appvoid commited on
Commit
35e4a25
·
verified ·
1 Parent(s): 95594e5

Qwen90M step 835000; complete resumable checkpoint

Browse files
Files changed (6) hide show
  1. complete.json +5 -5
  2. coverage.json +49 -49
  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": 830000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "148b9cd485ba37f4e9cce26a66ff0cccc598fbe722764f410999681d873c39a5",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "f6ee31a07eec8634f2a4579290acbcefb3cab5d7972f2153a10c3041f0c32cb5",
12
  "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
13
  "performance.json": "eb45b2d9ecb051b6b7e592c18f2a0bfebf4278147b783ba131f7a9a4453a9ec7",
14
- "shard_scheduler.json": "d229c2adbc90c3e90d75f2ded34df13e9ded5f7d2aa7cf78783ee6b128d17bdd",
15
- "coverage.json": "537e5ac8f9bc339cc5920f3da24a226dfbd2e65ffea7c0b2d60b4a270d3159d5",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"
 
1
  {
2
+ "step": 835000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "434bc5e66416d00f38daf4d1c20bd1d7d620b0fc06f97108edac543c845dc140",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "8fc02097e06a5d8dde4c92534118b36500d860341974727dbdf1235db028d4da",
12
  "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
13
  "performance.json": "eb45b2d9ecb051b6b7e592c18f2a0bfebf4278147b783ba131f7a9a4453a9ec7",
14
+ "shard_scheduler.json": "15ec84a118722aa0da30b1eedddde0fa6f0d20effd89f99c89d67b1b955f9b9e",
15
+ "coverage.json": "2268abb420311a79a47ad9499884d787f6b2b1069f26033253f7830b5909fc05",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 830000,
3
  "stats": {
4
- "climbmix/blocks": 29880000,
5
- "climbmix/targets": 49950066776,
6
- "climbmix/input_tokens": 49953231760,
7
- "ultra_qa/blocks": 11205000,
8
- "ultra_qa/targets": 18279978645,
9
- "ultra_qa/input_tokens": 18280013833,
10
- "cortex/blocks": 11205000,
11
- "cortex/targets": 426071773,
12
- "cortex/input_tokens": 17811149534,
13
- "ultra_style/blocks": 11205000,
14
- "ultra_style/targets": 17722842692,
15
- "ultra_style/input_tokens": 17723772365,
16
- "openmath/blocks": 7470000,
17
- "openmath/targets": 11542438616,
18
- "openmath/input_tokens": 11546284486,
19
- "rewrite6/blocks": 3735000,
20
- "rewrite6/targets": 6122985883,
21
- "rewrite6/input_tokens": 6129699061
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 4092,
26
- "row": 49,
27
  "start": {
28
- "file": 198,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 7,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
- "offset": 2048,
37
- "documents": 16760881
38
  },
39
  "rewrite6": {
40
- "page": 28,
41
- "row": 636,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 114688
47
  }
48
  },
49
  "epoch": 46,
50
  "offset": 0,
51
- "documents": 9563908
52
  },
53
  "openmath": {
54
- "page": 2557,
55
- "row": 2242,
56
  "start": {
57
- "file": 23,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 430,
61
- "row": 522
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 10475714
67
  },
68
  "ultra_qa": {
69
- "page": 1105,
70
- "row": 3374,
71
  "start": {
72
  "file": 8,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
- "row_group": 5,
76
- "row": 46357
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
- "documents": 4529454
82
  },
83
  "ultra_style": {
84
- "page": 1687,
85
- "row": 3474,
86
  "start": {
87
  "file": 10,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
- "row_group": 0,
91
- "row": 60537
92
  }
93
  },
94
  "epoch": 0,
95
- "offset": 2048,
96
- "documents": 6913426
97
  },
98
  "cortex": {
99
- "page": 217307,
100
- "row": 113,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 27815409
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 835000,
3
  "stats": {
4
+ "climbmix/blocks": 30060000,
5
+ "climbmix/targets": 50250362065,
6
+ "climbmix/input_tokens": 50253546093,
7
+ "ultra_qa/blocks": 11272500,
8
+ "ultra_qa/targets": 18390084306,
9
+ "ultra_qa/input_tokens": 18390119687,
10
+ "cortex/blocks": 11272500,
11
+ "cortex/targets": 428641870,
12
+ "cortex/input_tokens": 17918505142,
13
+ "ultra_style/blocks": 11272500,
14
+ "ultra_style/targets": 17829475019,
15
+ "ultra_style/input_tokens": 17830410407,
16
+ "openmath/blocks": 7515000,
17
+ "openmath/targets": 11612035358,
18
+ "openmath/input_tokens": 11615904568,
19
+ "rewrite6/blocks": 3757500,
20
+ "rewrite6/targets": 6159911854,
21
+ "rewrite6/input_tokens": 6166665342
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 4116,
26
+ "row": 3220,
27
  "start": {
28
+ "file": 199,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 20,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
+ "offset": 0,
37
+ "documents": 16862356
38
  },
39
  "rewrite6": {
40
+ "page": 42,
41
+ "row": 790,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 172032
47
  }
48
  },
49
  "epoch": 46,
50
  "offset": 0,
51
+ "documents": 9621406
52
  },
53
  "openmath": {
54
+ "page": 2572,
55
+ "row": 4072,
56
  "start": {
57
+ "file": 24,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 55,
61
+ "row": 313
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 10538984
67
  },
68
  "ultra_qa": {
69
+ "page": 1112,
70
+ "row": 1914,
71
  "start": {
72
  "file": 8,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
+ "row_group": 6,
76
+ "row": 10507
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
+ "documents": 4556666
82
  },
83
  "ultra_style": {
84
+ "page": 1698,
85
+ "row": 306,
86
  "start": {
87
  "file": 10,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
+ "row_group": 1,
91
+ "row": 20264
92
  }
93
  },
94
  "epoch": 0,
95
+ "offset": 0,
96
+ "documents": 6955314
97
  },
98
  "cortex": {
99
+ "page": 218617,
100
+ "row": 105,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 27983081
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 830000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "148b9cd485ba37f4e9cce26a66ff0cccc598fbe722764f410999681d873c39a5", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "f6ee31a07eec8634f2a4579290acbcefb3cab5d7972f2153a10c3041f0c32cb5", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "eb45b2d9ecb051b6b7e592c18f2a0bfebf4278147b783ba131f7a9a4453a9ec7", "shard_scheduler.json": "d229c2adbc90c3e90d75f2ded34df13e9ded5f7d2aa7cf78783ee6b128d17bdd", "coverage.json": "537e5ac8f9bc339cc5920f3da24a226dfbd2e65ffea7c0b2d60b4a270d3159d5", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000830000"}
 
