appvoid commited on
Commit
5394e54
·
verified ·
1 Parent(s): 359c442

Qwen90M step 955000; complete resumable checkpoint

Browse files
Files changed (6) hide show
  1. complete.json +5 -5
  2. coverage.json +51 -51
  3. latest.json +1 -1
  4. model.safetensors +1 -1
  5. shard_scheduler.json +241 -241
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 950000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "2e6800a411ea52928020ac842716553e33b1e7f874b99a56f4a6d2a6bbbab114",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "8ff2ea6bb34408b69ebbd7193690d1ba5ccb5272bc038889553875b472839158",
12
  "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
13
  "performance.json": "a7562f2130d78c752d5494512bcbf6d4c62627eee4ff6943c08f7615044b2de4",
14
- "shard_scheduler.json": "b552fa7d9f27edf138aa08049c5810dc5f2939827c4512222fbf3c1ba1750c42",
15
- "coverage.json": "b37ea7e378a52a99096460a590042a103f46dd59f031c3a7ade67958795c5436",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"
 
1
  {
2
+ "step": 955000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "5efa97481adb165836eab13ec2e1d8440d32bc3f84314194d651cebe4c6e7d73",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "af9b8f970077d4bd0f492458ce4f2991fcef79425d6050931c1082434f1da813",
12
  "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
13
  "performance.json": "a7562f2130d78c752d5494512bcbf6d4c62627eee4ff6943c08f7615044b2de4",
14
+ "shard_scheduler.json": "235d1326e974dcf3b57d55e9bb793f73668d9a67f58d4b3bce1cf5d4eec5e79a",
15
+ "coverage.json": "471d0cc109ff51b50b59b2ac675520cec412cadab32f02434f02213ec1811ae9",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 950000,
3
  "stats": {
4
- "climbmix/blocks": 34200000,
5
- "climbmix/targets": 57171948898,
6
- "climbmix/input_tokens": 57175571843,
7
- "ultra_qa/blocks": 12825000,
8
- "ultra_qa/targets": 20922958426,
9
- "ultra_qa/input_tokens": 20922998603,
10
- "cortex/blocks": 12825000,
11
- "cortex/targets": 487674525,
12
- "cortex/input_tokens": 20386543238,
13
- "ultra_style/blocks": 12825000,
14
- "ultra_style/targets": 20285002197,
15
- "ultra_style/input_tokens": 20286064763,
16
- "openmath/blocks": 8550000,
17
- "openmath/targets": 13211092129,
18
- "openmath/input_tokens": 13215494259,
19
- "rewrite6/blocks": 4275000,
20
- "rewrite6/targets": 7008276799,
21
- "rewrite6/input_tokens": 7015960317
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 4683,
26
- "row": 3857,
27
  "start": {
28
- "file": 226,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 57,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
- "offset": 8192,
37
- "documents": 19185425
38
  },
39
  "rewrite6": {
40
- "page": 14,
41
- "row": 2553,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 57344
47
  }
48
  },
49
  "epoch": 53,
50
- "offset": 0,
51
- "documents": 10946309
52
  },
53
  "openmath": {
54
- "page": 2927,
55
- "row": 1756,
56
  "start": {
57
  "file": 27,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 199,
61
- "row": 446
62
  }
63
  },
64
  "epoch": 0,
65
- "offset": 0,
66
- "documents": 11990748
67
  },
68
  "ultra_qa": {
69
- "page": 1265,
70
- "row": 2680,
71
  "start": {
72
- "file": 9,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
- "row_group": 7,
76
