appvoid commited on
Commit
0cf64c3
·
verified ·
1 Parent(s): 59f226e

Qwen90M step 585000; 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 +235 -235
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 580000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "86da1ad7765ecdf63d4c21ed8d3083e166f543b8c584d4e1ad8585460b37a1c2",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "1330a58909395b69dd9e933634473173544669f01ca72451c097b4fe93ebb964",
12
  "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
13
  "performance.json": "b7cb16c0913422fe54a766da97356133a598f4da7e691939f51028a8bce75667",
14
- "shard_scheduler.json": "4c7cecc17496ee854ce7e9db16f9fc5fe111120cc8b45911284ec118a736e4bb",
15
- "coverage.json": "e8919debdd5781a63f54d50ff90e0325881c51f855c15fdd97780f35b3c257bc",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 585000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "2b447c14881264552ab420a8708a59aa948aecc2bafff1047401930b2ef52981",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "92d7ebabc11fd9cdaa8f8eeb61f310d2c4a21ca61ea41e252ac9c8c24070f6e8",
12
  "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
13
  "performance.json": "b7cb16c0913422fe54a766da97356133a598f4da7e691939f51028a8bce75667",
14
+ "shard_scheduler.json": "0c12b8bca40ac662ed2a947c032a8d65d4cbe1bb3c37e3f37268d66836f48587",
15
+ "coverage.json": "cf63695a4947eed2d72d78bf215c300aafb9cd742c85d0c9a809906a58a8ff57",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 580000,
3
  "stats": {
4
- "climbmix/blocks": 20880000,
5
- "climbmix/targets": 34903766814,
6
- "climbmix/input_tokens": 34905978789,
7
- "ultra_qa/blocks": 7830000,
8
- "ultra_qa/targets": 12773998197,
9
- "ultra_qa/input_tokens": 12774022841,
10
- "cortex/blocks": 7830000,
11
- "cortex/targets": 297744461,
12
- "cortex/input_tokens": 12446411636,
13
- "ultra_style/blocks": 7830000,
14
- "ultra_style/targets": 12384628359,
15
- "ultra_style/input_tokens": 12385277999,
16
- "openmath/blocks": 5220000,
17
- "openmath/targets": 8065835620,
18
- "openmath/input_tokens": 8068521894,
19
- "rewrite6/blocks": 2610000,
20
- "rewrite6/targets": 4278710471,
21
- "rewrite6/input_tokens": 4283401756
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 2859,
26
- "row": 2966,
27
  "start": {
28
- "file": 138,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 29,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
- "offset": 10240,
37
- "documents": 11713430
38
  },
39
  "rewrite6": {
40
- "page": 26,
41
- "row": 3974,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 106496
47
  }
48
  },
49
  "epoch": 32,
50
  "offset": 0,
51
- "documents": 6683398
52
  },
53
  "openmath": {
54
- "page": 1786,
55
- "row": 3533,
56
  "start": {
57
  "file": 16,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 329,
61
- "row": 56
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 2048,
66
- "documents": 7318989
67
  },
68
  "ultra_qa": {
69
- "page": 772,
70
- "row": 2970,
71
  "start": {
72
  "file": 6,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
- "row_group": 0,
76
- "row": 44355
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
