appvoid commited on
Commit
66082e4
·
verified ·
1 Parent(s): c363ad2

Qwen90M step 540000; complete resumable checkpoint

Browse files
Files changed (6) hide show
  1. complete.json +5 -5
  2. coverage.json +52 -46
  3. latest.json +1 -1
  4. model.safetensors +1 -1
  5. shard_scheduler.json +246 -246
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 535000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "746a63165b4986046476ae84c2713e36ed3fa23737687659143ada4b831f1654",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "2b45969c25feb44bef2bbc28e1a234c7ba36a717f6375ea0dba3053855528f94",
12
  "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
13
  "performance.json": "3070ebbdf9b2b0fd6a0a237819043aa2c2a35c74a97edfa223e5c324471af4af",
14
- "shard_scheduler.json": "dc3b4c9c9c5634b43786ba061a7ead1b9ed759c5b1eacff95382e91a9ef4f364",
15
- "coverage.json": "7998ce90c7aacd5ffa1b3b45381b7a23cce87145599cd5217ab5d192996cc5db",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 540000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "b38e10d38ed7ada06841ff8982da7d676daf791bf507a6cf07a68d71c4a4078d",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "043647c8c26e8fed077bb197406f27a2662b7f2ed8742ee4065fa206a57818d0",
12
  "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
13
  "performance.json": "3070ebbdf9b2b0fd6a0a237819043aa2c2a35c74a97edfa223e5c324471af4af",
14
+ "shard_scheduler.json": "de12fe13ce1f77cd03863bd480ed4c92361aad4e582dff7ec3d7a242555cb0c8",
15
+ "coverage.json": "c6a7d6cc92b6319e9b013aeaaaf6b0e9a72597bf84b77a74dadf487e898d80d9",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,101 +1,107 @@
1
  {
2
- "step": 535000,
3
  "stats": {
4
- "climbmix/blocks": 19260000,
5
- "climbmix/targets": 32195111082,
6
- "climbmix/input_tokens": 32197151068,
7
- "ultra_qa/blocks": 7222500,
8
- "ultra_qa/targets": 11782590681,
9
- "ultra_qa/input_tokens": 11782613431,
10
- "cortex/blocks": 7222500,
11
- "cortex/targets": 274657059,
12
- "cortex/input_tokens": 11480802169,
13
- "ultra_style/blocks": 7222500,
14
- "ultra_style/targets": 11423290047,
15
- "ultra_style/input_tokens": 11423888845,
16
- "openmath/blocks": 4815000,
17
- "openmath/targets": 7439991699,
18
- "openmath/input_tokens": 7442470149,
19
- "rewrite6/blocks": 2407500,
20
- "rewrite6/targets": 3946780306,
21
- "rewrite6/input_tokens": 3951107292
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 2637,
26
- "row": 4073,
27
  "start": {
28
- "file": 127,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 51,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 0,
37
- "documents": 10805225
38
  },
39
  "rewrite6": {
40
- "page": 0,
41
- "row": 2348,
42
- "start": null,
 
 
 
 
 
 
43
  "epoch": 30,
44
  "offset": 2048,
45
- "documents": 6164468
46
  },
47
  "openmath": {
48
- "page": 1648,
49
- "row": 1424,
50
  "start": {
51
  "file": 15,
52
  "reader": {
53
  "format": "parquet-row-v1",
54
- "row_group": 200,
55
- "row": 458
56
  }
57
  },
58
  "epoch": 0,
59
  "offset": 0,
60
- "documents": 6751632
61
  },
62
  "ultra_qa": {
63
- "page": 712,
64
- "row": 3263,
65
  "start": {
66
  "file": 5,
67
  "reader": {
68
  "format": "parquet-row-v1",
69
- "row_group": 4,
70
- "row": 60132
71
  }
72
  },
73
  "epoch": 0,
74
- "offset": 6144,
75
- "documents": 2919615
76
  },
77
  "ultra_style": {
78
- "page": 1087,
79
- "row": 3342,
80
  "start": {
81
  "file": 6,
82
  "reader": {
83
  "format": "parquet-row-v1",
84
  "row_group": 4,
85
- "row": 2030
86
  }
87
  },
88
  "epoch": 0,
89
  "offset": 4096,
90
- "documents": 4455694
91
  },
92
  "cortex": {
93
- "page": 140074,
94
- "row": 15,
95
  "start": null,
96
  "epoch": 0,
97
  "offset": 0,
98
- "documents": 17929487
99
  }
100
  }
101
  }
 