- "row": 52881
77
  }
78
  },
79
  "epoch": 0,
80
- "offset": 4096,
81
- "documents": 5184120
82
  },
83
  "ultra_style": {
84
- "page": 1931,
85
- "row": 1294,
86
  "start": {
87
  "file": 11,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 4,
91
- "row": 35497
92
  }
93
  },
94
  "epoch": 0,
95
- "offset": 0,
96
- "documents": 7910670
97
  },
98
  "cortex": {
99
- "page": 248725,
100
- "row": 90,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 31836890
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 955000,
3
  "stats": {
4
+ "climbmix/blocks": 34380000,
5
+ "climbmix/targets": 57473020907,
6
+ "climbmix/input_tokens": 57476662882,
7
+ "ultra_qa/blocks": 12892500,
8
+ "ultra_qa/targets": 21033125499,
9
+ "ultra_qa/input_tokens": 21033165912,
10
+ "cortex/blocks": 12892500,
11
+ "cortex/targets": 490243480,
12
+ "cortex/input_tokens": 20493820360,
13
+ "ultra_style/blocks": 12892500,
14
+ "ultra_style/targets": 20391585022,
15
+ "ultra_style/input_tokens": 20392653584,
16
+ "openmath/blocks": 8595000,
17
+ "openmath/targets": 13280438962,
18
+ "openmath/input_tokens": 13284864104,
19
+ "rewrite6/blocks": 4297500,
20
+ "rewrite6/targets": 7045113171,
21
+ "rewrite6/input_tokens": 7052837408
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 4708,
26
+ "row": 2416,
27
  "start": {
28
+ "file": 227,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 75,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
+ "offset": 0,
37
+ "documents": 19286384
38
  },
39
  "rewrite6": {
40
+ "page": 28,
41
+ "row": 3208,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 114688
47
  }
48
  },
49
  "epoch": 53,
50
+ "offset": 2048,
51
+ "documents": 11004308
52
  },
53
  "openmath": {
54
+ "page": 2942,
55
+ "row": 3402,
56
  "start": {
57
  "file": 27,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 260,
61
+ "row": 886
62
  }
63
  },
64
  "epoch": 0,
65
+ "offset": 2048,
66
+ "documents": 12053834
67
  },
68
  "ultra_qa": {
69
+ "page": 1272,
70
+ "row": 1325,
71
  "start": {
72
+ "file": 10,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
+ "row_group": 0,
76
+ "row": 13836
77
  }
78
  },
79
  "epoch": 0,
80
+ "offset": 2048,
81
+ "documents": 5211437
82
  },
83
  "ultra_style": {
84
+ "page": 1941,
85
+ "row": 2726,
86
  "start": {
87
  "file": 11,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 4,
91
+ "row": 76457
92
  }
93
  },
94
  "epoch": 0,
95
+ "offset": 4096,
96
+ "documents": 7953062
97
  },
98
  "cortex": {
99
+ "page": 250033,
100
+ "row": 97,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 32004321
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 950000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "2e6800a411ea52928020ac842716553e33b1e7f874b99a56f4a6d2a6bbbab114", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "8ff2ea6bb34408b69ebbd7193690d1ba5ccb5272bc038889553875b472839158", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "a7562f2130d78c752d5494512bcbf6d4c62627eee4ff6943c08f7615044b2de4", "shard_scheduler.json": "b552fa7d9f27edf138aa08049c5810dc5f2939827c4512222fbf3c1ba1750c42", "coverage.json": "b37ea7e378a52a99096460a590042a103f46dd59f031c3a7ade67958795c5436", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000950000"}
 
