appvoid commited on
Commit
7f6d6b5
·
verified ·
1 Parent(s): c045ba5

Qwen90M step 240000; complete resumable checkpoint

Browse files
Files changed (7) hide show
  1. complete.json +6 -6
  2. coverage.json +46 -46
  3. latest.json +1 -1
  4. model.safetensors +1 -1
  5. performance.json +3 -3
  6. shard_scheduler.json +96 -336
  7. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 237163,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "2b20b64846b5ec5f1a1266ad70e66bdcf4161c507eb849f41d4457b347e1fef8",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "3817012a1ce6a6ac292529de34664fef83f2285d26044dba167d3aaa20957a39",
12
  "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
13
- "performance.json": "7f72f0a0e868a4baec06f238f922d0c5d7e397a51386c11ac32acc776d4d3522",
14
- "shard_scheduler.json": "7642e374dd789aa75440578f87a3db0e21df171d6322760efedf97e6eb520c03",
15
- "coverage.json": "70e3c6c517ccfd837f7cb477fd294e1034ff341fbf2983c3135f90eb9c011c05",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 240000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "65898f02c82a8377d9bda1d5b03ff50aeed1638370f2d8e06f763970f456a634",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "93447def639b9eb557ef085795282a30d7afe7ce82e2b6d060081b6e8ae7533d",
12
  "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
13
+ "performance.json": "ac15a69eb79c2ac9ef8406edc12d9ae8abafb561b0f74b54b23d5920ee4fe3f7",
14
+ "shard_scheduler.json": "2c0ed3cddd320bf86cf19e2dddc35b552898c1782e20beeea5fb498238b66008",
15
+ "coverage.json": "aafdacf1fbbe0595d411cba8e2eec7aa9696122beadc4ad6618a4d163528507a",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 237163,
3
  "stats": {
4
- "climbmix/blocks": 8537867,
5
- "climbmix/targets": 14271950160,
6
- "climbmix/input_tokens": 14272854649,
7
- "ultra_qa/blocks": 3201701,
8
- "ultra_qa/targets": 5223297877,
9
- "ultra_qa/input_tokens": 5223307978,
10
- "cortex/blocks": 3201700,
11
- "cortex/targets": 121802303,
12
- "cortex/input_tokens": 5089120155,
13
- "ultra_style/blocks": 3201701,
14
- "ultra_style/targets": 5064464924,
15
- "ultra_style/input_tokens": 5064730515,
16
- "openmath/blocks": 2134467,
17
- "openmath/targets": 3298047515,
18
- "openmath/input_tokens": 3299146987,
19
- "rewrite6/blocks": 1067234,
20
- "rewrite6/targets": 1749558797,
21
- "rewrite6/input_tokens": 1751477016
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 1169,
26
- "row": 1585,
27
  "start": {
28
- "file": 56,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 48,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
- "documents": 4789809
38
  },
39
  "rewrite6": {
40
- "page": 15,
41
- "row": 1043,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 61440
47
  }
48
  },
49
  "epoch": 13,
50
  "offset": 0,
51
- "documents": 2732735
52
  },
53
  "openmath": {
54
- "page": 730,
55
- "row": 4073,
56
  "start": {
57
  "file": 6,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 370,
61
- "row": 180
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 2994153
67
  },
68
  "ultra_qa": {
69
- "page": 315,
70
- "row": 4076,
71
  "start": {
72
  "file": 2,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
- "row_group": 3,
76
- "row": 57255
77
  }
78
  },
79
  "epoch": 0,
80
- "offset": 2048,
81
- "documents": 1294316
82
  },
83
  "ultra_style": {
84
- "page": 482,
85
- "row": 1439,
86
  "start": {
87
  "file": 2,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 7,
91
- "row": 8799
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
- "documents": 1975711
97
  },
98
  "cortex": {
99
- "page": 62090,
100
- "row": 107,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 7947627
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 240000,
3
  "stats": {
4
+ "climbmix/blocks": 8640000,
5
+ "climbmix/targets": 14442418448,
6
+ "climbmix/input_tokens": 14443333774,
7
+ "ultra_qa/blocks": 3240000,
8
+ "ultra_qa/targets": 5285705457,
9
+ "ultra_qa/input_tokens": 5285715652,
10
+ "cortex/blocks": 3240000,
11
+ "cortex/targets": 123256375,
12
+ "cortex/input_tokens": 5149964923,
13
+ "ultra_style/blocks": 3240000,
14
+ "ultra_style/targets": 5125067906,
15
+ "ultra_style/input_tokens": 5125336525,
16
+ "openmath/blocks": 2160000,
17
+ "openmath/targets": 3337488563,
18
+ "openmath/input_tokens": 3338601175,
19
+ "rewrite6/blocks": 1080000,
20
+ "rewrite6/targets": 1770500519,
21
+ "rewrite6/input_tokens": 1772442074
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 1183,
26
+ "row": 1723,
27
  "start": {
28
+ "file": 57,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 21,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
+ "documents": 4847291
38
  },
39
  "rewrite6": {
40
+ "page": 23,
41
+ "row": 1439,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 94208
47
  }
48
  },
49
  "epoch": 13,
50
  "offset": 0,
51
+ "documents": 2765899
52
  },
53
  "openmath": {
54
+ "page": 739,
55
+ "row": 3019,
56
  "start": {
57
  "file": 6,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 407,
61
+ "row": 44
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 3029963
67
  },
68
  "ultra_qa": {
69
+ "page": 319,
70
+ "row": 3147,
71
  "start": {
72
  "file": 2,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
+ "row_group": 4,
76
+ "row": 9056
77
  }
78
  },
79
  "epoch": 0,
80
+ "offset": 4096,
81
+ "documents": 1309771
82
  },
83
  "ultra_style": {
84
+ "page": 488,
85
+ "row": 260,
86
  "start": {
87
  "file": 2,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 7,
91
+ "row": 33375
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
+ "documents": 1999108
97
  },
98
  "cortex": {
99
+ "page": 62833,
100
+ "row": 5,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 8042629
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 237163, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "2b20b64846b5ec5f1a1266ad70e66bdcf4161c507eb849f41d4457b347e1fef8", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "3817012a1ce6a6ac292529de34664fef83f2285d26044dba167d3aaa20957a39", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "7f72f0a0e868a4baec06f238f922d0c5d7e397a51386c11ac32acc776d4d3522", "shard_scheduler.json": "7642e374dd789aa75440578f87a3db0e21df171d6322760efedf97e6eb520c03", "coverage.json": "70e3c6c517ccfd837f7cb477fd294e1034ff341fbf2983c3135f90eb9c011c05", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000237163"}
 
