appvoid commited on
Commit
65f2f5b
·
verified ·
1 Parent(s): 295e04a

Qwen90M step 640000; 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 +249 -249
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 635000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "aee83f47a1033d9e4d5e4cf38362bb2e073804129f1b99ebd3e435b22ca3f85c",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "4a95f31929a892cce3d118cf8d73e75858c30e57cad5a3ec036a4b699b0452b6",
12
  "training_config.json": "6c3c3ccff89dabb5098a22bc207c2f3abaee8b3bffec3f6cf6ad361a63b1a7fb",
13
  "performance.json": "539995c1e4203016bfc6259363445c7c25dfb51604d74cd18d2cc7ad51bfb71e",
14
- "shard_scheduler.json": "b4388fd71b8b8bba9cd013a154900d67eb794933f8eaad21782341991d4d34b5",
15
- "coverage.json": "73ad8690a37dbe8bda15b94f41cc2fef8466db1737bf7d9eba4081b860f31f0c",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "33a459803c8cffa8929a90def998b4ad2c7c52fd1245365a9ee946b5552d845a"
 
1
  {
2
+ "step": 640000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "ebc22606ea875e024664b2b5ba82033f919f8d1a96f22dfebda33a7cb5296c83",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "2e88f84323ca4bdc4a8a0fa1302e353f3a442dcb7967001b5b87a164915124f0",
12
  "training_config.json": "6c3c3ccff89dabb5098a22bc207c2f3abaee8b3bffec3f6cf6ad361a63b1a7fb",
13
  "performance.json": "539995c1e4203016bfc6259363445c7c25dfb51604d74cd18d2cc7ad51bfb71e",
14
+ "shard_scheduler.json": "bf475a77a1c1a0c722213e8028f3d9758aacd623f550f0da5d55fe94219c3d81",
15
+ "coverage.json": "33bc21ae73a3caaa60d3061ffb19116a58b5eaf2262d82489a7714a0458299ba",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "33a459803c8cffa8929a90def998b4ad2c7c52fd1245365a9ee946b5552d845a"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 635000,
3
  "stats": {
4
- "climbmix/blocks": 22860000,
5
- "climbmix/targets": 38215482464,
6
- "climbmix/input_tokens": 38217903896,
7
- "ultra_qa/blocks": 8572500,
8
- "ultra_qa/targets": 13985741427,
9
- "ultra_qa/input_tokens": 13985768433,
10
- "cortex/blocks": 8572500,
11
- "cortex/targets": 325969268,
12
- "cortex/input_tokens": 13626603924,
13
- "ultra_style/blocks": 8572500,
14
- "ultra_style/targets": 13558916090,
15
- "ultra_style/input_tokens": 13559627493,
16
- "openmath/blocks": 5715000,
17
- "openmath/targets": 8830719600,
18
- "openmath/input_tokens": 8833661278,
19
- "rewrite6/blocks": 2857500,
20
- "rewrite6/targets": 4684417110,
21
- "rewrite6/input_tokens": 4689553106
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 3130,
26
- "row": 2106,
27
  "start": {
28
- "file": 151,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 42,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
- "documents": 12822586
38
  },
39
  "rewrite6": {
40
- "page": 31,
41
- "row": 886,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 126976
47
  }
48
  },
49
  "epoch": 35,
50
  "offset": 2048,
51
- "documents": 7317002
52
  },
53
  "openmath": {
54
- "page": 1956,
55
- "row": 1912,
56
  "start": {
57
  "file": 18,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 152,
61
- "row": 76
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 8013688
67
  },
68
  "ultra_qa": {
69
- "page": 846,
70
- "row": 64,
71
  "start": {
72
  "file": 6,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 5,
76
- "row": 24839
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
- "documents": 3465280
82
  },
83
  "ultra_style": {
84
- "page": 1291,
85
- "row": 1369,
86
  "start": {
87
  "file": 7,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
- "row_group": 5,
91
- "row": 66925
92
  }
93
  },
94
  "epoch": 0,
95
- "offset": 2048,
96
- "documents": 5289305
97
  },
98
  "cortex": {
99
- "page": 166257,
100
- "row": 71,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 21280967
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 640000,
3
  "stats": {
4
+ "climbmix/blocks": 23040000,
5
+ "climbmix/targets": 38516782526,
6
+ "climbmix/input_tokens": 38519223105,
7
+ "ultra_qa/blocks": 8640000,
8
+ "ultra_qa/targets": 14095882783,
9
+ "ultra_qa/input_tokens": 14095910016,
10
+ "cortex/blocks": 8640000,
11
+ "cortex/targets": 328530053,
12
+ "cortex/input_tokens": 13733804797,
13
+ "ultra_style/blocks": 8640000,
14
+ "ultra_style/targets": 13665830277,
15
+ "ultra_style/input_tokens": 13666547133,
16
+ "openmath/blocks": 5760000,
17
+ "openmath/targets": 8900224915,
18
+ "openmath/input_tokens": 8903189812,
19
+ "rewrite6/blocks": 2880000,
20
+ "rewrite6/targets": 4721409048,
21
+ "rewrite6/input_tokens": 4726585429
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 3155,
26
+ "row": 394,
27
  "start": {
28
+ "file": 152,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 60,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
+ "documents": 12923274
38
  },
39
  "rewrite6": {
40
+ "page": 45,
41
+ "row": 1043,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 184320
47
  }
48
  },
49
  "epoch": 35,
50
  "offset": 2048,
51
+ "documents": 7374503
52
  },
53
  "openmath": {
54
+ "page": 1971,
55
+ "row": 3641,
56
  "start": {
57
  "file": 18,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 213,
61
+ "row": 516
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 8076857
67
  },
68
  "ultra_qa": {
69
+ "page": 852,
70
+ "row": 2758,
71
  "start": {
72
  "file": 6,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 5,
76
+ "row": 49415
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
+ "documents": 3492550
82
  },
83
  "ultra_style": {
84
+ "page": 1301,
85
+ "row": 1747,
86
  "start": {
87
  "file": 7,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
+ "row_group": 6,
91
+ "row": 23328
92
  }
93
  },
94
  "epoch": 0,
95
+ "offset": 0,
96
+ "documents": 5330643
97
  },
98
  "cortex": {
99
+ "page": 167566,
100
+ "row": 31,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 21448479
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 635000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "aee83f47a1033d9e4d5e4cf38362bb2e073804129f1b99ebd3e435b22ca3f85c", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "4a95f31929a892cce3d118cf8d73e75858c30e57cad5a3ec036a4b699b0452b6", "training_config.json": "6c3c3ccff89dabb5098a22bc207c2f3abaee8b3bffec3f6cf6ad361a63b1a7fb", "performance.json": "539995c1e4203016bfc6259363445c7c25dfb51604d74cd18d2cc7ad51bfb71e", "shard_scheduler.json": "b4388fd71b8b8bba9cd013a154900d67eb794933f8eaad21782341991d4d34b5", "coverage.json": "73ad8690a37dbe8bda15b94f41cc2fef8466db1737bf7d9eba4081b860f31f0c", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "33a459803c8cffa8929a90def998b4ad2c7c52fd1245365a9ee946b5552d845a", "folder": "step-000000635000"}
 
