appvoid commited on
Commit
cdd53bb
·
verified ·
1 Parent(s): 8a7be40

Qwen90M step 515000; complete resumable checkpoint

Browse files
Files changed (6) hide show
  1. complete.json +5 -5
  2. coverage.json +47 -47
  3. latest.json +1 -1
  4. model.safetensors +1 -1
  5. shard_scheduler.json +243 -243
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 510000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "6f5d47ff878370918d779712398d1df72e8136b909f0f516e626f1deaa459e61",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "bc66272882613b497b87b905193beb979dbc37c617486fbe288516fb56b1ed9c",
12
  "training_config.json": "42b690efbefcf434fbcb690b27d17a2af5e3b65c3c6aab71efa1d83d1c6a6905",
13
  "performance.json": "234744063e094dfd7b525da4e181ac493fac400c830f871c7e8c59127ed10dc9",
14
- "shard_scheduler.json": "8e1dcb9a6b41bfa5fb3adfa01b8d6110beb427eb27007634bc7f22a838f0dfc5",
15
- "coverage.json": "d8336715320aedee9a1a94d413b7034cb41f2357aa8f3ab987da371fbc49a337",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 515000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "a5f425d4581d70ba7b0c3a42632d0397ad924a8ed9116534dcbe7528d217d892",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "ef09d114016bc2e115e2f936e0e83f4c16079f978a8f1e5a828b8b2f3204fe22",
12
  "training_config.json": "42b690efbefcf434fbcb690b27d17a2af5e3b65c3c6aab71efa1d83d1c6a6905",
13
  "performance.json": "234744063e094dfd7b525da4e181ac493fac400c830f871c7e8c59127ed10dc9",
14
+ "shard_scheduler.json": "ba84c281f6b706297b3d60285a57ade76d674a22362d7ffd41755519d1b1e1eb",
15
+ "coverage.json": "ab9b69f5c3f1e54e5942d2dbadb0d1da6152c604c5616a3c283412918e4e3e24",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 510000,
3
  "stats": {
4
- "climbmix/blocks": 18360000,
5
- "climbmix/targets": 30689271681,
6
- "climbmix/input_tokens": 30691216792,
7
- "ultra_qa/blocks": 6885000,
8
- "ultra_qa/targets": 11231922686,
9
- "ultra_qa/input_tokens": 11231944353,
10
- "cortex/blocks": 6885000,
11
- "cortex/targets": 261847115,
12
- "cortex/input_tokens": 10944570168,
13
- "ultra_style/blocks": 6885000,
14
- "ultra_style/targets": 10889862082,
15
- "ultra_style/input_tokens": 10890432997,
16
- "openmath/blocks": 4590000,
17
- "openmath/targets": 7092338265,
18
- "openmath/input_tokens": 7094701522,
19
- "rewrite6/blocks": 2295000,
20
- "rewrite6/targets": 3762269811,
21
- "rewrite6/input_tokens": 3766394881
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 2515,
26
- "row": 756,
27
  "start": {
28
- "file": 121,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 56,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 0,
37
- "documents": 10302196
38
  },
39
  "rewrite6": {
40
- "page": 30,
41
- "row": 2578,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 122880
47
  }
48
  },
49
  "epoch": 28,
50
  "offset": 0,
51
- "documents": 5876770
52
  },
53
  "openmath": {
54
- "page": 1571,
55
- "row": 2225,
56
  "start": {
57
  "file": 14,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 321,
61
- "row": 716
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 6437041
67
  },
68
  "ultra_qa": {
69
- "page": 679,
70
- "row": 1959,
71
  "start": {
72
  "file": 5,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
- "row_group": 2,
76
- "row": 54023
