appvoid commited on
Commit
2e5f789
·
verified ·
1 Parent(s): 8a4da96

Qwen90M step 65000; 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 +248 -248
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 60000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "a86621d512304974e4d03bef262524ec95edf0e52d000b45faebb196bf0ad813",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "09ea04c8a1fd2002db31adafc765f00a01b0176621104639d3a70a8010a648f4",
12
  "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
13
  "performance.json": "d14cb208816158eb276e72d732c3b9e6358b2a75509678412180fc272350709f",
14
- "shard_scheduler.json": "502ba4fb1878cbdcbe2c170d26ff10458bed187e056b37f6ad7fa2afaa31cc12",
15
- "coverage.json": "558c9d8fbc6f9cdc4f9f8ba67a6152d0f15e7f96f74c310caabb313dfccd148e",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 65000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "e33da1dad32ae2433103ec21715c2d542c98dc7e7caa2f82997e62b6fe8342d2",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "e3e3dd01665bea5cf593b73f4417e92f7e07fdb4e1949ec1cfe309333ab911ee",
12
  "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
13
  "performance.json": "d14cb208816158eb276e72d732c3b9e6358b2a75509678412180fc272350709f",
14
+ "shard_scheduler.json": "fb85ef21cda081906908c01fe4de71830aa788514ed6534050e7da541b10e3e2",
15
+ "coverage.json": "f8f337b02bacae59e68d5dbc5eb4b6339c704cf53f376b91e8950e701871a03e",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 60000,
3
  "stats": {
4
- "climbmix/blocks": 2160000,
5
- "climbmix/targets": 3609663868,
6
- "climbmix/input_tokens": 3609892580,
7
- "ultra_qa/blocks": 810000,
8
- "ultra_qa/targets": 1321742315,
9
- "ultra_qa/input_tokens": 1321744914,
10
- "cortex/blocks": 810000,
11
- "cortex/targets": 30814012,
12
- "cortex/input_tokens": 1287629180,
13
- "ultra_style/blocks": 810000,
14
- "ultra_style/targets": 1281312518,
15
- "ultra_style/input_tokens": 1281379443,
16
- "openmath/blocks": 540000,
17
- "openmath/targets": 834463745,
18
- "openmath/input_tokens": 834742126,
19
- "rewrite6/blocks": 270000,
20
- "rewrite6/targets": 442578795,
21
- "rewrite6/input_tokens": 443064258
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 295,
26
- "row": 3347,
27
  "start": {
28
- "file": 14,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 22,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
- "documents": 1211667
38
  },
39
  "rewrite6": {
40
- "page": 18,
41
- "row": 1596,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 73728
47
  }
48
  },
49
  "epoch": 3,
50
- "offset": 2048,
51
- "documents": 691536
52
  },
53
  "openmath": {
54
- "page": 184,
55
- "row": 3842,
56
  "start": {
57
  "file": 1,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 317,
61
- "row": 14
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 757506
67
  },
68
  "ultra_qa": {
69
- "page": 79,
70
- "row": 4028,
71
  "start": {
72
  "file": 0,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 5,
76
- "row": 847
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
- "documents": 327612
82
  },
83
  "ultra_style": {
84
- "page": 121,
85
- "row": 3563,
86
  "start": {
87
  "file": 0,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
- "row_group": 5,
91
- "row": 69812
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
- "documents": 499179
97
  },
98
  "cortex": {
99
- "page": 15706,
100
- "row": 87,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 2010455
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 65000,
3
  "stats": {
4
+ "climbmix/blocks": 2340000,
5
+ "climbmix/targets": 3910377700,
6
+ "climbmix/input_tokens": 3910625311,
7
+ "ultra_qa/blocks": 877500,
8
+ "ultra_qa/targets": 1431685811,
9
+ "ultra_qa/input_tokens": 1431688622,
10
+ "cortex/blocks": 877500,
11
+ "cortex/targets": 33383362,
12
+ "cortex/input_tokens": 1394824403,
13
+ "ultra_style/blocks": 877500,
14
+ "ultra_style/targets": 1387878431,
15
+ "ultra_style/input_tokens": 1387950984,
16
+ "openmath/blocks": 585000,
17
+ "openmath/targets": 903936284,
18
+ "openmath/input_tokens": 904237672,
19
+ "rewrite6/blocks": 292500,
20
+ "rewrite6/targets": 479420381,
21
+ "rewrite6/input_tokens": 479946209
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 320,
26
+ "row": 1789,
27
  "start": {
28
+ "file": 15,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 39,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
+ "documents": 1312509
38
  },
39
  "rewrite6": {
40
+ "page": 32,
41
+ "row": 1861,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 131072
47
  }
48
  },
49
  "epoch": 3,
50
+ "offset": 0,
51
+ "documents": 749145
52
  },
53
  "openmath": {
54
+ "page": 200,
55
+ "row": 1224,
56
  "start": {
57
  "file": 1,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 382,
61
+ "row": 550
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 820424
67
  },
68
  "ultra_qa": {
69
+ "page": 86,
70
+ "row": 2631,
71
  "start": {
72
  "file": 0,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 5,
76
+ "row": 29519
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
+ "documents": 354887
82
  },
83
  "ultra_style": {
84
+ "page": 132,
85
+ "row": 470,
86
  "start": {
87
  "file": 0,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
+ "row_group": 6,
91
+ "row": 30335
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
+ "documents": 541142
97
  },
98
  "cortex": {
99
+ "page": 17014,
100
+ "row": 1,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 2177793
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 60000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "a86621d512304974e4d03bef262524ec95edf0e52d000b45faebb196bf0ad813", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "09ea04c8a1fd2002db31adafc765f00a01b0176621104639d3a70a8010a648f4", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "d14cb208816158eb276e72d732c3b9e6358b2a75509678412180fc272350709f", "shard_scheduler.json": "502ba4fb1878cbdcbe2c170d26ff10458bed187e056b37f6ad7fa2afaa31cc12", "coverage.json": "558c9d8fbc6f9cdc4f9f8ba67a6152d0f15e7f96f74c310caabb313dfccd148e", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000060000"}
 