1
+ {"step": 640000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "ebc22606ea875e024664b2b5ba82033f919f8d1a96f22dfebda33a7cb5296c83", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "2e88f84323ca4bdc4a8a0fa1302e353f3a442dcb7967001b5b87a164915124f0", "training_config.json": "6c3c3ccff89dabb5098a22bc207c2f3abaee8b3bffec3f6cf6ad361a63b1a7fb", "performance.json": "539995c1e4203016bfc6259363445c7c25dfb51604d74cd18d2cc7ad51bfb71e", "shard_scheduler.json": "bf475a77a1c1a0c722213e8028f3d9758aacd623f550f0da5d55fe94219c3d81", "coverage.json": "33bc21ae73a3caaa60d3061ffb19116a58b5eaf2262d82489a7714a0458299ba", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "33a459803c8cffa8929a90def998b4ad2c7c52fd1245365a9ee946b5552d845a", "folder": "step-000000640000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:aee83f47a1033d9e4d5e4cf38362bb2e073804129f1b99ebd3e435b22ca3f85c
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ebc22606ea875e024664b2b5ba82033f919f8d1a96f22dfebda33a7cb5296c83
3
  size 360622744
shard_scheduler.json CHANGED
@@ -22,8 +22,8 @@
22
  "tokens_per_second": 598315.063954628,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
- "required_token_positions_per_second": 76226.1169229518,
26
- "required_pages_per_second_upper_estimate": 0.005273499637688854
27
  },
28
  {
29
  "source": "rewrite6",
@@ -36,8 +36,8 @@
36
  "tokens_per_second": 8533.957213384612,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
- "required_token_positions_per_second": 9528.264615368975,
40
- "required_pages_per_second_upper_estimate": 0.003989285421663871
41
  },
42
  {
43
  "source": "openmath",
@@ -50,8 +50,8 @@
50
  "tokens_per_second": 380144.35761751566,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
- "required_token_positions_per_second": 19056.52923073795,
54
- "required_pages_per_second_upper_estimate": 0.003968972484291576
55
  },
56
  {
57
  "source": "ultra_qa",
@@ -64,8 +64,8 @@
64
  "tokens_per_second": 43742.16556775435,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
- "required_token_positions_per_second": 28584.793846106924,
68
- "required_pages_per_second_upper_estimate": 0.0017136147446211187
69
  },
70
  {
71
  "source": "ultra_style",
@@ -78,8 +78,8 @@
78
  "tokens_per_second": 22018.86276523721,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
- "required_token_positions_per_second": 28584.793846106924,
82
- "required_pages_per_second_upper_estimate": 0.002222341990314369
83
  },
84
  {
85
  "source": "cortex",
@@ -92,8 +92,8 @@
92
  "tokens_per_second": 910315.184299418,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
  "estimated_pages_for_buffer": 23,
95
- "required_token_positions_per_second": 28584.793846106924,
96
- "required_pages_per_second_upper_estimate": 0.3666550434974785
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,410 +158,410 @@
158
  "prefetch_batches": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
- "producer_batch_seconds": 0.27300006673436883,
162
- "producer_p95_seconds": 0.3025267820012232,
163
- "gpu_update_seconds": 0.9672275458360701,
164
- "producer_batches_per_second": 3.663002767589093,
165
- "gpu_batches_per_second": 1.033882879271807,
166
- "producer_headroom": 3.5429571772859307,
167
  "sustainable": true,
168
- "buffer_seconds": 60.93533538767242,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 78,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 339322.7889448642,
174
- "upload_bytes_per_second": 5346631.938425806,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 72.81077147731251,
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": 0.9240188730000227,
184
- "bytes": 2599575,
 
