appvoid commited on
Commit
4459913
·
verified ·
1 Parent(s): 7c16afe

Qwen90M step 470000; 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 +237 -237
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 465000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "a36c516961ecbdb4d135f7aec065aa33ba895b425e41adf7a7130615b0e7c534",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "5d3c17f33f3432629d704e2ea2debd635921d926d1bb7e925d71c43aa06fd061",
12
  "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
13
  "performance.json": "58f8b6f79b191aa11b2bedb4594ca2897b6fdad2419ea7c9210688ac8022e7cb",
14
- "shard_scheduler.json": "c18cd8a1a78ebd1e9db154aa776adb8bac1bfd7543f18628291873cd4bfa08f0",
15
- "coverage.json": "f26a841a3097414f1b1263238f23d6076c33c6431b047f162e8bf7038c932c5f",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 470000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "e476e5303342f155469c3f56a579382f525ddf8789c7fe638747144789d71895",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "629577df31a18ce0968102b0e3f539c35489cebe6a6ddc046c8ab547af131faa",
12
  "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
13
  "performance.json": "58f8b6f79b191aa11b2bedb4594ca2897b6fdad2419ea7c9210688ac8022e7cb",
14
+ "shard_scheduler.json": "0d21f56fc9331c0f8678684332a3f56087d5f2a22aa6813b19d0321679297611",
15
+ "coverage.json": "e5989284b0b1414d692537fe449a887c54c77be9ca79006a1b7f013bdd941e1d",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 465000,
3
  "stats": {
4
- "climbmix/blocks": 16740000,
5
- "climbmix/targets": 27982214264,
6
- "climbmix/input_tokens": 27983987016,
7
- "ultra_qa/blocks": 6277500,
8
- "ultra_qa/targets": 10240630112,
9
- "ultra_qa/input_tokens": 10240649816,
10
- "cortex/blocks": 6277500,
11
- "cortex/targets": 238761321,
12
- "cortex/input_tokens": 9979079341,
13
- "ultra_style/blocks": 6277500,
14
- "ultra_style/targets": 9929669455,
15
- "ultra_style/input_tokens": 9930189433,
16
- "openmath/blocks": 4185000,
17
- "openmath/targets": 6466843653,
18
- "openmath/input_tokens": 6468998513,
19
- "rewrite6/blocks": 2092500,
20
- "rewrite6/targets": 3430367981,
21
- "rewrite6/input_tokens": 3434128818
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 2292,
26
- "row": 3018,
27
  "start": {
28
- "file": 110,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 77,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
- "offset": 0,
37
- "documents": 9391050
38
  },
39
  "rewrite6": {
40
- "page": 4,
41
- "row": 985,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 16384
47
  }
48
  },
49
  "epoch": 26,
50
  "offset": 2048,
51
- "documents": 5357873
52
  },
53
  "openmath": {
54
- "page": 1432,
55
- "row": 3738,
56
  "start": {
57
  "file": 13,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 189,
61
- "row": 22
62
  }
63
  },
64
  "epoch": 0,
65
- "offset": 2048,
66
- "documents": 5869210
67
  },
68
  "ultra_qa": {
69
- "page": 619,
70
- "row": 2226,
71
  "start": {
72
  "file": 4,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 7,
76
- "row": 4726
77
  }
78
  },
79
  "epoch": 0,
80
- "offset": 2048,
81
- "documents": 2537650
82
  },
83
  "ultra_style": {
84
- "page": 945,
85
- "row": 653,
86
  "start": {
87
  "file": 5,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 5,
91
- "row": 21316
92
  }
93
  },
94
  "epoch": 0,
95
- "offset": 2048,
96
- "documents": 3871373
97
  },
98
  "cortex": {
99
- "page": 121754,
100
- "row": 94,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 15584606
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 470000,
3
  "stats": {
4
+ "climbmix/blocks": 16920000,
5
+ "climbmix/targets": 28282840249,
6
+ "climbmix/input_tokens": 28284632204,
7
+ "ultra_qa/blocks": 6345000,
8
+ "ultra_qa/targets": 10350703632,
9
+ "ultra_qa/input_tokens": 10350723557,
10
+ "cortex/blocks": 6345000,
11
+ "cortex/targets": 241332160,
12
+ "cortex/input_tokens": 10086409319,
13
+ "ultra_style/blocks": 6345000,
14
+ "ultra_style/targets": 10036596659,
15
+ "ultra_style/input_tokens": 10037122085,
16
+ "openmath/blocks": 4230000,
17
+ "openmath/targets": 6536344953,
18
+ "openmath/input_tokens": 6538523066,
19
+ "rewrite6/blocks": 2115000,
20
+ "rewrite6/targets": 3467211486,
21
+ "rewrite6/input_tokens": 3471013006
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 2317,
26
+ "row": 1972,
27
  "start": {
28
+ "file": 112,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 11,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
+ "offset": 2048,
37
+ "documents": 9492404
38
  },
39
  "rewrite6": {
40
+ "page": 18,
41
+ "row": 1545,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 73728
47
  }
48
  },
49
  "epoch": 26,
50
  "offset": 2048,
51
+ "documents": 5415777
52
  },
53
  "openmath": {
54
+ "page": 1448,
55
+ "row": 1399,
56
  "start": {
57
  "file": 13,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 254,
61
+ "row": 558
62
  }
63
  },
64
  "epoch": 0,
65
+ "offset": 0,
66
+ "documents": 5932407
67
  },
68
  "ultra_qa": {
69
+ "page": 626,
70
+ "row": 855,
71
  "start": {
72
  "file": 4,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 7,
76
+ "row": 33398
77
  }
78
  },
79
  "epoch": 0,
80
+ "offset": 4096,
81
+ "documents": 2564951
82
  },
83
  "ultra_style": {
84
+ "page": 955,
85
+ "row": 1084,
86
  "start": {
87
  "file": 5,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 5,
91
+ "row": 62276
92
  }
93
  },
94
  "epoch": 0,
95
+ "offset": 0,
96
+ "documents": 3912764
97
  },
98
  "cortex": {
99
+ "page": 123063,
100
+ "row": 96,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 15752160
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 465000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "a36c516961ecbdb4d135f7aec065aa33ba895b425e41adf7a7130615b0e7c534", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "5d3c17f33f3432629d704e2ea2debd635921d926d1bb7e925d71c43aa06fd061", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "58f8b6f79b191aa11b2bedb4594ca2897b6fdad2419ea7c9210688ac8022e7cb", "shard_scheduler.json": "c18cd8a1a78ebd1e9db154aa776adb8bac1bfd7543f18628291873cd4bfa08f0", "coverage.json": "f26a841a3097414f1b1263238f23d6076c33c6431b047f162e8bf7038c932c5f", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000465000"}
 
