimate": 0.001700203370169813
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": 27920.35889937375,
82
+ "required_pages_per_second_upper_estimate": 0.0021493865165739087
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": 21,
95
+ "required_token_positions_per_second": 27920.35889937375,
96
+ "required_pages_per_second_upper_estimate": 0.34667741410002545
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
 
158
  "prefetch_batches": 61,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 688946,
161
+ "producer_batch_seconds": 0.10249432328106423,
162
+ "producer_p95_seconds": 0.39124788300250657,
163
+ "gpu_update_seconds": 0.9902451504883822,
164
+ "producer_batches_per_second": 9.756637909182132,
165
+ "gpu_batches_per_second": 1.0098509439877659,
166
+ "producer_headroom": 9.661463374638716,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.40495417979132,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 84,
172
  "max_pending_pages": 4096,
173
+ "generation_bytes_per_second": 328429.3339946892,
174
+ "upload_bytes_per_second": 4846781.735776416,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 70.47328963165768,
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": 1.207307631000731,
184
+ "bytes": 6036390,
185
+ "pages": 36,
186
  "shards": 2
187
  },
188
  {
189
+ "seconds": 1.257034783000563,
190
+ "bytes": 2415905,
191
+ "pages": 31,
192
+ "shards": 3
193
  },
194
  {
195
+ "seconds": 1.2379260390007403,
196
+ "bytes": 10087351,
197
+ "pages": 39,
198
+ "shards": 3
199
+ },
200
+ {
201
+ "seconds": 1.0402059740008553,
202
+ "bytes": 6293747,
203
  "pages": 28,
204
  "shards": 2
205
  },
206
  {
207
+ "seconds": 1.1847501439988264,
208
+ "bytes": 2572744,
209
+ "pages": 33,
210
+ "shards": 2
211
+ },
212
+ {
213
+ "seconds": 1.268939194000268,
214
+ "bytes": 16767391,
215
+ "pages": 42,
216
  "shards": 4
217
  },
218
  {
219
+ "seconds": 1.2854500899993582,
220
+ "bytes": 7334524,
221
+ "pages": 31,
222
+ "shards": 3
223
  },
224
  {
225
+ "seconds": 1.0401804770008312,
226
+ "bytes": 1254870,
227
+ "pages": 34,
228
+ "shards": 1
229
  },
230
  {
231
+ "seconds": 1.428139841002121,
232
+ "bytes": 6304931,
233
+ "pages": 37,
234
  "shards": 2
235
  },
236
  {
237
+ "seconds": 0.987766936996195,
238
+ "bytes": 2378061,
239
  "pages": 29,
240
  "shards": 2
241
  },
242
  {
243
+ "seconds": 1.0273188179999124,
244
+ "bytes": 9379394,
245
+ "pages": 33,
246
  "shards": 3
247
  },
248
  {
249
+ "seconds": 1.4798621720037772,
250
+ "bytes": 13649971,
251
+ "pages": 37,
252
+ "shards": 4
253
  },
254
  {
255
+ "seconds": 0.8202600639997399,
256
+ "bytes": 1256174,
257
+ "pages": 35,
258
+ "shards": 1
259
  },
260
  {
261
+ "seconds": 1.0791088629994192,
262
+ "bytes": 10296860,
263
+ "pages": 35,
264
  "shards": 3
265
  },
266
  {
267
+ "seconds": 1.2926067090011202,
268
+ "bytes": 2391669,
