appvoid commited on
Commit
7ae3381
·
verified ·
1 Parent(s): 835f7a7

Qwen90M step 505000; complete resumable checkpoint

Browse files
Files changed (6) hide show
  1. complete.json +5 -5
  2. coverage.json +48 -48
  3. latest.json +1 -1
  4. model.safetensors +1 -1
  5. shard_scheduler.json +239 -239
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 500000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "b32b5405757b9ffe78293bedfdb956dae40e8488a1c520be2cde0c5aed95c84e",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "442d5a63837b6ac40b1f2659fb1a8dca8d13b6c8efafd35b7dd88a3f95501c4e",
12
  "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
13
  "performance.json": "234744063e094dfd7b525da4e181ac493fac400c830f871c7e8c59127ed10dc9",
14
- "shard_scheduler.json": "28346349a8e1d8ef3ce4b1c69a6243a3f826cee646b53fd277baf701c2210e6e",
15
- "coverage.json": "19aa4f8d54332a074a778e21560e24dcd35db2f280ad9c717dbc173725ed6762",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 505000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "e9099c8a8b03dc9572c97e934128304524a8a458a8efa4493e025952c068ceab",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "b8e924ba03c8c90f3a051947fd0eb5994e5b786f87c64a7022243d5bfe8f0021",
12
  "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
13
  "performance.json": "234744063e094dfd7b525da4e181ac493fac400c830f871c7e8c59127ed10dc9",
14
+ "shard_scheduler.json": "58df431fac6f1fc091d15d01fbea31a13160b04a17771a23da2d9dd054165338",
15
+ "coverage.json": "a92539a3090e14fb0565a5f8b4bf84185ba11d618ff984ee9c273425d8c514dd",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 500000,
3
  "stats": {
4
- "climbmix/blocks": 18000000,
5
- "climbmix/targets": 30087292818,
6
- "climbmix/input_tokens": 30089199837,
7
- "ultra_qa/blocks": 6750000,
8
- "ultra_qa/targets": 11011602216,
9
- "ultra_qa/input_tokens": 11011623449,
10
- "cortex/blocks": 6750000,
11
- "cortex/targets": 256714249,
12
- "cortex/input_tokens": 10729951557,
13
- "ultra_style/blocks": 6750000,
14
- "ultra_style/targets": 10676414105,
15
- "ultra_style/input_tokens": 10676973780,
16
- "openmath/blocks": 4500000,
17
- "openmath/targets": 6953343610,
18
- "openmath/input_tokens": 6955660259,
19
- "rewrite6/blocks": 2250000,
20
- "rewrite6/targets": 3688563295,
21
- "rewrite6/input_tokens": 3692607180
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 2465,
26
- "row": 3704,
27
  "start": {
28
- "file": 119,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 22,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
- "offset": 0,
37
- "documents": 10100344
38
  },
39
  "rewrite6": {
40
- "page": 2,
41
- "row": 1645,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 8192
47
  }
48
  },
49
  "epoch": 28,
50
  "offset": 0,
51
- "documents": 5761149
52
  },
53
  "openmath": {
54
- "page": 1540,
55
- "row": 2675,
56
  "start": {
57
  "file": 14,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 194,
61
- "row": 740
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 6310515
67
  },
68
  "ultra_qa": {
69
- "page": 666,
70
- "row": 700,
71
  "start": {
72
  "file": 5,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 2,
76
- "row": 775
77
  }
78