1
  {
2
+ "step": 540000,
3
  "stats": {
4
+ "climbmix/blocks": 19440000,
5
+ "climbmix/targets": 32495587833,
6
+ "climbmix/input_tokens": 32497646947,
7
+ "ultra_qa/blocks": 7290000,
8
+ "ultra_qa/targets": 11892698357,
9
+ "ultra_qa/input_tokens": 11892721312,
10
+ "cortex/blocks": 7290000,
11
+ "cortex/targets": 277235924,
12
+ "cortex/input_tokens": 11588137812,
13
+ "ultra_style/blocks": 7290000,
14
+ "ultra_style/targets": 11529724861,
15
+ "ultra_style/input_tokens": 11530329511,
16
+ "openmath/blocks": 4860000,
17
+ "openmath/targets": 7509510059,
18
+ "openmath/input_tokens": 7512011367,
19
+ "rewrite6/blocks": 2430000,
20
+ "rewrite6/targets": 3983637088,
21
+ "rewrite6/input_tokens": 3988004561
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 2662,
26
+ "row": 3055,
27
  "start": {
28
+ "file": 128,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 68,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 0,
37
+ "documents": 10906607
38
  },
39
  "rewrite6": {
40
+ "page": 14,
41
+ "row": 2627,
42
+ "start": {
43
+ "file": 0,
44
+ "reader": {
45
+ "format": "jsonl-row-v1",
46
+ "row": 57344
47
+ }
48
+ },
49
  "epoch": 30,
50
  "offset": 2048,
51
+ "documents": 6222091
52
  },
53
  "openmath": {
54
+ "page": 1663,
55
+ "row": 2767,
56
  "start": {
57
  "file": 15,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 261,
61
+ "row": 898
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 6814415
67
  },
68
  "ultra_qa": {
69
+ "page": 719,
70
+ "row": 1872,
71
  "start": {
72
  "file": 5,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
+ "row_group": 5,
76
+ "row": 24371
77
  }
78
  },
79
  "epoch": 0,
80
+ "offset": 2048,
81
+ "documents": 2946896
82
  },
83
  "ultra_style": {
84
+ "page": 1098,
85
+ "row": 616,
86
  "start": {
87
  "file": 6,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 4,
91
+ "row": 47086
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 4096,
96
+ "documents": 4498024
97
  },
98
  "cortex": {
99
+ "page": 141384,
100
+ "row": 37,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 18097189
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 535000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "746a63165b4986046476ae84c2713e36ed3fa23737687659143ada4b831f1654", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "2b45969c25feb44bef2bbc28e1a234c7ba36a717f6375ea0dba3053855528f94", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "3070ebbdf9b2b0fd6a0a237819043aa2c2a35c74a97edfa223e5c324471af4af", "shard_scheduler.json": "dc3b4c9c9c5634b43786ba061a7ead1b9ed759c5b1eacff95382e91a9ef4f364", "coverage.json": "7998ce90c7aacd5ffa1b3b45381b7a23cce87145599cd5217ab5d192996cc5db", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000535000"}
 
