appvoid commited on
Commit
92a2631
·
verified ·
1 Parent(s): dbd97f4

Qwen90M step 225000; 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 +245 -245
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 220000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "498fb63be761dda04e4ae3ce457299bd3d432b3bdea8411d8a28b5ef0b7d194b",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "796925d7192e7f222c8c88ea2d27a48542f3cbdadf89ca004b7822e1a82d3d36",
12
  "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf",
13
  "performance.json": "7f72f0a0e868a4baec06f238f922d0c5d7e397a51386c11ac32acc776d4d3522",
14
- "shard_scheduler.json": "7a64512af661fdaf88001f0c4d3af7a439cbeb5dc84c5e6b0d6fd966ff4d9a43",
15
- "coverage.json": "2672288392f9e38924ab0573a5d641849c393b2a73175abb1c4c6167c4492336",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 225000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "8026e56ede4b6be87fe5a0a8df3ee3a51ac084bd9dae00c82e2b3e41fd9a844c",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "c3941f290352b013b70ffc29d16e39df4aacbd1a763c6261ae4bbca696edc601",
12
  "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf",
13
  "performance.json": "7f72f0a0e868a4baec06f238f922d0c5d7e397a51386c11ac32acc776d4d3522",
14
+ "shard_scheduler.json": "2e841a817820d99fd0c9bb30efa85bd8f3246df3a2689387bff17be63f08f2a2",
15
+ "coverage.json": "2e342a84546df2ccad31948e71307f1c47ca262201e1551430aefb1fb67fb347",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 220000,
3
  "stats": {
4
- "climbmix/blocks": 7920000,
5
- "climbmix/targets": 13238818073,
6
- "climbmix/input_tokens": 13239656864,
7
- "ultra_qa/blocks": 2970000,
8
- "ultra_qa/targets": 4845561521,
9
- "ultra_qa/input_tokens": 4845570884,
10
- "cortex/blocks": 2970000,
11
- "cortex/targets": 113003359,
12
- "cortex/input_tokens": 4720874700,
13
- "ultra_style/blocks": 2970000,
14
- "ultra_style/targets": 4697944483,
15
- "ultra_style/input_tokens": 4698190714,
16
- "openmath/blocks": 1980000,
17
- "openmath/targets": 3059241505,
18
- "openmath/input_tokens": 3060261058,
19
- "rewrite6/blocks": 990000,
20
- "rewrite6/targets": 1622934859,
21
- "rewrite6/input_tokens": 1624714314
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 1084,
26
- "row": 2938,
27
  "start": {
28
- "file": 52,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 37,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
- "documents": 4443002
38
  },
39
  "rewrite6": {
40
- "page": 17,
41
- "row": 543,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 69632
47
  }
48
  },
49
  "epoch": 12,
50
  "offset": 0,
51
- "documents": 2535023
52
  },
53
  "openmath": {
54
- "page": 677,
55
- "row": 4056,
56
  "start": {
57
  "file": 6,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 153,
61
- "row": 92
62
  }
63
  },
64
  "epoch": 0,
65
- "offset": 0,
66
- "documents": 2777048
67
  },
68
  "ultra_qa": {
69
- "page": 293,
70
- "row": 482,
71
  "start": {
72
  "file": 2,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 2,
76
- "row": 31730
77
  }
78
  },
79
  "epoch": 0,
80
- "offset": 4096,
81
- "documents": 1200610