1
+ {"step": 65000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "e33da1dad32ae2433103ec21715c2d542c98dc7e7caa2f82997e62b6fe8342d2", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "e3e3dd01665bea5cf593b73f4417e92f7e07fdb4e1949ec1cfe309333ab911ee", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "d14cb208816158eb276e72d732c3b9e6358b2a75509678412180fc272350709f", "shard_scheduler.json": "fb85ef21cda081906908c01fe4de71830aa788514ed6534050e7da541b10e3e2", "coverage.json": "f8f337b02bacae59e68d5dbc5eb4b6339c704cf53f376b91e8950e701871a03e", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000065000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:a86621d512304974e4d03bef262524ec95edf0e52d000b45faebb196bf0ad813
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e33da1dad32ae2433103ec21715c2d542c98dc7e7caa2f82997e62b6fe8342d2
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": 75551.01885780718,
26
- "required_pages_per_second_upper_estimate": 0.006548570832288048
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": 9443.877357225898,
40
- "required_pages_per_second_upper_estimate": 0.0037673038763466962
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": 18887.754714451796,
54
- "required_pages_per_second_upper_estimate": 0.004369349640058767
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": 28331.632071677694,
68
- "required_pages_per_second_upper_estimate": 0.0017252477485795504
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": 28331.632071677694,
82
- "required_pages_per_second_upper_estimate": 0.002181047463854873
83
  },
84
  {
85
  "source": "cortex",
@@ -92,8 +92,8 @@
92
  "tokens_per_second": 1015177.8395898147,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
  "estimated_pages_for_buffer": 22,
95
- "required_token_positions_per_second": 28331.632071677694,
96
- "required_pages_per_second_upper_estimate": 0.3517840504572767
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,410 +158,410 @@
158
  "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 688946,
161
- "producer_batch_seconds": 0.09136518590602805,
162
- "producer_p95_seconds": 0.29452696799853584,
163
- "gpu_update_seconds": 0.9758703603820585,
164
- "producer_batches_per_second": 10.945088001337089,
165
- "gpu_batches_per_second": 1.0247262757406574,
166
- "producer_headroom": 10.68098697227817,
167
  "sustainable": true,
168
- "buffer_seconds": 60.50396234368763,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 87,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 384384.7746639473,
174
- "upload_bytes_per_second": 6418706.338644549,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 82.48002462328733,
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": 1.1064888179989794,
184
- "bytes": 5541578,
185
- "pages": 33,
 