- "documents": 3165082
82
  },
83
  "ultra_style": {
84
- "page": 1179,
85
- "row": 1648,
86
  "start": {
87
  "file": 7,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 0,
91
- "row": 34590
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
- "documents": 4830832
97
  },
98
  "cortex": {
99
- "page": 151857,
100
- "row": 6,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 19437702
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 585000,
3
  "stats": {
4
+ "climbmix/blocks": 21060000,
5
+ "climbmix/targets": 35204728526,
6
+ "climbmix/input_tokens": 35206959792,
7
+ "ultra_qa/blocks": 7897500,
8
+ "ultra_qa/targets": 12884016978,
9
+ "ultra_qa/input_tokens": 12884041832,
10
+ "cortex/blocks": 7897500,
11
+ "cortex/targets": 300307032,
12
+ "cortex/input_tokens": 12553655276,
13
+ "ultra_style/blocks": 7897500,
14
+ "ultra_style/targets": 12491928833,
15
+ "ultra_style/input_tokens": 12492584132,
16
+ "openmath/blocks": 5265000,
17
+ "openmath/targets": 8135369806,
18
+ "openmath/input_tokens": 8138079413,
19
+ "rewrite6/blocks": 2632500,
20
+ "rewrite6/targets": 4315618350,
21
+ "rewrite6/input_tokens": 4320349938
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 2884,
26
+ "row": 2144,
27
  "start": {
28
+ "file": 139,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 47,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
+ "offset": 2048,
37
+ "documents": 11815008
38
  },
39
  "rewrite6": {
40
+ "page": 41,
41
+ "row": 2,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 167936
47
  }
48
  },
49
  "epoch": 32,
50
  "offset": 0,
51
+ "documents": 6740866
52
  },
53
  "openmath": {
54
+ "page": 1802,
55
+ "row": 1223,
56
  "start": {
57
  "file": 16,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 394,
61
+ "row": 592
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 2048,
66
+ "documents": 7382215
67
  },
68
  "ultra_qa": {
69
+ "page": 779,
70
+ "row": 1531,
71
  "start": {
72
  "file": 6,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
+ "row_group": 1,
76
+ "row": 8520
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
+ "documents": 3192315
82
  },
83
  "ultra_style": {
84
+ "page": 1189,
85
+ "row": 2100,
86
  "start": {
87
  "file": 7,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 0,
91
+ "row": 75550
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
+ "documents": 4872244
97
  },
98
  "cortex": {
99
+ "page": 153166,
100
+ "row": 102,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 19605350
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 580000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "86da1ad7765ecdf63d4c21ed8d3083e166f543b8c584d4e1ad8585460b37a1c2", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "1330a58909395b69dd9e933634473173544669f01ca72451c097b4fe93ebb964", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "b7cb16c0913422fe54a766da97356133a598f4da7e691939f51028a8bce75667", "shard_scheduler.json": "4c7cecc17496ee854ce7e9db16f9fc5fe111120cc8b45911284ec118a736e4bb", "coverage.json": "e8919debdd5781a63f54d50ff90e0325881c51f855c15fdd97780f35b3c257bc", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000580000"}
 