 
 
 
 
 
185
  "pages": 36,
186
- "shards": 2
187
  },
188
  {
189
- "seconds": 1.3233367040011217,
190
- "bytes": 5989613,
191
- "pages": 28,
192
- "shards": 2
193
  },
194
  {
195
- "seconds": 1.4566520219996164,
196
- "bytes": 7624893,
197
- "pages": 38,
 
 
 
 
 
 
198
  "shards": 3
199
  },
200
  {
201
- "seconds": 1.1967464800000016,
202
- "bytes": 10837616,
203
- "pages": 33,
204
  "shards": 3
205
  },
206
  {
207
- "seconds": 1.1927219149984012,
208
- "bytes": 6126779,
209
- "pages": 36,
210
  "shards": 2
211
  },
212
  {
213
- "seconds": 1.3350182510002924,
214
- "bytes": 7769901,
215
- "pages": 39,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
216
  "shards": 3
217
  },
218
  {
219
- "seconds": 2.9742520910003805,
220
- "bytes": 5474226,
221
  "pages": 29,
222
  "shards": 2
223
  },
224
  {
225
- "seconds": 1.198957466000138,
226
- "bytes": 7513050,
 
 
 
 
 
 
227
  "pages": 39,
228
  "shards": 3
229
  },
230
  {
231
- "seconds": 3.0025762599980226,
232
- "bytes": 963365,
233
- "pages": 26,
234
  "shards": 1
235
  },
236
  {
237
- "seconds": 1.1665040549996775,
238
- "bytes": 11979568,
239
- "pages": 40,
240
- "shards": 3
241
  },
242
  {
243
- "seconds": 1.0769242890019086,
244
- "bytes": 7547252,
245
- "pages": 30,
246
- "shards": 3
247
  },
248
  {
249
- "seconds": 0.9740304760016443,
250
- "bytes": 5265802,
251
- "pages": 40,
252
  "shards": 2
253
  },
254
  {
255
- "seconds": 1.0207744180006557,
256
- "bytes": 5943379,
257
- "pages": 28,
258
- "shards": 2
259
  },
260
  {
261
- "seconds": 1.2191226570030267,
262
- "bytes": 2638070,
263
- "pages": 37,
264
  "shards": 2
265
  },
266
  {
267
- "seconds": 1.1678417209986947,
268
- "bytes": 10292515,
269
- "pages": 32,
270
  "shards": 3
271
  },
272
  {
273
- "seconds": 0.9083037529999274,
274
- "bytes": 719591,
275
- "pages": 20,
276
  "shards": 1
277
  },
278
  {
279
- "seconds": 1.395833352999034,
280
- "bytes": 13910666,
281
- "pages": 50,
282
- "shards": 4
283
- },
284
- {
285
- "seconds": 0.8840833290014416,
286
- "bytes": 6069784,
287
- "pages": 36,
288
  "shards": 2
289
  },
290
  {
291
- "seconds": 1.1806990879995283,
292
- "bytes": 12168625,
293
- "pages": 35,
294
  "shards": 4
295
  },
296
  {
297
- "seconds": 0.9337092679998023,
298
- "bytes": 919576,
299
- "pages": 27,
300
- "shards": 1
301
- },
302
- {
303
- "seconds": 1.2493900669978757,
304
- "bytes": 6142029,
305
- "pages": 37,
306
  "shards": 2
307
  },
308
  {
309
- "seconds": 0.9163747099992179,
310
- "bytes": 2165474,
311
- "pages": 22,
312
  "shards": 2
313
  },
314
  {
315
- "seconds": 1.0913281839966658,
316
- "bytes": 12193503,
317
- "pages": 46,
318
  "shards": 3
319
  },
320
  {
321
- "seconds": 1.1941587969995453,
322
- "bytes": 6022783,
323
  "pages": 31,
324
- "shards": 3
325
  },
326
  {
327
- "seconds": 0.9699245899973903,
328
- "bytes": 5017188,
329
- "pages": 6,
330
- "shards": 2
331
  },
332
  {
333
- "seconds": 0.8813080820000323,
334
- "bytes": 1258193,
335
  "pages": 35,
336
  "shards": 1
337
  },
338
  {
339
- "seconds": 1.1000836000021081,
340
- "bytes": 6163961,
341
- "pages": 36,
342
  "shards": 3
343
  },
344
  {
345
- "seconds": 1.1141700879998098,
346
- "bytes": 11677393,
347
- "pages": 38,
348
- "shards": 4
349
  },
350
  {
351
- "seconds": 0.971025193000969,
352
- "bytes": 6720640,
353
- "pages": 27,
354
  "shards": 2
355
  },
356
  {
357
- "seconds": 1.2639794309980061,