1
+ {"step": 955000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "5efa97481adb165836eab13ec2e1d8440d32bc3f84314194d651cebe4c6e7d73", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "af9b8f970077d4bd0f492458ce4f2991fcef79425d6050931c1082434f1da813", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "a7562f2130d78c752d5494512bcbf6d4c62627eee4ff6943c08f7615044b2de4", "shard_scheduler.json": "235d1326e974dcf3b57d55e9bb793f73668d9a67f58d4b3bce1cf5d4eec5e79a", "coverage.json": "471d0cc109ff51b50b59b2ac675520cec412cadab32f02434f02213ec1811ae9", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000955000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:2e6800a411ea52928020ac842716553e33b1e7f874b99a56f4a6d2a6bbbab114
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5efa97481adb165836eab13ec2e1d8440d32bc3f84314194d651cebe4c6e7d73
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": 76668.71420286849,
26
- "required_pages_per_second_upper_estimate": 0.005304119544480668
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": 9583.589275358561,
40
- "required_pages_per_second_upper_estimate": 0.004012448701491235
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": 19167.178550717123,
54
- "required_pages_per_second_upper_estimate": 0.003992017819624442
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": 28750.767826075684,
68
- "required_pages_per_second_upper_estimate": 0.0017235646312926524
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": 28750.767826075684,
82
- "required_pages_per_second_upper_estimate": 0.0022352457372145557
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": 28750.767826075684,
96
- "required_pages_per_second_upper_estimate": 0.3687839795035426
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": 0.10526186014067207,
162
- "producer_p95_seconds": 0.27886092599874246,
163
- "gpu_update_seconds": 0.9616438826000493,
164
- "producer_batches_per_second": 9.500117123748325,
165
- "gpu_batches_per_second": 1.039885989079705,
166
- "producer_headroom": 9.135729516036552,
167
  "sustainable": true,
168
- "buffer_seconds": 60.583564603803104,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 82,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 357854.62686666014,
174
- "upload_bytes_per_second": 4807959.195163555,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 76.78727249622204,
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": 1.4908248419997108,
184
- "bytes": 6812358,
185
  "pages": 30,
186
- "shards": 2
187
- },
188
- {
189
- "seconds": 1.521788887999719,
190
- "bytes": 7493690,
191
- "pages": 39,
192
  "shards": 3
193
  },
194
  {
195
- "seconds": 1.3138526910006476,
196
- "bytes": 6032843,
197
- "pages": 36,
198
  "shards": 2
199
  },
200
  {
201
- "seconds": 1.2020050520004588,
202
- "bytes": 6391888,
203
- "pages": 29,
204
- "shards": 3
205
- },
206
- {
207
- "seconds": 1.6235010039999906,
208
- "bytes": 6476306,
209
- "pages": 37,
210
  "shards": 2