1
+ {"step": 585000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "2b447c14881264552ab420a8708a59aa948aecc2bafff1047401930b2ef52981", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "92d7ebabc11fd9cdaa8f8eeb61f310d2c4a21ca61ea41e252ac9c8c24070f6e8", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "b7cb16c0913422fe54a766da97356133a598f4da7e691939f51028a8bce75667", "shard_scheduler.json": "0c12b8bca40ac662ed2a947c032a8d65d4cbe1bb3c37e3f37268d66836f48587", "coverage.json": "cf63695a4947eed2d72d78bf215c300aafb9cd742c85d0c9a809906a58a8ff57", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000585000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:86da1ad7765ecdf63d4c21ed8d3083e166f543b8c584d4e1ad8585460b37a1c2
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2b447c14881264552ab420a8708a59aa948aecc2bafff1047401930b2ef52981
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": 76844.82269293365,
26
- "required_pages_per_second_upper_estimate": 0.0053163031384513715
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": 9605.602836616707,
40
- "required_pages_per_second_upper_estimate": 0.004021665319894588
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": 19211.205673233413,
54
- "required_pages_per_second_upper_estimate": 0.004001187508171285
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": 28816.808509850118,
68
- "required_pages_per_second_upper_estimate": 0.0017275236694466443
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": 28816.808509850118,
82
- "required_pages_per_second_upper_estimate": 0.0022403801098957493
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": 28816.808509850118,
96
- "required_pages_per_second_upper_estimate": 0.3696310784860394
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,409 +158,409 @@
158
  "prefetch_batches": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 704566,
161
- "producer_batch_seconds": 0.0661616753671126,
162
- "producer_p95_seconds": 0.24438537600144628,
163
- "gpu_update_seconds": 0.9594400431452845,
164
- "producer_batches_per_second": 15.11449029141539,
165
- "gpu_batches_per_second": 1.0422746133481668,
166
- "producer_headroom": 14.501447217314565,
167
  "sustainable": true,
168
- "buffer_seconds": 60.444722718152924,
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": 406720.83803339687,
174
- "upload_bytes_per_second": 5366265.038987921,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 87.2728238654273,
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": 1.2003755249988899,
184
- "bytes": 10273667,
185
- "pages": 44,
186
  "shards": 3
187
  },
188
  {
189
- "seconds": 1.0975800400010485,
190
- "bytes": 2683103,
191
- "pages": 36,
192
  "shards": 2
193
  },
194
  {
195
- "seconds": 1.1853528239989828,
196
- "bytes": 5804023,
197
  "pages": 28,
198
- "shards": 2
199
  },
200
  {
201
- "seconds": 1.138313076000486,
202
- "bytes": 10599364,
203
- "pages": 39,
204
  "shards": 3
205
  },
206
  {
207
- "seconds": 0.9668768979991,
208
- "bytes": 2353349,
209
- "pages": 29,
210
  "shards": 2
211
  },
212
  {
213
- "seconds": 1.4940221720007685,
214
- "bytes": 12237361,
215
- "pages": 41,
216
- "shards": 3
217
  },
218
  {
219
- "seconds": 0.8858112369998707,
220
- "bytes": 2416252,
221
- "pages": 28,
222
  "shards": 3
223
  },
224
  {
225
- "seconds": 1.2992984630000137,
226
- "bytes": 10017513,
227
- "pages": 40,
228
- "shards": 3
229
  },
230
  {
231
- "seconds": 4.1799396490005165,
232
- "bytes": 5932760,
233
- "pages": 32,
234
- "shards": 2
235
  },
236
  {
237
- "seconds": 0.8289485320001404,
238
- "bytes": 2513176,
239
- "pages": 33,
240
  "shards": 2
241
  },
242
  {
243
- "seconds": 0.9106573579992983,
244
- "bytes": 5965787,
245
  "pages": 36,
246
  "shards": 2
247
  },
248
  {
249
- "seconds": 1.575513838999541,
250