1
+ {"step": 470000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "e476e5303342f155469c3f56a579382f525ddf8789c7fe638747144789d71895", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "629577df31a18ce0968102b0e3f539c35489cebe6a6ddc046c8ab547af131faa", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "58f8b6f79b191aa11b2bedb4594ca2897b6fdad2419ea7c9210688ac8022e7cb", "shard_scheduler.json": "0d21f56fc9331c0f8678684332a3f56087d5f2a22aa6813b19d0321679297611", "coverage.json": "e5989284b0b1414d692537fe449a887c54c77be9ca79006a1b7f013bdd941e1d", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000470000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:a36c516961ecbdb4d135f7aec065aa33ba895b425e41adf7a7130615b0e7c534
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e476e5303342f155469c3f56a579382f525ddf8789c7fe638747144789d71895
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": 76285.71196533572,
26
- "required_pages_per_second_upper_estimate": 0.005277622561000523
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": 9535.713995666965,
40
- "required_pages_per_second_upper_estimate": 0.003992404321633889
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": 19071.42799133393,
54
- "required_pages_per_second_upper_estimate": 0.00397207550321698
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": 28607.141987000894,
68
- "required_pages_per_second_upper_estimate": 0.0017149544815440768
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": 28607.141987000894,
82
- "required_pages_per_second_upper_estimate": 0.002224079459969792
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": 28607.141987000894,
96
- "required_pages_per_second_upper_estimate": 0.36694170145330224
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": 0.21675075976563107,
162
- "producer_p95_seconds": 0.2752362049977819,
163
- "gpu_update_seconds": 0.9664719395094858,
164
- "producer_batches_per_second": 4.613593978084705,
165
- "gpu_batches_per_second": 1.0346911887659467,
166
- "producer_headroom": 4.4589091201088085,
167
  "sustainable": true,
168
- "buffer_seconds": 60.887732189097605,
169
  "buffer_drain_seconds": null,
170
- "shard_target_mib": 32,
171
- "upload_pages_per_commit": 169,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 614227.4902624867,
174
- "upload_bytes_per_second": 4817176.597645194,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 65.89946046307541,
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": 1.1430494310006907,
184
- "bytes": 10900083,
185
- "pages": 37,
186
- "shards": 3
187
  },
188
  {
189
- "seconds": 1.591609420000168,
190
- "bytes": 11822162,
191
- "pages": 35,
192
- "shards": 3
193
  },
194
  {
195
- "seconds": 0.9329451240009803,
196
- "bytes": 2485686,
197
- "pages": 29,
198
  "shards": 2
199
  },
200
  {
201
- "seconds": 4.417064211000252,
202
- "bytes": 6221614,
203
- "pages": 37,
204
  "shards": 2
205
  },
206
  {
207
- "seconds": 1.0521548180004174,
208
- "bytes": 6267595,
209
- "pages": 39,
210
- "shards": 3
211
  },
212
  {
213
- "seconds": 0.9623302439977124,
214
- "bytes": 6111937,
215
- "pages": 28,
216
  "shards": 2
217
  },
218
  {
219
- "seconds": 1.1251186529989354,
220
- "bytes": 6276648,
221