1
+ {"step": 540000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "b38e10d38ed7ada06841ff8982da7d676daf791bf507a6cf07a68d71c4a4078d", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "043647c8c26e8fed077bb197406f27a2662b7f2ed8742ee4065fa206a57818d0", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "3070ebbdf9b2b0fd6a0a237819043aa2c2a35c74a97edfa223e5c324471af4af", "shard_scheduler.json": "de12fe13ce1f77cd03863bd480ed4c92361aad4e582dff7ec3d7a242555cb0c8", "coverage.json": "c6a7d6cc92b6319e9b013aeaaaf6b0e9a72597bf84b77a74dadf487e898d80d9", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000540000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:746a63165b4986046476ae84c2713e36ed3fa23737687659143ada4b831f1654
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b38e10d38ed7ada06841ff8982da7d676daf791bf507a6cf07a68d71c4a4078d
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": 76499.76095720792,
26
- "required_pages_per_second_upper_estimate": 0.00529243096692033
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": 9562.47011965099,
40
- "required_pages_per_second_upper_estimate": 0.004003606552014596
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": 19124.94023930198,
54
- "required_pages_per_second_upper_estimate": 0.003983220693255846
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": 28687.41035895297,
68
- "required_pages_per_second_upper_estimate": 0.0017197664478798965
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": 28687.41035895297,
82
- "required_pages_per_second_upper_estimate": 0.002230319972825811
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": 28687.41035895297,
96
- "required_pages_per_second_upper_estimate": 0.36797129794324046
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,409 +158,409 @@
158
  "prefetch_batches": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 704566,
161
- "producer_batch_seconds": 0.08245735881254745,
162
- "producer_p95_seconds": 0.19908109500102,
163
- "gpu_update_seconds": 0.9637677174081841,
164
- "producer_batches_per_second": 12.127480365619363,
165
- "gpu_batches_per_second": 1.03759441402463,
166
- "producer_headroom": 11.688074069885543,
167
  "sustainable": true,
168
- "buffer_seconds": 60.7173661967156,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 87,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 322888.0851025507,
174
- "upload_bytes_per_second": 4628498.063736584,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 69.28426661307708,
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": 1.011093388000063,
184
- "bytes": 1199201,
185
- "pages": 34,
186
- "shards": 1
187
- },
188
- {
189
- "seconds": 1.0240878049999083,
190
- "bytes": 7205273,
191
- "pages": 28,
192
- "shards": 4
193
- },
194
- {
195
- "seconds": 1.1896957620001558,
196
- "bytes": 10758587,
197
- "pages": 42,
198
  "shards": 3
199
  },
200
  {
201
- "seconds": 0.9081728179999118,
202
- "bytes": 2358450,
203
- "pages": 28,
204
  "shards": 2
205
  },
206
  {
207
- "seconds": 1.152849502999743,
208
- "bytes": 6119990,
209
- "pages": 36,
210
- "shards": 2
211
  },
212
  {
213
- "seconds": 1.0355856849996599,
214
- "bytes": 6832540,
215
- "pages": 32,
216
  "shards": 2
217
  },
218
  {
219
- "seconds": 20.55054765299974,
220
- "bytes": 11612692,
221
  "pages": 35,
222
- "shards": 4
223
  },
224
  {
225
- "seconds": 1.2162191410006926,
226
- "bytes": 6401722,
227
- "pages": 45,
228
- "shards": 2
229
  },
230
  {
231
- "seconds": 1.3737986060004914,
232
- "bytes": 7819084,
233
- "pages": 35,
234
  "shards": 3
235
  },
236
  {
237
- "seconds": 1.079906239000593,
238
- "bytes": 5671187,
239
- "pages": 35,
 
 
 
 
 
 
 
 
 
 
 
 
240
  "shards": 2
241
  },
242
  {
243
- "seconds": 1.1921040119996178,
244
- "bytes": 5850261,
 
 
 
 
 