77
  }
78
  },
79
  "epoch": 0,
80
- "offset": 2048,
81
- "documents": 2783143
82
  },
83
  "ultra_style": {
84
- "page": 1036,
85
- "row": 3816,
86
  "start": {
87
  "file": 6,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
- "row_group": 1,
91
- "row": 48587
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
- "documents": 4247272
97
  },
98
  "cortex": {
99
- "page": 133533,
100
- "row": 97,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 17092321
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 515000,
3
  "stats": {
4
+ "climbmix/blocks": 18540000,
5
+ "climbmix/targets": 30990604912,
6
+ "climbmix/input_tokens": 30992568790,
7
+ "ultra_qa/blocks": 6952500,
8
+ "ultra_qa/targets": 11342093677,
9
+ "ultra_qa/input_tokens": 11342115570,
10
+ "cortex/blocks": 6952500,
11
+ "cortex/targets": 264409093,
12
+ "cortex/input_tokens": 11051859128,
13
+ "ultra_style/blocks": 6952500,
14
+ "ultra_style/targets": 10996098074,
15
+ "ultra_style/input_tokens": 10996674542,
16
+ "openmath/blocks": 4635000,
17
+ "openmath/targets": 7161856281,
18
+ "openmath/input_tokens": 7164242696,
19
+ "rewrite6/blocks": 2317500,
20
+ "rewrite6/targets": 3799257439,
21
+ "rewrite6/input_tokens": 3803422876
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 2539,
26
+ "row": 2373,
27
  "start": {
28
+ "file": 122,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 70,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 0,
37
+ "documents": 10402117
38
  },
39
  "rewrite6": {
40
+ "page": 44,
41
+ "row": 2713,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 180224
47
  }
48
  },
49
  "epoch": 28,
50
  "offset": 0,
51
+ "documents": 5934249
52
  },
53
  "openmath": {
54
+ "page": 1586,
55
+ "row": 3895,
56
  "start": {
57
  "file": 14,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 383,
61
+ "row": 156
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 6500151
67
  },
68
  "ultra_qa": {
69
+ "page": 686,
70
+ "row": 590,
71
  "start": {
72
  "file": 5,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
+ "row_group": 3,
76
+ "row": 18153
77
  }
78
  },
79
  "epoch": 0,
80
+ "offset": 4096,
81
+ "documents": 2810446
82
  },
83
  "ultra_style": {
84
+ "page": 1047,
85
+ "row": 806,
86
  "start": {
87
  "file": 6,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
+ "row_group": 2,
91
+ "row": 8236
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
+ "documents": 4289318
97
  },
98
  "cortex": {
99
+ "page": 134842,
100
+ "row": 79,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 17259855
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 510000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "6f5d47ff878370918d779712398d1df72e8136b909f0f516e626f1deaa459e61", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "bc66272882613b497b87b905193beb979dbc37c617486fbe288516fb56b1ed9c", "training_config.json": "42b690efbefcf434fbcb690b27d17a2af5e3b65c3c6aab71efa1d83d1c6a6905", "performance.json": "234744063e094dfd7b525da4e181ac493fac400c830f871c7e8c59127ed10dc9", "shard_scheduler.json": "8e1dcb9a6b41bfa5fb3adfa01b8d6110beb427eb27007634bc7f22a838f0dfc5", "coverage.json": "d8336715320aedee9a1a94d413b7034cb41f2357aa8f3ab987da371fbc49a337", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000510000"}
 