269
+ "pages": 32,
270
+ "shards": 2
271
  },
272
  {
273
+ "seconds": 1.033323986994219,
274
+ "bytes": 6001182,
275
+ "pages": 33,
276
+ "shards": 2
277
  },
278
  {
279
+ "seconds": 3.9172589509980753,
280
+ "bytes": 5890320,
281
+ "pages": 32,
282
  "shards": 2
283
  },
284
  {
285
+ "seconds": 1.3426188599987654,
286
+ "bytes": 12956908,
287
+ "pages": 37,
288
  "shards": 4
289
  },
290
  {
291
+ "seconds": 1.2346246180022717,
292
+ "bytes": 6315095,
293
+ "pages": 36,
294
  "shards": 2
295
  },
296
  {
297
+ "seconds": 3.684220068003924,
298
+ "bytes": 2396656,
299
+ "pages": 29,
300
+ "shards": 2
301
  },
302
  {
303
+ "seconds": 1.1850955770059954,
304
+ "bytes": 6553717,
305
+ "pages": 37,
306
+ "shards": 2
307
  },
308
  {
309
+ "seconds": 1.714512132006348,
310
+ "bytes": 6167274,
311
+ "pages": 33,
312
+ "shards": 2
313
  },
314
  {
315
+ "seconds": 1.2202410780009814,
316
+ "bytes": 6552255,
317
+ "pages": 33,
318
+ "shards": 3
319
  },
320
  {
321
+ "seconds": 1.1842839309974806,
322
+ "bytes": 12073153,
323
+ "pages": 39,
324
+ "shards": 3
325
  },
326
  {
327
+ "seconds": 0.9938246039964724,
328
+ "bytes": 235889,
329
+ "pages": 8,
330
  "shards": 1
331
  },
332
  {
333
+ "seconds": 1.3448229619971244,
334
+ "bytes": 7303500,
335
+ "pages": 31,
336
  "shards": 3
337
  },
338
  {
339
+ "seconds": 1.1307224229967687,
340
+ "bytes": 5856127,
341
+ "pages": 37,
342
  "shards": 2
343
  },
344
  {
345
+ "seconds": 1.2856477119930787,
346
+ "bytes": 6096134,
347
+ "pages": 29,
348
  "shards": 2
349
  },
350
  {
351
+ "seconds": 1.163583995999943,
352
+ "bytes": 2620539,
353
+ "pages": 37,
354
  "shards": 2
355
  },
356
  {
357
+ "seconds": 1.2240531869974802,
358
+ "bytes": 11650354,
359
+ "pages": 31,
360
+ "shards": 3
361
+ },
362
+ {
363
+ "seconds": 1.2925304770033108,
364
+ "bytes": 12240192,
365
+ "pages": 41,
366
  "shards": 4
367
  },
368
  {
369
+ "seconds": 0.9206129219965078,
370
+ "bytes": 852268,
371
+ "pages": 27,
372
+ "shards": 1
373
  },
374
  {
375
+ "seconds": 1.2395337650013971,
376
+ "bytes": 6262507,
377
+ "pages": 38,
378
  "shards": 3
379
  },
380
  {
381
+ "seconds": 1.0192943180009024,
382
+ "bytes": 2301578,
383
+ "pages": 28,
384
  "shards": 2
385
  },
386
  {
387
+ "seconds": 2.074865057998977,
388
+ "bytes": 9907182,
389
+ "pages": 39,
390
+ "shards": 3
391
  },
392
  {
393
+ "seconds": 1.498775358995772,
394
+ "bytes": 5765034,
395
  "pages": 28,
396
  "shards": 2
397
  },
398
  {
399
+ "seconds": 1.3661099450037,
400
+ "bytes": 8604864,
401
+ "pages": 41,
402
+ "shards": 3
403
+ },
404
+ {
405
+ "seconds": 1.0984054020009353,
406
+ "bytes": 6085035,
407
  "pages": 29,
408
  "shards": 2
409
  },
410
  {
411
+ "seconds": 1.4394773240055656,
412
+ "bytes": 12207803,
413
+ "pages": 42,
414
  "shards": 4
415
  },
416
  {
417
+ "seconds": 0.8468366760062054,
418
+ "bytes": 871307,
 