211
  },
212
  {
213
- "seconds": 3.9122833619985613,
214
- "bytes": 6289054,
215
- "pages": 37,
216
- "shards": 2
217
  },
218
  {
219
- "seconds": 1.8021375259995693,
220
- "bytes": 8272426,
221
  "pages": 35,
222
- "shards": 3
223
  },
224
  {
225
- "seconds": 1.257975387001352,
226
- "bytes": 10890364,
227
  "pages": 29,
228
  "shards": 3
229
  },
230
  {
231
- "seconds": 1.6343959760015423,
232
- "bytes": 2597952,
233
- "pages": 37,
234
- "shards": 2
235
- },
236
- {
237
- "seconds": 1.3388132420004695,
238
- "bytes": 5864353,
239
- "pages": 35,
240
  "shards": 2
241
  },
242
  {
243
- "seconds": 1.5614973880001344,
244
- "bytes": 5966551,
245
- "pages": 30,
246
  "shards": 2
247
  },
248
  {
249
- "seconds": 3.168707209999411,
250
- "bytes": 6305368,
251
- "pages": 39,
252
- "shards": 4
253
- },
254
- {
255
- "seconds": 1.9134112159990764,
256
- "bytes": 11699377,
257
- "pages": 28,
258
  "shards": 3
259
  },
260
  {
261
- "seconds": 0.9410961429966846,
262
- "bytes": 1236103,
263
- "pages": 35,
264
- "shards": 1
265
  },
266
  {
267
- "seconds": 1.262457610002457,
268
- "bytes": 7619870,
269
- "pages": 38,
270
- "shards": 3
271
  },
272
  {
273
- "seconds": 1.496031877999485,
274
- "bytes": 10106377,
275
- "pages": 30,
276
  "shards": 3
277
  },
278
  {
279
- "seconds": 1.0667447009982425,
280
- "bytes": 2533809,
281
- "pages": 37,
282
- "shards": 2
283
  },
284
  {
285
- "seconds": 1.2533954940008698,
286
- "bytes": 6219875,
287
- "pages": 36,
288
- "shards": 2
289
  },
290
  {
291
- "seconds": 1.2242291910006315,
292
- "bytes": 6745128,
293
- "pages": 29,
294
- "shards": 2
295
  },
296
  {
297
- "seconds": 1.3833632300011232,
298
- "bytes": 11874470,
299
- "pages": 36,
300
- "shards": 4
301
  },
302
  {
303
- "seconds": 1.439283288000297,
304
- "bytes": 6018491,
305
- "pages": 36,
306
  "shards": 2
307
  },
308
  {
309
- "seconds": 0.925738228001137,
310
- "bytes": 863516,
311
- "pages": 29,
312
  "shards": 1
313
  },
314
  {
315
- "seconds": 1.338383825001074,
316
- "bytes": 7489532,
317
- "pages": 36,
318
  "shards": 3
319
  },
320
  {
321
- "seconds": 1.0948180369996408,
322
- "bytes": 10021223,
323
- "pages": 34,
324
  "shards": 3
325
  },
326
  {
327
- "seconds": 1.3469298159980099,
328
- "bytes": 8271893,
329
- "pages": 34,
330
  "shards": 3
331
  },
332
  {
333
- "seconds": 1.2550481799989939,
334
- "bytes": 6009424,
335
- "pages": 28,
336
  "shards": 2
337
  },
338
  {
339
- "seconds": 0.9456898729986278,
340
- "bytes": 975770,
341
- "pages": 34,
342
- "shards": 1
343
- },
344
- {
345
- "seconds": 1.7259603879974748,
346
- "bytes": 12302933,
347
  "pages": 37,
348
- "shards": 4
349
  },
350
  {
351
- "seconds": 1.298020391001046,
352
- "bytes": 6312632,
353
- "pages": 38,
354
- "shards": 2
355
  },
356
  {
357
- "seconds": 1.6241299210014404,
358
- "bytes": 2390001,
359
  "pages": 28,
360
  "shards": 2
361
  },
362
  {
363
- "seconds": 1.5602491789977648,
364
- "bytes": 11841247,
365
- "pages": 32,
 