82
  },
83
  "ultra_style": {
84
- "page": 447,
85
- "row": 1254,
86
  "start": {
87
  "file": 2,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 5,
91
- "row": 36808
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
- "documents": 1832166
97
  },
98
  "cortex": {
99
- "page": 57598,
100
- "row": 126,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 7372670
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 225000,
3
  "stats": {
4
+ "climbmix/blocks": 8100000,
5
+ "climbmix/targets": 13540352549,
6
+ "climbmix/input_tokens": 13541210224,
7
+ "ultra_qa/blocks": 3037500,
8
+ "ultra_qa/targets": 4955605158,
9
+ "ultra_qa/input_tokens": 4955614747,
10
+ "cortex/blocks": 3037500,
11
+ "cortex/targets": 115559566,
12
+ "cortex/input_tokens": 4828157787,
13
+ "ultra_style/blocks": 3037500,
14
+ "ultra_style/targets": 4804534476,
15
+ "ultra_style/input_tokens": 4804786332,
16
+ "openmath/blocks": 2025000,
17
+ "openmath/targets": 3128842388,
18
+ "openmath/input_tokens": 3129885291,
19
+ "rewrite6/blocks": 1012500,
20
+ "rewrite6/targets": 1659779143,
21
+ "rewrite6/input_tokens": 1661599123
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 1109,
26
+ "row": 563,
27
  "start": {
28
+ "file": 53,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 55,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
+ "documents": 4543027
38
  },
39
  "rewrite6": {
40
+ "page": 31,
41
+ "row": 995,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 126976
47
  }
48
  },
49
  "epoch": 12,
50
  "offset": 0,
51
+ "documents": 2592819
52
  },
53
  "openmath": {
54
+ "page": 693,
55
+ "row": 1819,
56
  "start": {
57
  "file": 6,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 218,
61
+ "row": 628
62
  }
63
  },
64
  "epoch": 0,
65
+ "offset": 4096,
66
+ "documents": 2840347
67
  },
68
  "ultra_qa": {
69
+ "page": 299,
70
+ "row": 3200,
71
  "start": {
72
  "file": 2,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 2,
76
+ "row": 56306
77
  }
78
  },
79
  "epoch": 0,
80
+ "offset": 2048,
81
+ "documents": 1227904
82
  },
83
  "ultra_style": {
84
+ "page": 457,
85
+ "row": 2213,
86
  "start": {
87
  "file": 2,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 5,
91
+ "row": 77768
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
+ "documents": 1874085
97
  },
98
  "cortex": {
99
+ "page": 58907,
100
+ "row": 79,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 7540175
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 220000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "498fb63be761dda04e4ae3ce457299bd3d432b3bdea8411d8a28b5ef0b7d194b", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "796925d7192e7f222c8c88ea2d27a48542f3cbdadf89ca004b7822e1a82d3d36", "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf", "performance.json": "7f72f0a0e868a4baec06f238f922d0c5d7e397a51386c11ac32acc776d4d3522", "shard_scheduler.json": "7a64512af661fdaf88001f0c4d3af7a439cbeb5dc84c5e6b0d6fd966ff4d9a43", "coverage.json": "2672288392f9e38924ab0573a5d641849c393b2a73175abb1c4c6167c4492336", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000220000"}
 