  },
79
  "epoch": 0,
80
- "offset": 4096,
81
- "documents": 2728636
82
  },
83
  "ultra_style": {
84
- "page": 1016,
85
- "row": 2694,
86
  "start": {
87
  "file": 6,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
- "row_group": 0,
91
- "row": 51884
92
  }
93
  },
94
  "epoch": 0,
95
- "offset": 4096,
96
- "documents": 4164230
97
  },
98
  "cortex": {
99
- "page": 130914,
100
- "row": 26,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 16757018
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 505000,
3
  "stats": {
4
+ "climbmix/blocks": 18180000,
5
+ "climbmix/targets": 30388475981,
6
+ "climbmix/input_tokens": 30390402108,
7
+ "ultra_qa/blocks": 6817500,
8
+ "ultra_qa/targets": 11121758760,
9
+ "ultra_qa/input_tokens": 11121780202,
10
+ "cortex/blocks": 6817500,
11
+ "cortex/targets": 259282200,
12
+ "cortex/input_tokens": 10837250165,
13
+ "ultra_style/blocks": 6817500,
14
+ "ultra_style/targets": 10783164447,
15
+ "ultra_style/input_tokens": 10783729590,
16
+ "openmath/blocks": 4545000,
17
+ "openmath/targets": 7022833549,
18
+ "openmath/input_tokens": 7025173354,
19
+ "rewrite6/blocks": 2272500,
20
+ "rewrite6/targets": 3725426609,
21
+ "rewrite6/input_tokens": 3729511162
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 2490,
26
+ "row": 2231,
27
  "start": {
28
+ "file": 120,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 39,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
+ "offset": 10240,
37
+ "documents": 10201271
38
  },
39
  "rewrite6": {
40
+ "page": 16,
41
+ "row": 2127,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 65536
47
  }
48
  },
49
  "epoch": 28,
50
  "offset": 0,
51
+ "documents": 5818975
52
  },
53
  "openmath": {
54
+ "page": 1556,
55
+ "row": 257,
56
  "start": {
57
  "file": 14,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 260,
61
+ "row": 276
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 6373633
67
  },
68
  "ultra_qa": {
69
+ "page": 672,
70
+ "row": 3323,
71
  "start": {
72
  "file": 5,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 2,
76
+ "row": 25351
77
  }
78
  },
79
  "epoch": 0,
80
+ "offset": 0,
81
+ "documents": 2755835
82
  },
83
  "ultra_style": {
84
+ "page": 1026,
85
+ "row": 2979,
86
  "start": {
87
  "file": 6,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
+ "row_group": 1,
91
+ "row": 7627
92
  }
93
  },
94
  "epoch": 0,
95
+ "offset": 0,
96
+ "documents": 4205475
97
  },
98
  "cortex": {
99
+ "page": 132224,
100
+ "row": 58,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 16924730
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 500000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "b32b5405757b9ffe78293bedfdb956dae40e8488a1c520be2cde0c5aed95c84e", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "442d5a63837b6ac40b1f2659fb1a8dca8d13b6c8efafd35b7dd88a3f95501c4e", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "234744063e094dfd7b525da4e181ac493fac400c830f871c7e8c59127ed10dc9", "shard_scheduler.json": "28346349a8e1d8ef3ce4b1c69a6243a3f826cee646b53fd277baf701c2210e6e", "coverage.json": "19aa4f8d54332a074a778e21560e24dcd35db2f280ad9c717dbc173725ed6762", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000500000"}
 