1
+ {"step": 240000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "65898f02c82a8377d9bda1d5b03ff50aeed1638370f2d8e06f763970f456a634", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "93447def639b9eb557ef085795282a30d7afe7ce82e2b6d060081b6e8ae7533d", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "ac15a69eb79c2ac9ef8406edc12d9ae8abafb561b0f74b54b23d5920ee4fe3f7", "shard_scheduler.json": "2c0ed3cddd320bf86cf19e2dddc35b552898c1782e20beeea5fb498238b66008", "coverage.json": "aafdacf1fbbe0595d411cba8e2eec7aa9696122beadc4ad6618a4d163528507a", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000240000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:2b20b64846b5ec5f1a1266ad70e66bdcf4161c507eb849f41d4457b347e1fef8
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:65898f02c82a8377d9bda1d5b03ff50aeed1638370f2d8e06f763970f456a634
3
  size 360622744
performance.json CHANGED
@@ -49,11 +49,11 @@
49
  "selected": {
50
  "microbatch": 86,
51
  "mode": "compiled",
52
- "microbatch_seconds": 0.9016621779999809,
53
  "peak_gb": 85.39913082122803,
54
  "fits_budget": true,
55
- "estimated_update_seconds": 0.9578150019999612,
56
- "input_tokens_per_second": 195337.01678679453
57
  },
58
  "measurements": [
59
  {
 