- "bytes": 18103820,
251
- "pages": 39,
252
- "shards": 5
253
  },
254
  {
255
- "seconds": 0.8145781550010724,
256
- "bytes": 930729,
257
- "pages": 26,
258
  "shards": 1
259
  },
260
  {
261
- "seconds": 1.3511633730013273,
262
- "bytes": 6255780,
263
  "pages": 36,
264
  "shards": 2
265
  },
266
  {
267
- "seconds": 1.031140414001129,
268
- "bytes": 2762787,
269
- "pages": 38,
270
- "shards": 2
271
  },
272
  {
273
- "seconds": 1.0752210870014096,
274
- "bytes": 5610482,
275
- "pages": 19,
276
- "shards": 2
277
  },
278
  {
279
- "seconds": 1.425927022999531,
280
- "bytes": 10079783,
281
- "pages": 46,
282
  "shards": 3
283
  },
284
  {
285
- "seconds": 1.1401308700005757,
286
- "bytes": 8309975,
287
- "pages": 34,
288
- "shards": 3
289
  },
290
  {
291
- "seconds": 1.2373918849989423,
292
- "bytes": 6097119,
293
- "pages": 33,
294
  "shards": 2
295
  },
296
  {
297
- "seconds": 1.1592294109977956,
298
- "bytes": 6491968,
299
- "pages": 35,
300
- "shards": 3
301
  },
302
  {
303
- "seconds": 1.6240387940015353,
304
- "bytes": 5730323,
305
- "pages": 27,
306
  "shards": 2
307
  },
308
  {
309
- "seconds": 1.3926623539991851,
310
- "bytes": 6408754,
311
  "pages": 36,
312
  "shards": 2
313
  },
314
  {
315
- "seconds": 1.260699045000365,
316
- "bytes": 2670941,
317
- "pages": 37,
318
  "shards": 2
319
  },
320
  {
321
- "seconds": 1.4827766799971869,
322
- "bytes": 15946512,
323
- "pages": 34,
324
- "shards": 4
325
  },
326
  {
327
- "seconds": 1.33145204000175,
328
- "bytes": 7515329,
329
- "pages": 38,
330
  "shards": 3
331
  },
332
  {
333
- "seconds": 0.9712299829989206,
334
- "bytes": 867774,
335
  "pages": 27,
336
  "shards": 1
337
  },
338
  {
339
- "seconds": 1.1689961140000378,
340
- "bytes": 10341916,
341
- "pages": 38,
342
  "shards": 3
343
  },
344
  {
345
- "seconds": 4.051943458998721,
346
- "bytes": 2586599,
347
  "pages": 37,
348
- "shards": 2
349
- },
350
- {
351
- "seconds": 0.9690040009991208,
352
- "bytes": 5902816,
353
- "pages": 29,
354
- "shards": 2
355
- },
356
- {
357
- "seconds": 0.9907326490028936,
358
- "bytes": 186236,
359
- "pages": 7,
360
- "shards": 1
361
- },
362
- {
363
- "seconds": 1.1790409999994154,
364
- "bytes": 12018438,
365
- "pages": 30,
366
  "shards": 3
367
  },
368
  {
369
- "seconds": 1.0833680409996305,
370
- "bytes": 2366981,
371
- "pages": 28,
372
- "shards": 2
373
- },
374
- {
375
- "seconds": 1.2835563219996402,
376
- "bytes": 10401632,
377
- "pages": 38,
378
  "shards": 3
379
  },
380
  {
381
- "seconds": 0.9978380930006097,
382
- "bytes": 912994,
383
- "pages": 28,
384
  "shards": 1
385
  },
386
  {
387
- "seconds": 1.1661826020026638,
388
- "bytes": 7439707,
389
- "pages": 37,
390
  "shards": 3
391
  },
392
  {
393
- "seconds": 1.0849376400001347,
394
- "bytes": 10733196,
395
- "pages": 39,
396
  "shards": 3
397
  },
398
  {
399
- "seconds": 1.2397475959987787,
400
- "bytes": 8159623,
401
  "pages": 29,
402
  "shards": 3
403
  },
404
  {
405
- "seconds": 1.2689961300020514,
406
- "bytes": 6434173,
407
  "pages": 36,
408
  "shards": 2
409
  },
410
  {
411
- "seconds": 1.3497664780006744,
412
- "bytes": 7459096,
413
- "pages": 39,
414
- "shards": 4
415
  },
416
  {
417
- "seconds": 1.0805805290001445,
418
- "bytes": 1076352,
419
- "pages": 29,
420
- "shards": 1
421
  },
422
  {
423
- "seconds": 1.1792509940023592,
424
- "bytes": 9839539,
425
  "pages": 37,
426
- "shards": 3
427
  },
428
  {
429
- "seconds": 1.0725541200008593,
430
- "bytes": 2601868,
431
- "pages": 38,
432
- "shards": 2
433
  },
434
  {
435
- "seconds": 1.330655336001655,
436
- "bytes": 11531295,
437
- "pages": 28,
438
- "shards": 3
439
  },
440
  {
441
- "seconds": 1.5732356919979793,
442
- "bytes": 10605987,
443
- "pages": 41,
444
  "shards": 3
445
  },
446
  {
447
- "seconds": 0.9885955910031043,
448
- "bytes": 2380823,
 