 
245
  "pages": 28,
246
  "shards": 2
247
  },
248
  {
249
- "seconds": 0.9688384539995241,
250
- "bytes": 8315017,
251
  "pages": 35,
252
- "shards": 3
253
  },
254
  {
255
- "seconds": 1.0787890209994657,
256
- "bytes": 6174644,
257
- "pages": 35,
258
- "shards": 2
259
  },
260
  {
261
- "seconds": 1.431994874999873,
262
- "bytes": 7632897,
263
  "pages": 39,
264
  "shards": 3
265
  },
266
  {
267
- "seconds": 1.237401497999599,
268
- "bytes": 5052734,
269
- "pages": 33,
270
  "shards": 2
271
  },
272
  {
273
- "seconds": 1.0236900689997128,
274
- "bytes": 6421100,
275
- "pages": 37,
276
- "shards": 2
277
  },
278
  {
279
- "seconds": 0.9477791999997862,
280
- "bytes": 2414897,
281
- "pages": 27,
282
  "shards": 2
283
  },
284
  {
285
- "seconds": 1.0987587250001525,
286
- "bytes": 11899150,
287
- "pages": 30,
288
- "shards": 3
289
- },
290
- {
291
- "seconds": 1.4687352659993849,
292
- "bytes": 10700369,
293
- "pages": 42,
294
  "shards": 3
295
  },
296
  {
297
- "seconds": 1.1858354729993152,
298
- "bytes": 2742758,
299
- "pages": 37,
300
- "shards": 2
301
  },
302
  {
303
- "seconds": 1.046531567000784,
304
- "bytes": 5904437,
305
- "pages": 28,
306
- "shards": 2
307
  },
308
  {
309
- "seconds": 1.4689589510007863,
310
- "bytes": 7230449,
311
- "pages": 37,
312
- "shards": 3
313
  },
314
  {
315
- "seconds": 1.0110029770003166,
316
- "bytes": 6117097,
317
- "pages": 37,
318
  "shards": 2
319
  },
320
  {
321
- "seconds": 1.1878640209997684,
322
- "bytes": 11776096,
323
- "pages": 38,
324
  "shards": 3
325
  },
326
  {
327
- "seconds": 1.224845865000134,
328
- "bytes": 2065038,
329
- "pages": 19,
330
  "shards": 2
331
  },
332
  {
333
- "seconds": 0.9610936170001878,
334
- "bytes": 5798019,
335
- "pages": 27,
336
- "shards": 2
337
  },
338
  {
339
- "seconds": 1.1097549460000664,
340
- "bytes": 5080154,
341
- "pages": 36,
342
  "shards": 2
343
  },
344
  {
345
- "seconds": 1.1122178149998945,
346
- "bytes": 7607293,
347
- "pages": 38,
348
- "shards": 3
349
  },
350
  {
351
- "seconds": 1.0907007900004828,
352
- "bytes": 6092133,
353
  "pages": 36,
354
  "shards": 2
355
  },
356
  {
357
- "seconds": 0.8937167260000933,
358
- "bytes": 2214439,
359
- "pages": 24,
360
  "shards": 2
361
  },
362
  {
363
- "seconds": 1.3153861170003438,
364
- "bytes": 17177113,
365
- "pages": 40,
366
- "shards": 4
367
  },
368
  {
369
- "seconds": 0.9797320220004622,
370
- "bytes": 1103154,
371
- "pages": 35,
372
  "shards": 2
373
  },
374
  {
375
- "seconds": 1.195222697000645,
376
- "bytes": 7323521,
377
- "pages": 29,
378
  "shards": 3
379
  },
380
  {
381
- "seconds": 1.158046825999918,
382
- "bytes": 6457008,
383
  "pages": 36,
384
  "shards": 2
385
  },
386
  {
387
- "seconds": 1.1688710530006574,
388
- "bytes": 6589194,
389
- "pages": 40,
390
- "shards": 3
391
- },
392
- {
393
- "seconds": 1.0017207049995704,
394
- "bytes": 5856566,
395
- "pages": 27,
396
  "shards": 2
397
  },
398
  {
399
- "seconds": 1.1244407199992565,
400
- "bytes": 12346633,