- "pages": 36,
222
  "shards": 2
223
  },
224
  {
225
- "seconds": 1.272974397001235,
226
- "bytes": 8064738,
227
- "pages": 28,
228
- "shards": 3
229
- },
230
- {
231
- "seconds": 1.1465204630003427,
232
- "bytes": 10636216,
233
- "pages": 42,
234
  "shards": 3
235
  },
236
  {
237
- "seconds": 0.9501123180016293,
238
- "bytes": 1078327,
239
- "pages": 34,
240
- "shards": 1
241
  },
242
  {
243
- "seconds": 3.8270818399978452,
244
- "bytes": 7309721,
245
- "pages": 35,
246
- "shards": 4
247
  },
248
  {
249
- "seconds": 1.1971988570003305,
250
- "bytes": 10569382,
251
- "pages": 35,
252
- "shards": 3
253
  },
254
  {
255
- "seconds": 0.8849704640015261,
256
- "bytes": 2587633,
257
- "pages": 35,
258
  "shards": 2
259
  },
260
  {
261
- "seconds": 2.1376686900002824,
262
- "bytes": 11886724,
263
  "pages": 33,
264
  "shards": 3
265
  },
266
  {
267
- "seconds": 1.1407382979996328,
268
- "bytes": 5980776,
269
- "pages": 35,
270
  "shards": 2
271
  },
272
  {
273
- "seconds": 0.9373651969981438,
274
- "bytes": 2246840,
275
- "pages": 25,
276
  "shards": 2
277
  },
278
  {
279
- "seconds": 2.0274646729994856,
280
- "bytes": 10144494,
281
- "pages": 44,
282
- "shards": 3
283
  },
284
  {
285
- "seconds": 2.5764055170002393,
286
- "bytes": 7638759,
287
  "pages": 39,
288
  "shards": 3
289
  },
290
  {
291
- "seconds": 0.8547688260005089,
292
- "bytes": 858247,
293
- "pages": 26,
294
  "shards": 1
295
  },
296
  {
297
- "seconds": 1.4054743929991673,
298
- "bytes": 16596361,
299
  "pages": 32,
300
- "shards": 4
301
- },
302
- {
303
- "seconds": 3.107019715000206,
304
- "bytes": 2583987,
305
- "pages": 34,
306
  "shards": 2
307
  },
308
  {
309
- "seconds": 1.155031209000299,
310
- "bytes": 6268449,
311
- "pages": 36,
312
- "shards": 2
313
  },
314
  {
315
- "seconds": 0.8907613930023217,
316
- "bytes": 6205142,
317
  "pages": 37,
318
  "shards": 2
319
  },
320
  {
321
- "seconds": 1.1056236590011395,
322
- "bytes": 6648792,
323
- "pages": 34,
324
  "shards": 3
325
  },
326
  {
327
- "seconds": 1.084016504999454,
328
- "bytes": 5907243,
329
- "pages": 27,
330
  "shards": 2
331
  },
332
  {
333
- "seconds": 1.0631858130000182,
334
- "bytes": 6774777,
335
- "pages": 30,
336
  "shards": 2
337
  },
338
  {
339
- "seconds": 1.0792735680006444,
340
- "bytes": 7390991,
341
- "pages": 38,
342
  "shards": 3
343
  },
344
  {
345
- "seconds": 1.0645652359999076,
346
- "bytes": 834051,
347
- "pages": 23,
348
- "shards": 1
349
- },
350
- {
351
- "seconds": 1.34710112600078,
352
- "bytes": 12646439,
353
- "pages": 42,
354
- "shards": 4
355
  },
356
  {
357
- "seconds": 1.121954108999489,
358
- "bytes": 5958053,
359
- "pages": 34,
360
  "shards": 2
361
  },
362
  {
363
- "seconds": 0.8286497560002317,
364
- "bytes": 1212322,
365
- "pages": 35,
366
- "shards": 1
367
  },
368
  {
369
- "seconds": 1.1948913660016842,
370
- "bytes": 13257800,
371
- "pages": 36,
372
- "shards": 4
373
  },
374
  {
375
- "seconds": 1.0794521610005177,
376
- "bytes": 9877213,
377
- "pages": 34,
378
- "shards": 3
379
  },
380
  {
381
- "seconds": 1.2257234370008518,
382
- "bytes": 2680179,
383
  "pages": 38,
384
  "shards": 2
385
  },
386
  {
387
- "seconds": 1.3023044059991662,
388
- "bytes": 5788392,
389
- "pages": 27,
 