1
+ {"step": 225000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "8026e56ede4b6be87fe5a0a8df3ee3a51ac084bd9dae00c82e2b3e41fd9a844c", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "c3941f290352b013b70ffc29d16e39df4aacbd1a763c6261ae4bbca696edc601", "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf", "performance.json": "7f72f0a0e868a4baec06f238f922d0c5d7e397a51386c11ac32acc776d4d3522", "shard_scheduler.json": "2e841a817820d99fd0c9bb30efa85bd8f3246df3a2689387bff17be63f08f2a2", "coverage.json": "2e342a84546df2ccad31948e71307f1c47ca262201e1551430aefb1fb67fb347", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000225000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:498fb63be761dda04e4ae3ce457299bd3d432b3bdea8411d8a28b5ef0b7d194b
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8026e56ede4b6be87fe5a0a8df3ee3a51ac084bd9dae00c82e2b3e41fd9a844c
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": 74218.2968781337,
26
- "required_pages_per_second_upper_estimate": 0.006433053868842927
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": 9277.287109766712,
40
- "required_pages_per_second_upper_estimate": 0.0037008485358890667
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": 18554.574219533424,
54
- "required_pages_per_second_upper_estimate": 0.0042922741963358395
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": 27831.861329300136,
68
- "required_pages_per_second_upper_estimate": 0.0016948143324631963
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": 27831.861329300136,
82
- "required_pages_per_second_upper_estimate": 0.002142573728652689
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": 27831.861329300136,
96
- "required_pages_per_second_upper_estimate": 0.34557857046202534
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": 0.3495573840859265,
162
- "producer_p95_seconds": 0.33007302099940716,
163
- "gpu_update_seconds": 0.993393854362641,
164
- "producer_batches_per_second": 2.8607606233664478,
165
- "gpu_batches_per_second": 1.0066500770146172,
166
- "producer_headroom": 2.8418620220548676,
167
  "sustainable": true,
168
- "buffer_seconds": 60.5970251161211,
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": 384535.0366229114,
174
- "upload_bytes_per_second": 2874108.1431070296,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 82.51226734176166,
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": 1.1491692820018216,
184
- "bytes": 11269048,
185
- "pages": 45,
186
- "shards": 3
187
- },
188
- {
189
- "seconds": 1.0205925500013109,
190
- "bytes": 2657105,
191
  "pages": 36,
192
  "shards": 2
193
  },
194
  {
195
- "seconds": 1.0683662000010372,
196
- "bytes": 6645182,
197
- "pages": 35,
198
- "shards": 2
199
- },
200
- {
201
- "seconds": 1.0938033080019522,
202
- "bytes": 6719573,
203
- "pages": 31,
204
- "shards": 2
205
  },
206
  {
207
- "seconds": 1.3891484439991473,
208
- "bytes": 11561860,
209
- "pages": 34,
210
  "shards": 4
211
  },
212
  {
213
- "seconds": 1.268800834997819,
214
- "bytes": 6327049,
215
  "pages": 36,
216
  "shards": 2
217
  },
218
  {
219
- "seconds": 1.941673843000899,
220
- "bytes": 1008328,
221
- "pages": 30,
222
- "shards": 1
223
  },
224
  {
225
- "seconds": 1.119837071000802,
226
- "bytes": 7224601,
227
  "pages": 32,
228
  "shards": 3
229
  },
230
  {
231
- "seconds": 1.2717098209977848,
232
- "bytes": 11010766,
233
- "pages": 39,
234
- "shards": 3
235
  },
236
  {
237
- "seconds": 1.1728442359999462,
238
- "bytes": 8097014,
239
- "pages": 29,
240
  "shards": 3
241
  },
242
  {
243
- "seconds": 1.3505811920003907,
244
- "bytes": 5904996,
245
- "pages": 27,
246
  "shards": 2
247
  },
248
  {
249
- "seconds": 0.9558858569980657,
250
- "bytes": 1199021,
251
  "pages": 33,
252
  "shards": 1
253
  },
254
  {
255
- "seconds": 1.896848256001249,
256
- "bytes": 11337881,
257
- "pages": 42,
258
- "shards": 4
259
  },
260
  {
261
- "seconds": 1.1246725650016742,
262
- "bytes": 878351,
263
- "pages": 26,
264
- "shards": 1
265
  },
266
  {
267
- "seconds": 1.1035908939993533,
268
- "bytes": 6360663,
269
- "pages": 35,
270
  "shards": 2
271
  },
272
  {
273
- "seconds": 0.9343603500019526,
274
- "bytes": 2109863,
275
- "pages": 20,
276
  "shards": 3
277
  },
278
  {
279
- "seconds": 1.3331431920014438,
280
- "bytes": 16665888,
281
- "pages": 43,
282
- "shards": 4
283
  },
284
  {
285
- "seconds": 1.675639938999666,
286
- "bytes": 7647558,
287
- "pages": 37,
288
  "shards": 3
289
  },
290
  {
291
- "seconds": 0.968150565000542,
292
- "bytes": 1141925,
293
- "pages": 34,
294
- "shards": 1
295
- },
296
- {
297
- "seconds": 1.0775915699996403,
298
- "bytes": 5849379,
299
- "pages": 27,
300
- "shards": 2
301
  },
302
  {
303
- "seconds": 1.3228814079993754,
304
- "bytes": 11625452,
305
  "pages": 35,
306
- "shards": 4
307
  },
308
  {
309
- "seconds": 1.1623286959984398,
310
- "bytes": 7130450,
311
- "pages": 38,
312
  "shards": 2
313
  },
314
  {
315
- "seconds": 1.6525030769989826,
316
- "bytes": 8063307,
317
- "pages": 38,
 