358
- "bytes": 7601943,
359
- "pages": 40,
360
  "shards": 3
361
  },
362
  {
363
- "seconds": 1.1646632219999447,
364
- "bytes": 931521,
365
- "pages": 28,
366
- "shards": 1
367
- },
368
- {
369
- "seconds": 1.2082039209999493,
370
- "bytes": 10963858,
371
  "pages": 35,
372
- "shards": 3
373
  },
374
  {
375
- "seconds": 1.5337231229968893,
376
- "bytes": 7962330,
377
- "pages": 40,
378
  "shards": 3
379
  },
380
  {
381
- "seconds": 1.0160535649993108,
382
- "bytes": 1074791,
383
  "pages": 31,
384
- "shards": 1
385
- },
386
- {
387
- "seconds": 1.7156042310016346,
388
- "bytes": 13288348,
389
- "pages": 39,
390
- "shards": 4
391
  },
392
  {
393
- "seconds": 1.1277745370025514,
394
- "bytes": 5019078,
395
- "pages": 31,
396
  "shards": 2
397
  },
398
  {
399
- "seconds": 1.113636714999302,
400
- "bytes": 5851090,
401
- "pages": 31,
402
  "shards": 2
403
  },
404
  {
405
- "seconds": 1.4330363720000605,
406
- "bytes": 7577948,
407
- "pages": 40,
408
  "shards": 3
409
  },
410
  {
411
- "seconds": 0.8521174329980568,
412
- "bytes": 865962,
413
- "pages": 27,
414
- "shards": 1
415
- },
416
- {
417
- "seconds": 1.0794645450005191,
418
- "bytes": 10910956,
419
  "pages": 37,
420
- "shards": 3
421
  },
422
  {
423
- "seconds": 1.1706882670005143,
424
- "bytes": 13271387,
425
- "pages": 35,
426
- "shards": 4
427
  },
428
  {
429
- "seconds": 0.8168831449984282,
430
- "bytes": 1231448,
431
- "pages": 36,
432
  "shards": 1
433
  },
434
  {
435
- "seconds": 1.1650056679973204,
436
- "bytes": 7481151,
437
- "pages": 30,
438
- "shards": 3
439
- },
440
- {
441
- "seconds": 0.9575125699993805,
442
- "bytes": 1008516,
443
- "pages": 30,
444
- "shards": 1
445
  },
446
  {
447
- "seconds": 1.5694887619974907,
448
- "bytes": 10308114,
449
- "pages": 44,
450
- "shards": 3
451
  },
452
  {
453
- "seconds": 1.0191816589976952,
454
- "bytes": 7135611,
455
- "pages": 30,
456
- "shards": 3
457
  },
458
  {
459
- "seconds": 1.1397755830003007,
460
- "bytes": 6913587,
461
- "pages": 33,
462
  "shards": 2
463
  },
464
  {
465
- "seconds": 1.4925930749996041,
466
- "bytes": 12021057,
467
- "pages": 40,
468
  "shards": 4
469
  },
470
  {
471
- "seconds": 1.3497725520028325,
472
- "bytes": 6087212,
473
- "pages": 37,
474
  "shards": 2
475
  },
476
  {
477
- "seconds": 0.784925423002278,
478
- "bytes": 818137,
479
- "pages": 26,
480
- "shards": 1
481
  },
482
  {
483
- "seconds": 1.161659340999904,
484
- "bytes": 8019004,
485
- "pages": 40,
486
- "shards": 3
487
  },
488
  {
489
- "seconds": 1.0203383750013018,
490
- "bytes": 5011407,
491
- "pages": 29,
492
  "shards": 2
493
  },
494
  {
495
- "seconds": 1.1040821489987138,
496
- "bytes": 11902249,
497
- "pages": 35,
498
- "shards": 3
499
  },
500
  {
501
- "seconds": 1.1162976039995556,
502
- "bytes": 7875044,
503
- "pages": 39,
504
- "shards": 3
505
  },
506
  {
507
- "seconds": 1.1005420340006822,
508
- "bytes": 1290047,
509
- "pages": 35,
510
- "shards": 1
511
  },
512
  {
513
- "seconds": 1.504874573998677,
514
- "bytes": 12022604,
515
- "pages": 34,
516
  "shards": 4
517
  },
518
  {
519
- "seconds": 1.3143507820022933,
520
- "bytes": 918492,
521
- "pages": 27,
522
  "shards": 2
523
  },
524
  {
525
- "seconds": 1.5125747689999116,
526
- "bytes": 6481526,
527
- "pages": 37,
528
  "shards": 2
529
  },
530
  {
531
- "seconds": 1.7741823629985447,
532
- "bytes": 13400964,
533
- "pages": 38,
 