401
- "pages": 40,
402
  "shards": 3
403
  },
404
  {
405
- "seconds": 0.9011206420000235,
406
- "bytes": 2301339,
407
- "pages": 25,
408
  "shards": 2
409
  },
410
  {
411
- "seconds": 1.4037659219993657,
412
- "bytes": 10492690,
413
- "pages": 42,
414
- "shards": 3
415
  },
416
  {
417
- "seconds": 0.9956328799999028,
418
- "bytes": 1276481,
419
- "pages": 36,
420
- "shards": 1
421
  },
422
  {
423
- "seconds": 1.4530193070004316,
424
- "bytes": 7690650,
425
- "pages": 38,
426
- "shards": 3
427
  },
428
  {
429
- "seconds": 1.0416680760008603,
430
- "bytes": 5628205,
431
- "pages": 24,
432
- "shards": 2
433
  },
434
  {
435
- "seconds": 1.0915457699993567,
436
- "bytes": 13214903,
437
- "pages": 39,
438
- "shards": 4
439
  },
440
  {
441
- "seconds": 1.4333472569996957,
442
- "bytes": 6210705,
443
- "pages": 36,
444
  "shards": 2
445
  },
446
  {
447
- "seconds": 1.232096203000765,
448
- "bytes": 1201273,
449
- "pages": 35,
450
- "shards": 1
451
  },
452
  {
453
- "seconds": 1.2199994860002334,
454
- "bytes": 7355179,
455
- "pages": 26,
456
  "shards": 3
457
  },
458
  {
459
- "seconds": 1.1187948159986263,
460
- "bytes": 10435269,
461
- "pages": 45,
462
  "shards": 3
463
  },
464
  {
465
- "seconds": 0.8605066399995849,
466
- "bytes": 757336,
467
- "pages": 24,
468
- "shards": 1
469
  },
470
  {
471
- "seconds": 1.4440002779992938,
472
- "bytes": 13477323,
473
- "pages": 45,
474
- "shards": 4
475
  },
476
  {
477
- "seconds": 1.0930595740010176,
478
- "bytes": 6005049,
479
- "pages": 30,
480
  "shards": 2
481
  },
482
  {
483
- "seconds": 1.1409170160004578,
484
- "bytes": 7751245,
485
- "pages": 38,
486
  "shards": 3
487
  },
488
  {
489
- "seconds": 1.385997141000189,
490
- "bytes": 5854774,
491
- "pages": 28,
492
- "shards": 2
493
  },
494
  {
495
- "seconds": 0.8962630369987892,
496
- "bytes": 1124581,
497
- "pages": 35,
498
- "shards": 1
499
  },
500
  {
501
- "seconds": 1.1323460120001982,
502
- "bytes": 7094094,
503
- "pages": 25,
504
- "shards": 3
505
  },
506
  {
507
- "seconds": 1.3368760009998368,
508
- "bytes": 10726853,
509
- "pages": 39,
510
- "shards": 3
511
  },
512
  {
513
- "seconds": 0.9546346269999049,
514
- "bytes": 5999756,
515
- "pages": 29,
516
  "shards": 2
517
  },
518
  {
519
- "seconds": 1.1835868019988993,
520
- "bytes": 7840439,
521
- "pages": 38,
522
  "shards": 3
523
  },
524
  {
525
- "seconds": 1.00542208799925,
526
- "bytes": 5360697,
527
  "pages": 33,
528
  "shards": 2
529
  },
530
  {
531
- "seconds": 1.2967056479992607,
532
- "bytes": 7612762,
533
- "pages": 38,
534
  "shards": 3
535
  },
536
  {
537
- "seconds": 0.9550784620005288,
538
- "bytes": 883731,
539
- "pages": 26,
540
- "shards": 1
541
  },
542
  {
543
- "seconds": 1.1562814480002999,
544
- "bytes": 11922123,
545
- "pages": 40,
546
- "shards": 3
547
  },
548
  {
549
- "seconds": 1.2267546230013977,
550
- "bytes": 11778233,
551
- "pages": 33,
552
  "shards": 4
553
  },
554
  {
555
- "seconds": 0.9305929850015673,
556
- "bytes": 1085853,
 