1
+ {"step": 515000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "a5f425d4581d70ba7b0c3a42632d0397ad924a8ed9116534dcbe7528d217d892", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "ef09d114016bc2e115e2f936e0e83f4c16079f978a8f1e5a828b8b2f3204fe22", "training_config.json": "42b690efbefcf434fbcb690b27d17a2af5e3b65c3c6aab71efa1d83d1c6a6905", "performance.json": "234744063e094dfd7b525da4e181ac493fac400c830f871c7e8c59127ed10dc9", "shard_scheduler.json": "ba84c281f6b706297b3d60285a57ade76d674a22362d7ffd41755519d1b1e1eb", "coverage.json": "ab9b69f5c3f1e54e5942d2dbadb0d1da6152c604c5616a3c283412918e4e3e24", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000515000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:6f5d47ff878370918d779712398d1df72e8136b909f0f516e626f1deaa459e61
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a5f425d4581d70ba7b0c3a42632d0397ad924a8ed9116534dcbe7528d217d892
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": 75832.22366015913,
26
- "required_pages_per_second_upper_estimate": 0.005246249187810568
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": 9479.027957519891,
40
- "required_pages_per_second_upper_estimate": 0.003968671061200793
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": 18958.055915039782,
54
- "required_pages_per_second_upper_estimate": 0.003948463089547618
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": 28437.083872559673,
68
- "required_pages_per_second_upper_estimate": 0.0017047597572470315
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": 28437.083872559673,
82
- "required_pages_per_second_upper_estimate": 0.002210858189578581
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": 22,
95
- "required_token_positions_per_second": 28437.083872559673,
96
- "required_pages_per_second_upper_estimate": 0.36476037855542737
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,410 +158,410 @@
158
  "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 704566,
161
- "producer_batch_seconds": 0.08594841671902032,
162
- "producer_p95_seconds": 0.3017171040009998,
163
- "gpu_update_seconds": 0.9722515896462541,
164
- "producer_batches_per_second": 11.63488564622623,
165
- "gpu_batches_per_second": 1.0285403599739467,
166
- "producer_headroom": 11.312036064895837,
167
  "sustainable": true,
168
- "buffer_seconds": 60.279598558067754,
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": 339437.1945102421,
174
- "upload_bytes_per_second": 6151997.828176973,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 72.83532024841735,
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": 0.7940784490019723,
184
- "bytes": 968621,
185
- "pages": 29,
186
  "shards": 1
187
  },
188
  {
189
- "seconds": 1.324730037998961,
190
- "bytes": 13434552,
191
- "pages": 42,
192
- "shards": 4
193
- },
194
- {
195
- "seconds": 1.516792244001408,
196
- "bytes": 10349687,
197
- "pages": 29,
198
  "shards": 3
199
  },
200
  {
201
- "seconds": 1.0428093720001925,
202
- "bytes": 2636885,
203
  "pages": 37,
204
  "shards": 2
205
  },
206
  {
207
- "seconds": 1.006605718001083,
208
- "bytes": 6127707,
209
- "pages": 28,
210
  "shards": 2
211
  },
212
  {
213
- "seconds": 1.116981245999341,
214
- "bytes": 6115742,
 