1
+ {"step": 505000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "e9099c8a8b03dc9572c97e934128304524a8a458a8efa4493e025952c068ceab", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "b8e924ba03c8c90f3a051947fd0eb5994e5b786f87c64a7022243d5bfe8f0021", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "234744063e094dfd7b525da4e181ac493fac400c830f871c7e8c59127ed10dc9", "shard_scheduler.json": "58df431fac6f1fc091d15d01fbea31a13160b04a17771a23da2d9dd054165338", "coverage.json": "a92539a3090e14fb0565a5f8b4bf84185ba11d618ff984ee9c273425d8c514dd", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000505000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b32b5405757b9ffe78293bedfdb956dae40e8488a1c520be2cde0c5aed95c84e
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e9099c8a8b03dc9572c97e934128304524a8a458a8efa4493e025952c068ceab
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": 76297.59507963236,
26
- "required_pages_per_second_upper_estimate": 0.005278444662420188
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": 9537.199384954045,
40
- "required_pages_per_second_upper_estimate": 0.0039930262231099335
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": 19074.39876990809,
54
- "required_pages_per_second_upper_estimate": 0.003972694238049278
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": 28611.598154862135,
68
- "required_pages_per_second_upper_estimate": 0.0017152216220031838
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": 28611.598154862135,
82
- "required_pages_per_second_upper_estimate": 0.0022244259074203927
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": 28611.598154862135,
96
- "required_pages_per_second_upper_estimate": 0.36699886038996593
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": 704566,
161
- "producer_batch_seconds": 0.13252937544521615,
162
- "producer_p95_seconds": 0.25005536200114875,
163
- "gpu_update_seconds": 0.9663214144960867,
164
- "producer_batches_per_second": 7.545496963526938,
165
- "gpu_batches_per_second": 1.0348523638187983,
166
- "producer_headroom": 7.291375298871277,
167
  "sustainable": true,
168
- "buffer_seconds": 60.87824911325346,
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": 321426.0121983039,
174
- "upload_bytes_per_second": 5933281.998072939,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 68.970539803141,
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": 1.1922360469998239,
184
- "bytes": 10997251,
185
- "pages": 39,
186
- "shards": 3
187
- },
188
- {
189
- "seconds": 1.0726386229998752,
190
- "bytes": 2419938,
191
- "pages": 28,
192
  "shards": 2
193
  },
194
  {
195
- "seconds": 1.240178467000078,
196
- "bytes": 6103593,
197
- "pages": 36,
198
- "shards": 2
199
  },
200
  {
201
- "seconds": 0.8772162379991642,
202
- "bytes": 2610307,
203
- "pages": 37,
204
  "shards": 2
205
  },
206
  {
207
- "seconds": 1.183507477000603,
208
- "bytes": 15672429,
209
- "pages": 34,
210
- "shards": 4
211
- },
212
- {
213
- "seconds": 1.1264393319997907,
214
- "bytes": 6331952,
215
  "pages": 36,
216
  "shards": 2
217
  },
218
  {
219
- "seconds": 0.856032101000892,
220
- "bytes": 2311911,
221
  "pages": 29,
222
  "shards": 2
223
  },
224
  {
225
- "seconds": 1.5313736190000782,
226