 
 
 
 
 
186
  "shards": 2
187
  },
188
  {
189
- "seconds": 0.9643621880004503,
190
- "bytes": 7410728,
191
- "pages": 34,
 
 
 
 
 
 
192
  "shards": 3
193
  },
194
  {
195
- "seconds": 0.8883736560001125,
196
- "bytes": 6293157,
197
- "pages": 36,
198
  "shards": 2
199
  },
200
  {
201
- "seconds": 0.9868076469992957,
202
- "bytes": 927962,
203
- "pages": 26,
204
- "shards": 1
205
  },
206
  {
207
- "seconds": 1.2888746650005487,
208
- "bytes": 18374728,
209
- "pages": 40,
210
- "shards": 5
211
  },
212
  {
213
- "seconds": 1.1437757880012214,
214
- "bytes": 6056576,
215
- "pages": 36,
216
- "shards": 2
217
  },
218
  {
219
- "seconds": 1.102731920998849,
220
- "bytes": 2598092,
221
- "pages": 36,
222
- "shards": 2
223
  },
224
  {
225
- "seconds": 1.1366644360005012,
226
- "bytes": 5837515,
227
- "pages": 27,
228
  "shards": 2
229
  },
230
  {
231
- "seconds": 0.8290303609992407,
232
- "bytes": 5336251,
233
- "pages": 39,
234
- "shards": 2
235
  },
236
  {
237
- "seconds": 1.0252480340004695,
238
- "bytes": 7371694,
239
- "pages": 29,
240
  "shards": 3
241
  },
242
  {
243
- "seconds": 1.1864866850010003,
244
- "bytes": 11535625,
245
  "pages": 35,
246
- "shards": 3
247
  },
248
  {
249
- "seconds": 0.9566465869993408,
250
- "bytes": 1137762,
251
- "pages": 35,
252
  "shards": 2
253
  },
254
  {
255
- "seconds": 1.0941519280004286,
256
- "bytes": 11375746,
257
- "pages": 34,
258
- "shards": 4
259
  },
260
  {
261
- "seconds": 0.7203741030007222,
262
- "bytes": 1052913,
263
  "pages": 34,
264
  "shards": 1
265
  },
266
  {
267
- "seconds": 1.0976181589994667,
268
- "bytes": 7695785,
269
- "pages": 37,
270
- "shards": 3
271
  },
272
  {
273
- "seconds": 1.024126566000632,
274
- "bytes": 5953181,
275
  "pages": 28,
276
  "shards": 2
277
  },
278
  {
279
- "seconds": 1.452807390000089,
280
- "bytes": 11684781,
281
- "pages": 40,
282
  "shards": 3
283
  },
284
  {
285
- "seconds": 0.8971567329990648,
286
- "bytes": 7415356,
287
- "pages": 29,
288
- "shards": 3
289
  },
290
  {
291
- "seconds": 0.9046226490008848,
292
- "bytes": 1040439,
293
- "pages": 34,
294
- "shards": 1
295
  },
296
  {
297
- "seconds": 1.043625083999359,
298
- "bytes": 6212867,
299
- "pages": 35,
300
- "shards": 2
301
  },
302
  {
303
- "seconds": 1.3019511900001817,
304
- "bytes": 12069002,
305
- "pages": 36,
306
- "shards": 4
307
  },
308
  {
309
- "seconds": 0.8234439619991463,
310
- "bytes": 1074472,
311
- "pages": 32,
312
  "shards": 1
313
  },
314
  {
315
- "seconds": 1.0881272910010011,
316
- "bytes": 13091178,
317
- "pages": 33,
318
  "shards": 4
319
  },
320
  {
321
- "seconds": 1.0808892179993563,
322
- "bytes": 6450126,
 