 
 
 
 
 
534
  "shards": 4
535
  },
536
  {
537
- "seconds": 1.2863249110014294,
538
- "bytes": 4967385,
539
- "pages": 31,
540
- "shards": 2
541
  },
542
  {
543
- "seconds": 1.6688778630013985,
544
- "bytes": 7600081,
545
- "pages": 39,
546
  "shards": 3
547
  },
548
  {
549
- "seconds": 1.2229772380014765,
550
- "bytes": 1028666,
551
- "pages": 30,
552
- "shards": 1
553
  },
554
  {
555
- "seconds": 1.4301322219980648,
556
- "bytes": 6037828,
557
  "pages": 34,
558
  "shards": 2
559
- },
560
- {
561
- "seconds": 1.568642534999526,
562
- "bytes": 11958930,
563
- "pages": 32,
564
- "shards": 4
565
  }
566
  ],
567
  "cache_key": {
 
22
  "tokens_per_second": 598315.063954628,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
+ "required_token_positions_per_second": 77349.77256644783,
26
+ "required_pages_per_second_upper_estimate": 0.00535123674234621
27
  },
28
  {
29
  "source": "rewrite6",
 
36
  "tokens_per_second": 8533.957213384612,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
+ "required_token_positions_per_second": 9668.72157080598,
40
+ "required_pages_per_second_upper_estimate": 0.004048091815830584
41
  },
42
  {
43
  "source": "openmath",
 