 
 
 
 
 
557
  "pages": 35,
558
- "shards": 2
559
  },
560
  {
561
- "seconds": 1.4772485220000817,
562
- "bytes": 7565247,
563
- "pages": 36,
564
  "shards": 3
565
  }
566
  ],
 
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": 76448.66075247723,
26
+ "required_pages_per_second_upper_estimate": 0.005288895736188249
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": 9556.082594059653,
40
+ "required_pages_per_second_upper_estimate": 0.004000932228436205
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": 19112.165188119307,
54
+ "required_pages_per_second_upper_estimate": 0.00398055998699525
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": 28668.24778217896,
68
+ "required_pages_per_second_upper_estimate": 0.0017186176806618552
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": 28668.24778217896,
82
+ "required_pages_per_second_upper_estimate": 0.0022288301667688993
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": 28668.24778217896,
96
+ "required_pages_per_second_upper_estimate": 0.3677255009835554
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": 0.3263542902969334,
162
+ "producer_p95_seconds": 0.2078722389996983,
163
+ "gpu_update_seconds": 0.9644119239539579,
164
+ "producer_batches_per_second": 3.064154600480815,
165
+ "gpu_batches_per_second": 1.0369013231401534,
166
+ "producer_headroom": 2.9551072335420745,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.757951209099346,
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": 393197.65477415506,
174
+ "upload_bytes_per_second": 5782840.946028825,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 84.37106354158868,
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": 1.4530193070004316,
184
+ "bytes": 7690650,
185
+ "pages": 38,
 
 
 
 
 
 
 
 
 
 
 
 
186
  "shards": 3
187
  },
188
  {
189
+ "seconds": 1.0416680760008603,
190
+ "bytes": 5628205,
191
+ "pages": 24,
192
  "shards": 2
193
  },
194
  {
195
+ "seconds": 1.0915457699993567,
196
+ "bytes": 13214903,
197
+ "pages": 39,
198
+ "shards": 4
199
  },
200
  {
201
+ "seconds": 1.4333472569996957,
202
+ "bytes": 6210705,
203
+ "pages": 36,
204
  "shards": 2
205
  },
206
  {
207
+ "seconds": 1.232096203000765,
208
+ "bytes": 1201273,
209
  "pages": 35,
210
+ "shards": 1
211
  },
212
  {
213
+ "seconds": 1.2199994860002334,
214
+ "bytes": 7355179,
215
+ "pages": 26,
216
+ "shards": 3
217
  },
218
  {
219
+ "seconds": 1.1187948159986263,
220
+ "bytes": 10435269,
221
+ "pages": 45,
222
  "shards": 3
223
  },
224
  {
225
+ "seconds": 0.8605066399995849,
226
+ "bytes": 757336,
227
+ "pages": 24,
228
+ "shards": 1
229
+ },
230
+ {
231
+ "seconds": 1.4440002779992938,
232
+ "bytes": 13477323,
233
+ "pages": 45,
234
+ "shards": 4
235
+ },
236
+ {
237
+ "seconds": 1.0930595740010176,
238
+ "bytes": 6005049,
239
+ "pages": 30,
240
  "shards": 2
241
  },
242
  {
243
+ "seconds": 1.1409170160004578,
244
+ "bytes": 7751245,
245
+ "pages": 38,
246
+ "shards": 3
247
+ },
248
+ {
249
+ "seconds": 1.385997141000189,
250
+ "bytes": 5854774,
251
  "pages": 28,
252
  "shards": 2
253
  },
254
  {
255
+ "seconds": 0.8962630369987892,
256
+ "bytes": 1124581,
257
  "pages": 35,
258
+ "shards": 1
259
  },
260
  {
261
+ "seconds": 1.1323460120001982,
262
+ "bytes": 7094094,
263
+ "pages": 25,
264
+ "shards": 3
265
  },
266
  {
267
+ "seconds": 1.3368760009998368,
268
+ "bytes": 10726853,
269
  "pages": 39,
270
  "shards": 3
271
  },
272
  {
273
+ "seconds": 0.9546346269999049,
274
+ "bytes": 5999756,
275
+ "pages": 29,
276
  "shards": 2
277
  },
278
  {
279
+ "seconds": 1.1835868019988993,
280
+ "bytes": 7840439,
281
+ "pages": 38,
282
+ "shards": 3
283
  },
284
  {
285
+ "seconds": 1.00542208799925,
286
+ "bytes": 5360697,
287
+ "pages": 33,
288
  "shards": 2
289
  },
290
  {
291
+ "seconds": 1.2967056479992607,
292
+ "bytes": 7612762,
293
+ "pages": 38,
 