 
 
 
 
 
390
  "shards": 2
391
  },
392
  {
393
- "seconds": 1.0160050430022238,
394
- "bytes": 1159604,
395
- "pages": 32,
396
- "shards": 1
397
  },
398
  {
399
- "seconds": 1.8482585339988873,
400
- "bytes": 12054418,
401
- "pages": 40,
402
- "shards": 4
403
  },
404
  {
405
- "seconds": 2.080120522998186,
406
- "bytes": 11568040,
407
- "pages": 30,
408
  "shards": 3
409
  },
410
  {
411
- "seconds": 0.9286353159986902,
412
- "bytes": 2598650,
413
- "pages": 37,
414
  "shards": 2
415
  },
416
  {
417
- "seconds": 1.0528518130013254,
418
- "bytes": 5770251,
419
- "pages": 27,
420
  "shards": 2
421
  },
422
  {
423
- "seconds": 1.3281948670010024,
424
- "bytes": 5971602,
425
- "pages": 35,
426
  "shards": 2
427
  },
428
  {
429
- "seconds": 1.442292184001417,
430
- "bytes": 7645104,
431
- "pages": 40,
432
  "shards": 4
433
  },
434
  {
435
- "seconds": 1.229033948999131,
436
- "bytes": 6359372,
437
- "pages": 38,
438
  "shards": 2
439
  },
440
  {
441
- "seconds": 0.9523748860010528,
442
- "bytes": 6784780,
443
- "pages": 33,
444
  "shards": 2
445
  },
446
  {
447
- "seconds": 1.1795219960004033,
448
- "bytes": 10882665,
449
- "pages": 33,
450
- "shards": 4
451
  },
452
  {
453
- "seconds": 1.0230853959983506,
454
- "bytes": 6145347,
455
  "pages": 37,
456
  "shards": 2
457
  },
458
  {
459
- "seconds": 1.223426783002651,
460
- "bytes": 2322033,
461
- "pages": 29,
462
  "shards": 2
463
  },
464
  {
465
- "seconds": 1.2023178439994808,
466
- "bytes": 11142768,
467
- "pages": 40,
468
  "shards": 3
469
  },
470
  {
471
- "seconds": 1.175540739997814,
472
- "bytes": 6085764,
473
- "pages": 28,
474
  "shards": 2
475
  },
476
  {
477
- "seconds": 1.1963462499988964,
478
- "bytes": 8278514,
479
- "pages": 33,
480
  "shards": 3
481
  },
482
  {
483
- "seconds": 0.9674637549978797,
484
- "bytes": 6111470,
485
- "pages": 36,
486
  "shards": 2
487
  },
488
  {
489
- "seconds": 1.2372247309976956,
490
- "bytes": 2619840,
491
- "pages": 37,
492
  "shards": 2
493
  },
494
  {
495
- "seconds": 2.7946544099977473,
496
- "bytes": 9699006,
497
- "pages": 33,
498
  "shards": 3
499
  },
500
  {
501
- "seconds": 1.143838866999431,
502
- "bytes": 6205051,
503
- "pages": 37,
 
 
 
 
 
 
504
  "shards": 2
505
  },
506
  {
507
- "seconds": 3.739576465002756,
508
- "bytes": 2400719,
509
- "pages": 29,
510
  "shards": 2
511
  },
512
  {
513
- "seconds": 1.58055241799957,
514
- "bytes": 15825794,
515
- "pages": 36,
516
- "shards": 4
517
  },
518
  {
519
- "seconds": 1.061300241999561,
520
- "bytes": 7669220,
521
- "pages": 39,
522
- "shards": 3
523
  },
524
  {
525
- "seconds": 1.3649325709993718,
526
- "bytes": 1043801,
527
- "pages": 32,
528
  "shards": 1
529
  },
530
  {
531
- "seconds": 1.102126538000448,
532
- "bytes": 5859413,
533
- "pages": 32,
534
- "shards": 2
535
  },
536
  {
537
- "seconds": 1.2403170809993753,
538
- "bytes": 7520295,