50
  "tokens_per_second": 380144.35761751566,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
+ "required_token_positions_per_second": 19337.44314161196,
54
+ "required_pages_per_second_upper_estimate": 0.004027479443728623
55
  },
56
  {
57
  "source": "ultra_qa",
 
64
  "tokens_per_second": 43742.16556775435,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
+ "required_token_positions_per_second": 29006.16471241794,
68
+ "required_pages_per_second_upper_estimate": 0.001738875284660405
69
  },
70
  {
71
  "source": "ultra_style",
 
78
  "tokens_per_second": 22018.86276523721,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
+ "required_token_positions_per_second": 29006.16471241794,
82
+ "required_pages_per_second_upper_estimate": 0.0022551017217554846
83
  },
84
  {
85
  "source": "cortex",
 
92
  "tokens_per_second": 910315.184299418,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
  "estimated_pages_for_buffer": 23,
95
+ "required_token_positions_per_second": 29006.16471241794,
96
+ "required_pages_per_second_upper_estimate": 0.3720599365377296
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
 
158
  "prefetch_batches": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
+ "producer_batch_seconds": 0.10940042345293932,
162
+ "producer_p95_seconds": 0.3257678099980694,
163
+ "gpu_update_seconds": 0.9531766875806063,
164
+ "producer_batches_per_second": 9.140732443601273,
165
+ "gpu_batches_per_second": 1.049123434332246,
166
+ "producer_headroom": 8.712733072652442,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.05013131757819,
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": 425795.8908965641,
174
+ "upload_bytes_per_second": 5026975.921434505,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 91.36588616535848,
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": 1.1706882670005143,
184
+ "bytes": 13271387,
185
+ "pages": 35,
186
+ "shards": 4
187
+ },
188
+ {
189
+ "seconds": 0.8168831449984282,
190
+ "bytes": 1231448,
191
  "pages": 36,
192
+ "shards": 1
193
  },
194
  {
195
+ "seconds": 1.1650056679973204,
196
+ "bytes": 7481151,
197
+ "pages": 30,
198
+ "shards": 3
199
  },
200
  {
201
+ "seconds": 0.9575125699993805,
202
+ "bytes": 1008516,
203
+ "pages": 30,
204
+ "shards": 1
205
+ },
206
+ {
207
+ "seconds": 1.5694887619974907,
208
+ "bytes": 10308114,
209
+ "pages": 44,
210
  "shards": 3
211
  },
212
  {
213
+ "seconds": 1.0191816589976952,
214
+ "bytes": 7135611,
215
+ "pages": 30,
216
  "shards": 3
217
  },
218
  {
219
+ "seconds": 1.1397755830003007,
220
+ "bytes": 6913587,
221
+ "pages": 33,
222
  "shards": 2
223
  },
224
  {
225
+ "seconds": 1.4925930749996041,
226
+ "bytes": 12021057,
227
+ "pages": 40,
228
+ "shards": 4
229
+ },
230
+ {
231
+ "seconds": 1.3497725520028325,
232
+ "bytes": 6087212,
233
+ "pages": 37,
234
+ "shards": 2
235
+ },
236
+ {
237
+ "seconds": 0.784925423002278,
238
+ "bytes": 818137,
239
+ "pages": 26,
240
+ "shards": 1
241
+ },
242
+ {
243
+ "seconds": 1.161659340999904,
244
+ "bytes": 8019004,
245
+ "pages": 40,
246
  "shards": 3
247
  },
248
  {
249
+ "seconds": 1.0203383750013018,
250
+ "bytes": 5011407,
251
  "pages": 29,
252
  "shards": 2
253
  },
254
  {
255
+ "seconds": 1.1040821489987138,
256
+ "bytes": 11902249,
257
+ "pages": 35,
258
+ "shards": 3
259
+ },
260
+ {
261
+ "seconds": 1.1162976039995556,
262
+ "bytes": 7875044,
263
  "pages": 39,
264
  "shards": 3
265
  },
266
  {
267
+ "seconds": 1.1005420340006822,
268
+ "bytes": 1290047,
269
+ "pages": 35,
270
  "shards": 1
271
  },
272
  {
273
+ "seconds": 1.504874573998677,
274
+ "bytes": 12022604,
275
+ "pages": 34,
276
+ "shards": 4
277
  },
278
  {
279
+ "seconds": 1.3143507820022933,
280
+ "bytes": 918492,
281
+ "pages": 27,
282
+ "shards": 2
283
  },
284
  {
285
+ "seconds": 1.5125747689999116,
286
+ "bytes": 6481526,
287
+ "pages": 37,
288
  "shards": 2
289
  },
290
  {
291
+ "seconds": 1.7741823629985447,
292
+ "bytes": 13400964,
293
+ "pages": 38,
294
+ "shards": 4
295
  },
296
  {
297
+ "seconds": 1.2863249110014294,
298
+ "bytes": 4967385,
299
+ "pages": 31,
300
  "shards": 2
301
  },
302
  {
303
+ "seconds": 1.6688778630013985,
304
+ "bytes": 7600081,
305
+ "pages": 39,
306
  "shards": 3
307
  },
308
  {
309
+ "seconds": 1.2229772380014765,
310
+ "bytes": 1028666,
311
+ "pages": 30,
312
  "shards": 1
313
  },
314
  {
315
+ "seconds": 1.4301322219980648,
316
+ "bytes": 6037828,
317
+ "pages": 34,
 