 
 
 
 
 
366
  "shards": 3
367
  },
368
  {
369
- "seconds": 1.3097548330006248,
370
- "bytes": 6099339,
371
- "pages": 40,
372
- "shards": 2
373
  },
374
  {
375
- "seconds": 1.2393856140006392,
376
- "bytes": 7638444,
377
- "pages": 39,
378
  "shards": 3
379
  },
380
  {
381
- "seconds": 1.58254548900004,
382
- "bytes": 5975386,
383
- "pages": 28,
384
  "shards": 2
385
  },
386
  {
387
- "seconds": 0.9414393419974658,
388
- "bytes": 1125185,
389
- "pages": 34,
390
  "shards": 1
391
  },
392
  {
393
- "seconds": 1.356874795001204,
394
- "bytes": 11516865,
395
  "pages": 35,
396
  "shards": 4
397
  },
398
  {
399
- "seconds": 0.9571627120021731,
400
- "bytes": 762513,
401
- "pages": 23,
402
  "shards": 1
403
  },
404
  {
405
- "seconds": 1.4426862469990738,
406
- "bytes": 13691203,
407
- "pages": 49,
408
  "shards": 4
409
  },
410
  {
411
- "seconds": 2.0410720630025025,
412
- "bytes": 6120614,
413
- "pages": 34,
414
- "shards": 2
415
  },
416
  {
417
- "seconds": 1.370815718000813,
418
- "bytes": 5076360,
419
- "pages": 34,
420
  "shards": 2
421
  },
422
  {
423
- "seconds": 1.1959222149998823,
424
- "bytes": 7412774,
425
- "pages": 30,
426
  "shards": 3
427
  },
428
  {
429
- "seconds": 1.488441301000421,
430
- "bytes": 6011041,
431
- "pages": 37,
432
- "shards": 2
433
  },
434
  {
435
- "seconds": 1.0648157450013969,
436
- "bytes": 668578,
437
- "pages": 20,
438
- "shards": 2
439
  },
440
  {
441
- "seconds": 1.9570103289988765,
442
- "bytes": 18490744,
443
- "pages": 47,
444
- "shards": 5
445
  },
446
  {
447
- "seconds": 0.8552882639996824,
448
- "bytes": 1175820,
449
  "pages": 35,
450
  "shards": 1
451
  },
452
  {
453
- "seconds": 1.1064946840015182,
454
- "bytes": 7293995,
455
- "pages": 29,
456
  "shards": 3
457
  },
458
  {
459
- "seconds": 1.736566771996877,
460
- "bytes": 6191401,
461
- "pages": 38,
462
  "shards": 2
463
  },
464
  {
465
- "seconds": 1.1331273229989165,
466
- "bytes": 4688984,
467
- "pages": 31,
468
  "shards": 2
469
  },
470
  {
471
- "seconds": 1.3685046969985706,
472
- "bytes": 7267431,
473
- "pages": 27,
474
- "shards": 3
475
  },
476
  {
477
- "seconds": 1.4039977110005566,
478
- "bytes": 12182760,
479
- "pages": 48,
480
- "shards": 3
481
  },
482
  {
483
- "seconds": 1.015954297999997,
484
- "bytes": 2619113,
485
- "pages": 37,
486
- "shards": 2
487
  },
488
  {
489
- "seconds": 1.3503473790005955,
490
- "bytes": 10477535,
491
- "pages": 34,
492
  "shards": 3
493
  },
494
  {
495
- "seconds": 0.9818654609989608,
496
- "bytes": 963058,
497
- "pages": 27,
498
  "shards": 1
499
  },
500
  {
501
- "seconds": 1.1789819430014177,
502
- "bytes": 7759537,
503
- "pages": 39,
504
  "shards": 3
505
  },
506
  {
507
- "seconds": 1.2661478619993431,
508
- "bytes": 11756535,
509
- "pages": 35,
 
 
 
 
 
 
510
  "shards": 3
511
  },
512
  {
513
- "seconds": 1.2254978609998943,
514
- "bytes": 7186634,
515
  "pages": 35,
516
  "shards": 3
517
  },
518
  {
519
- "seconds": 1.2788624959976005,
520
- "bytes": 6210071,
521
  "pages": 38,
522
  "shards": 2
523
  },
524
  {
525
- "seconds": 1.412427707000461,
526
- "bytes": 887856,
527
- "pages": 26,
528
  "shards": 1
529
  },
530
  {
531
- "seconds": 1.2455697520017566,
532
- "bytes": 7910132,
533
- "pages": 40,
 
 
 
 
 
 
 
 
 
 
 
 
534
  "shards": 3
535
  },
536
  {
537
- "seconds": 1.3195631069975207,
538
- "bytes": 10098822,
539
- "pages": 36,
540
  "shards": 3
541
  },
542
  {
543
- "seconds": 1.6743710149967228,
544
- "bytes": 8331470,
545
  "pages": 33,
 
 
 
 
 
 
546
  "shards": 3
547
  },
548
  {
549
- "seconds": 1.060381152001355,
550
- "bytes": 6181629,
551
- "pages": 30,
552
- "shards": 2
553
  },
554
  {
555
- "seconds": 1.3323454749988741,
556
- "bytes": 6231984,
557
- "pages": 37,
558
- "shards": 2
559
  },
560
  {
561
- "seconds": 1.3298568760001217,
562
- "bytes": 7493698,
563
- "pages": 36,
564
- "shards": 3
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": 76445.90440437887,
26
+ "required_pages_per_second_upper_estimate": 0.005288705045631197
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": 9555.73805054736,
40
+ "required_pages_per_second_upper_estimate": 0.0040007879752624945
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": 19111.47610109472,
54
+ "required_pages_per_second_upper_estimate": 0.003980416468340475
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": 28667.214151642078,
68
+ "required_pages_per_second_upper_estimate": 0.0017185557160894315
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": 28667.214151642078,
82
+ "required_pages_per_second_upper_estimate": 0.002228749806541117
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": 28667.214151642078,
96
+ "required_pages_per_second_upper_estimate": 0.36771224268085423
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": 0.10924110078121885,
162
+ "producer_p95_seconds": 0.32970197800023016,
163
+ "gpu_update_seconds": 0.9644466969740868,
164
+ "producer_batches_per_second": 9.154063743853484,
165
+ "gpu_batches_per_second": 1.0368639377764062,
166
+ "producer_headroom": 8.828606541649735,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.76014190936747,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 79,
172
  "max_pending_pages": 4096,
173
+ "generation_bytes_per_second": 340180.5052731563,
174
+ "upload_bytes_per_second": 5134244.0499124015,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 72.99481743474976,
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": 1.1959222149998823,
184
+ "bytes": 7412774,
185
  "pages": 30,
 