539
- "pages": 35,
540
  "shards": 3
541
  },
542
  {
543
- "seconds": 1.3349149159985245,
544
- "bytes": 6151647,
545
- "pages": 37,
546
  "shards": 2
547
  },
548
  {
549
- "seconds": 1.3665102609993482,
550
- "bytes": 11866902,
551
- "pages": 32,
552
  "shards": 3
553
  },
554
  {
555
- "seconds": 1.1544846530014183,
556
- "bytes": 2585533,
557
- "pages": 37,
558
  "shards": 2
559
  },
560
  {
561
- "seconds": 1.3021024750014476,
562
- "bytes": 5911875,
563
- "pages": 28,
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": 77031.49068797404,
26
+ "required_pages_per_second_upper_estimate": 0.005329217263477687
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": 9628.936335996756,
40
+ "required_pages_per_second_upper_estimate": 0.004031434568826139
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": 19257.87267199351,
54
+ "required_pages_per_second_upper_estimate": 0.004010907013321496
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": 28886.809007990265,
68
+ "required_pages_per_second_upper_estimate": 0.0017317200924255743
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": 28886.809007990265,
82
+ "required_pages_per_second_upper_estimate": 0.002245822340726494
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": 28886.809007990265,
96
+ "required_pages_per_second_upper_estimate": 0.3705289697155022
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": 0.10171361467956785,
162
+ "producer_p95_seconds": 0.3008065889989666,
163
+ "gpu_update_seconds": 0.9571150621846947,
164
+ "producer_batches_per_second": 9.831525535203294,
165
+ "gpu_batches_per_second": 1.0448064600690923,
166
+ "producer_headroom": 9.409901173996515,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.29824891763576,
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": 320066.6025585662,
174
+ "upload_bytes_per_second": 5515946.4320974825,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 68.6788421398901,
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": 1.3281948670010024,
184
+ "bytes": 5971602,
185
+ "pages": 35,
186
+ "shards": 2
187
  },
188
  {
189
+ "seconds": 1.442292184001417,
190
+ "bytes": 7645104,
191
+ "pages": 40,
192
+ "shards": 4
193
  },
194
  {
195
+ "seconds": 1.229033948999131,
196
+ "bytes": 6359372,
197
+ "pages": 38,
198
  "shards": 2
199
  },
200
  {
201
+ "seconds": 0.9523748860010528,
202
+ "bytes": 6784780,
203
+ "pages": 33,
204
  "shards": 2
205
  },
206
  {
207
+ "seconds": 1.1795219960004033,
208
+ "bytes": 10882665,
209
+ "pages": 33,
210
+ "shards": 4
211
  },
212
  {
213
+ "seconds": 1.0230853959983506,
214
+ "bytes": 6145347,
215
+ "pages": 37,
216
  "shards": 2
217
  },
218
  {
219
+ "seconds": 1.223426783002651,
220
+ "bytes": 2322033,
221
+ "pages": 29,
222
  "shards": 2
223
  },
224
  {
225
+ "seconds": 1.2023178439994808,
226
+ "bytes": 11142768,
227
+ "pages": 40,
 