1
+ {"step": 835000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "434bc5e66416d00f38daf4d1c20bd1d7d620b0fc06f97108edac543c845dc140", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "8fc02097e06a5d8dde4c92534118b36500d860341974727dbdf1235db028d4da", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "eb45b2d9ecb051b6b7e592c18f2a0bfebf4278147b783ba131f7a9a4453a9ec7", "shard_scheduler.json": "15ec84a118722aa0da30b1eedddde0fa6f0d20effd89f99c89d67b1b955f9b9e", "coverage.json": "2268abb420311a79a47ad9499884d787f6b2b1069f26033253f7830b5909fc05", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000835000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:148b9cd485ba37f4e9cce26a66ff0cccc598fbe722764f410999681d873c39a5
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:434bc5e66416d00f38daf4d1c20bd1d7d620b0fc06f97108edac543c845dc140
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": 77014.49798515823,
26
- "required_pages_per_second_upper_estimate": 0.005328041668868387
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": 9626.812248144779,
40
- "required_pages_per_second_upper_estimate": 0.0040305452575985145
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": 19253.624496289558,
54
- "required_pages_per_second_upper_estimate": 0.00401002223035429
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": 28880.436744434337,
68
- "required_pages_per_second_upper_estimate": 0.0017313380849552794
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": 28880.436744434337,
82
- "required_pages_per_second_upper_estimate": 0.002245326925263646
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": 28880.436744434337,
96
- "required_pages_per_second_upper_estimate": 0.3704472331606103
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,408 +158,408 @@
158
  "prefetch_batches": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
- "producer_batch_seconds": 0.10640840249212147,
162
- "producer_p95_seconds": 0.2958205579998321,
163
- "gpu_update_seconds": 0.9573262428355817,
164
- "producer_batches_per_second": 9.397754092530809,
165
- "gpu_batches_per_second": 1.0445759817865428,
166
- "producer_headroom": 8.99671661649523,
167
  "sustainable": true,
168
- "buffer_seconds": 60.31155329864164,
169
  "buffer_drain_seconds": null,
170
- "shard_target_mib": 16,
171
- "upload_pages_per_commit": 83,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 311300.6360700732,
174
- "upload_bytes_per_second": 3156742.7663847003,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 66.79786979271552,
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": 1.913415155999246,
184
- "bytes": 2704493,
185
- "pages": 37,
186
  "shards": 2
187
  },
188
  {
189
- "seconds": 1.6119605509993562,
190
- "bytes": 6041654,
191
- "pages": 28,
192
- "shards": 2
193
  },
194
  {
195
- "seconds": 1.6828175019982154,
196
- "bytes": 10664698,
197
- "pages": 40,
198
  "shards": 3
199
  },
200
  {
201
- "seconds": 1.565426788998593,
202
- "bytes": 2415576,
203
- "pages": 30,
204
  "shards": 2
205
  },
206
  {
207
- "seconds": 1.5388405060002697,
208
- "bytes": 6065532,
209
- "pages": 36,
210
- "shards": 2
211
  },
212
  {
213
- "seconds": 1.3961634170009347,
214
- "bytes": 973137,
215
  "pages": 28,
216
- "shards": 1
217
  },
218
  {
219
- "seconds": 2.0744263130000036,
220
- "bytes": 17948904,
221
- "pages": 44,
222
- "shards": 5
223
  },
224
  {
225
- "seconds": 1.631489851999504,
226
- "bytes": 5794503,
227
- "pages": 28,
228
- "shards": 2
229
  },
230
  {
231
- "seconds": 1.5992371809988981,
232
- "bytes": 2589704,
233
  "pages": 37,
234
  "shards": 2
235
  },
236
  {
237
- "seconds": 1.4273169570005848,
238
- "bytes": 6151606,
239
- "pages": 36,
240
  "shards": 2
241
  },
242
  {
243
- "seconds": 1.4220550049976737,
244
- "bytes": 4624242,
245
- "pages": 30,
 