49
  "selected": {
50
  "microbatch": 86,
51
  "mode": "compiled",
52
+ "microbatch_seconds": 0.8988038599999868,
53
  "peak_gb": 85.39913082122803,
54
  "fits_budget": true,
55
+ "estimated_update_seconds": 0.9552281219999941,
56
+ "input_tokens_per_second": 195958.21495470946
57
  },
58
  "measurements": [
59
  {
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": 75472.3444404108,
26
- "required_pages_per_second_upper_estimate": 0.006541751533186617
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": 9434.04305505135,
40
- "required_pages_per_second_upper_estimate": 0.0037633808261733487
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": 18868.0861101027,
54
- "required_pages_per_second_upper_estimate": 0.004364799654598218
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": 28302.129165154052,
68
- "required_pages_per_second_upper_estimate": 0.0017234511763620484
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": 28302.129165154052,
82
- "required_pages_per_second_upper_estimate": 0.002178776248441411
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": 28302.129165154052,
96
- "required_pages_per_second_upper_estimate": 0.35141772309812946
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -157,410 +157,170 @@
157
  "selected": {
158
  "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
- "batch_memory_estimate_bytes": 688946,
161
- "producer_batch_seconds": 0.12409712750809376,
162
- "producer_p95_seconds": 0.4301016780009377,
163
- "gpu_update_seconds": 0.9768876340950552,
164
- "producer_batches_per_second": 8.05820424759452,
165
- "gpu_batches_per_second": 1.0236591856609538,
166
- "producer_headroom": 7.871960082487336,
167
  "sustainable": true,
168
- "buffer_seconds": 60.56703331389342,
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": 379406.6242124819,
174
- "upload_bytes_per_second": 5575493.562052467,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 81.41182942181437,
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": 1.3212683820020175,
184
- "bytes": 7578218,
185
- "pages": 38,
186
- "shards": 3
187
- },
188
- {
189
- "seconds": 0.9010331350000342,
190
- "bytes": 973266,
191
- "pages": 27,
192
  "shards": 1
193
  },
194
  {
195
- "seconds": 1.851456552001764,
196
- "bytes": 16129409,
197
- "pages": 39,
198
- "shards": 4
199
- },
200
- {
201
- "seconds": 1.138110815998516,
202
- "bytes": 2483869,
203
- "pages": 32,
204
- "shards": 2
205
- },
206
- {
207
- "seconds": 1.5748863410044578,
208
- "bytes": 6345410,
209
- "pages": 37,
210
- "shards": 2
211
- },
212
- {
213
- "seconds": 1.1974535199988168,
214
- "bytes": 6123455,
215
- "pages": 28,
216
- "shards": 2
217
- },
218
- {
219
- "seconds": 1.302649115998065,
220
- "bytes": 6509625,
221
- "pages": 37,
222
- "shards": 3
223
- },
224
- {
225
- "seconds": 1.073786435998045,
226
- "bytes": 5871200,
227
- "pages": 28,
228
- "shards": 2
229
- },
230
- {
231
- "seconds": 0.9513727629964706,
232
- "bytes": 2591189,
233
- "pages": 36,
234
- "shards": 2
235
- },
236
- {
237
- "seconds": 1.2582520649957587,
238
- "bytes": 11695145,
239
- "pages": 30,
240
- "shards": 3
241
- },
242
- {
243
- "seconds": 1.135505132995604,
244
- "bytes": 10769907,
245
- "pages": 36,
246
- "shards": 3
247
- },
248
- {
249
- "seconds": 1.0033940869980142,
250
- "bytes": 2504989,
251
- "pages": 37,
252
- "shards": 2
253
- },
254
- {
255
- "seconds": 1.0786675720009953,
256
- "bytes": 5961901,
257
- "pages": 28,
258
- "shards": 2
259
- },
260
- {
261
- "seconds": 0.8719808540045051,
262
- "bytes": 2734178,
263
- "pages": 38,
264
  "shards": 2
265
  },
266
  {
267
- "seconds": 1.0888003759973799,
268
- "bytes": 5559059,
269
- "pages": 23,
270
- "shards": 2
271
  },
272
  {
273
- "seconds": 1.564358691997768,
274
- "bytes": 15793806,
275
- "pages": 43,
276
  "shards": 4
277
  },
278
  {
279
- "seconds": 0.8641434079981991,
280
- "bytes": 2665706,
281
- "pages": 37,
282
- "shards": 2
283
  },
284
  {
285
- "seconds": 0.9742672610009322,
286
- "bytes": 6080743,
287
- "pages": 28,
288
  "shards": 2
289
  },
290
  {
291
- "seconds": 1.1234689760021865,
292
- "bytes": 10516427,
293
- "pages": 39,
294
  "shards": 3
295
  },
296
  {
297
- "seconds": 1.2163533749990165,
298
- "bytes": 2238236,
299
- "pages": 28,
300
- "shards": 2
301
- },
302
- {
303
- "seconds": 0.9348791060037911,
304
- "bytes": 6426893,
305
- "pages": 37,
306
- "shards": 2
307
- },
308
- {
309
- "seconds": 1.129189448998659,
310
- "bytes": 8174843,
311
- "pages": 30,
312
  "shards": 3
313
  },
314
  {
315
- "seconds": 1.1518778129975544,
316
- "bytes": 10667449,
317
- "pages": 40,
318
  "shards": 3
319
  },
320
  {
321
- "seconds": 1.092219940997893,
322
- "bytes": 846588,
323
- "pages": 27,
324
  "shards": 1
325
  },
326
  {