 
 
 
 
 
228
  "shards": 3
229
  },
230
  {
231
+ "seconds": 1.175540739997814,
232
+ "bytes": 6085764,
233
+ "pages": 28,
234
+ "shards": 2
235
  },
236
  {
237
+ "seconds": 1.1963462499988964,
238
+ "bytes": 8278514,
239
+ "pages": 33,
240
+ "shards": 3
241
  },
242
  {
243
+ "seconds": 0.9674637549978797,
244
+ "bytes": 6111470,
245
+ "pages": 36,
246
+ "shards": 2
247
  },
248
  {
249
+ "seconds": 1.2372247309976956,
250
+ "bytes": 2619840,
251
+ "pages": 37,
252
  "shards": 2
253
  },
254
  {
255
+ "seconds": 2.7946544099977473,
256
+ "bytes": 9699006,
257
  "pages": 33,
258
  "shards": 3
259
  },
260
  {
261
+ "seconds": 1.143838866999431,
262
+ "bytes": 6205051,
263
+ "pages": 37,
264
  "shards": 2
265
  },
266
  {
267
+ "seconds": 3.739576465002756,
268
+ "bytes": 2400719,
269
+ "pages": 29,
270
  "shards": 2
271
  },
272
  {
273
+ "seconds": 1.58055241799957,
274
+ "bytes": 15825794,
275
+ "pages": 36,
276
+ "shards": 4
277
  },
278
  {
279
+ "seconds": 1.061300241999561,
280
+ "bytes": 7669220,
281
  "pages": 39,
282
  "shards": 3
283
  },
284
  {
285
+ "seconds": 1.3649325709993718,
286
+ "bytes": 1043801,
287
+ "pages": 32,
288
  "shards": 1
289
  },
290
  {
291
+ "seconds": 1.102126538000448,
292
+ "bytes": 5859413,
293
  "pages": 32,
 
 
 
 
 