 
 
 
 
 
323
  "pages": 36,
324
  "shards": 2
325
  },
326
  {
327
- "seconds": 0.8513057910004136,
328
- "bytes": 341762,
329
- "pages": 9,
330
- "shards": 1
331
  },
332
  {
333
- "seconds": 1.0242165819981892,
334
- "bytes": 5354140,
335
  "pages": 32,
336
  "shards": 2
337
  },
338
  {
339
- "seconds": 0.845649806999063,
340
- "bytes": 8068507,
341
- "pages": 38,
342
  "shards": 3
343
  },
344
  {
345
- "seconds": 0.9696599159997277,
346
- "bytes": 6181281,
347
- "pages": 35,
348
- "shards": 2
349
  },
350
  {
351
- "seconds": 1.0534396790008032,
352
- "bytes": 2069289,
353
- "pages": 21,
354
  "shards": 2
355
  },
356
  {
357
- "seconds": 1.112150704000669,
358
- "bytes": 16576455,
359
- "pages": 44,
360
- "shards": 4
361
  },
362
  {
363
- "seconds": 0.7744886969994695,
364
- "bytes": 1014054,
365
  "pages": 34,
366
- "shards": 1
367
  },
368
  {
369
- "seconds": 0.9544328729989502,
370
- "bytes": 7444287,
371
- "pages": 37,
372
- "shards": 3
373
  },
374
  {
375
- "seconds": 1.0936733489998005,
376
- "bytes": 876493,
377
- "pages": 25,
378
- "shards": 1
379
  },
380
  {
381
- "seconds": 1.364482594999572,
382
- "bytes": 12283477,
383
- "pages": 38,
384
  "shards": 4
385
  },
386
  {
387
- "seconds": 1.1743630839991965,
388
- "bytes": 5760431,
389
- "pages": 28,
390
- "shards": 2
391
- },
392
- {
393
- "seconds": 0.9926189179968787,
394
- "bytes": 7099055,
395
- "pages": 43,
396
- "shards": 2
397
  },
398
  {
399
- "seconds": 1.0609590189997107,
400
- "bytes": 7221766,
401
- "pages": 30,
402
  "shards": 3
403
  },
404
  {
405
- "seconds": 0.9498501529997156,
406
- "bytes": 6032509,
407
- "pages": 30,
408
  "shards": 2
409
  },
410
  {
411
- "seconds": 1.024946067998826,
412
- "bytes": 5141639,
413
- "pages": 39,
414
- "shards": 2
415
  },
416
  {
417
- "seconds": 1.2369695059969672,
418
- "bytes": 7463645,
419
  "pages": 35,
420
- "shards": 3
421
  },
422
  {
423
- "seconds": 0.9742800499989244,
424
- "bytes": 772407,
425
- "pages": 25,
426
  "shards": 1
427
  },
428
  {
429
- "seconds": 1.406286874000216,
430
- "bytes": 18261180,
431
- "pages": 44,
432
- "shards": 5
433
- },
434
- {
435
- "seconds": 1.0190579750014876,
436
- "bytes": 5885831,
437
- "pages": 35,
438
- "shards": 2
439
- },
440
- {
441
- "seconds": 0.8101317319997179,
442
- "bytes": 713378,
443
- "pages": 25,