 
 
 
 
 
186
  "shards": 3
187
  },
188
  {
189
+ "seconds": 1.488441301000421,
190
+ "bytes": 6011041,
191
+ "pages": 37,
192
  "shards": 2
193
  },
194
  {
195
+ "seconds": 1.0648157450013969,
196
+ "bytes": 668578,
197
+ "pages": 20,
 
 
 
 
 
 
198
  "shards": 2
199
  },
200
  {
201
+ "seconds": 1.9570103289988765,
202
+ "bytes": 18490744,
203
+ "pages": 47,
204
+ "shards": 5
205
  },
206
  {
207
+ "seconds": 0.8552882639996824,
208
+ "bytes": 1175820,
209
  "pages": 35,
210
+ "shards": 1
211
  },
212
  {
213
+ "seconds": 1.1064946840015182,
214
+ "bytes": 7293995,
215
  "pages": 29,
216
  "shards": 3
217
  },
218
  {
219
+ "seconds": 1.736566771996877,
220
+ "bytes": 6191401,
221
+ "pages": 38,
 
 
 
 
 
 
222
  "shards": 2
223
  },
224
  {
225
+ "seconds": 1.1331273229989165,
226
+ "bytes": 4688984,
227
+ "pages": 31,
228
  "shards": 2
229
  },
230
  {
231
+ "seconds": 1.3685046969985706,
232
+ "bytes": 7267431,
233
+ "pages": 27,
 
 
 
 
 
 
234
  "shards": 3
235
  },
236
  {
237
+ "seconds": 1.4039977110005566,
238
+ "bytes": 12182760,
239
+ "pages": 48,
240
+ "shards": 3
241
  },
242
  {
243
+ "seconds": 1.015954297999997,
244
+ "bytes": 2619113,
245
+ "pages": 37,
246
+ "shards": 2
247
  },
248
  {
249
+ "seconds": 1.3503473790005955,
250
+ "bytes": 10477535,
251
+ "pages": 34,
252
  "shards": 3
253
  },
254
  {
255
+ "seconds": 0.9818654609989608,
256
+ "bytes": 963058,
257
+ "pages": 27,
258
+ "shards": 1
259
  },
260
  {
261
+ "seconds": 1.1789819430014177,
262
+ "bytes": 7759537,
263
+ "pages": 39,
264
+ "shards": 3
265
  },
266
  {
267
+ "seconds": 1.2661478619993431,
268
+ "bytes": 11756535,
269
+ "pages": 35,
270
+ "shards": 3
271
  },
272
  {
273
+ "seconds": 1.2254978609998943,
274
+ "bytes": 7186634,
275
+ "pages": 35,
276
+ "shards": 3
277
  },
278
  {
279
+ "seconds": 1.2788624959976005,
280
+ "bytes": 6210071,
281
+ "pages": 38,
282
  "shards": 2
283
  },
284
  {
285
+ "seconds": 1.412427707000461,
286
+ "bytes": 887856,
287
+ "pages": 26,
288
  "shards": 1
289
  },
290
  {
291
+ "seconds": 1.2455697520017566,
292
+ "bytes": 7910132,
293
+ "pages": 40,
294
  "shards": 3
295
  },
296
  {
297
+ "seconds": 1.3195631069975207,
298
+ "bytes": 10098822,
299
+ "pages": 36,
300
  "shards": 3
301
  },
302
  {
303
+ "seconds": 1.6743710149967228,
304
+ "bytes": 8331470,
305
+ "pages": 33,
306
  "shards": 3
307
  },
308
  {
309
+ "seconds": 1.060381152001355,
310
+ "bytes": 6181629,
311
+ "pages": 30,
312
  "shards": 2
313
  },
314
  {
315
+ "seconds": 1.3323454749988741,
316
+ "bytes": 6231984,
 