 
294
  "shards": 2
295
  },
296
  {
297
+ "seconds": 1.2403170809993753,
298
+ "bytes": 7520295,
299
+ "pages": 35,
300
+ "shards": 3
301
  },
302
  {
303
+ "seconds": 1.3349149159985245,
304
+ "bytes": 6151647,
305
  "pages": 37,
306
  "shards": 2
307
  },
308
  {
309
+ "seconds": 1.3665102609993482,
310
+ "bytes": 11866902,
311
+ "pages": 32,
312
  "shards": 3
313
  },
314
  {
315
+ "seconds": 1.1544846530014183,
316
+ "bytes": 2585533,
317
+ "pages": 37,
318
  "shards": 2
319
  },
320
  {
321
+ "seconds": 1.3021024750014476,
322
+ "bytes": 5911875,
323
+ "pages": 28,
324
  "shards": 2
325
  },
326
  {
327
+ "seconds": 1.0596985519987356,
328
+ "bytes": 6289450,
329
+ "pages": 35,
330
  "shards": 3
331
  },
332
  {
333
+ "seconds": 1.1195497589978913,
334
+ "bytes": 6194218,
335
+ "pages": 37,
336
+ "shards": 2
 
 
 
 
 
 
337
  },
338
  {
339
+ "seconds": 1.2359527980006533,
340
+ "bytes": 6034339,
341
+ "pages": 29,
342
  "shards": 2
343
  },
344
  {
345
+ "seconds": 1.2140417409973452,
346
+ "bytes": 13269645,
347
+ "pages": 40,
348
+ "shards": 4
349
  },
350
  {
351
+ "seconds": 1.3757806130015524,
352
+ "bytes": 5911161,
353
+ "pages": 29,
354
+ "shards": 2
355
  },
356
  {
357
+ "seconds": 1.0065856220026035,
358
+ "bytes": 6244496,
359
+ "pages": 36,
360
+ "shards": 2
361
  },
362
  {
363
+ "seconds": 0.9367298329998448,
364
+ "bytes": 2789154,
365
  "pages": 38,
366
  "shards": 2
367
  },
368
  {
369
+ "seconds": 1.218474695997429,
370
+ "bytes": 10329696,
371
+ "pages": 35,
372
+ "shards": 4
373
+ },
374
+ {
375
+ "seconds": 0.9623371359994053,
376
+ "bytes": 2372191,
377
+ "pages": 29,
378
  "shards": 2
379
  },
380
  {
381
+ "seconds": 1.2589823859998432,
382
+ "bytes": 12076255,
383
+ "pages": 34,
384
+ "shards": 3
385
  },
386
  {
387
+ "seconds": 1.1457863180003187,
388
+ "bytes": 6106429,
389
+ "pages": 36,
390
+ "shards": 2
391
  },
392
  {
393
+ "seconds": 0.9636820249979792,
394
+ "bytes": 6954429,
395
+ "pages": 37,
396
  "shards": 3
397
  },
398
  {
399
+ "seconds": 1.3229683749996184,
400
+ "bytes": 6064094,
401
+ "pages": 34,
402
  "shards": 2
403
  },
404
  {
405
+ "seconds": 0.8354571769996255,
406
+ "bytes": 2427677,
407
+ "pages": 30,
408
  "shards": 2
409
  },
410
  {
411
+ "seconds": 1.1609949890007556,
412
+ "bytes": 6207129,
413
+ "pages": 38,
414
  "shards": 2
415
  },
416
  {
417
+ "seconds": 1.1704257509991294,
418
+ "bytes": 15475495,
419
+ "pages": 35,
420
  "shards": 4
421
  },
422
  {
423
+ "seconds": 1.1073925720011175,
424
+ "bytes": 2592720,
425
+ "pages": 36,
426
  "shards": 2
427
  },
428
  {
429
+ "seconds": 0.9348203119989194,
430
+ "bytes": 5613969,
431
+ "pages": 28,
432
  "shards": 2
433
  },
434
  {
435
+ "seconds": 1.1180898200000229,
436
+ "bytes": 7687731,
437