 
 
 
 
 
246
  "shards": 2
247
  },
248
  {
249
- "seconds": 2.047509969001112,
250
- "bytes": 7597965,
251
- "pages": 31,
252
  "shards": 3
253
  },
254
  {
255
- "seconds": 1.836660891000065,
256
- "bytes": 12233872,
257
- "pages": 41,
258
  "shards": 3
259
  },
260
  {
261
- "seconds": 1.7772512390001793,
262
- "bytes": 2617112,
263
- "pages": 37,
264
  "shards": 2
265
  },
266
  {
267
- "seconds": 1.5108418660020106,
268
- "bytes": 11046966,
269
- "pages": 31,
270
  "shards": 3
271
  },
272
  {
273
- "seconds": 1.4716807509976206,
274
- "bytes": 1094934,
275
- "pages": 36,
 
 
 
 
 
 
 
 
 
 
 
 
276
  "shards": 1
277
  },
278
  {
279
- "seconds": 1.608142429999134,
280
- "bytes": 7266564,
281
- "pages": 31,
282
  "shards": 3
283
  },
284
  {
285
- "seconds": 1.734035388999473,
286
- "bytes": 6027691,
287
  "pages": 29,
288
  "shards": 2
289
  },
290
  {
291
- "seconds": 1.5976459840021562,
292
- "bytes": 11304870,
293
- "pages": 44,
294
  "shards": 3
295
  },
296
  {
297
- "seconds": 1.606354307001311,
298
- "bytes": 7330991,
299
- "pages": 32,
300
- "shards": 3
301
  },
302
  {
303
- "seconds": 1.5559647560003214,
304
- "bytes": 6263907,
305
  "pages": 37,
306
  "shards": 2
307
  },
308
  {
309
- "seconds": 1.8874185050008236,
310
- "bytes": 2469051,
311
- "pages": 30,
312
  "shards": 2
313
  },
314
  {
315
- "seconds": 24.49332847400001,
316
- "bytes": 9957694,
317
- "pages": 37,
318
- "shards": 3
319
  },
320
  {
321
- "seconds": 1.9510848309982975,
322
- "bytes": 11992256,
323
- "pages": 40,
324
- "shards": 3
325
  },
326
  {
327
- "seconds": 1.68195884799934,
328
- "bytes": 2652603,
329
  "pages": 36,
330
- "shards": 3
331
- },
332
- {
333
- "seconds": 1.9732827289990382,
334
- "bytes": 10394345,
335
- "pages": 35,
336
- "shards": 3
337
- },
338
- {
339
- "seconds": 1.4147246300017287,
340
- "bytes": 2402371,
341
- "pages": 27,
342
  "shards": 2
343
  },
344
  {
345
- "seconds": 2.1062088899998344,
346
- "bytes": 6113756,
347
- "pages": 37,
348
  "shards": 2
349
  },
350
  {
351
- "seconds": 1.8720440979996056,
352
- "bytes": 6184562,
353
  "pages": 37,
354
- "shards": 2
355
  },
356
  {
357
- "seconds": 1.9332000800022797,
358
- "bytes": 8278775,
359
  "pages": 32,
360
- "shards": 3
361
  },
362
  {
363
- "seconds": 2.060914223002328,
364
- "bytes": 10411136,
365
- "pages": 38,
366
  "shards": 3
367
  },
368
  {
369
- "seconds": 1.9575544710023678,
370
- "bytes": 7322575,
371
- "pages": 31,
372
  "shards": 3
373
  },
374
  {
375
- "seconds": 1.413499946000229,
376
- "bytes": 1206249,
377
- "pages": 35,
378
- "shards": 1
379
- },
380
- {
381
- "seconds": 2.568047996999667,
382
- "bytes": 10001629,
383
- "pages": 35,
384
  "shards": 3
385
  },
386