444
- "shards": 2
445
  },
446
  {
447
- "seconds": 1.03739196099923,
448
- "bytes": 7802740,
449
- "pages": 38,
450
  "shards": 3
451
  },
452
  {
453
- "seconds": 1.1620692619981128,
454
- "bytes": 6079962,
455
- "pages": 34,
456
  "shards": 2
457
  },
458
  {
459
- "seconds": 0.8016836730021168,
460
- "bytes": 4650466,
461
- "pages": 32,
462
  "shards": 2
463
  },
464
  {
465
- "seconds": 1.2261960830001044,
466
- "bytes": 13273679,
467
- "pages": 37,
468
- "shards": 4
469
- },
470
- {
471
- "seconds": 0.7825120570014406,
472
- "bytes": 1093241,
473
- "pages": 34,
474
- "shards": 1
475
  },
476
  {
477
- "seconds": 1.4853517330011528,
478
- "bytes": 11092996,
479
- "pages": 24,
480
- "shards": 4
481
  },
482
  {
483
- "seconds": 0.8748619340003643,
484
- "bytes": 6384371,
485
- "pages": 42,
486
- "shards": 2
487
  },
488
  {
489
- "seconds": 0.9224505880010838,
490
- "bytes": 1117951,
491
  "pages": 34,
492
  "shards": 1
493
  },
494
  {
495
- "seconds": 1.0108905939996475,
496
- "bytes": 7845615,
497
- "pages": 36,
498
  "shards": 3
499
  },
500
  {
501
- "seconds": 1.27624071000173,
502
- "bytes": 16075425,
503
- "pages": 35,
504
- "shards": 4
505
  },
506
  {
507
- "seconds": 0.8457192399982887,
508
- "bytes": 2628229,
509
  "pages": 36,
510
  "shards": 2
511
  },
512
  {
513
- "seconds": 0.9226963619985327,
514
- "bytes": 5611919,
515
- "pages": 27,
516
  "shards": 2
517
  },
518
  {
519
- "seconds": 0.7924021649996575,
520
- "bytes": 1122256,
521
- "pages": 34,
522
- "shards": 1
523
  },
524
  {
525
- "seconds": 1.1671633029982331,
526
- "bytes": 11363880,
527
- "pages": 40,
528
- "shards": 4
529
  },
530
  {
531
- "seconds": 0.9065282010014926,
532
- "bytes": 6177542,
533
- "pages": 28,
534
  "shards": 2
535
  },
536
  {
537
- "seconds": 1.0971184539994283,
538
- "bytes": 8482200,
539
- "pages": 38,
540
- "shards": 3
541
  },
542
  {
543
- "seconds": 1.1112909670009685,
544
- "bytes": 6011206,
545
- "pages": 35,
546
- "shards": 2
547
  },
548
  {
549
- "seconds": 0.9119941530007054,
550
- "bytes": 5747516,
551
- "pages": 29,
552
- "shards": 2
553
  },
554
  {
555
- "seconds": 1.016742787000112,
556
- "bytes": 7467282,
557
  "pages": 37,
558
  "shards": 3
559
  },
560
  {
561
- "seconds": 0.8858778819994768,
562
- "bytes": 5985432,
563
- "pages": 27,
564
  "shards": 2
 