327
- "seconds": 1.1457775259987102,
328
- "bytes": 7463584,
329
- "pages": 39,
330
- "shards": 3
331
- },
332
- {
333
- "seconds": 0.950148095995246,
334
- "bytes": 5837048,
335
- "pages": 28,
336
  "shards": 2
337
  },
338
  {
339
- "seconds": 0.8334624849958345,
340
- "bytes": 871180,
341
- "pages": 26,
342
- "shards": 1
343
- },
344
- {
345
- "seconds": 1.5179614139997284,
346
- "bytes": 17372933,
347
- "pages": 43,
348
- "shards": 5
349
- },
350
- {
351
- "seconds": 0.9750086150015704,
352
- "bytes": 928101,
353
- "pages": 28,
354
- "shards": 1
355
- },
356
- {
357
- "seconds": 1.0726397860053112,
358
- "bytes": 7787197,
359
- "pages": 38,
360
- "shards": 3
361
- },
362
- {
363
- "seconds": 0.9909355400013737,
364
- "bytes": 5704571,
365
- "pages": 21,
366
  "shards": 3
367
  },
368
  {
369
- "seconds": 1.0439709420024883,
370
- "bytes": 6002810,
371
- "pages": 44,
372
  "shards": 2
373
  },
374
  {
375
- "seconds": 1.7808462090033572,
376
- "bytes": 7399793,
377
- "pages": 30,
378
- "shards": 3
379
- },
380
- {
381
- "seconds": 1.2099171500012744,
382
- "bytes": 12004299,
383
- "pages": 41,
384
  "shards": 3
385
  },
386
  {
387
- "seconds": 0.7972624179965351,
388
- "bytes": 919254,
389
- "pages": 26,
390
- "shards": 1
391
- },
392
- {
393
- "seconds": 1.5076305830007186,
394
- "bytes": 11558434,
395
- "pages": 35,
396
- "shards": 4
397
- },
398
- {
399
- "seconds": 0.8286585169989849,
400
- "bytes": 1181916,
401
  "pages": 35,
402
- "shards": 1
403
- },
404
- {
405
- "seconds": 1.184618612001941,
406
- "bytes": 7236384,
407
- "pages": 30,
408
- "shards": 3
409
- },
410
- {
411
- "seconds": 1.0676331659997231,
412
- "bytes": 5948596,
413
- "pages": 34,
414
  "shards": 2
415
  },
416
  {
417
- "seconds": 1.030731860002561,
418
- "bytes": 4760298,
419
- "pages": 25,
420
- "shards": 2
421
- },
422
- {
423
- "seconds": 1.582079469000746,
424
- "bytes": 13247311,
425
- "pages": 40,
426
- "shards": 4
427
- },
428
- {
429
- "seconds": 1.0703849589990568,
430
- "bytes": 1241496,
431
- "pages": 35,
432
- "shards": 1
433
- },
434
- {
435
- "seconds": 1.07058091199724,
436
- "bytes": 5784118,
437
- "pages": 26,
438
  "shards": 2
439
  },
440
  {
441
- "seconds": 2.4059382019986515,
442
- "bytes": 12510715,
443
- "pages": 47,
444
  "shards": 4
445
  },
446
  {
447
- "seconds": 0.7931236949953018,
448
- "bytes": 1028533,
449
- "pages": 27,
450
  "shards": 1
451
  },
452
  {
453
- "seconds": 2.0690407029978815,
454
- "bytes": 7202195,
455
- "pages": 30,
456
- "shards": 3
457
- },
458
- {
459
- "seconds": 1.457639926004049,
460
- "bytes": 7115441,
461
- "pages": 39,
462
- "shards": 2
463
- },
464
- {
465
- "seconds": 1.29948361699644,
466
- "bytes": 9783282,
467
- "pages": 32,
468
- "shards": 3
469
- },
470
- {
471
- "seconds": 1.0601626389980083,
472
- "bytes": 7547567,
473
- "pages": 29,
474
- "shards": 3
475
- },
476
- {
477
- "seconds": 1.118395095996675,
478
- "bytes": 1248062,
479
  "pages": 35,
480
- "shards": 1
481
- },
482
- {
483
- "seconds": 0.9650491309948848,
484
- "bytes": 5725695,
485
- "pages": 29,
486
  "shards": 2
487
  },
488
  {
489
- "seconds": 1.1484884869933012,
490
- "bytes": 6640918,
491
- "pages": 32,
492
  "shards": 3
493
  },
494
  {
495
- "seconds": 1.1328601960049127,
496
- "bytes": 11867525,
497
  "pages": 34,
498
  "shards": 3
499
  },
500
  {
501
- "seconds": 1.0216524100032984,
502
- "bytes": 7553152,
503
- "pages": 40,
504
- "shards": 3
505
- },
506
- {
507
- "seconds": 0.8235062559979269,
508
- "bytes": 854296,
509
- "pages": 26,
510
- "shards": 1
511
- },
512
- {
513
- "seconds": 1.2072945199979586,
514
- "bytes": 10797475,
515
- "pages": 40,
516
- "shards": 3
517
- },
518
- {
519
- "seconds": 1.3132626469960087,
520
- "bytes": 2377167,
521
- "pages": 28,
522
- "shards": 2
523
- },
524
- {
525
- "seconds": 0.8975952609980595,
526
- "bytes": 5947609,
527
- "pages": 37,
528
- "shards": 2
529
- },
530
- {
531
- "seconds": 1.3013317000004463,
532
- "bytes": 11518633,
533
  "pages": 35,
534
- "shards": 3
535
- },
536
- {
537
- "seconds": 1.1227065580023918,
538
- "bytes": 6254929,
539
- "pages": 31,
540
- "shards": 3
541
- },
542
- {
543
- "seconds": 1.5046920809982112,
544
- "bytes": 5774998,
545
- "pages": 29,
546
  "shards": 2
547
  },
548
  {
549
- "seconds": 0.8562071050037048,
550
- "bytes": 2686484,
551
- "pages": 37,
552
- "shards": 3
553
- },
554
- {
555
- "seconds": 1.2383149749948643,
556
- "bytes": 6204112,
557
- "pages": 30,
558
  "shards": 2
559
  },
560
  {
561
- "seconds": 1.4738188229966909,
562
- "bytes": 10769021,
563
- "pages": 39,
564
  "shards": 3
565
  }
566
  ],
@@ -606,7 +366,7 @@
606
  },
607
  "cache_reused": true,
608
  "resume_probe_seconds": [
609
- 3.044028898000022,
610
- 0.01595121099995822
611
  ]
612
  }
 