 
 
 
 
 
 
 
 
 
 
 
215
  "pages": 37,
216
  "shards": 2
217
  },
218
  {
219
- "seconds": 1.009036264000315,
220
- "bytes": 6506928,
221
- "pages": 34,
222
  "shards": 3
223
  },
224
  {
225
- "seconds": 1.1265610020018357,
226
- "bytes": 11819263,
227
  "pages": 35,
228
- "shards": 3
229
  },
230
  {
231
- "seconds": 1.0389532609988237,
232
- "bytes": 7773400,
233
- "pages": 39,
234
  "shards": 3
235
  },
236
  {
237
- "seconds": 0.9040890260002925,
238
- "bytes": 1006858,
239
- "pages": 30,
240
- "shards": 1
241
  },
242
  {
243
- "seconds": 1.4120586079989152,
244
- "bytes": 10016675,
245
  "pages": 37,
 
 
 
 
 
 
246
  "shards": 3
247
  },
248
  {
249
- "seconds": 1.0292591729994456,
250
- "bytes": 2310644,
251
- "pages": 28,
252
  "shards": 2
253
  },
254
  {
255
- "seconds": 1.012728962003166,
256
- "bytes": 6259160,
257
  "pages": 37,
258
  "shards": 2
259
  },
260
  {
261
- "seconds": 0.9536746940029843,
262
- "bytes": 6477995,
263
- "pages": 23,
264
- "shards": 2
265
  },
266
  {
267
- "seconds": 1.2943796040017332,
268
- "bytes": 12517532,
269
- "pages": 41,
270
- "shards": 4
271
  },
272
  {
273
- "seconds": 1.182365384996956,
274
- "bytes": 6177943,
275
- "pages": 36,
276
- "shards": 2
277
  },
278
  {
279
- "seconds": 0.8839036259996647,
280
- "bytes": 2483088,
281
- "pages": 29,
282
  "shards": 2
283
  },
284
  {
285
- "seconds": 1.165388396999333,
286
- "bytes": 6302666,
287
- "pages": 36,
288
- "shards": 3
289
  },
290
  {
291
- "seconds": 1.2500933689989324,
292
- "bytes": 4989357,
293
- "pages": 30,
294
  "shards": 2
295
  },
296
  {
297
- "seconds": 1.2095713759990758,
298
- "bytes": 7770394,
299
- "pages": 34,
300
  "shards": 3
301
  },
302
  {
303
- "seconds": 1.1173389949981356,
304
- "bytes": 12086613,
305
- "pages": 41,
306
  "shards": 3
307
  },
308
  {
309
- "seconds": 0.9005163840010937,
310
- "bytes": 1099069,
311
- "pages": 30,
312
- "shards": 1
313
- },
314
- {
315
- "seconds": 1.1803624109998054,
316
- "bytes": 11418078,
317
- "pages": 40,
318
- "shards": 4
319
- },
320
- {
321
- "seconds": 1.0558311360000516,
322
- "bytes": 6077678,
323
- "pages": 30,
324
  "shards": 2
325
  },
326
  {
327
- "seconds": 1.0529501199998776,
328
- "bytes": 2460568,
329
- "pages": 35,
330
  "shards": 2
331
  },
332
  {
333
- "seconds": 1.1411234580009477,
334
- "bytes": 10341495,
335
- "pages": 29,
336
  "shards": 3
337
  },
338
  {
339
- "seconds": 1.1926653939990501,
340
- "bytes": 7075984,
341
- "pages": 38,
342
  "shards": 2
343
  },
344
  {
345
- "seconds": 1.1058772389988007,
346
- "bytes": 7919154,
347
- "pages": 39,
348
- "shards": 3
349
- },
350
- {
351
- "seconds": 1.395315304001997,
352
- "bytes": 5693810,
353
- "pages": 28,
354
  "shards": 2
355
  },
356
  {
357
- "seconds": 1.0359600679985306,
358
- "bytes": 6649243,
359
- "pages": 34,
360
  "shards": 3
361
  },
362
  {
363
- "seconds": 0.9754138270000112,
364
- "bytes": 5934822,
365
- "pages": 37,
366
- "shards": 2
367
  },
368
  {
369
- "seconds": 0.7901811400006409,
370
- "bytes": 1002646,
371
- "pages": 28,
372
- "shards": 1
373
  },
374
  {
375
- "seconds": 1.1965479969985608,
376
- "bytes": 13159361,
377
- "pages": 38,
378
- "shards": 4
379
  },
380
  {
381
- "seconds": 1.0418553449999308,
382
- "bytes": 11343588,
383
- "pages": 42,
384
  "shards": 3
385
  },
386
  {
387
- "seconds": 0.967730065996875,
388
- "bytes": 2342541,
389
- "pages": 28,
390
  "shards": 2
391
  },
392
  {
393
- "seconds": 1.1295750149984087,
394
- "bytes": 5944567,
395
- "pages": 37,
396
- "shards": 2
397
  },
398
  {
399
- "seconds": 1.1342636209992634,
400
- "bytes": 6342296,
401
- "pages": 37,
402
  "shards": 2
403
  },
404
  {
405
- "seconds": 1.3789669100005995,
406
- "bytes": 6252708,
407
- "pages": 25,
408
- "shards": 3
409
  },
410
  {
411
- "seconds": 1.107486043998506,
412
- "bytes": 11895475,
413
  "pages": 41,
414
  "shards": 3
415
  },
416
  {
417
- "seconds": 1.5734916069995961,
418
- "bytes": 7449351,
419
- "pages": 39,
420
- "shards": 3
421
- },
422
- {
423
- "seconds": 0.9707876079992275,
424
- "bytes": 899756,
425
  "pages": 27,
426
  "shards": 1
427
  },
428
  {
429
- "seconds": 1.1750758769994718,
430
- "bytes": 10630556,
431
- "pages": 35,
 
 
 
 
 