 
 
 
 
 
318
  "shards": 3
319
  },
320
  {
321
- "seconds": 1.0988706719981565,
322
- "bytes": 771997,
323
- "pages": 25,
324
- "shards": 1
325
  },
326
  {
327
- "seconds": 1.993406634999701,
328
- "bytes": 10476570,
329
- "pages": 39,
330
  "shards": 3
331
  },
332
  {
333
- "seconds": 0.8121612659997481,
334
- "bytes": 2238306,
335
- "pages": 27,
336
  "shards": 2
337
  },
338
  {
339
- "seconds": 1.133708954002941,
340
- "bytes": 6148950,
341
- "pages": 35,
342
- "shards": 2
343
  },
344
  {
345
- "seconds": 1.2174438740003097,
346
- "bytes": 5854390,
347
- "pages": 27,
348
  "shards": 2
349
  },
350
  {
351
- "seconds": 1.2782237920000625,
352
- "bytes": 7965202,
353
- "pages": 25,
354
  "shards": 3
355
  },
356
  {
357
- "seconds": 1.4521252480008116,
358
- "bytes": 10591031,
359
- "pages": 45,
360
- "shards": 3
361
  },
362
  {
363
- "seconds": 0.8998503880029602,
364
- "bytes": 1160065,
365
- "pages": 33,
366
- "shards": 1
367
  },
368
  {
369
- "seconds": 2.0495857230016554,
370
- "bytes": 7463969,
371
- "pages": 30,
372
  "shards": 3
373
  },
374
  {
375
- "seconds": 1.0575754780002171,
376
- "bytes": 10536411,
377
- "pages": 39,
378
  "shards": 3
379
  },
380
  {
381
- "seconds": 3.8299359349985025,
382
- "bytes": 2330234,
383
- "pages": 28,
384
  "shards": 2
385
  },
386
  {
387
- "seconds": 1.1309446140003274,
388
- "bytes": 12087341,
389
- "pages": 41,
390
- "shards": 3
391
  },
392
  {
393
- "seconds": 2.7144779629998084,
394
- "bytes": 851325,
395
- "pages": 25,
396
  "shards": 1
397
  },
398
  {
399
- "seconds": 1.4230423680019157,
400
- "bytes": 11644215,
401
- "pages": 41,
402
  "shards": 4
403
  },
404
  {
405
- "seconds": 1.8787968369979353,
406
- "bytes": 5636855,
407
- "pages": 27,
408
- "shards": 2
409
  },
410
  {
411
- "seconds": 0.8648194890010927,
412
- "bytes": 1161752,
413
- "pages": 33,
414
- "shards": 1
415
  },
416
  {
417
- "seconds": 1.3708638380012417,
418
- "bytes": 7781670,
419
- "pages": 39,
420
  "shards": 3
421
  },
422
  {
423
- "seconds": 2.350306128999364,
424
- "bytes": 11141333,
425
- "pages": 32,
426
- "shards": 3
427
  },
428
  {
429
- "seconds": 1.3496742860006634,
430
- "bytes": 7704664,
431
- "pages": 35,
432
  "shards": 3
433
  },
434
  {
435
- "seconds": 1.1927491759997793,
436
- "bytes": 6167341,
437
  "pages": 36,
438
  "shards": 2
439
  },
440
  {
441
- "seconds": 0.8386335059985868,
442
- "bytes": 1124320,
443
- "pages": 34,
444
- "shards": 1
445
  },
446
  {
447
- "seconds": 34.149081805000606,
448
- "bytes": 11274982,
449
- "pages": 35,
450
  "shards": 4
451
  },
452
  {
453
- "seconds": 0.9564227560003928,
454
- "bytes": 6333046,
455
- "pages": 36,
456
- "shards": 2
457
  },
458
  {
459
- "seconds": 1.5801954900016426,
460
- "bytes": 8475194,
 