- "bytes": 6009990,
227
- "pages": 37,
228
  "shards": 3
229
  },
230
  {
231
- "seconds": 1.0109308320006676,
232
- "bytes": 5084502,
233
- "pages": 28,
234
- "shards": 2
235
  },
236
  {
237
- "seconds": 1.1290311129996553,
238
- "bytes": 7603249,
239
- "pages": 38,
240
- "shards": 3
241
  },
242
  {
243
- "seconds": 1.4725222399993072,
244
- "bytes": 12055069,
245
  "pages": 36,
246
- "shards": 3
247
  },
248
  {
249
- "seconds": 1.2178526319985394,
250
- "bytes": 7463726,
251
- "pages": 32,
252
  "shards": 3
253
  },
254
  {
255
- "seconds": 1.3964880730000004,
256
- "bytes": 6495644,
257
- "pages": 37,
258
- "shards": 2
259
- },
260
- {
261
- "seconds": 1.2645150809985353,
262
- "bytes": 6000960,
263
  "pages": 35,
264
  "shards": 2
265
  },
266
  {
267
- "seconds": 0.9461197429991444,
268
- "bytes": 2470605,
269
- "pages": 29,
270
- "shards": 2
271
  },
272
  {
273
- "seconds": 1.2941537009992317,
274
- "bytes": 6225550,
275
  "pages": 37,
276
  "shards": 2
277
  },
278
  {
279
- "seconds": 1.2335469089994149,
280
- "bytes": 11703737,
281
- "pages": 35,
282
  "shards": 3
283
  },
284
  {
285
- "seconds": 1.1459508749994711,
286
- "bytes": 7841640,
287
- "pages": 34,
288
  "shards": 3
289
  },
290
  {
291
- "seconds": 1.0024716899988562,
292
- "bytes": 5917383,
293
- "pages": 36,
294
- "shards": 2
295
- },
296
- {
297
- "seconds": 1.1789036200007104,
298
- "bytes": 2313669,
299
- "pages": 29,
300
- "shards": 2
301
- },
302
- {
303
- "seconds": 1.5398991070014745,
304
- "bytes": 10300581,
305
- "pages": 40,
306
- "shards": 3
307
  },
308
  {
309
- "seconds": 0.9979469910012995,
310
- "bytes": 6506963,
311
  "pages": 36,
312
  "shards": 2
313
  },
314
  {
315
- "seconds": 1.1795985310000106,
316
- "bytes": 8204299,
317
- "pages": 31,
318
  "shards": 3
319
  },
320
  {
321
- "seconds": 0.9598508790004416,
322
- "bytes": 6041026,
323
- "pages": 37,
324
- "shards": 2
325
- },
326
- {
327
- "seconds": 0.9930732779994287,
328
- "bytes": 6062259,
329
- "pages": 32,
330
  "shards": 3
331
  },
332
  {
333
- "seconds": 1.0064647640010662,
334
- "bytes": 5698061,
335
  "pages": 28,
336
  "shards": 2
337
  },
338
  {
339
- "seconds": 1.1021013760000642,
340
- "bytes": 5987765,
341
  "pages": 37,
342
  "shards": 2
343
  },
344
  {
345
- "seconds": 1.0322518390003097,
346
- "bytes": 2308031,
347
- "pages": 30,
348
  "shards": 2
349
  },
350
  {
351
- "seconds": 1.3352466599990294,
352
- "bytes": 15653746,
353
- "pages": 42,
354
  "shards": 4
355
  },
356
  {
357
- "seconds": 1.1028008130015223,
358
- "bytes": 1043427,
359
- "pages": 31,
360
- "shards": 1
361
- },
362
- {
363
- "seconds": 0.9759294559989939,
364
- "bytes": 7445734,
365
- "pages": 34,
366
- "shards": 3
367
- },
368
- {
369
- "seconds": 1.0883692229999724,
370
- "bytes": 10848676,
371
- "pages": 33,
372
- "shards": 3
373
  },
374
  {
375
- "seconds": 1.0698365700009163,
376
- "bytes": 2697855,
377
- "pages": 38,
378
  "shards": 2
379
  },
380
  {
381
- "seconds": 0.9394886710015271,
382
- "bytes": 6144198,
383
- "pages": 36,
384
  "shards": 2
385
  },
386
  {
387
- "seconds": 1.0910024349996092,
388
- "bytes": 11590114,
389
- "pages": 31,
390
  "shards": 3
391
  },
392
  {
393
- "seconds": 1.1193175999997038,
394
- "bytes": 7119189,
395
  "pages": 35,
 