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": 75700.32285632046,
26
+ "required_pages_per_second_upper_estimate": 0.006561512124471654
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": 9462.540357040058,
40
+ "required_pages_per_second_upper_estimate": 0.003774748826009278
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": 18925.080714080115,
54
+ "required_pages_per_second_upper_estimate": 0.004377984353157622
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": 28387.62107112017,
68
+ "required_pages_per_second_upper_estimate": 0.0017286571848940193
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": 28387.62107112017,
82
+ "required_pages_per_second_upper_estimate": 0.002185357652019425
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": 28387.62107112017,
96
+ "required_pages_per_second_upper_estimate": 0.3524792464472251
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
 
157
  "selected": {
158
  "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
+ "batch_memory_estimate_bytes": 690670,
161
+ "producer_batch_seconds": 0.2644437009218734,
162
+ "producer_p95_seconds": 0.16535117300009006,
163
+ "gpu_update_seconds": 0.9739456480249902,
164
+ "producer_batches_per_second": 3.781523237323915,
165
+ "gpu_batches_per_second": 1.026751340824659,
166
+ "producer_headroom": 3.6829980998969996,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.38463017754939,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 83,
172
  "max_pending_pages": 4096,
173
+ "generation_bytes_per_second": 516494.2735893789,
174
+ "upload_bytes_per_second": 4500469.795094228,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 110.82764774094606,
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": 0.8557939730000044,
184
+ "bytes": 46834,
185
+ "pages": 1,
 
 
 
 
 