 
 
 
 
 
449
  "pages": 28,
450
  "shards": 2
451
  },
452
  {
453
- "seconds": 0.9730731119998381,
454
- "bytes": 6090581,
455
- "pages": 37,
456
  "shards": 2
457
  },
458
  {
459
- "seconds": 1.1632558510027593,
460
- "bytes": 7851564,
461
- "pages": 38,
462
  "shards": 3
463
  },
464
  {
465
- "seconds": 1.0146159530013392,
466
- "bytes": 5636542,
467
- "pages": 31,
468
- "shards": 2
469
  },
470
  {
471
- "seconds": 1.0724632399978873,
472
- "bytes": 12061908,
473
  "pages": 38,
 
 
 
 
 
 
 
 
 
 
 
 
474
  "shards": 3
475
  },
476
  {
477
- "seconds": 0.9588815639981476,
478
- "bytes": 2343317,
479
- "pages": 28,
480
  "shards": 2
481
  },
482
  {
483
- "seconds": 1.1246145429977332,
484
- "bytes": 6175304,
485
- "pages": 36,
486
  "shards": 2
487
  },
488
  {
489
- "seconds": 1.2072375589996227,
490
- "bytes": 11138063,
491
- "pages": 36,
492
  "shards": 4
493
  },
494
  {
495
- "seconds": 0.9064097570008016,
496
- "bytes": 1171302,
497
  "pages": 35,
498
- "shards": 1
499
  },
500
  {
501
- "seconds": 0.9584431869989203,
502
- "bytes": 6298065,
503
- "pages": 36,
504
  "shards": 2
505
  },
506
  {
507
- "seconds": 0.972795835998113,
508
- "bytes": 8254778,
509
- "pages": 31,
510
- "shards": 3
511
  },
512
  {
513
- "seconds": 1.3151332330016885,
514
- "bytes": 10337557,
515
- "pages": 35,
 
 
 
 
 
 
516
  "shards": 3
517
  },
518
  {
519
- "seconds": 1.029109121998772,
520
- "bytes": 7643162,
521
- "pages": 40,
522
  "shards": 3
523
  },
524
  {
525
- "seconds": 0.8460896450014843,
526
- "bytes": 884554,
527
- "pages": 26,
528
- "shards": 1
529
  },
530
  {
531
- "seconds": 1.0631770070031052,
532
- "bytes": 6064387,
533
- "pages": 32,
534
  "shards": 2
535
  },
536
  {
537
- "seconds": 1.2800072050013114,
538
- "bytes": 12000660,
539
- "pages": 43,
540
  "shards": 4
541
  },
542
  {
543
- "seconds": 1.2026952199994412,
544
- "bytes": 6686271,
545
- "pages": 28,
546
  "shards": 2
547
  },
548
  {
549
- "seconds": 1.3011887590000697,
550
- "bytes": 6177443,
551
- "pages": 36,
552
  "shards": 2
553
  },
554
  {
555
- "seconds": 1.0062806160021864,
556
- "bytes": 2394266,
557
- "pages": 29,
558
- "shards": 2
559
  },
560
  {
561
- "seconds": 1.1657910530011577,
562
- "bytes": 10819661,
563
- "pages": 44,
564
  "shards": 3
565
  }
566
  ],
 