 
 
 
 
 
396
  "shards": 3
397
  },
398
  {
399
- "seconds": 1.3181563210000604,
400
- "bytes": 6116642,
401
- "pages": 36,
402
  "shards": 2
403
  },
404
  {
405
- "seconds": 0.95897385599892,
406
- "bytes": 2279278,
407
- "pages": 29,
408
- "shards": 3
409
  },
410
  {
411
- "seconds": 1.0730731740004558,
412
- "bytes": 6315193,
413
  "pages": 36,
414
  "shards": 2
415
  },
416
  {
417
- "seconds": 1.276775304999319,
418
- "bytes": 10081435,
 
 
 
 
 
 
 
 
 
 
 
 
419
  "pages": 36,
420
  "shards": 3
421
  },
422
  {
423
- "seconds": 1.1636577470017073,
424
- "bytes": 8230638,
425
  "pages": 32,
426
  "shards": 3
427
  },
428
  {
429
- "seconds": 1.2417163159989286,
430
- "bytes": 6099161,
431
  "pages": 37,
432
  "shards": 2
433
  },
434
  {
435
- "seconds": 1.1390252959990903,
436
- "bytes": 7728259,
437
- "pages": 38,
438
- "shards": 3
439
  },
440
  {
441
- "seconds": 1.331549176999033,
442
- "bytes": 5708517,
443
- "pages": 35,
444
- "shards": 2
445
  },
446
  {
447
- "seconds": 1.0187081299991405,
448
- "bytes": 6222643,
449
  "pages": 36,
450
  "shards": 2
451
  },
452
  {
453
- "seconds": 1.2927204380011972,
454
- "bytes": 2373334,
455
  "pages": 29,
456
- "shards": 2
457
  },
458
  {
459
- "seconds": 1.0165509960006602,
460
- "bytes": 11776139,
461
- "pages": 33,
462
- "shards": 3
463
  },
464
  {
465
- "seconds": 1.1994448559999,
466
- "bytes": 10381245,
467
- "pages": 35,
468
  "shards": 3
469
  },
470
  {
471
- "seconds": 1.1291984180006693,
472
- "bytes": 2752026,
473
  "pages": 37,
474
  "shards": 2
475
  },
476
  {
477
- "seconds": 0.8884318259988504,
478
- "bytes": 6185512,
479
- "pages": 36,
480
  "shards": 2
481
  },
482
  {
483
- "seconds": 1.246648587000891,
484
- "bytes": 6489173,
485
- "pages": 30,
 
 
 
 
 
 
486
  "shards": 3
487
  },
488
  {
489
- "seconds": 1.0390239320004184,
490
- "bytes": 6087872,
491
  "pages": 35,
492
- "shards": 2
493
  },
494
  {
495
- "seconds": 1.244901927000683,
496
- "bytes": 11604745,
497
- "pages": 30,
498
  "shards": 3
499
  },
500
  {
501
- "seconds": 1.1130494770004589,
502
- "bytes": 2612508,
 
 
 
 
 
 
503
  "pages": 37,
 
 
 
 
 
 
504
  "shards": 2
505
  },
506
  {
507
- "seconds": 0.9836850789997698,
508
- "bytes": 10345276,
509
- "pages": 38,
510
- "shards": 3
511
  },
512
  {
513
- "seconds": 1.1697606620000442,
514
- "bytes": 7536660,
515
- "pages": 38,
516
- "shards": 3
517
  },
518
  {
519
- "seconds": 0.8494176380008867,
520
- "bytes": 947075,
521
- "pages": 27,
522
- "shards": 1
523
  },
524
  {
525
- "seconds": 1.246135413999582,
526
- "bytes": 6441176,
527
  "pages": 36,
528
  "shards": 2
529
  },
530
  {
531
- "seconds": 0.9687062859993603,
532
- "bytes": 7942604,
533
- "pages": 25,
534
- "shards": 3
535
  },
536
  {
537
- "seconds": 1.493546093999612,
538
- "bytes": 10942153,
539
- "pages": 44,
540
  "shards": 3
541
  },
542
  {
543
- "seconds": 1.185240850001719,
544
- "bytes": 5667842,
545
- "pages": 28,
546
  "shards": 2
547
  },
548
  {
549
- "seconds": 0.9565514620007889,
550
- "bytes": 2665835,
551
- "pages": 37,
552
- "shards": 2
553
  },
554
  {
555
- "seconds": 1.161401912002475,
556
- "bytes": 5953132,
557
- "pages": 33,
558
- "shards": 2
559
  },
560
  {
561
- "seconds": 1.9063561590010067,
562
- "bytes": 11430108,
563
- "pages": 37,
 
 
 
 
 