  {
387
- "seconds": 1.5792092789997696,
388
- "bytes": 2645048,
389
- "pages": 36,
390
  "shards": 2
391
  },
392
  {
393
- "seconds": 2.121729968999716,
394
- "bytes": 11759509,
395
- "pages": 29,
396
- "shards": 3
397
  },
398
  {
399
- "seconds": 1.7052124439978797,
400
- "bytes": 6177071,
401
- "pages": 37,
402
  "shards": 2
403
  },
404
  {
405
- "seconds": 2.175897373002954,
406
- "bytes": 7190309,
407
- "pages": 36,
408
  "shards": 3
409
  },
410
  {
411
- "seconds": 1.8149941150004452,
412
- "bytes": 6150813,
413
- "pages": 37,
414
- "shards": 2
415
  },
416
  {
417
- "seconds": 1.5219117590022506,
418
- "bytes": 2396663,
419
  "pages": 28,
420
  "shards": 2
421
  },
422
  {
423
- "seconds": 1.5374675419989217,
424
- "bytes": 6254376,
425
  "pages": 36,
426
- "shards": 2
427
- },
428
- {
429
- "seconds": 1.5695442620017275,
430
- "bytes": 11600439,
431
- "pages": 29,
432
- "shards": 3
433
  },
434
  {
435
- "seconds": 1.7552413789999264,
436
- "bytes": 6467958,
437
- "pages": 38,
438
- "shards": 3
439
  },
440
  {
441
- "seconds": 1.6743759019991558,
442
- "bytes": 6275115,
443
- "pages": 37,
444
  "shards": 2
445
  },
446
  {
447
- "seconds": 2.272818253000878,
448
- "bytes": 7589463,
449
- "pages": 39,
450
- "shards": 3
451
  },
452
  {
453
- "seconds": 1.5132595840004797,
454
- "bytes": 5701329,
455
- "pages": 28,
456
- "shards": 2
457
  },
458
  {
459
- "seconds": 1.5889989900024375,
460
- "bytes": 6430899,
461
- "pages": 37,
462
  "shards": 2
463
  },
464
  {
465
- "seconds": 1.6179743659995438,
466
- "bytes": 8234542,
467
- "pages": 31,
468
  "shards": 3
469
  },
470
  {
471
- "seconds": 1.6133322549976583,
472
- "bytes": 6059251,
473
- "pages": 37,
474
- "shards": 2
475
  },
476
  {
477
- "seconds": 1.6795359120005742,
478
- "bytes": 5921876,
479
- "pages": 24,
480
- "shards": 2
481
  },
482
  {
483
- "seconds": 1.7577662710027653,
484
- "bytes": 6866333,
485
- "pages": 46,
486
- "shards": 3
487
  },
488
  {
489
- "seconds": 1.93685000899859,
490
- "bytes": 6023833,
491
- "pages": 33,
492
  "shards": 2
493
  },
494
  {
495
- "seconds": 1.8011142590003146,
496
- "bytes": 7499768,
497
- "pages": 34,
498
- "shards": 3
499
  },
500
  {
501
- "seconds": 1.6402012090002245,
502
- "bytes": 10865294,
503
- "pages": 22,
504
  "shards": 3
505
  },
506
  {
507
- "seconds": 1.7843240430011065,
508
- "bytes": 6397585,
509
- "pages": 42,
510
  "shards": 2
511
  },
512
  {
513
- "seconds": 1.4411041519997525,
514
- "bytes": 2727332,
515
- "pages": 38,
516
  "shards": 3
517
  },
518
  {
519
- "seconds": 1.913498541998706,
520
- "bytes": 6177731,
521
- "pages": 28,
522
  "shards": 2
523
  },
524
  {
525
- "seconds": 2.387397704002069,
526
- "bytes": 11768966,
 