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": 76453.60105814453,
26
+ "required_pages_per_second_upper_estimate": 0.005289237517997402
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": 9556.700132268066,
40
+ "required_pages_per_second_upper_estimate": 0.0040011907787884035
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": 19113.400264536132,
54
+ "required_pages_per_second_upper_estimate": 0.003980817220841719
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": 28670.1003968042,
68
+ "required_pages_per_second_upper_estimate": 0.0017187287420798595
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": 28670.1003968042,
82
+ "required_pages_per_second_upper_estimate": 0.0022289741994072207
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": 28670.1003968042,
96
+ "required_pages_per_second_upper_estimate": 0.36774926433478533
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": 0.12654482389845612,
162
+ "producer_p95_seconds": 0.2847857099986868,
163
+ "gpu_update_seconds": 0.9643496052452565,
164
+ "producer_batches_per_second": 7.902338232359738,
165
+ "gpu_batches_per_second": 1.0369683303242259,
166
+ "producer_headroom": 7.620616754890611,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.75402513045116,
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": 354955.35036680783,
174
+ "upload_bytes_per_second": 5937659.932236937,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 76.16515525105645,
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": 1.1792509940023592,
184
+ "bytes": 9839539,
185
+ "pages": 37,
186
  "shards": 3
187
  },
188
  {
189
+ "seconds": 1.0725541200008593,
190
+ "bytes": 2601868,
191
+ "pages": 38,
192
  "shards": 2
193
  },
194
  {
195
+ "seconds": 1.330655336001655,
196
+ "bytes": 11531295,
197
  "pages": 28,
198
+ "shards": 3
199
  },
200
  {
201
+ "seconds": 1.5732356919979793,
202
+ "bytes": 10605987,
203
+ "pages": 41,
204
  "shards": 3
205
  },
206
  {
207
+ "seconds": 0.9885955910031043,
208
+ "bytes": 2380823,
209
+ "pages": 28,
210
  "shards": 2
211
  },
212
  {
213
+ "seconds": 0.9730731119998381,
214
+ "bytes": 6090581,
215
+ "pages": 37,
216
+ "shards": 2
217
  },
218
  {
219
+ "seconds": 1.1632558510027593,
220
+ "bytes": 7851564,
221
+ "pages": 38,
222
  "shards": 3
223
  },
224
  {
225
+ "seconds": 1.0146159530013392,
226
+ "bytes": 5636542,
227
+ "pages": 31,
228
+ "shards": 2
229
  },
230
  {
231
+ "seconds": 1.0724632399978873,
232
+ "bytes": 12061908,
233
+ "pages": 38,
234
+ "shards": 3
235
  },
236
  {
237
+ "seconds": 0.9588815639981476,
238
+ "bytes": 2343317,
239
+ "pages": 28,
240
  "shards": 2
241
  },
242
  {
243
+ "seconds": 1.1246145429977332,
244
+ "bytes": 6175304,
245
  "pages": 36,
246
  "shards": 2
247
  },
248
  {
249
+ "seconds": 1.2072375589996227,
250
+ "bytes": 11138063,
251
+ "pages": 36,
252
+ "shards": 4
253
  },
254
  {
255
+ "seconds": 0.9064097570008016,
256
+ "bytes": 1171302,
257
+ "pages": 35,
258
  "shards": 1
259
  },
260
  {
261
+ "seconds": 0.9584431869989203,
262
+ "bytes": 6298065,
263
  "pages": 36,
264
  "shards": 2
265
  },
266
  {
267
+ "seconds": 0.972795835998113,
268
+ "bytes": 8254778,
269
+ "pages": 31,
270
+ "shards": 3
271
  },
272
  {
273
+ "seconds": 1.3151332330016885,
274
+ "bytes": 10337557,
275
+ "pages": 35,
276
+ "shards": 3
277
  },
278
  {
279
+ "seconds": 1.029109121998772,
280
+ "bytes": 7643162,
281
+ "pages": 40,
282
  "shards": 3
283
  },
284
  {
285
+ "seconds": 0.8460896450014843,
286
+ "bytes": 884554,
287
+ "pages": 26,
288
+ "shards": 1
289
  },
290
  {
291
+ "seconds": 1.0631770070031052,
292
+ "bytes": 6064387,
293
+ "pages": 32,
294
  "shards": 2
295
  },
296
  {
297
+ "seconds": 1.2800072050013114,
298
+ "bytes": 12000660,
299
+ "pages": 43,
300
+ "shards": 4
301
  },
302
  {
303
+ "seconds": 1.2026952199994412,
304
+ "bytes": 6686271,
305
+ "pages": 28,
306
  "shards": 2
307
  },
308
  {
309
+ "seconds": 1.3011887590000697,
310
+ "bytes": 6177443,
311
  "pages": 36,
312
  "shards": 2
313
  },
314
  {
315
+ "seconds": 1.0062806160021864,
316
+ "bytes": 2394266,
317
+ "pages": 29,
318
  "shards": 2
319
  },
320
  {
321
+ "seconds": 1.1657910530011577,
322
+ "bytes": 10819661,
323
+ "pages": 44,
324
+ "shards": 3
325
  },
326
  {
327
+ "seconds": 1.1555115019982622,
328
+ "bytes": 7463395,
329
+ "pages": 30,
330
  "shards": 3
331
  },
332
  {
333
+ "seconds": 0.8593025919981301,
334
+ "bytes": 778515,
335
  "pages": 27,
336
  "shards": 1
337
  },
338
  {
339
+ "seconds": 1.26428677900185,
340
+ "bytes": 11594708,
341
+ "pages": 31,
342
  "shards": 3
343
  },
344
  {
345
+ "seconds": 1.083468554999854,
346
+ "bytes": 6825645,
347
  "pages": 37,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
348
  "shards": 3
349
  },
350
  {
351
+ "seconds": 1.0758820400005789,
352
+ "bytes": 6474447,
353
+ "pages": 37,
 