 
432
  "shards": 3
433
  },
434
  {
435
- "seconds": 1.1581052289984655,
436
- "bytes": 2607511,
437
- "pages": 37,
438
  "shards": 2
439
  },
440
  {
441
- "seconds": 0.9523471370011976,
442
- "bytes": 5723812,
443
- "pages": 24,
444
  "shards": 2
445
  },
446
  {
447
- "seconds": 1.186567714001285,
448
- "bytes": 11920154,
449
- "pages": 44,
450
- "shards": 3
451
  },
452
  {
453
- "seconds": 0.9970581140005379,
454
- "bytes": 6389566,
455
- "pages": 36,
 
 
 
 
 
 
456
  "shards": 3
457
  },
458
  {
459
- "seconds": 1.1336718419988756,
460
- "bytes": 6075801,
461
  "pages": 37,
462
  "shards": 2
463
  },
464
  {
465
- "seconds": 1.0171541380004783,
466
- "bytes": 7685246,
467
- "pages": 31,
468
- "shards": 3
469
- },
470
- {
471
- "seconds": 0.8166221829997085,
472
- "bytes": 1223594,
473
- "pages": 35,
474
  "shards": 2
475
  },
476
  {
477
- "seconds": 1.285200268997869,
478
- "bytes": 10185249,
479
- "pages": 22,
480
- "shards": 3
481
  },
482
  {
483
- "seconds": 1.1744725909993576,
484
- "bytes": 8758656,
485
- "pages": 47,
486
  "shards": 3
487
  },
488
  {
489
- "seconds": 1.0896003289999499,
490
- "bytes": 6195866,
491
  "pages": 37,
492
  "shards": 2
493
  },
494
  {
495
- "seconds": 1.1199947279965272,
496
- "bytes": 7293697,
497
- "pages": 31,
498
- "shards": 3
499
  },
500
  {
501
- "seconds": 1.0610679239980527,
502
- "bytes": 5725336,
503
- "pages": 35,
504
- "shards": 2
505
  },
506
  {
507
- "seconds": 0.9264001560004544,
508
- "bytes": 6158826,
509
  "pages": 37,
510
  "shards": 2
511
  },
512
  {
513
- "seconds": 1.2382891359993664,
514
- "bytes": 13055404,
515
- "pages": 34,
516
- "shards": 4
517
  },
518
  {
519
- "seconds": 0.9811806699981389,
520
- "bytes": 1058761,
521
- "pages": 33,
522
- "shards": 1
523
  },
524
  {
525
- "seconds": 1.215036166999198,
526
- "bytes": 9780503,
527
- "pages": 30,
528
- "shards": 3
529
  },
530
  {
531
- "seconds": 0.9811340400010522,
532
- "bytes": 2572134,
533
- "pages": 37,
534
- "shards": 2
535
  },
536
  {
537
- "seconds": 1.036350621998281,
538
- "bytes": 6380141,
539
- "pages": 38,
540
  "shards": 2
541
  },
542
  {
543
- "seconds": 0.9442703610002354,
544
- "bytes": 2120655,
545
- "pages": 20,
546
  "shards": 2
547
  },
548
  {
549
- "seconds": 1.0825822449987754,
550
- "bytes": 10431134,
551
- "pages": 32,
552
  "shards": 3
553
  },
554
  {
555
- "seconds": 1.0676360670004215,
556
- "bytes": 12466956,
557
- "pages": 47,
 
 
 
 
 
 
558
  "shards": 3
559
  },
560
  {
561
- "seconds": 0.9538209790007386,
562
- "bytes": 2685432,
563
- "pages": 38,
564
- "shards": 2
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": 75955.3176128496,
26
+ "required_pages_per_second_upper_estimate": 0.005254765113075015
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": 9494.4147016062,
40
+ "required_pages_per_second_upper_estimate": 0.003975113169637976
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": 18988.8294032124,
54
+ "required_pages_per_second_upper_estimate": 0.003954872395582516
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": 28483.244104818597,
68
+ "required_pages_per_second_upper_estimate": 0.0017075269926883603
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": 28483.244104818597,
82
+ "required_pages_per_second_upper_estimate": 0.002214446944599308
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": 22,
95
+ "required_token_positions_per_second": 28483.244104818597,
96
+ "required_pages_per_second_upper_estimate": 0.365352472451849
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
 