 
 
 
 
 
527
  "pages": 40,
528
- "shards": 4
529
  },
530
  {
531
- "seconds": 1.479657814001257,
532
- "bytes": 836473,
533
- "pages": 27,
534
  "shards": 1
535
  },
536
  {
537
- "seconds": 1.9911568649986293,
538
- "bytes": 11961950,
539
- "pages": 42,
540
  "shards": 3
541
  },
542
  {
543
- "seconds": 1.958790611999575,
544
- "bytes": 2439502,
545
- "pages": 29,
546
- "shards": 2
547
  },
548
  {
549
- "seconds": 1.8466758049980854,
550
- "bytes": 10680095,
551
- "pages": 40,
552
- "shards": 3
553
  },
554
  {
555
- "seconds": 1.7479712950007524,
556
- "bytes": 6009903,
557
- "pages": 28,
558
- "shards": 2
559
  },
560
  {
561
- "seconds": 2.2005035470028815,
562
- "bytes": 2644353,
563
  "pages": 37,
564
  "shards": 2
565
  }
 
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": 76407.9111220439,
26
+ "required_pages_per_second_upper_estimate": 0.0052860765822027025
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": 9550.988890255487,
40
+ "required_pages_per_second_upper_estimate": 0.003998799601022032
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": 19101.977780510973,
54
+ "required_pages_per_second_upper_estimate": 0.00397843821865044
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": 28652.966670766462,
68
+ "required_pages_per_second_upper_estimate": 0.001717701601365582
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": 28652.966670766462,
82
+ "required_pages_per_second_upper_estimate": 0.0022276421275710828
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": 28652.966670766462,
96
+ "required_pages_per_second_upper_estimate": 0.3675294912939349
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": 0.10879093923438177,
162
+ "producer_p95_seconds": 0.32612680799502414,
163
+ "gpu_update_seconds": 0.964926261133309,
164
+ "producer_batches_per_second": 9.191941967203505,
165
+ "gpu_batches_per_second": 1.0363486209044583,
166
+ "producer_headroom": 8.86954619496803,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.79035445139847,
169
  "buffer_drain_seconds": null,
170
+ "shard_target_mib": 32,
171
+ "upload_pages_per_commit": 165,
172
  "max_pending_pages": 4096,
173
+ "generation_bytes_per_second": 666855.3792614836,
174
+ "upload_bytes_per_second": 3927796.2312513823,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 71.54582039539042,
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": 1.5374675419989217,
184
+ "bytes": 6254376,
185
+ "pages": 36,
186
  "shards": 2
187
  },
188
  {
189
+ "seconds": 1.5695442620017275,
190
+ "bytes": 11600439,
191
+ "pages": 29,
192
+ "shards": 3
193
  },
194
  {
195
+ "seconds": 1.7552413789999264,
196
+ "bytes": 6467958,
197
+ "pages": 38,
198
  "shards": 3
199
  },
200
  {
201
+ "seconds": 1.6743759019991558,
202
+ "bytes": 6275115,