 
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": 76710.58553286047,
26
+ "required_pages_per_second_upper_estimate": 0.005307016300244384
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": 9588.823191607558,
40
+ "required_pages_per_second_upper_estimate": 0.004014640032928091
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": 19177.646383215117,
54
+ "required_pages_per_second_upper_estimate": 0.003994197993078467
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": 28766.469574822677,
68
+ "required_pages_per_second_upper_estimate": 0.0017245059271548526
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": 28766.469574822677,
82
+ "required_pages_per_second_upper_estimate": 0.0022364664791149408
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": 28766.469574822677,
96
+ "required_pages_per_second_upper_estimate": 0.36898538467724473
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": 704566,
161
+ "producer_batch_seconds": 0.0882111206484808,
162
+ "producer_p95_seconds": 0.2659228590018756,
163
+ "gpu_update_seconds": 0.9611189836168983,
164
+ "producer_batches_per_second": 11.336439132033885,
165
+ "gpu_batches_per_second": 1.0404539053393618,
166
+ "producer_headroom": 10.89566685641524,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.55049596786459,
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": 393537.5578407053,
174
+ "upload_bytes_per_second": 5534601.836951126,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 84.44399882713194,
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": 1.2417163159989286,
184
+ "bytes": 6099161,
185
+ "pages": 37,
 
 
 
 
 
 
186
  "shards": 2
187
  },
188
  {
189
+ "seconds": 1.1390252959990903,
190
+ "bytes": 7728259,
191
+ "pages": 38,
192
+ "shards": 3
193
  },
194
  {
195
+ "seconds": 1.331549176999033,
196
+ "bytes": 5708517,
197
+ "pages": 35,
198
  "shards": 2
199
  },
200
  {
201
+ "seconds": 1.0187081299991405,
202
+ "bytes": 6222643,
 
 
 
 
 
 
203
  "pages": 36,
204
  "shards": 2
205
  },
206
  {
207
+ "seconds": 1.2927204380011972,
208
+ "bytes": 2373334,
209
  "pages": 29,
210
  "shards": 2
211
  },
212
  {
213
+ "seconds": 1.0165509960006602,
214
+ "bytes": 11776139,
215
+ "pages": 33,
216
  "shards": 3
217
  },
218
  {
219
+ "seconds": 1.1994448559999,
220
+ "bytes": 10381245,
221
+ "pages": 35,
222
+ "shards": 3
223
  },
224
  {
225
+ "seconds": 1.1291984180006693,
226
+ "bytes": 2752026,
227
+ "pages": 37,
228
+ "shards": 2
229
  },
230
  {
231
+ "seconds": 0.8884318259988504,
232
+ "bytes": 6185512,
233
  "pages": 36,
234
+ "shards": 2
235
  },
236
  {
237
+ "seconds": 1.246648587000891,
238
+ "bytes": 6489173,
239
+ "pages": 30,
240
  "shards": 3
241
  },
242
  {
243
+ "seconds": 1.0390239320004184,
244
+ "bytes": 6087872,
 
 
 
 
 
 
245
  "pages": 35,
246
  "shards": 2
247
  },
248
  {
249
+ "seconds": 1.244901927000683,
250
+ "bytes": 11604745,
251
+ "pages": 30,
252
+ "shards": 3
253
  },
254
  {
255
+ "seconds": 1.1130494770004589,
256
+ "bytes": 2612508,
257
  "pages": 37,
258
  "shards": 2
259
  },
260
  {
261
+ "seconds": 0.9836850789997698,
262
+ "bytes": 10345276,
263
+ "pages": 38,
264
  "shards": 3
265
  },
266
  {
267
+ "seconds": 1.1697606620000442,
268
+ "bytes": 7536660,
269
+ "pages": 38,
270
  "shards": 3
271
  },
272
  {
273
+ "seconds": 0.8494176380008867,
274
+ "bytes": 947075,
275
+ "pages": 27,
276
+ "shards": 1
 
 
 
 
 
 
 
 
 
 
 
 
277
  },
278
  {
279
+ "seconds": 1.246135413999582,
280
+ "bytes": 6441176,
281
  "pages": 36,
282
  "shards": 2
283
  },
284
  {
285
+ "seconds": 0.9687062859993603,
286
+ "bytes": 7942604,
287
+ "pages": 25,
288
  "shards": 3
289
  },
290
  {
291
+ "seconds": 1.493546093999612,
292
+ "bytes": 10942153,
293
+ "pages": 44,
 
 
 
 
 