 
 
 
 
 
565
  }
566
  ],
567
  "cache_key": {
 
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": 75748.15161257412,
26
+ "required_pages_per_second_upper_estimate": 0.006565657799842846
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": 9468.518951571765,
40
+ "required_pages_per_second_upper_estimate": 0.0037771337767559298
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": 18937.03790314353,
54
+ "required_pages_per_second_upper_estimate": 0.0043807504384081025
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": 28405.556854715294,
68
+ "required_pages_per_second_upper_estimate": 0.0017297493800131785
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": 28405.556854715294,
82
+ "required_pages_per_second_upper_estimate": 0.0021867383982900044
83
  },
84
  {
85
  "source": "cortex",
 
92
  "tokens_per_second": 1015177.8395898147,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
  "estimated_pages_for_buffer": 22,
95
+ "required_token_positions_per_second": 28405.556854715294,
96
+ "required_pages_per_second_upper_estimate": 0.3527019488522703
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
 
158
  "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 688946,
161
+ "producer_batch_seconds": 0.11656077062494319,
162
+ "producer_p95_seconds": 0.29636224299974856,
163
+ "gpu_update_seconds": 0.9733306810850448,
164
+ "producer_batches_per_second": 8.579215757055119,
165
+ "gpu_batches_per_second": 1.0274000598493669,
166
+ "producer_headroom": 8.350413915990007,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.34650222727278,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 86,
172
  "max_pending_pages": 4096,
173
+ "generation_bytes_per_second": 512873.7689111375,
174
+ "upload_bytes_per_second": 4674834.667993678,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 110.05077171803086,
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": 1.406286874000216,
184
+ "bytes": 18261180,
185
+ "pages": 44,
186
+ "shards": 5
187
+ },
188
+ {
189
+ "seconds": 1.0190579750014876,
190
+ "bytes": 5885831,
191
+ "pages": 35,
192
  "shards": 2
193
  },
194
  {
195
+ "seconds": 0.8101317319997179,
196
+ "bytes": 713378,
197
+ "pages": 25,
198
+ "shards": 2
199
+ },
200
+ {
201
+ "seconds": 1.03739196099923,
202
+ "bytes": 7802740,
203
+ "pages": 38,
204
  "shards": 3
205
  },
206
  {
207
+ "seconds": 1.1620692619981128,
208
+ "bytes": 6079962,
209
+ "pages": 34,
210
  "shards": 2
211
  },
212
  {
213
+ "seconds": 0.8016836730021168,
214
+ "bytes": 4650466,
215
+ "pages": 32,
216
+ "shards": 2
217
  },
218
  {
219
+ "seconds": 1.2261960830001044,
220
+ "bytes": 13273679,
221
+ "pages": 37,
222
+ "shards": 4
223
  },
224
  {
225
+ "seconds": 0.7825120570014406,
226
+ "bytes": 1093241,
227
+ "pages": 34,
228
+ "shards": 1
229
  },
230
  {
231
+ "seconds": 1.4853517330011528,
232
+ "bytes": 11092996,
233
+ "pages": 24,
234
+ "shards": 4
235
  },
236
  {
237
+ "seconds": 0.8748619340003643,
238
+ "bytes": 6384371,
239
+ "pages": 42,
240
  "shards": 2
241
  },
242
  {
243
+ "seconds": 0.9224505880010838,
244
+ "bytes": 1117951,
245
+ "pages": 34,