 
 
 
 
 
317
  "pages": 37,
318
+ "shards": 2
319
  },
320
  {
321
+ "seconds": 1.3298568760001217,
322
+ "bytes": 7493698,
323
+ "pages": 36,
324
+ "shards": 3
325
  },
326
  {
327
+ "seconds": 1.2196332230014377,
328
+ "bytes": 5954986,
329
  "pages": 28,
330
  "shards": 2
331
  },
332
  {
333
+ "seconds": 1.0563156829994114,
334
+ "bytes": 932363,
335
+ "pages": 27,
336
+ "shards": 1
337
+ },
338
+ {
339
+ "seconds": 1.3606180970018613,
340
+ "bytes": 7382890,
341
+ "pages": 28,
342
  "shards": 3
343
  },
344
  {
345
+ "seconds": 1.5418479689978994,
346
+ "bytes": 10959545,
347
+ "pages": 43,
348
+ "shards": 3
349
  },
350
  {
351
+ "seconds": 1.2728104929992696,
352
+ "bytes": 7484326,
353
+ "pages": 33,
354
  "shards": 3
355
  },
356
  {
357
+ "seconds": 1.4130519919999642,
358
+ "bytes": 6156286,
359
+ "pages": 35,
360
  "shards": 2
361
  },
362
  {
363
+ "seconds": 0.9332762750018446,
364
+ "bytes": 1158586,
365
+ "pages": 35,
366
  "shards": 1
367
  },
368
  {
369
+ "seconds": 1.577455007001845,
370
+ "bytes": 11622766,
371
  "pages": 35,
372
  "shards": 4
373
  },
374
  {
375
+ "seconds": 1.0520612039981643,
376
+ "bytes": 811209,
377
+ "pages": 24,
378
  "shards": 1
379
  },
380
  {
381
+ "seconds": 1.4460338999997475,
382
+ "bytes": 11954296,
383
+ "pages": 39,
384
  "shards": 4
385
  },
386
  {
387
+ "seconds": 1.3092962549999356,
388
+ "bytes": 7432740,
389
+ "pages": 31,
390
+ "shards": 3
391
  },
392
  {
393
+ "seconds": 1.1428929149988107,
394
+ "bytes": 5656732,
395
+ "pages": 36,
396
  "shards": 2
397
  },
398
  {
399
+ "seconds": 1.3837304669978039,
400
+ "bytes": 7550021,
401
+ "pages": 39,
402
  "shards": 3
403
  },
404
  {
405
+ "seconds": 1.4462293939977826,
406
+ "bytes": 810671,
407
+ "pages": 26,
408
+ "shards": 1
409
  },
410
  {
411
+ "seconds": 1.2543598109987215,
412
+ "bytes": 12226155,
413
+ "pages": 42,
414
+ "shards": 3
415
  },
416
  {
417
+ "seconds": 1.268981666999025,
418
+ "bytes": 11497273,
419
+ "pages": 33,
420
+ "shards": 4
421
  },
422
  {
423
+ "seconds": 1.0184783190015878,
424
+ "bytes": 1341723,
425
  "pages": 35,
426
  "shards": 1
427
  },
428
  {
429
+ "seconds": 1.3343794770007662,
430
+ "bytes": 7400925,
431
+ "pages": 31,
432
  "shards": 3
433
  },
434
  {
435
+ "seconds": 1.8426530299984734,
436
+ "bytes": 5834868,
437
+ "pages": 33,
438
  "shards": 2
439
  },
440
  {
441
+ "seconds": 1.0089386899999226,
442
+ "bytes": 4885086,
443
+ "pages": 33,
444
  "shards": 2
445
  },
446
  {
447
+ "seconds": 1.4905629579989181,
448
+ "bytes": 13725537,