 
 
 
 
 
419
  "pages": 27,
420
+ "shards": 1
421
  },
422
  {
423
+ "seconds": 1.2028667299964582,
424
+ "bytes": 6129254,
425
  "pages": 38,
426
  "shards": 2
427
  },
428
  {
429
+ "seconds": 1.6012624290015083,
430
+ "bytes": 8291413,
431
+ "pages": 33,
432
+ "shards": 3
433
  },
434
  {
435
+ "seconds": 1.3447834969992982,
436
+ "bytes": 10108218,
437
+ "pages": 32,
438
  "shards": 3
439
  },
440
  {
441
+ "seconds": 1.1404380909953034,
442
+ "bytes": 7821336,
443
+ "pages": 39,
444
  "shards": 3
445
  },
446
  {
447
+ "seconds": 1.2159917039971333,
448
+ "bytes": 870322,
449
+ "pages": 29,
450
+ "shards": 1
451
  },
452
  {
453
+ "seconds": 1.047054365000804,
454
+ "bytes": 6014250,
455
+ "pages": 35,
456
+ "shards": 2
457
  },
458
  {
459
+ "seconds": 1.1608161260010093,
460
+ "bytes": 6519078,
461
+ "pages": 32,
462
  "shards": 3
463
  },
464
  {
465
+ "seconds": 1.5200698819971876,
466
+ "bytes": 6082599,
467
+ "pages": 34,
 
 
 
 
 
 
468
  "shards": 2
469
  },
470
  {
471
+ "seconds": 1.4744428199992399,
472
+ "bytes": 11695751,
473
+ "pages": 38,
474
+ "shards": 3
475
  },
476
  {
477
+ "seconds": 1.3584106329944916,
478
+ "bytes": 2483287,
479
+ "pages": 36,
480
+ "shards": 2
481
  },
482
  {
483
+ "seconds": 1.1772192700009327,
484
+ "bytes": 10757429,
485
  "pages": 34,
486
+ "shards": 3
487
  },
488
  {
489
+ "seconds": 2.1233109069944476,
490
+ "bytes": 2398655,
491
+ "pages": 29,
492
  "shards": 2
493
  },
494
  {
495
+ "seconds": 1.0890797859974555,
496
+ "bytes": 6225655,
497
+ "pages": 38,
498
  "shards": 2
499
  },
500
  {
501
+ "seconds": 3.324964125000406,
502
+ "bytes": 5816918,
503
+ "pages": 28,
504
+ "shards": 2
505
  },
506
  {
507
+ "seconds": 1.304179352002393,
508
+ "bytes": 11867177,
509
  "pages": 37,
510
  "shards": 4
511
  },
512
  {
513
+ "seconds": 1.3306422289970214,
514
+ "bytes": 6277468,
515
+ "pages": 38,
516
+ "shards": 2
517
+ },
518
+ {
519
+ "seconds": 0.9982111080025788,
520
+ "bytes": 877422,
521
+ "pages": 26,
522
  "shards": 1
523
  },
524
  {
525
+ "seconds": 1.6349642249988392,
526
+ "bytes": 7479267,
527
+ "pages": 33,
528
  "shards": 3
529
  },
530
  {
531
+ "seconds": 1.1783448380010668,
532
+ "bytes": 10185058,
533
+ "pages": 38,
534
+ "shards": 3
535
  },
536
  {
537
+ "seconds": 0.9609743260007235,
538
+ "bytes": 2529461,
539
  "pages": 33,
540
  "shards": 2
541
  },
542
  {
543
+ "seconds": 1.5562968439990073,
544
+ "bytes": 11835464,
545
+ "pages": 39,
546
+ "shards": 3
 
 
 
 
 
 
547
  },
548
  {
549
+ "seconds": 0.8645425069989869,
550
+ "bytes": 693988,
551
+ "pages": 20,
552
+ "shards": 1
553
  },
554
  {
555
+ "seconds": 1.3442884890027926,
556
+ "bytes": 12581350,
557
+ "pages": 41,
558
+ "shards": 4
559
  },
560
  {
561
+ "seconds": 1.4273276979947696,
562
+ "bytes": 6232998,
563
  "pages": 37,
 
 
 
 
 
 
 
 
 
 
 
 
564
  "shards": 3
565
  }
566
  ],
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:003d5c36c105e74cd2f29e7718b5727a4a7950ad08dd430bb87b3b0378f6d788
3
- size 721770879
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ba31b41d133fff78caf1b22d96d15106f826e2e03648c9dbff5ca60049e9ff2e
3
+ size 721763199