246
+ "shards": 1
247
  },
248
  {
249
+ "seconds": 1.0108905939996475,
250
+ "bytes": 7845615,
251
+ "pages": 36,
252
  "shards": 3
253
  },
254
  {
255
+ "seconds": 1.27624071000173,
256
+ "bytes": 16075425,
257
  "pages": 35,
258
+ "shards": 4
259
  },
260
  {
261
+ "seconds": 0.8457192399982887,
262
+ "bytes": 2628229,
263
+ "pages": 36,
264
  "shards": 2
265
  },
266
  {
267
+ "seconds": 0.9226963619985327,
268
+ "bytes": 5611919,
269
+ "pages": 27,
270
+ "shards": 2
271
  },
272
  {
273
+ "seconds": 0.7924021649996575,
274
+ "bytes": 1122256,
275
  "pages": 34,
276
  "shards": 1
277
  },
278
  {
279
+ "seconds": 1.1671633029982331,
280
+ "bytes": 11363880,
281
+ "pages": 40,
282
+ "shards": 4
283
  },
284
  {
285
+ "seconds": 0.9065282010014926,
286
+ "bytes": 6177542,
287
  "pages": 28,
288
  "shards": 2
289
  },
290
  {
291
+ "seconds": 1.0971184539994283,
292
+ "bytes": 8482200,
293
+ "pages": 38,
294
  "shards": 3
295
  },
296
  {
297
+ "seconds": 1.1112909670009685,
298
+ "bytes": 6011206,
299
+ "pages": 35,
300
+ "shards": 2
301
  },
302
  {
303
+ "seconds": 0.9119941530007054,
304
+ "bytes": 5747516,
305
+ "pages": 29,
306
+ "shards": 2
307
  },
308
  {
309
+ "seconds": 1.016742787000112,
310
+ "bytes": 7467282,
311
+ "pages": 37,
312
+ "shards": 3
313
  },
314
  {
315
+ "seconds": 0.8858778819994768,
316
+ "bytes": 5985432,
317
+ "pages": 27,
318
+ "shards": 2
319
  },
320
  {
321
+ "seconds": 0.7830291239988583,
322
+ "bytes": 1207474,
323
+ "pages": 35,
324
  "shards": 1
325
  },
326
  {
327
+ "seconds": 1.2103096939972602,
328
+ "bytes": 12749523,
329
+ "pages": 22,
330
  "shards": 4
331
  },
332
  {
333
+ "seconds": 0.8322358400000667,
334
+ "bytes": 4694947,
335
+ "pages": 29,
336
+ "shards": 2
337
+ },
338
+ {
339
+ "seconds": 0.8593272979996982,
340
+ "bytes": 6225685,
341
  "pages": 36,
342
  "shards": 2
343
  },
344
  {
345
+ "seconds": 0.9578338470018934,
346
+ "bytes": 2306768,
347
+ "pages": 27,
348
+ "shards": 2
349
  },
350
  {
351
+ "seconds": 0.8912073430001328,
352
+ "bytes": 6007286,
353
  "pages": 32,
354
  "shards": 2
355
  },
356
  {
357
+ "seconds": 1.088943154001754,
358
+ "bytes": 10590685,
359
+ "pages": 36,
360
  "shards": 3
361
  },
362
  {
363
+ "seconds": 0.9887308050019783,
364
+ "bytes": 8510426,
365
+ "pages": 40,
366
+ "shards": 3
367
  },
368
  {
369
+ "seconds": 1.0075569920009002,
370
+ "bytes": 5909692,
371
+ "pages": 30,
372
  "shards": 2
373
  },
374
  {
375
+ "seconds": 0.7540812590013957,
376
+ "bytes": 2508470,
377
+ "pages": 31,
378
+ "shards": 2
379
  },
380
  {
381
+ "seconds": 1.3975003769992327,
382
+ "bytes": 10375015,
383
  "pages": 34,
384
+ "shards": 4
385
  },
386
  {
387
+ "seconds": 1.0634488309988228,
388
+ "bytes": 6386527,
389
+ "pages": 36,
390
+ "shards": 2
391
  },
392
  {
393
+ "seconds": 14.660640815000079,
394
+ "bytes": 2712693,
395
+ "pages": 36,
396
+ "shards": 2
397
  },
398
  {
399
+ "seconds": 3.9525413480005227,
400
+ "bytes": 15333188,
401
+ "pages": 34,
402
  "shards": 4
403
  },
404
  {
405
+ "seconds": 2.024789704999421,
406
+ "bytes": 1170220,
407
+ "pages": 34,
408
+ "shards": 1
 