 
 
 
 
 
354
  "shards": 3
355
  },
356
  {
357
+ "seconds": 0.8884251060007955,
358
+ "bytes": 1122963,
359
+ "pages": 34,
360
  "shards": 1
361
  },
362
  {
363
+ "seconds": 1.0163737470029446,
364
+ "bytes": 7245209,
365
+ "pages": 27,
366
  "shards": 3
367
  },
368
  {
369
+ "seconds": 1.4228277920010441,
370
+ "bytes": 10524529,
371
+ "pages": 42,
372
  "shards": 3
373
  },
374
  {
375
+ "seconds": 1.1392915559990797,
376
+ "bytes": 8097285,
377
  "pages": 29,
378
  "shards": 3
379
  },
380
  {
381
+ "seconds": 1.2682354990029125,
382
+ "bytes": 6131054,
383
  "pages": 36,
384
  "shards": 2
385
  },
386
  {
387
+ "seconds": 1.2669715150004777,
388
+ "bytes": 6065248,
389
+ "pages": 33,
390
+ "shards": 2
391
  },
392
  {
393
+ "seconds": 1.0634998090026784,
394
+ "bytes": 7147043,
395
+ "pages": 34,
396
+ "shards": 3
397
  },
398
  {
399
+ "seconds": 3.4036351569993712,
400
+ "bytes": 6159372,
401
  "pages": 37,
402
+ "shards": 2
403
  },
404
  {
405
+ "seconds": 0.9252640429986059,
406
+ "bytes": 1101958,
407
+ "pages": 31,
408
+ "shards": 1
409
  },
410
  {
411
+ "seconds": 1.268539120999776,
412
+ "bytes": 13168093,
413
+ "pages": 33,
414
+ "shards": 4
415
  },
416
  {
417
+ "seconds": 1.1610033780016238,
418
+ "bytes": 10490094,
419
+ "pages": 39,
420
  "shards": 3
421
  },
422
  {
423
+ "seconds": 1.0719913379980426,
424
+ "bytes": 2751265,
425
+ "pages": 37,
426
+ "shards": 2
427
+ },
428
+ {
429
+ "seconds": 1.124440429997776,
430
+ "bytes": 5889635,
431
  "pages": 28,
432
  "shards": 2
433
  },
434
  {
435
+ "seconds": 1.0922325869978522,
436
+ "bytes": 6027814,
437
+ "pages": 36,
438
  "shards": 2
439
  },
440
  {
441
+ "seconds": 1.0737022979992616,
442
+ "bytes": 6591396,
443
+ "pages": 31,
444
  "shards": 3
445
  },
446
  {
447
+ "seconds": 1.1665754960013146,
448
+ "bytes": 11803743,
449
+ "pages": 35,
450
+ "shards": 3
451
  },
452
  {
453
+ "seconds": 0.962029407000955,
454
+ "bytes": 2665696,
455
  "pages": 38,
456
+ "shards": 2
457
+ },
458
+ {
459
+ "seconds": 1.1741747699998086,
460
+ "bytes": 5919198,
461
+ "pages": 26,
462
+ "shards": 2
463
+ },
464
+ {
465
+ "seconds": 1.178343194998888,
466
+ "bytes": 10619543,