 
 
 
 
 
294
  "shards": 3
295
  },
296
  {
297
+ "seconds": 0.9550784620005288,
298
+ "bytes": 883731,
299
+ "pages": 26,
300
+ "shards": 1
301
  },
302
  {
303
+ "seconds": 1.1562814480002999,
304
+ "bytes": 11922123,
305
+ "pages": 40,
306
+ "shards": 3
307
  },
308
  {
309
+ "seconds": 1.2267546230013977,
310
+ "bytes": 11778233,
311
+ "pages": 33,
312
+ "shards": 4
313
  },
314
  {
315
+ "seconds": 0.9305929850015673,
316
+ "bytes": 1085853,
317
+ "pages": 35,
318
  "shards": 2
319
  },
320
  {
321
+ "seconds": 1.4772485220000817,
322
+ "bytes": 7565247,
323
+ "pages": 36,
324
  "shards": 3
325
  },
326
  {
327
+ "seconds": 1.035932449000029,
328
+ "bytes": 5952411,
329
+ "pages": 30,
330
  "shards": 2
331
  },
332
  {
333
+ "seconds": 0.7514245589991333,
334
+ "bytes": 257606,
335
+ "pages": 8,
336
+ "shards": 1
337
  },
338
  {
339
+ "seconds": 0.9524619530002383,
340
+ "bytes": 4638620,
341
+ "pages": 26,
342
  "shards": 2
343
  },
344
  {
345
+ "seconds": 1.3739831450002384,
346
+ "bytes": 13357834,
347
+ "pages": 40,
348
+ "shards": 4
349
  },
350
  {
351
+ "seconds": 1.1055001610002364,
352
+ "bytes": 6014129,
353
  "pages": 36,
354
  "shards": 2
355
  },
356
  {
357
+ "seconds": 0.9146980179993989,
358
+ "bytes": 2516554,
359
+ "pages": 34,
360
  "shards": 2
361
  },
362
  {
363
+ "seconds": 1.1637235170001077,
364
+ "bytes": 10485207,
365
+ "pages": 34,
366
+ "shards": 3
367
  },
368
  {
369
+ "seconds": 1.0128666689997772,
370
+ "bytes": 6455974,
371
+ "pages": 36,
372
  "shards": 2
373
  },
374
  {
375
+ "seconds": 1.0892136950005806,
376
+ "bytes": 8273419,
377
+ "pages": 32,
378
  "shards": 3
379
  },
380
  {
381
+ "seconds": 1.3893709189997026,
382
+ "bytes": 6184789,
383
  "pages": 36,
384
  "shards": 2
385
  },
386
  {
387
+ "seconds": 1.226723518999279,
388
+ "bytes": 5211595,
389
+ "pages": 30,
 
 
 
 
 