158
  "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 704566,
161
+ "producer_batch_seconds": 0.3101006119218823,
162
+ "producer_p95_seconds": 0.3071302360003756,
163
+ "gpu_update_seconds": 0.9706759489282579,
164
+ "producer_batches_per_second": 3.2247598410154406,
165
+ "gpu_batches_per_second": 1.0302099285597004,
166
+ "producer_headroom": 3.130196818743401,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.181908833551994,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 80,
172
  "max_pending_pages": 4096,
173
+ "generation_bytes_per_second": 525135.3264741183,
174
+ "upload_bytes_per_second": 6072452.474192158,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 112.68181653659498,
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": 0.9707876079992275,
184
+ "bytes": 899756,
185
+ "pages": 27,
186
  "shards": 1
187
  },
188
  {
189
+ "seconds": 1.1750758769994718,
190
+ "bytes": 10630556,
191
+ "pages": 35,
 
 
 
 
 
 
192
  "shards": 3
193
  },
194
  {
195
+ "seconds": 1.1581052289984655,
196
+ "bytes": 2607511,
197
  "pages": 37,
198
  "shards": 2
199
  },
200
  {
201
+ "seconds": 0.9523471370011976,
202
+ "bytes": 5723812,
203
+ "pages": 24,
204
  "shards": 2
205
  },
206
  {
207
+ "seconds": 1.186567714001285,
208
+ "bytes": 11920154,
209
+ "pages": 44,
210
+ "shards": 3
211
+ },
212
+ {
213
+ "seconds": 0.9970581140005379,
214
+ "bytes": 6389566,
215
+ "pages": 36,
216
+ "shards": 3
217
+ },
218
+ {
219
+ "seconds": 1.1336718419988756,
220
+ "bytes": 6075801,
221
  "pages": 37,
222
  "shards": 2
223
  },
224
  {
225
+ "seconds": 1.0171541380004783,
226
+ "bytes": 7685246,
227
+ "pages": 31,
228
  "shards": 3
229
  },
230
  {
231
+ "seconds": 0.8166221829997085,
232
+ "bytes": 1223594,
233
  "pages": 35,
234
+ "shards": 2
235
  },
236
  {
237
+ "seconds": 1.285200268997869,
238
+ "bytes": 10185249,
239
+ "pages": 22,
240
  "shards": 3
241
  },
242
  {
243
+ "seconds": 1.1744725909993576,
244
+ "bytes": 8758656,
245
+ "pages": 47,
246
+ "shards": 3
247
  },
248
  {
249
+ "seconds": 1.0896003289999499,
250
+ "bytes": 6195866,
251
  "pages": 37,
252
+ "shards": 2
253
+ },
254
+ {
255
+ "seconds": 1.1199947279965272,
256
+ "bytes": 7293697,
257
+ "pages": 31,
258
  "shards": 3
259
  },
260
  {
261
+ "seconds": 1.0610679239980527,
262
+ "bytes": 5725336,
263
+ "pages": 35,
264
  "shards": 2
265
  },
266
  {
267
+ "seconds": 0.9264001560004544,
268
+ "bytes": 6158826,
269
  "pages": 37,
270
  "shards": 2
271
  },
272
  {
273
+ "seconds": 1.2382891359993664,
274
+ "bytes": 13055404,
275
+ "pages": 34,
276
+ "shards": 4
277
  },
278
  {
279
+ "seconds": 0.9811806699981389,
280
+ "bytes": 1058761,
281
+ "pages": 33,
282
+ "shards": 1
283
  },
284
  {
285
+ "seconds": 1.215036166999198,
286
+ "bytes": 9780503,
287
+ "pages": 30,
288
+ "shards": 3
289
  },
290
  {
291
+ "seconds": 0.9811340400010522,
292
+ "bytes": 2572134,
293
+ "pages": 37,
294
  "shards": 2
295
  },
296
  {
297
+ "seconds": 1.036350621998281,
298
+ "bytes": 6380141,
299
+ "pages": 38,
300
+ "shards": 2
301
  },
302
  {
303
+ "seconds": 0.9442703610002354,
304
+ "bytes": 2120655,
305
+ "pages": 20,
306
  "shards": 2
307
  },
308
  {
309
+ "seconds": 1.0825822449987754,
310
+ "bytes": 10431134,
311
+ "pages": 32,
312
  "shards": 3
313
  },
314
  {
315
+ "seconds": 1.0676360670004215,
316
+ "bytes": 12466956,
317
+ "pages": 47,
318
  "shards": 3
319
  },
320
  {
321
+ "seconds": 0.9538209790007386,
322
+ "bytes": 2685432,
323
+ "pages": 38,
 