 
294
  "shards": 3
295
  },
296
  {
297
+ "seconds": 1.185240850001719,
298
+ "bytes": 5667842,
299
  "pages": 28,
300
  "shards": 2
301
  },
302
  {
303
+ "seconds": 0.9565514620007889,
304
+ "bytes": 2665835,
305
  "pages": 37,
306
  "shards": 2
307
  },
308
  {
309
+ "seconds": 1.161401912002475,
310
+ "bytes": 5953132,
311
+ "pages": 33,
312
  "shards": 2
313
  },
314
  {
315
+ "seconds": 1.9063561590010067,
316
+ "bytes": 11430108,
317
+ "pages": 37,
318
  "shards": 4
319
  },
320
  {
321
+ "seconds": 1.1905037959986657,
322
+ "bytes": 6976659,
323
+ "pages": 37,
324
+ "shards": 2
 
 
 
 
 
 
 
 
 
 
 
 
325
  },
326
  {
327
+ "seconds": 1.609685786999762,
328
+ "bytes": 5649551,
329
+ "pages": 20,
330
  "shards": 2
331
  },
332
  {
333
+ "seconds": 0.9700031609972939,
334
+ "bytes": 2540670,
335
+ "pages": 37,
336
  "shards": 2
337
  },
338
  {
339
+ "seconds": 1.1479466919990955,
340
+ "bytes": 10073767,
341
+ "pages": 30,
342
  "shards": 3
343
  },
344
  {
345
+ "seconds": 0.986695225998119,
346
+ "bytes": 1193548,
347
  "pages": 35,
348
+ "shards": 2
349
+ },
350
+ {
351
+ "seconds": 1.203395697997621,
352
+ "bytes": 7988495,
353
+ "pages": 39,
354
  "shards": 3
355
  },
356
  {
357
+ "seconds": 0.9280091689979599,
358
+ "bytes": 6513685,
359
+ "pages": 23,
360
  "shards": 2
361
  },
362
  {
363
+ "seconds": 4.352831413998501,
364
+ "bytes": 12282177,
365
+ "pages": 40,
366
+ "shards": 4
367
  },
368
  {
369
+ "seconds": 1.176046257998678,
370
+ "bytes": 6195296,
371
  "pages": 36,
372
  "shards": 2
373
  },
374
  {
375
+ "seconds": 0.9275896610015479,
376
+ "bytes": 1143867,
377
+ "pages": 35,
378
+ "shards": 1
379
+ },
380
+ {
381
+ "seconds": 3.292675117001636,
382
+ "bytes": 7453711,
383
+ "pages": 34,
384
+ "shards": 3
385
+ },
386
+ {
387
+ "seconds": 1.383926528000302,
388
+ "bytes": 10496353,
389
  "pages": 36,
390
  "shards": 3
391
  },
392
  {
393
+ "seconds": 1.054139716001373,
394
+ "bytes": 8196073,
395
  "pages": 32,
396
  "shards": 3
397
  },
398
  {
399
+ "seconds": 1.1359631509985775,
400
+ "bytes": 6053377,
401
  "pages": 37,
402
  "shards": 2
403
  },
404
  {
405
+ "seconds": 0.9480016679990513,
406
+ "bytes": 5853547,
407
+ "pages": 32,
408
+ "shards": 2
409
  },
410
  {
411
+ "seconds": 0.9646089939997182,
412
+ "bytes": 7151890,
413
+ "pages": 37,
414
+ "shards": 3
415
  },
416
  {
417
+ "seconds": 1.181685737999942,
418
+ "bytes": 6229682,
419
  "pages": 36,
420
  "shards": 2
421
  },
422
  {
423
+ "seconds": 0.7940784490019723,
424
+ "bytes": 968621,
425
  "pages": 29,
426
+ "shards": 1
427
  },
428
  {
429
+ "seconds": 1.324730037998961,
430
+ "bytes": 13434552,
431
+ "pages": 42,
432
+ "shards": 4
433
  },
434
  {
435
+ "seconds": 1.516792244001408,
436