 
 
 
 
 
461
  "pages": 35,
462
- "shards": 4
463
  },
464
  {
465
- "seconds": 1.4765983690012945,
466
- "bytes": 10921563,
467
  "pages": 32,
468
- "shards": 3
469
  },
470
  {
471
- "seconds": 0.8757540010010416,
472
- "bytes": 1242501,
473
- "pages": 34,
474
- "shards": 1
475
  },
476
  {
477
- "seconds": 1.202817702996981,
478
- "bytes": 7607811,
479
- "pages": 38,
480
- "shards": 3
481
  },
482
  {
483
- "seconds": 1.1735024220033665,
484
- "bytes": 5786171,
485
- "pages": 28,
486
  "shards": 2
487
  },
488
  {
489
- "seconds": 1.022158355001011,
490
- "bytes": 1083307,
491
- "pages": 33,
492
- "shards": 1
493
  },
494
  {
495
- "seconds": 33.35392404899903,
496
- "bytes": 16956987,
497
- "pages": 37,
498
- "shards": 5
499
  },
500
  {
501
- "seconds": 1.4096410159982042,
502
- "bytes": 6047627,
503
- "pages": 45,
504
- "shards": 2
505
  },
506
  {
507
- "seconds": 1.516586848996667,
508
- "bytes": 2356968,
509
- "pages": 27,
510
  "shards": 2
511
  },
512
  {
513
- "seconds": 1.1747013890017115,
514
- "bytes": 10632168,
515
- "pages": 34,
516
  "shards": 3
517
  },
518
  {
519
- "seconds": 0.9600440670001262,
520
- "bytes": 1151300,
521
- "pages": 35,
522
- "shards": 1
523
  },
524
  {
525
- "seconds": 1.3565457900003821,
526
- "bytes": 7973344,
527
- "pages": 38,
528
- "shards": 3
529
  },
530
  {
531
- "seconds": 1.262600800000655,
532
- "bytes": 11885196,
533
- "pages": 29,
534
- "shards": 3
535
  },
536
  {
537
- "seconds": 0.9526699870002631,
538
- "bytes": 6951757,
539
- "pages": 35,
540
  "shards": 3
541
  },
542
  {
543
- "seconds": 1.125239683999098,
544
- "bytes": 6131144,
545
- "pages": 36,
546
  "shards": 2
547
  },
548
  {
549
- "seconds": 0.9603508889995283,
550
- "bytes": 1054047,
551
- "pages": 35,
552
- "shards": 1
553
  },
554
  {
555
- "seconds": 1.33804497500023,
556
- "bytes": 7455487,
557
- "pages": 29,
558
  "shards": 3
559
  },
560
  {
561
- "seconds": 1.2401924379992124,
562
- "bytes": 10747728,
563
- "pages": 36,
 
 
 
 
 