 
 
 
 
 
318
  "shards": 2
319
  },
320
  {
321
+ "seconds": 1.568642534999526,
322
+ "bytes": 11958930,
323
+ "pages": 32,
324
  "shards": 4
325
  },
326
  {
327
+ "seconds": 1.4158421129977796,
328
+ "bytes": 6969172,
329
+ "pages": 34,
 
 
 
 
 
 
330
  "shards": 2
331
  },
332
  {
333
+ "seconds": 1.4889061510002648,
334
+ "bytes": 5773625,
335
+ "pages": 28,
336
  "shards": 2
337
  },
338
  {
339
+ "seconds": 1.7626394689978042,
340
+ "bytes": 7617654,
341
+ "pages": 40,
342
  "shards": 3
343
  },
344
  {
345
+ "seconds": 1.338195986001665,
346
+ "bytes": 5044470,
347
  "pages": 31,
348
+ "shards": 2
349
  },
350
  {
351
+ "seconds": 1.4609920459988643,
352
+ "bytes": 7624505,
353
+ "pages": 32,
354
+ "shards": 3
355
  },
356
  {
357
+ "seconds": 1.1959701189989573,
358
+ "bytes": 1136148,
359
  "pages": 35,
360
  "shards": 1
361
  },
362
  {
363
+ "seconds": 1.5574687859989353,
364
+ "bytes": 11997417,
365
+ "pages": 38,
366
  "shards": 3
367
  },
368
  {
369
+ "seconds": 1.3358469599988894,
370
+ "bytes": 7186422,
371
+ "pages": 30,
372
+ "shards": 3
373
  },
374
  {
375
+ "seconds": 1.527223689998209,
376
+ "bytes": 6243254,
377
+ "pages": 33,
378
  "shards": 2
379
  },
380
  {
381
+ "seconds": 1.4163199819995498,
382
+ "bytes": 7360891,
383
+ "pages": 36,
384
  "shards": 3
385
  },
386
  {
387
+ "seconds": 1.1906762940016051,
388
+ "bytes": 1163380,
 
 
 
 
 
 
389
  "pages": 35,
390
+ "shards": 1
391
  },
392
  {
393
+ "seconds": 1.3281080369997653,
394
+ "bytes": 9814293,
395
+ "pages": 33,
396
  "shards": 3
397
  },
398
  {
399
+ "seconds": 1.3582085800007917,
400
+ "bytes": 8243145,
401
  "pages": 31,
402
+ "shards": 3
 
 
 
 
 
 
403
  },
404
  {
405
+ "seconds": 1.29941092199806,
406
+ "bytes": 6125022,
407
+ "pages": 36,
408
  "shards": 2
409
  },
410
  {
411
+ "seconds": 1.4454559629994037,
412
+ "bytes": 6189473,
413
+ "pages": 37,
414
  "shards": 2
415
  },
416
  {
417
+ "seconds": 1.604715366000164,
418
+ "bytes": 6773206,
419
+ "pages": 29,
420
  "shards": 3
421
  },
422
  {
423
+ "seconds": 1.4528134669999417,
424
+ "bytes": 6484124,
 
 
 
 
 