+ "pages": 38,
438
+ "shards": 3
439
  },
440
  {
441
+ "seconds": 1.1014808299987635,
442
+ "bytes": 5892940,
443
  "pages": 37,
444
  "shards": 2
445
  },
446
  {
447
+ "seconds": 0.8753045230005227,
448
+ "bytes": 5867523,
449
+ "pages": 28,
450
  "shards": 2
451
  },
452
  {
453
+ "seconds": 1.8846877780015348,
454
+ "bytes": 8419016,
455
+ "pages": 36,
456
  "shards": 3
457
  },
458
  {
459
+ "seconds": 0.9963944320006703,
460
+ "bytes": 6337215,
461
+ "pages": 37,
462
  "shards": 2
463
  },
464
  {
465
+ "seconds": 1.0649026120008784,
466
+ "bytes": 9997467,
467
+ "pages": 31,
468
  "shards": 3
469
  },
470
  {
471
+ "seconds": 1.0817448720008542,
472
+ "bytes": 2541266,
473
+ "pages": 37,
474
  "shards": 2
475
  },
476
  {
477
+ "seconds": 1.4285129730014887,
478
+ "bytes": 6270013,
479
+ "pages": 36,
480
  "shards": 2
481
  },
482
  {
483
+ "seconds": 1.485799860998668,
484
+ "bytes": 7350955,
485
+ "pages": 31,
486
  "shards": 3
487
  },
488
  {
489
+ "seconds": 1.3855582350006443,
490
+ "bytes": 11926342,
491
+ "pages": 38,
492
+ "shards": 3
493
+ },
494
+ {
495
+ "seconds": 1.1727449809986865,
496
+ "bytes": 5685962,
497
+ "pages": 28,
498
  "shards": 2
499
  },
500
  {
501
+ "seconds": 1.2225738800007093,
502
+ "bytes": 2613097,
503
+ "pages": 37,
504
  "shards": 2
505
  },
506
  {
507
+ "seconds": 1.0333214540005429,
508
+ "bytes": 6358222,
509
+ "pages": 38,
510
+ "shards": 2
511
  },
512
  {
513
+ "seconds": 1.2548035990002973,
514
+ "bytes": 11160297,
515
+ "pages": 33,
516
+ "shards": 4
517
  },
518
  {
519
+ "seconds": 1.0187819040002069,
520
+ "bytes": 1294212,
521
+ "pages": 35,
522
  "shards": 1
523
  },
524
  {
525
+ "seconds": 1.2491005060001044,
526
+ "bytes": 11541986,
527
+ "pages": 30,
528
+ "shards": 3
529
  },
530
  {
531
+ "seconds": 1.6100285119973705,
532
+ "bytes": 7404435,
533
+ "pages": 39,
534
  "shards": 3
535
  },
536
  {
537
+ "seconds": 1.136316104002617,
538
+ "bytes": 5061136,
539
+ "pages": 30,
540
  "shards": 2
541
  },
542
  {
543
+ "seconds": 1.2786440780000703,
544
+ "bytes": 7516874,
545
+ "pages": 38,
546
  "shards": 3
547
  },
548
  {
549
+ "seconds": 0.9931387870019535,
550
+ "bytes": 6287159,
551
+ "pages": 38,
552
  "shards": 2
553
  },
554
  {
555
+ "seconds": 0.9191857780024293,
556
+ "bytes": 859128,
557
+ "pages": 26,
558
  "shards": 2
559
+ },
560
+ {
561
+ "seconds": 1.4843631869989622,
562
+ "bytes": 18084349,
563
+ "pages": 33,
564
+ "shards": 5
565
  }
566
  ],
567
  "cache_key": {
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:5d3c17f33f3432629d704e2ea2debd635921d926d1bb7e925d71c43aa06fd061
3
- size 721776127
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:629577df31a18ce0968102b0e3f539c35489cebe6a6ddc046c8ab547af131faa
3
+ size 721758463