 
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": 74741.43834005215,
26
+ "required_pages_per_second_upper_estimate": 0.006478398444872104
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": 9342.679792506518,
40
+ "required_pages_per_second_upper_estimate": 0.00372693465474171
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": 18685.359585013037,
54
+ "required_pages_per_second_upper_estimate": 0.00432252908889573
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": 28028.03937751956,
68
+ "required_pages_per_second_upper_estimate": 0.0017067605463330859
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": 28028.03937751956,
82
+ "required_pages_per_second_upper_estimate": 0.0021576760578602147
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": 28028.03937751956,
96
+ "required_pages_per_second_upper_estimate": 0.34801444525521885
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": 0.11895504780454758,
162
+ "producer_p95_seconds": 0.35881434900147724,
163
+ "gpu_update_seconds": 0.9864407434140978,
164
+ "producer_batches_per_second": 8.406536910002155,
165
+ "gpu_batches_per_second": 1.0137456372077387,
166
+ "producer_headroom": 8.292550519040578,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.17288534825997,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 81,
172
  "max_pending_pages": 4096,
173
+ "generation_bytes_per_second": 252760.82975083197,
174
+ "upload_bytes_per_second": 2930376.808544498,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 54.23659009355277,
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": 1.1927491759997793,
184
+ "bytes": 6167341,
 
 
 
 
 
 
185
  "pages": 36,
186
  "shards": 2
187
  },
188
  {
189
+ "seconds": 0.8386335059985868,
190
+ "bytes": 1124320,
191
+ "pages": 34,
192
+ "shards": 1
 
 
 
 
 
 
193
  },
194
  {
195
+ "seconds": 34.149081805000606,
196
+ "bytes": 11274982,
197
+ "pages": 35,
198
  "shards": 4
199
  },
200
  {
201
+ "seconds": 0.9564227560003928,
202
+ "bytes": 6333046,
203
  "pages": 36,
204
  "shards": 2
205
  },
206
  {
207
+ "seconds": 1.5801954900016426,
208
+ "bytes": 8475194,
209
+ "pages": 35,
210
+ "shards": 4
211
  },
212
  {
213
+ "seconds": 1.4765983690012945,
214
+ "bytes": 10921563,
215
  "pages": 32,
216
  "shards": 3
217
  },
218
  {
219
+ "seconds": 0.8757540010010416,
220
+ "bytes": 1242501,
221
+ "pages": 34,
222
+ "shards": 1
223
  },
224
  {
225
+ "seconds": 1.202817702996981,
226
+ "bytes": 7607811,
227
+ "pages": 38,
228
  "shards": 3
229
  },
230
  {
231
+ "seconds": 1.1735024220033665,
232
+ "bytes": 5786171,
233
+ "pages": 28,
234
  "shards": 2
235
  },
236
  {
237
+ "seconds": 1.022158355001011,
238
+ "bytes": 1083307,
239
  "pages": 33,
240
  "shards": 1
241
  },
242
  {
243
+ "seconds": 33.35392404899903,
244
+ "bytes": 16956987,
245
+ "pages": 37,
246
+ "shards": 5
247
  },
248
  {
249
+ "seconds": 1.4096410159982042,
250
+ "bytes": 6047627,
251
+ "pages": 45,
252
+ "shards": 2
253
  },
254
  {
255
+ "seconds": 1.516586848996667,
256
+ "bytes": 2356968,
257
+ "pages": 27,
258
  "shards": 2
259
  },
260
  {
261
+ "seconds": 1.1747013890017115,
262
+ "bytes": 10632168,
263
+ "pages": 34,
264
  "shards": 3
265
  },
266
  {
267
+ "seconds": 0.9600440670001262,
268
+ "bytes": 1151300,
269
+ "pages": 35,
270
+ "shards": 1
271
  },
272
  {
273
+ "seconds": 1.3565457900003821,
274
+ "bytes": 7973344,
275
+ "pages": 38,
276
  "shards": 3
277
  },
278
  {
279
+ "seconds": 1.262600800000655,
280
+ "bytes": 11885196,
281
+ "pages": 29,
282
+ "shards": 3
 
 
 
 
 