467
+ "pages": 42,
468
  "shards": 3
469
  },
470
  {
471
+ "seconds": 1.036487166002189,
472
+ "bytes": 2690894,
473
+ "pages": 37,
474
  "shards": 2
475
  },
476
  {
477
+ "seconds": 1.1432905639994715,
478
+ "bytes": 5745742,
479
+ "pages": 28,
480
  "shards": 2
481
  },
482
  {
483
+ "seconds": 1.232796011998289,
484
+ "bytes": 13189628,
485
+ "pages": 37,
486
  "shards": 4
487
  },
488
  {
489
+ "seconds": 0.943955818998802,
490
+ "bytes": 5281939,
491
  "pages": 35,
492
+ "shards": 2
493
  },
494
  {
495
+ "seconds": 1.0029552209998656,
496
+ "bytes": 6176365,
497
+ "pages": 37,
498
  "shards": 2
499
  },
500
  {
501
+ "seconds": 1.1109868420026032,
502
+ "bytes": 2592499,
503
+ "pages": 35,
504
+ "shards": 2
505
  },
506
  {
507
+ "seconds": 1.24596934100191,
508
+ "bytes": 5407802,
509
+ "pages": 21,
510
+ "shards": 2
511
+ },
512
+ {
513
+ "seconds": 1.361496919002093,
514
+ "bytes": 10822221,
515
+ "pages": 38,
516
  "shards": 3
517
  },
518
  {
519
+ "seconds": 1.3625388020009268,
520
+ "bytes": 8506674,
521
+ "pages": 41,
522
  "shards": 3
523
  },
524
  {
525
+ "seconds": 1.141177229997993,
526
+ "bytes": 6001543,
527
+ "pages": 29,
528
+ "shards": 2
529
  },
530
  {
531
+ "seconds": 0.9294679570011795,
532
+ "bytes": 2588442,
533
+ "pages": 35,
534
  "shards": 2
535
  },
536
  {
537
+ "seconds": 1.1605264330028149,
538
+ "bytes": 9613122,
539
+ "pages": 38,
540
  "shards": 4
541
  },
542
  {
543
+ "seconds": 1.0401736989988422,
544
+ "bytes": 6081990,
545
+ "pages": 37,
546
  "shards": 2
547
  },
548
  {
549
+ "seconds": 0.9755379390007874,
550
+ "bytes": 2340646,
551
+ "pages": 28,
552
  "shards": 2
553
  },
554
  {
555
+ "seconds": 1.4557110550013022,
556
+ "bytes": 16684165,
557
+ "pages": 36,
558
+ "shards": 4
559
  },
560
  {
561
+ "seconds": 1.2760817240014148,
562
+ "bytes": 7386323,
563
+ "pages": 39,
564
  "shards": 3
565
  }
566
  ],
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:1330a58909395b69dd9e933634473173544669f01ca72451c097b4fe93ebb964
3
- size 721773055
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:92d7ebabc11fd9cdaa8f8eeb61f310d2c4a21ca61ea41e252ac9c8c24070f6e8
3
+ size 721766911