 
 
 
 
 
 
 
 
 
 
 
324
  "shards": 2
325
  },
326
  {
327
+ "seconds": 0.9428314830001909,
328
+ "bytes": 5761904,
329
+ "pages": 20,
330
  "shards": 2
331
  },
332
  {
333
+ "seconds": 1.1127153689994884,
334
+ "bytes": 6912505,
335
+ "pages": 37,
336
  "shards": 3
337
  },
338
  {
339
+ "seconds": 1.2596074149987544,
340
+ "bytes": 6204319,
341
+ "pages": 37,
342
  "shards": 2
343
  },
344
  {
345
+ "seconds": 1.0030813440025668,
346
+ "bytes": 5714103,
347
+ "pages": 25,
 
 
 
 
 
 
348
  "shards": 2
349
  },
350
  {
351
+ "seconds": 1.237253362996853,
352
+ "bytes": 8511115,
353
+ "pages": 39,
354
  "shards": 3
355
  },
356
  {
357
+ "seconds": 1.2614193410008738,
358
+ "bytes": 9853756,
359
+ "pages": 38,
360
+ "shards": 3
361
  },
362
  {
363
+ "seconds": 1.0838980580010684,
364
+ "bytes": 8174651,
365
+ "pages": 40,
366
+ "shards": 3
367
  },
368
  {
369
+ "seconds": 1.1033362929993018,
370
+ "bytes": 798343,
371
+ "pages": 26,
372
+ "shards": 1
373
  },
374
  {
375
+ "seconds": 1.2149433689992293,
376
+ "bytes": 11402427,
377
+ "pages": 41,
378
  "shards": 3
379
  },
380
  {
381
+ "seconds": 0.7934381430022768,
382
+ "bytes": 2164331,
383
+ "pages": 22,
384
  "shards": 2
385
  },
386
  {
387
+ "seconds": 1.25598544499735,
388
+ "bytes": 11955544,
389
+ "pages": 38,
390
+ "shards": 3
391
  },
392
  {
393
+ "seconds": 1.0177433429998928,
394
+ "bytes": 6488753,
395
+ "pages": 38,
396
  "shards": 2
397
  },
398
  {
399
+ "seconds": 0.9433292390021961,
400
+ "bytes": 2349435,
401
+ "pages": 28,
402
+ "shards": 2
403
  },
404
  {
405
+ "seconds": 1.25489443799961,
406
+ "bytes": 9999755,
407
  "pages": 41,
408
  "shards": 3
409
  },
410
  {
411
+ "seconds": 0.9152432709997811,
412
+ "bytes": 861201,
 
 
 
 
 