 
283
  },
284
  {
285
+ "seconds": 0.9526699870002631,
286
+ "bytes": 6951757,
287
  "pages": 35,
288
+ "shards": 3
289
  },
290
  {
291
+ "seconds": 1.125239683999098,
292
+ "bytes": 6131144,
293
+ "pages": 36,
294
  "shards": 2
295
  },
296
  {
297
+ "seconds": 0.9603508889995283,
298
+ "bytes": 1054047,
299
+ "pages": 35,
300
+ "shards": 1
301
+ },
302
+ {
303
+ "seconds": 1.33804497500023,
304
+ "bytes": 7455487,
305
+ "pages": 29,
306
  "shards": 3
307
  },
308
  {
309
+ "seconds": 1.2401924379992124,
310
+ "bytes": 10747728,
311
+ "pages": 36,
312
+ "shards": 3
313
  },
314
  {
315
+ "seconds": 1.4399759389998508,
316
+ "bytes": 8283696,
317
+ "pages": 33,
318
  "shards": 3
319
  },
320
  {
321
+ "seconds": 1.058994241997425,
322
+ "bytes": 5472606,
323
+ "pages": 16,
324
  "shards": 2
325
  },
326
  {
327
+ "seconds": 0.7334500290016877,
328
+ "bytes": 861667,
329
+ "pages": 26,
330
+ "shards": 1
331
  },
332
  {
333
+ "seconds": 1.2821004600009474,
334
+ "bytes": 6367474,
335
+ "pages": 36,
336
  "shards": 2
337
  },
338
  {
339
+ "seconds": 1.168987963003019,
340
+ "bytes": 6662663,
341
+ "pages": 33,
342
  "shards": 3
343
  },
344
  {
345
+ "seconds": 3.650391375998879,
346
+ "bytes": 6163332,
347
+ "pages": 31,
348
+ "shards": 2
349
  },
350
  {
351
+ "seconds": 0.9980511270005081,
352
+ "bytes": 2638272,
353
+ "pages": 37,
354
+ "shards": 2
355
  },
356
  {
357
+ "seconds": 1.6268621890012582,
358
+ "bytes": 11807082,
359
+ "pages": 28,
360
  "shards": 3
361
  },
362
  {
363
+ "seconds": 3.1277496100010467,
364
+ "bytes": 10196563,
365
+ "pages": 38,
366
  "shards": 3
367
  },
368
  {
369
+ "seconds": 0.9736092290004308,
370
+ "bytes": 2727259,
371
+ "pages": 37,
372
  "shards": 2
373
  },
374
  {
375
+ "seconds": 1.154159282999899,
376
+ "bytes": 5704653,
377
+ "pages": 27,
378
+ "shards": 2
379
  },
380
  {
381
+ "seconds": 1.143835795999621,
382
+ "bytes": 1177702,
383
+ "pages": 35,
384
  "shards": 1
385
  },
386
  {
387
+ "seconds": 1.4552070569989155,
388
+ "bytes": 11480592,
389
+ "pages": 40,
390
  "shards": 4
391
  },
392
  {
393
+ "seconds": 1.264268914997956,
394
+ "bytes": 11538285,
395
+ "pages": 29,
396
+ "shards": 3
397
  },
398
  {
399
+ "seconds": 0.9019468660007988,
400
+ "bytes": 1029860,
401
+ "pages": 29,
402
+ "shards": 2
403
  },
404
  {
405
+ "seconds": 1.031183188999421,
406
+ "bytes": 7489997,
407
+ "pages": 36,
408
  "shards": 3
409
  },
410
  {
411
+ "seconds": 1.171489565000229,
412
+ "bytes": 5465773,
413
+ "pages": 35,
414
+ "shards": 2
415
  },
416
  {
417
+ "seconds": 1.024480873002176,
418
+ "bytes": 7386490,
419
+ "pages": 32,
420
  "shards": 3
421
  },
422
  {
423
+ "seconds": 1.0989981620004983,
424
+ "bytes": 6127153,
425
  "pages": 36,
426
  "shards": 2
427
  },
428
  {
429
+ "seconds": 1.2221524419983325,
430
+ "bytes": 6743871,
431
+ "pages": 31,