203
+ "pages": 37,
204
  "shards": 2
205
  },
206
  {
207
+ "seconds": 2.272818253000878,
208
+ "bytes": 7589463,
209
+ "pages": 39,
210
+ "shards": 3
211
  },
212
  {
213
+ "seconds": 1.5132595840004797,
214
+ "bytes": 5701329,
215
  "pages": 28,
216
+ "shards": 2
217
  },
218
  {
219
+ "seconds": 1.5889989900024375,
220
+ "bytes": 6430899,
221
+ "pages": 37,
222
+ "shards": 2
223
  },
224
  {
225
+ "seconds": 1.6179743659995438,
226
+ "bytes": 8234542,
227
+ "pages": 31,
228
+ "shards": 3
229
  },
230
  {
231
+ "seconds": 1.6133322549976583,
232
+ "bytes": 6059251,
233
  "pages": 37,
234
  "shards": 2
235
  },
236
  {
237
+ "seconds": 1.6795359120005742,
238
+ "bytes": 5921876,
239
+ "pages": 24,
240
  "shards": 2
241
  },
242
  {
243
+ "seconds": 1.7577662710027653,
244
+ "bytes": 6866333,
245
+ "pages": 46,
246
+ "shards": 3
247
+ },
248
+ {
249
+ "seconds": 1.93685000899859,
250
+ "bytes": 6023833,
251
+ "pages": 33,
252
  "shards": 2
253
  },
254
  {
255
+ "seconds": 1.8011142590003146,
256
+ "bytes": 7499768,
257
+ "pages": 34,
258
  "shards": 3
259
  },
260
  {
261
+ "seconds": 1.6402012090002245,
262
+ "bytes": 10865294,
263
+ "pages": 22,
264
  "shards": 3
265
  },
266
  {
267
+ "seconds": 1.7843240430011065,
268
+ "bytes": 6397585,
269
+ "pages": 42,
270
  "shards": 2
271
  },
272
  {
273
+ "seconds": 1.4411041519997525,
274
+ "bytes": 2727332,
275
+ "pages": 38,
276
  "shards": 3
277
  },
278
  {
279
+ "seconds": 1.913498541998706,
280
+ "bytes": 6177731,
281
+ "pages": 28,
282
+ "shards": 2
283
+ },
284
+ {
285
+ "seconds": 2.387397704002069,
286
+ "bytes": 11768966,
287
+ "pages": 40,
288
+ "shards": 4
289
+ },
290
+ {
291
+ "seconds": 1.479657814001257,
292
+ "bytes": 836473,
293
+ "pages": 27,
294
  "shards": 1
295
  },
296
  {
297
+ "seconds": 1.9911568649986293,
298
+ "bytes": 11961950,
299
+ "pages": 42,
300
  "shards": 3
301
  },
302
  {
303
+ "seconds": 1.958790611999575,
304
+ "bytes": 2439502,
305
  "pages": 29,
306
  "shards": 2
307
  },
308
  {
309
+ "seconds": 1.8466758049980854,
310
+ "bytes": 10680095,
311
+ "pages": 40,
312
  "shards": 3
313
  },
314
  {
315
+ "seconds": 1.7479712950007524,
316
+ "bytes": 6009903,
317
+ "pages": 28,
318
+ "shards": 2
319
  },
320
  {
321
+ "seconds": 2.2005035470028815,
322
+ "bytes": 2644353,
323
  "pages": 37,
324
  "shards": 2
325
  },
326
  {
327
+ "seconds": 1.5131899659972987,
328
+ "bytes": 5828306,
329
+ "pages": 28,
330
  "shards": 2
331
  },
332
  {
333
+ "seconds": 1.8049468739991426,
334
+ "bytes": 4697774,
335
+ "pages": 26,
336
+ "shards": 2
337
  },
338
  {
339
+ "seconds": 2.496297884001251,
340
+ "bytes": 13693480,
341
+ "pages": 42,
342
+ "shards": 4
343
  },
344
  {
345
+ "seconds": 1.8265930059969833,
346
+ "bytes": 5855746,
347
  "pages": 36,
 