 
425
  "pages": 37,
426
+ "shards": 2
427
  },
428
  {
429
+ "seconds": 1.400596306000807,
430
+ "bytes": 7658131,
431
+ "pages": 39,
432
+ "shards": 3
433
  },
434
  {
435
+ "seconds": 1.01357181700223,
436
+ "bytes": 644212,
437
+ "pages": 18,
438
  "shards": 1
439
  },
440
  {
441
+ "seconds": 1.5190746830012358,
442
+ "bytes": 16798507,
443
+ "pages": 50,
444
+ "shards": 4
 
 
 
 
 
 
445
  },
446
  {
447
+ "seconds": 1.2172141400005785,
448
+ "bytes": 2428421,
449
+ "pages": 28,
450
+ "shards": 2
451
  },
452
  {
453
+ "seconds": 1.3105249759973958,
454
+ "bytes": 6190634,
455
+ "pages": 37,
456
+ "shards": 2
457
  },
458
  {
459
+ "seconds": 1.3242703979994985,
460
+ "bytes": 5961993,
461
+ "pages": 28,
462
  "shards": 2
463
  },
464
  {
465
+ "seconds": 1.7475975829947856,
466
+ "bytes": 6736460,
467
+ "pages": 41,
468
  "shards": 4
469
  },
470
  {
471
+ "seconds": 1.445422744996904,
472
+ "bytes": 6026824,
473
+ "pages": 21,
474
  "shards": 2
475
  },
476
  {
477
+ "seconds": 1.5578130739959306,
478
+ "bytes": 13753578,
479
+ "pages": 49,
480
+ "shards": 4
481
  },
482
  {
483
+ "seconds": 1.2824578229992767,
484
+ "bytes": 5780729,
485
+ "pages": 32,
486
+ "shards": 2
487
  },
488
  {
489
+ "seconds": 1.419093739001255,
490
+ "bytes": 6340288,
491
+ "pages": 37,
492
  "shards": 2
493
  },
494
  {
495
+ "seconds": 1.3758785059981165,
496
+ "bytes": 2329519,
497
+ "pages": 30,
498
+ "shards": 2
499
  },
500
  {
501
+ "seconds": 1.3759045859987964,
502
+ "bytes": 6265583,
503
+ "pages": 37,
504
+ "shards": 2
505
  },
506
  {
507
+ "seconds": 1.4603665260001435,
508
+ "bytes": 2072280,
509
+ "pages": 20,
510
+ "shards": 2
511
  },
512
  {
513
+ "seconds": 1.5149503700013156,
514
+ "bytes": 16156661,
515
+ "pages": 43,
516
  "shards": 4
517
  },
518
  {
519
+ "seconds": 1.2590785899956245,
520
+ "bytes": 6075199,
521
+ "pages": 38,
522
  "shards": 2
523
  },
524
  {
525
+ "seconds": 1.3119273870033794,
526
+ "bytes": 2422074,
527
+ "pages": 29,
528
  "shards": 2
529
  },
530
  {
531
+ "seconds": 1.4814246340029058,
532
+ "bytes": 5980643,
533
+ "pages": 34,
534
+ "shards": 2
535
+ },
536
+ {
537
+ "seconds": 1.4380262719932944,
538
+ "bytes": 12216092,
539
+ "pages": 40,
540
  "shards": 4
541
  },
542
  {
543
+ "seconds": 1.1287659260051441,
544
+ "bytes": 715029,
545
+ "pages": 21,
546
+ "shards": 1
547
  },
548
  {
549
+ "seconds": 1.51385825300531,
550
+ "bytes": 12055505,
551
+ "pages": 47,
552
  "shards": 3
553
  },
554
  {
555
+ "seconds": 1.379372390001663,
556
+ "bytes": 7458410,
557
+ "pages": 31,
558
+ "shards": 3
559
  },
560
  {
561
+ "seconds": 1.3072389689987176,
562
+ "bytes": 4796842,
563
  "pages": 34,
564
  "shards": 2
 
 
 
 
 
 
565
  }
566
  ],
567
  "cache_key": {
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:4a95f31929a892cce3d118cf8d73e75858c30e57cad5a3ec036a4b699b0452b6
3
- size 721772159
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2e88f84323ca4bdc4a8a0fa1302e353f3a442dcb7967001b5b87a164915124f0
3
+ size 721776191