449
+ "pages": 44,
450
+ "shards": 4
451
  },
452
  {
453
+ "seconds": 1.3237662440005806,
454
+ "bytes": 894969,
455
+ "pages": 26,
456
+ "shards": 1
457
  },
458
  {
459
+ "seconds": 1.5763037539982179,
460
+ "bytes": 10877110,
461
+ "pages": 36,
462
+ "shards": 3
463
  },
464
  {
465
+ "seconds": 1.3618358630010334,
466
+ "bytes": 7894203,
467
+ "pages": 37,
468
  "shards": 3
469
  },
470
  {
471
+ "seconds": 1.2813600879999285,
472
+ "bytes": 1187408,
473
+ "pages": 35,
474
  "shards": 1
475
  },
476
  {
477
+ "seconds": 1.2242876330019499,
478
+ "bytes": 7412020,
479
+ "pages": 31,
480
  "shards": 3
481
  },
482
  {
483
+ "seconds": 1.1913529729972652,
484
+ "bytes": 5109029,
485
+ "pages": 25,
486
+ "shards": 2
487
+ },
488
+ {
489
+ "seconds": 1.8349382990018057,
490
+ "bytes": 11979278,
491
+ "pages": 43,
492
  "shards": 3
493
  },
494
  {
495
+ "seconds": 1.191631180998229,
496
+ "bytes": 7251240,
497
  "pages": 35,
498
  "shards": 3
499
  },
500
  {
501
+ "seconds": 1.1317712890013354,
502
+ "bytes": 6287146,
503
  "pages": 38,
504
  "shards": 2
505
  },
506
  {
507
+ "seconds": 0.9238542080020125,
508
+ "bytes": 660547,
509
+ "pages": 18,
510
  "shards": 1
511
  },
512
  {
513
+ "seconds": 1.3592490819974046,
514
+ "bytes": 11983959,
515
+ "pages": 49,
516
+ "shards": 4
517
+ },
518
+ {
519
+ "seconds": 1.0659817249979824,
520
+ "bytes": 6806858,
521
+ "pages": 32,
522
+ "shards": 2
523
+ },
524
+ {
525
+ "seconds": 1.3951733760004572,
526
+ "bytes": 7437431,
527
+ "pages": 31,
528
  "shards": 3
529
  },
530
  {
531
+ "seconds": 1.5938844269985566,
532
+ "bytes": 10568768,
533
+ "pages": 34,
534
  "shards": 3
535
  },
536
  {
537
+ "seconds": 2.7108459590017446,
538
+ "bytes": 1157065,
539
  "pages": 33,
540
+ "shards": 1
541
+ },
542
+ {
543
+ "seconds": 1.3873778990018764,
544
+ "bytes": 7922754,
545
+ "pages": 39,
546
  "shards": 3
547
  },
548
  {
549
+ "seconds": 1.5672195150000334,
550
+ "bytes": 991660,
551
+ "pages": 27,
552
+ "shards": 1
553
  },
554
  {
555
+ "seconds": 1.7213281089971133,
556
+ "bytes": 11058082,
557
+ "pages": 32,
558
+ "shards": 4
559
  },
560
  {
561
+ "seconds": 1.1525909759984643,
562
+ "bytes": 12461215,
563
+ "pages": 46,
564
+ "shards": 4
565
  }
566
  ],
567
  "cache_key": {
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:8ff2ea6bb34408b69ebbd7193690d1ba5ccb5272bc038889553875b472839158
3
- size 721758527
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:af9b8f970077d4bd0f492458ce4f2991fcef79425d6050931c1082434f1da813
3
+ size 721766911