 
 
 
 
 
409
  },
410
  {
411
+ "seconds": 1.2876270810011192,
412
+ "bytes": 7494644,
413
+ "pages": 37,
414
  "shards": 3
415
  },
416
  {
417
+ "seconds": 0.8472818170012033,
418
+ "bytes": 6269698,
419
+ "pages": 35,
420
  "shards": 2
421
  },
422
  {
423
+ "seconds": 0.9710190369987686,
424
+ "bytes": 6743419,
425
+ "pages": 30,
426
+ "shards": 3
427
  },
428
  {
429
+ "seconds": 1.0164730639990012,
430
+ "bytes": 6038957,
431
  "pages": 35,
432
+ "shards": 2
433
  },
434
  {
435
+ "seconds": 0.9485057740021148,
436
+ "bytes": 880447,
437
+ "pages": 26,
438
  "shards": 1
439
  },
440
  {
441
+ "seconds": 1.0207911869983946,
442
+ "bytes": 13578331,
443
+ "pages": 42,
444
+ "shards": 4
 
 
 
 
 
 
 
 
 
 
 
 
445
  },
446
  {
447
+ "seconds": 1.2704945900004532,
448
+ "bytes": 9798343,
449
+ "pages": 31,
450
  "shards": 3
451
  },
452
  {
453
+ "seconds": 0.7868170459987596,
454
+ "bytes": 2512813,
455
+ "pages": 36,
456
  "shards": 2
457
  },
458
  {
459
+ "seconds": 1.0122808699998131,
460
+ "bytes": 6141256,
461
+ "pages": 36,
462
  "shards": 2
463
  },
464
  {
465
+ "seconds": 0.8550289739978325,
466
+ "bytes": 5835711,
467
+ "pages": 28,
468
+ "shards": 2
 
 
 
 
 
 
469
  },
470
  {
471
+ "seconds": 1.051623090999783,
472
+ "bytes": 7071478,
473
+ "pages": 33,
474
+ "shards": 3
475
  },
476
  {
477
+ "seconds": 1.1867464049973933,
478
+ "bytes": 12128568,
479
+ "pages": 40,
480
+ "shards": 3
481
  },
482
  {
483
+ "seconds": 0.8681122560010408,
484
+ "bytes": 1227302,
485
  "pages": 34,
486
  "shards": 1
487
  },
488
  {
489
+ "seconds": 1.205836907000048,
490
+ "bytes": 7335325,
491
+ "pages": 30,
492
  "shards": 3
493
  },
494
  {
495
+ "seconds": 1.085818534000282,
496
+ "bytes": 10112362,
497
+ "pages": 32,
498
+ "shards": 3
499
  },
500
  {
501
+ "seconds": 0.927039637001144,
502
+ "bytes": 2682847,
503
  "pages": 36,
504
  "shards": 2
505
  },
506
  {
507
+ "seconds": 0.8612236650005798,
508
+ "bytes": 6230072,
509
+ "pages": 34,
510
  "shards": 2
511
  },
512
  {
513
+ "seconds": 1.0601522930010105,
514
+ "bytes": 12072218,
515
+ "pages": 37,
516
+ "shards": 3
517
  },
518
  {
519
+ "seconds": 1.126622554998903,
520
+ "bytes": 6643361,
521
+ "pages": 34,
522
+ "shards": 3
523
  },
524
  {
525
+ "seconds": 0.8856843470020976,
526
+ "bytes": 5828359,
527
+ "pages": 27,
528
  "shards": 2
529
  },
530
  {
531
+ "seconds": 1.0851010650003445,
532
+ "bytes": 1256653,
533
+ "pages": 34,
534
+ "shards": 1
535
  },
536
  {
537
+ "seconds": 1.3800626919983188,
538
+ "bytes": 7827681,
539
+ "pages": 39,
540
+ "shards": 3
541
  },
542
  {
543
+ "seconds": 1.5110167620005086,
544
+ "bytes": 10255842,
545
+ "pages": 33,
546
+ "shards": 3
547
  },
548
  {
549
+ "seconds": 8.627843940001185,
550
+ "bytes": 8388512,
551
  "pages": 37,
552
  "shards": 3
553
  },
554
  {
555
+ "seconds": 1.2330220080002618,
556
+ "bytes": 6137294,
557
+ "pages": 36,
558
  "shards": 2
559
+ },
560
+ {
561
+ "seconds": 1.6848288789988146,
562
+ "bytes": 823777,
563
+ "pages": 25,
564
+ "shards": 1
565
  }
566
  ],
567
  "cache_key": {
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:09ea04c8a1fd2002db31adafc765f00a01b0176621104639d3a70a8010a648f4
3
- size 721780607
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e3e3dd01665bea5cf593b73f4417e92f7e07fdb4e1949ec1cfe309333ab911ee
3
+ size 721787647