 
186
  "shards": 1
187
  },
188
  {
189
+ "seconds": 1.4821450189999723,
190
+ "bytes": 5748205,
191
+ "pages": 24,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
192
  "shards": 2
193
  },
194
  {
195
+ "seconds": 1.3858465779999278,
196
+ "bytes": 775414,
197
+ "pages": 24,
198
+ "shards": 1
199
  },
200
  {
201
+ "seconds": 1.4328443340000376,
202
+ "bytes": 11483713,
203
+ "pages": 34,
204
  "shards": 4
205
  },
206
  {
207
+ "seconds": 0.909485917999973,
208
+ "bytes": 1076059,
209
+ "pages": 32,
210
+ "shards": 1
211
  },
212
  {
213
+ "seconds": 0.961313237000013,
214
+ "bytes": 6243214,
215
+ "pages": 35,
216
  "shards": 2
217
  },
218
  {
219
+ "seconds": 1.4556228910000755,
220
+ "bytes": 7655201,
221
+ "pages": 35,
222
  "shards": 3
223
  },
224
  {
225
+ "seconds": 1.1608069779999823,
226
+ "bytes": 11166576,
227
+ "pages": 32,
 
 
 
 
 
 
 
 
 
 
 
 
228
  "shards": 3
229
  },
230
  {
231
+ "seconds": 1.0247551580000618,
232
+ "bytes": 7489968,
233
+ "pages": 35,
234
  "shards": 3
235
  },
236
  {
237
+ "seconds": 0.7696208980000847,
238
+ "bytes": 1062863,
239
+ "pages": 33,
240
  "shards": 1
241
  },
242
  {
243
+ "seconds": 1.0704197729999123,
244
+ "bytes": 6206676,
245
+ "pages": 34,
 
 
 
 
 
 
246
  "shards": 2
247
  },
248
  {
249
+ "seconds": 2.8963105739999264,
250
+ "bytes": 7237253,
251
+ "pages": 32,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
252
  "shards": 3
253
  },
254
  {
255
+ "seconds": 1.3083305300001484,
256
+ "bytes": 5580567,
257
+ "pages": 35,
258
  "shards": 2
259
  },
260
  {
261
+ "seconds": 1.1845111140000881,
262
+ "bytes": 11699180,
263
+ "pages": 29,
 
 
 
 
 
 
264
  "shards": 3
265
  },
266
  {
267
+ "seconds": 0.859392975999981,
268
+ "bytes": 2445063,
 
 
 
 
 
 
 
 
 
 
 
 
269
  "pages": 35,
 
 
 
 
 
 
 
 
 
 
 
 
270
  "shards": 2
271
  },
272
  {
273
+ "seconds": 1.0558128229999966,
274
+ "bytes": 6261983,
275
+ "pages": 34,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
276
  "shards": 2
277
  },
278
  {
279
+ "seconds": 1.3157395409998571,
280
+ "bytes": 12201352,
281
+ "pages": 34,
282
  "shards": 4
283
  },
284
  {
285
+ "seconds": 2.8152504160002536,
286
+ "bytes": 1180754,
287
+ "pages": 33,
288
  "shards": 1
289
  },
290
  {
291
+ "seconds": 1.0497820410000713,
292
+ "bytes": 5934890,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
293
  "pages": 35,
 
 
 
 
 
 
294
  "shards": 2
295
  },
296
  {
297
+ "seconds": 1.250543007000033,
298
+ "bytes": 8195692,
299
+ "pages": 31,
300
  "shards": 3
301
  },
302
  {
303
+ "seconds": 1.6225744519997534,
304
+ "bytes": 9437094,
305
  "pages": 34,
306
  "shards": 3
307
  },
308
  {
309
+ "seconds": 1.1101808009998422,
310
+ "bytes": 6059135,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
311
  "pages": 35,
 
 
 
 
 
 
 
 
 
 
 
 
312
  "shards": 2
313
  },
314
  {
315
+ "seconds": 1.0066997960002482,
316
+ "bytes": 2574525,
317
+ "pages": 35,
 
 
 
 
 
 
318
  "shards": 2
319
  },
320
  {
321
+ "seconds": 3.0401188460000412,
322
+ "bytes": 10860861,
323
+ "pages": 35,
324
  "shards": 3
325
  }
326
  ],
 
366
  },
367
  "cache_reused": true,
368
  "resume_probe_seconds": [
369
+ 2.844574209999962,
370
+ 0.2503580399999805
371
  ]
372
  }
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:3817012a1ce6a6ac292529de34664fef83f2285d26044dba167d3aaa20957a39
3
- size 721779903
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:93447def639b9eb557ef085795282a30d7afe7ce82e2b6d060081b6e8ae7533d
3
+ size 721779519