 
413
  "pages": 27,
414
  "shards": 1
415
  },
416
  {
417
+ "seconds": 1.3237921879990608,
418
+ "bytes": 13505764,
419
+ "pages": 42,
420
+ "shards": 5
421
+ },
422
+ {
423
+ "seconds": 1.3861956760010798,
424
+ "bytes": 10089467,
425
+ "pages": 32,
426
  "shards": 3
427
  },
428
  {
429
+ "seconds": 1.1900551170001563,
430
+ "bytes": 2519802,
431
+ "pages": 34,
432
  "shards": 2
433
  },
434
  {
435
+ "seconds": 1.0123512749996735,
436
+ "bytes": 5925157,
437
+ "pages": 33,
438
  "shards": 2
439
  },
440
  {
441
+ "seconds": 0.84063605999836,
442
+ "bytes": 1225280,
443
+ "pages": 36,
444
+ "shards": 1
445
  },
446
  {
447
+ "seconds": 1.2510575380001683,
448
+ "bytes": 12042377,
449
+ "pages": 34,
450
+ "shards": 4
451
+ },
452
+ {
453
+ "seconds": 1.23156422700049,
454
+ "bytes": 11911377,
455
+ "pages": 32,
456
  "shards": 3
457
  },
458
  {
459
+ "seconds": 0.8838252950008609,
460
+ "bytes": 2597092,
461
  "pages": 37,
462
  "shards": 2
463
  },
464
  {
465
+ "seconds": 1.0776322439996875,
466
+ "bytes": 6282866,
467
+ "pages": 37,
 
 
 
 
 
 
468
  "shards": 2
469
  },
470
  {
471
+ "seconds": 1.0567811459986842,
472
+ "bytes": 4557138,
473
+ "pages": 29,
474
+ "shards": 2
475
  },
476
  {
477
+ "seconds": 1.3797404860015376,
478
+ "bytes": 7238899,
479
+ "pages": 30,
480
  "shards": 3
481
  },
482
  {
483
+ "seconds": 1.0933739200008858,
484
+ "bytes": 6120770,
485
  "pages": 37,
486
  "shards": 2
487
  },
488
  {
489
+ "seconds": 0.9898015409999061,
490
+ "bytes": 867435,
491
+ "pages": 29,
492
+ "shards": 1
493
  },
494
  {
495
+ "seconds": 1.4056065479999234,
496
+ "bytes": 17665783,
497
+ "pages": 34,
498
+ "shards": 5
499
  },
500
  {
501
+ "seconds": 1.0921460530007607,
502
+ "bytes": 6276750,
503
  "pages": 37,
504
  "shards": 2
505
  },
506
  {
507
+ "seconds": 1.0008894900020096,
508
+ "bytes": 2319506,
509
+ "pages": 29,
510
+ "shards": 2
511
  },
512
  {
513
+ "seconds": 1.162798567002028,
514
+ "bytes": 6350229,
515
+ "pages": 37,
516
+ "shards": 2
517
  },
518
  {
519
+ "seconds": 1.0111840479985403,
520
+ "bytes": 5239532,
521
+ "pages": 38,
522
+ "shards": 2
523
  },
524
  {
525
+ "seconds": 1.2082140299971798,
526
+ "bytes": 13368136,
527
+ "pages": 36,
528
+ "shards": 4
529
  },
530
  {
531
+ "seconds": 1.2696354079998855,
532
+ "bytes": 6070118,
533
+ "pages": 28,
534
  "shards": 2
535
  },
536
  {
537
+ "seconds": 0.9898074049997376,
538
+ "bytes": 2654933,
539
+ "pages": 38,
540
  "shards": 2
541
  },
542
  {
543
+ "seconds": 1.3211888400001044,
544
+ "bytes": 9995304,
545
+ "pages": 31,
546
  "shards": 3
547
  },
548
  {
549
+ "seconds": 0.8846473650010012,
550
+ "bytes": 1315669,
551
+ "pages": 36,
552
+ "shards": 1
553
+ },
554
+ {
555
+ "seconds": 1.6196549160013092,
556
+ "bytes": 7467278,
557
+ "pages": 37,
558
  "shards": 3
559
  },
560
  {
561
+ "seconds": 1.2983564580026723,
562
+ "bytes": 12027088,
563
+ "pages": 34,
564
+ "shards": 3
565
  }
566
  ],
567
  "cache_key": {
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:bc66272882613b497b87b905193beb979dbc37c617486fbe288516fb56b1ed9c
3
- size 721759487
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ef09d114016bc2e115e2f936e0e83f4c16079f978a8f1e5a828b8b2f3204fe22
3
+ size 721771839