432
+ "shards": 2
433
  },
434
  {
435
+ "seconds": 1.1651974790001987,
436
+ "bytes": 11511318,
437
+ "pages": 32,
438
  "shards": 4
439
  },
440
  {
441
+ "seconds": 0.8884790899974178,
442
+ "bytes": 1189196,
443
+ "pages": 35,
444
+ "shards": 1
445
  },
446
  {
447
+ "seconds": 1.1859263639998971,
448
+ "bytes": 7316047,
449
+ "pages": 30,
450
+ "shards": 3
451
+ },
452
+ {
453
+ "seconds": 1.2110801790004189,
454
+ "bytes": 6933944,
455
  "pages": 35,
456
+ "shards": 2
457
  },
458
  {
459
+ "seconds": 1.1756158520001918,
460
+ "bytes": 5451054,
461
  "pages": 32,
462
+ "shards": 2
463
  },
464
  {
465
+ "seconds": 1.2875092799986305,
466
+ "bytes": 13122964,
467
+ "pages": 36,
468
+ "shards": 4
469
  },
470
  {
471
+ "seconds": 1.006686715001706,
472
+ "bytes": 1247241,
473
+ "pages": 35,
474
+ "shards": 1
475
  },
476
  {
477
+ "seconds": 1.2502870179996535,
478
+ "bytes": 6125090,
479
+ "pages": 32,
480
  "shards": 2
481
  },
482
  {
483
+ "seconds": 1.1863440980014275,
484
+ "bytes": 11209339,
485
+ "pages": 32,
486
+ "shards": 4
487
  },
488
  {
489
+ "seconds": 0.8518053779989714,
490
+ "bytes": 1243189,
491
+ "pages": 34,
492
+ "shards": 1
493
  },
494
  {
495
+ "seconds": 1.3384703369993076,
496
+ "bytes": 7381158,
497
+ "pages": 30,
498
+ "shards": 3
499
  },
500
  {
501
+ "seconds": 0.9186418360004609,
502
+ "bytes": 6519513,
503
+ "pages": 25,
504
  "shards": 2
505
  },
506
  {
507
+ "seconds": 1.2462770620004449,
508
+ "bytes": 11030793,
509
+ "pages": 37,
510
  "shards": 3
511
  },
512
  {
513
+ "seconds": 0.8822051489987643,
514
+ "bytes": 2343640,
515
+ "pages": 29,
516
+ "shards": 2
517
  },
518
  {
519
+ "seconds": 1.0508419269972364,
520
+ "bytes": 6132224,
521
+ "pages": 36,
522
+ "shards": 2
523
  },
524
  {
525
+ "seconds": 1.237680772999738,
526
+ "bytes": 6534213,
527
+ "pages": 36,
528
+ "shards": 2
529
  },
530
  {
531
+ "seconds": 1.075308318999305,
532
+ "bytes": 6271877,
533
+ "pages": 31,
534
  "shards": 3
535
  },
536
  {
537
+ "seconds": 1.1617618540003605,
538
+ "bytes": 5816168,
539
+ "pages": 28,
540
  "shards": 2
541
  },
542
  {
543
+ "seconds": 1.0311880699991889,
544
+ "bytes": 6926958,
545
+ "pages": 34,
546
+ "shards": 2
547
  },
548
  {
549
+ "seconds": 1.0975180309978896,
550
+ "bytes": 7653776,
551
+ "pages": 37,
552
  "shards": 3
553
  },
554
  {
555
+ "seconds": 0.9772678519984765,
556
+ "bytes": 5704498,
557
+ "pages": 29,
558
+ "shards": 2
559
+ },
560
+ {
561
+ "seconds": 1.346531524999591,
562
+ "bytes": 7336381,
563
+ "pages": 29,
564
  "shards": 3
565
  }
566
  ],
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:796925d7192e7f222c8c88ea2d27a48542f3cbdadf89ca004b7822e1a82d3d36
3
- size 721771839
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c3941f290352b013b70ffc29d16e39df4aacbd1a763c6261ae4bbca696edc601
3
+ size 721759807