 
 
 
 
 
 
 
 
 
 
 
348
  "shards": 2
349
  },
350
  {
351
+ "seconds": 1.3942693540011533,
352
+ "bytes": 2663289,
353
+ "pages": 34,
354
  "shards": 2
355
  },
356
  {
357
+ "seconds": 1.5531271910003852,
358
+ "bytes": 10487346,
359
  "pages": 37,
360
+ "shards": 3
361
  },
362
  {
363
+ "seconds": 1.5084796250012005,
364
+ "bytes": 1063958,
365
  "pages": 32,
366
+ "shards": 1
367
  },
368
  {
369
+ "seconds": 1.9720517980022123,
370
+ "bytes": 7314841,
371
+ "pages": 27,
372
  "shards": 3
373
  },
374
  {
375
+ "seconds": 1.793492888999026,
376
+ "bytes": 11753421,
377
+ "pages": 38,
378
  "shards": 3
379
  },
380
  {
381
+ "seconds": 1.5141362069989555,
382
+ "bytes": 6622032,
383
+ "pages": 34,
 
 
 
 
 
 
384
  "shards": 3
385
  },
386
  {
387
+ "seconds": 1.5094903019999037,
388
+ "bytes": 6206747,
389
+ "pages": 38,
390
  "shards": 2
391
  },
392
  {
393
+ "seconds": 1.7292475489994104,
394
+ "bytes": 6419421,
395
+ "pages": 37,
396
+ "shards": 2
397
  },
398
  {
399
+ "seconds": 1.6914732460027153,
400
+ "bytes": 2366813,
401
+ "pages": 27,
402
  "shards": 2
403
  },
404
  {
405
+ "seconds": 1.8487111699978414,
406
+ "bytes": 10763047,
407
+ "pages": 31,
408
  "shards": 3
409
  },
410
  {
411
+ "seconds": 1.605599512000481,
412
+ "bytes": 8445757,
413
+ "pages": 39,
414
+ "shards": 3
415
  },
416
  {
417
+ "seconds": 1.6673135730015929,
418
+ "bytes": 6027005,
419
  "pages": 28,
420
  "shards": 2
421
  },
422
  {
423
+ "seconds": 1.4991452859976562,
424
+ "bytes": 1121499,
425
  "pages": 36,
426
+ "shards": 1
 
 
 
 
 