 
390
  "shards": 2
391
  },
392
  {
393
+ "seconds": 1.4460434569991776,
394
+ "bytes": 7613200,
395
+ "pages": 39,
396
  "shards": 3
397
  },
398
  {
399
+ "seconds": 1.303638638000848,
400
+ "bytes": 6027470,
401
+ "pages": 37,
402
  "shards": 2
403
  },
404
  {
405
+ "seconds": 1.9445103569996718,
406
+ "bytes": 2479005,
407
+ "pages": 31,
408
+ "shards": 2
409
  },
410
  {
411
+ "seconds": 1.5074067269997613,
412
+ "bytes": 15537281,
413
+ "pages": 35,
414
+ "shards": 4
415
  },
416
  {
417
+ "seconds": 0.9092938880003203,
418
+ "bytes": 1113077,
419
+ "pages": 34,
420
+ "shards": 1
421
  },
422
  {
423
+ "seconds": 1.0412252780006384,
424
+ "bytes": 7665353,
425
+ "pages": 39,
426
+ "shards": 3
427
  },
428
  {
429
+ "seconds": 1.1747847249989718,
430
+ "bytes": 10716254,
431
+ "pages": 34,
432
+ "shards": 3
433
  },
434
  {
435
+ "seconds": 1.0711626960001013,
436
+ "bytes": 2631436,
437
+ "pages": 37,
438
  "shards": 2
439
  },
440
  {
441
+ "seconds": 1.137612402000741,
442
+ "bytes": 5722821,
443
+ "pages": 28,
444
+ "shards": 2
445
  },
446
  {
447
+ "seconds": 1.4303736129986646,
448
+ "bytes": 11919616,
449
+ "pages": 42,
450
  "shards": 3
451
  },
452
  {
453
+ "seconds": 1.0499967649993778,
454
+ "bytes": 7310677,
455
+ "pages": 35,
456
  "shards": 3
457
  },
458
  {
459
+ "seconds": 1.0670534790006059,
460
+ "bytes": 5802765,
461
+ "pages": 28,
462
+ "shards": 2
463
  },
464
  {
465
+ "seconds": 0.9111343770000531,
466
+ "bytes": 2649643,
467
+ "pages": 36,
468
+ "shards": 2
469
  },
470
  {
471
+ "seconds": 1.2196053769985156,
472
+ "bytes": 6057143,
473
+ "pages": 37,
474
  "shards": 2
475
  },
476
  {
477
+ "seconds": 1.2347321990000637,
478
+ "bytes": 9643472,
479
+ "pages": 35,
480
  "shards": 3
481
  },
482
  {
483
+ "seconds": 1.1648727040010272,
484
+ "bytes": 8469052,
485
+ "pages": 35,
486
+ "shards": 3
487
  },
488
  {
489
+ "seconds": 1.1223673130007228,
490
+ "bytes": 6068911,
491
+ "pages": 36,
492
+ "shards": 2
493
  },
494
  {
495
+ "seconds": 1.1270621679996111,
496
+ "bytes": 5842224,
497
+ "pages": 28,
498
+ "shards": 2
499
  },
500
  {
501
+ "seconds": 1.3927226420000807,
502
+ "bytes": 7163586,
503
+ "pages": 36,
504
+ "shards": 4
505
  },
506
  {
507
+ "seconds": 1.230830958000297,
508
+ "bytes": 6020552,
509
+ "pages": 37,
510
  "shards": 2
511
  },
512
  {
513
+ "seconds": 1.2729082169989852,
514
+ "bytes": 7700586,
515
+ "pages": 39,
516
  "shards": 3
517
  },
518
  {
519
+ "seconds": 0.9891643420014589,
520
+ "bytes": 6883017,
521
  "pages": 33,
522
  "shards": 2
523
  },
524
  {
525
+ "seconds": 1.2779713299987634,
526
+ "bytes": 9455856,
527
+ "pages": 30,
528
  "shards": 3
529
  },
530
  {
531
+ "seconds": 1.159703276000073,
532
+ "bytes": 2570411,
533
+ "pages": 35,
534
+ "shards": 2
535
  },
536
  {
537
+ "seconds": 1.2064628600001015,
538
+ "bytes": 6189523,
539
+ "pages": 36,
540
+ "shards": 2
541
  },
542
  {
543
+ "seconds": 1.4404870660000597,
544
+ "bytes": 11907122,
545
+ "pages": 40,
546
  "shards": 4
547
  },
548
  {
549
+ "seconds": 0.842472636999446,
550
+ "bytes": 857193,
551
+ "pages": 26,
552
+ "shards": 1
553
+ },
554
+ {
555
+ "seconds": 1.0425618379995285,
556
+ "bytes": 11991809,
557
  "pages": 35,
558
+ "shards": 3
559
  },
560
  {
561
+ "seconds": 1.543527111000003,
562
+ "bytes": 7770797,
563
+ "pages": 39,
564
  "shards": 3
565
  }
566
  ],
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:2b45969c25feb44bef2bbc28e1a234c7ba36a717f6375ea0dba3053855528f94
3
- size 721759935
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:043647c8c26e8fed077bb197406f27a2662b7f2ed8742ee4065fa206a57818d0
3
+ size 721757375