+ "bytes": 10349687,
437
+ "pages": 29,
438
  "shards": 3
439
  },
440
  {
441
+ "seconds": 1.0428093720001925,
442
+ "bytes": 2636885,
443
  "pages": 37,
444
  "shards": 2
445
  },
446
  {
447
+ "seconds": 1.006605718001083,
448
+ "bytes": 6127707,
449
+ "pages": 28,
450
  "shards": 2
451
  },
452
  {
453
+ "seconds": 1.116981245999341,
454
+ "bytes": 6115742,
455
+ "pages": 37,
456
+ "shards": 2
457
+ },
458
+ {
459
+ "seconds": 1.009036264000315,
460
+ "bytes": 6506928,
461
+ "pages": 34,
462
  "shards": 3
463
  },
464
  {
465
+ "seconds": 1.1265610020018357,
466
+ "bytes": 11819263,
467
  "pages": 35,
468
+ "shards": 3
469
  },
470
  {
471
+ "seconds": 1.0389532609988237,
472
+ "bytes": 7773400,
473
+ "pages": 39,
474
  "shards": 3
475
  },
476
  {
477
+ "seconds": 0.9040890260002925,
478
+ "bytes": 1006858,
479
+ "pages": 30,
480
+ "shards": 1
481
+ },
482
+ {
483
+ "seconds": 1.4120586079989152,
484
+ "bytes": 10016675,
485
  "pages": 37,
486
+ "shards": 3
487
+ },
488
+ {
489
+ "seconds": 1.0292591729994456,
490
+ "bytes": 2310644,
491
+ "pages": 28,
492
  "shards": 2
493
  },
494
  {
495
+ "seconds": 1.012728962003166,
496
+ "bytes": 6259160,
497
+ "pages": 37,
498
+ "shards": 2
499
  },
500
  {
501
+ "seconds": 0.9536746940029843,
502
+ "bytes": 6477995,
503
+ "pages": 23,
504
+ "shards": 2
505
  },
506
  {
507
+ "seconds": 1.2943796040017332,
508
+ "bytes": 12517532,
509
+ "pages": 41,
510
+ "shards": 4
511
  },
512
  {
513
+ "seconds": 1.182365384996956,
514
+ "bytes": 6177943,
515
  "pages": 36,
516
  "shards": 2
517
  },
518
  {
519
+ "seconds": 0.8839036259996647,
520
+ "bytes": 2483088,
521
+ "pages": 29,
522
+ "shards": 2
523
  },
524
  {
525
+ "seconds": 1.165388396999333,
526
+ "bytes": 6302666,
527
+ "pages": 36,
528
  "shards": 3
529
  },
530
  {
531
+ "seconds": 1.2500933689989324,
532
+ "bytes": 4989357,
533
+ "pages": 30,
534
  "shards": 2
535
  },
536
  {
537
+ "seconds": 1.2095713759990758,
538
+ "bytes": 7770394,
539
+ "pages": 34,
540
+ "shards": 3
541
  },
542
  {
543
+ "seconds": 1.1173389949981356,
544
+ "bytes": 12086613,
545
+ "pages": 41,
546
+ "shards": 3
547
  },
548
  {
549
+ "seconds": 0.9005163840010937,
550
+ "bytes": 1099069,
551
+ "pages": 30,
552
+ "shards": 1
553
+ },
554
+ {
555
+ "seconds": 1.1803624109998054,
556
+ "bytes": 11418078,
557
+ "pages": 40,
558
  "shards": 4
559
+ },
560
+ {
561
+ "seconds": 1.0558311360000516,
562
+ "bytes": 6077678,
563
+ "pages": 30,
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 sha256:442d5a63837b6ac40b1f2659fb1a8dca8d13b6c8efafd35b7dd88a3f95501c4e
3
- size 721775615
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b8e924ba03c8c90f3a051947fd0eb5994e5b786f87c64a7022243d5bfe8f0021
3
+ size 721769151