 
427
  },
428
  {
429
+ "seconds": 2.2660662760026753,
430
+ "bytes": 11276737,
431
+ "pages": 36,
432
+ "shards": 4
433
  },
434
  {
435
+ "seconds": 1.8010207819970674,
436
+ "bytes": 5858104,
437
+ "pages": 32,
438
  "shards": 2
439
  },
440
  {
441
+ "seconds": 1.3410764899963397,
442
+ "bytes": 1019830,
443
+ "pages": 31,
444
+ "shards": 1
445
  },
446
  {
447
+ "seconds": 1.8211984559966368,
448
+ "bytes": 13136869,
449
+ "pages": 32,
450
+ "shards": 4
451
  },
452
  {
453
+ "seconds": 1.4420257410019985,
454
+ "bytes": 5442386,
455
+ "pages": 33,
456
  "shards": 2
457
  },
458
  {
459
+ "seconds": 2.6930293979967246,
460
+ "bytes": 5991867,
461
+ "pages": 29,
462
  "shards": 3
463
  },
464
  {
465
+ "seconds": 1.7729905640007928,
466
+ "bytes": 7619842,
467
+ "pages": 39,
468
+ "shards": 3
469
  },
470
  {
471
+ "seconds": 1.2987675089971162,
472
+ "bytes": 1223490,
473
+ "pages": 35,
474
+ "shards": 1
475
  },
476
  {
477
+ "seconds": 1.546362834997126,
478
+ "bytes": 11822887,
479
+ "pages": 33,
480
+ "shards": 4
481
  },
482
  {
483
+ "seconds": 1.5667999549987144,
484
+ "bytes": 6923893,
485
+ "pages": 35,
486
  "shards": 2
487
  },
488
  {
489
+ "seconds": 1.4826895270016394,
490
+ "bytes": 6163509,
491
+ "pages": 32,
492
+ "shards": 2
493
  },
494
  {
495
+ "seconds": 1.7263165549957193,
496
+ "bytes": 7465725,
497
+ "pages": 39,
498
  "shards": 3
499
  },
500
  {
501
+ "seconds": 2.012805174999812,
502
+ "bytes": 4595806,
503
+ "pages": 30,
504
  "shards": 2
505
  },
506
  {
507
+ "seconds": 1.6076512380022905,
508
+ "bytes": 7510724,
509
+ "pages": 40,
510
  "shards": 3
511
  },
512
  {
513
+ "seconds": 1.5883049809999648,
514
+ "bytes": 5665128,
515
+ "pages": 27,
516
  "shards": 2
517
  },
518
  {
519
+ "seconds": 2.0745964650050155,
520
+ "bytes": 11457524,
521
+ "pages": 35,
522
+ "shards": 3
523
+ },
524
+ {
525
+ "seconds": 1.6211695970050641,
526
+ "bytes": 7634724,
527
  "pages": 40,
528
+ "shards": 3
529
  },
530
  {
531
+ "seconds": 1.4634708860030514,
532
+ "bytes": 987981,
533
+ "pages": 29,
534
  "shards": 1
535
  },
536
  {
537
+ "seconds": 1.6518487119974452,
538
+ "bytes": 7534957,
539
+ "pages": 37,
540
  "shards": 3
541
  },
542
  {
543
+ "seconds": 1.5817108970004483,
544
+ "bytes": 10450177,
545
+ "pages": 33,
546
+ "shards": 3
547
  },
548
  {
549
+ "seconds": 1.328966004999529,
550
+ "bytes": 1184587,
551
+ "pages": 36,
552
+ "shards": 1
553
  },
554
  {
555
+ "seconds": 2.0095185120007955,
556
+ "bytes": 12961722,
557
+ "pages": 36,
558
+ "shards": 4
559
  },
560
  {
561
+ "seconds": 1.4560003849983332,
562
+ "bytes": 6473170,
563
  "pages": 37,
564
  "shards": 2
565
  }
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:f6ee31a07eec8634f2a4579290acbcefb3cab5d7972f2153a10c3041f0c32cb5
3
- size 721775359
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8fc02097e06a5d8dde4c92534118b36500d860341974727dbdf1235db028d4da
3
+ size 721778239