appvoid commited on
Commit
cbbfa44
·
verified ·
1 Parent(s): 2b05d3d

Qwen90M step 530000; complete resumable checkpoint

Browse files
Files changed (7) hide show
  1. complete.json +6 -6
  2. coverage.json +45 -45
  3. latest.json +1 -1
  4. model.safetensors +1 -1
  5. performance.json +3 -3
  6. shard_scheduler.json +100 -328
  7. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 525000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "555a0b375b5d8b6df17309058cb3fac983e079e022e3453c5bb40694dc0dfd3c",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "550740c374917e84e0c390d9c1282f579154291db3c66058d82c3775128b44da",
12
  "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
13
- "performance.json": "234744063e094dfd7b525da4e181ac493fac400c830f871c7e8c59127ed10dc9",
14
- "shard_scheduler.json": "ded67c5cf32a54e2366cba7242e168a5494026aca2d41bea5b359a7947e97553",
15
- "coverage.json": "8e20b86d8845516d961760222cf2e262fe8d2ed820d4ef25ec6904b4dc6e9d4b",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 530000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "3345a1d3d37a56145174811805daadbe6545c69564f12badd37ca9980383c189",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "4273939d74c5226bc57df9a209306193dff7da001a42417dfdf439575063da35",
12
  "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
13
+ "performance.json": "3070ebbdf9b2b0fd6a0a237819043aa2c2a35c74a97edfa223e5c324471af4af",
14
+ "shard_scheduler.json": "3e8b88313d070ffdb6572b9af938d708afe66786c34590df9f577e00e5a7ad71",
15
+ "coverage.json": "1f166ed6ce6875dfc663aa1359c72063b2c84099952c88e471aa37b08f50927f",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 525000,
3
  "stats": {
4
- "climbmix/blocks": 18900000,
5
- "climbmix/targets": 31592797042,
6
- "climbmix/input_tokens": 31594799016,
7
- "ultra_qa/blocks": 7087500,
8
- "ultra_qa/targets": 11562286610,
9
- "ultra_qa/input_tokens": 11562308928,
10
- "cortex/blocks": 7087500,
11
- "cortex/targets": 269544792,
12
- "cortex/input_tokens": 11266328102,
13
- "ultra_style/blocks": 7087500,
14
- "ultra_style/targets": 11209247634,
15
- "ultra_style/input_tokens": 11209834992,
16
- "openmath/blocks": 4725000,
17
- "openmath/targets": 7301172470,
18
- "openmath/input_tokens": 7303605185,
19
- "rewrite6/blocks": 2362500,
20
- "rewrite6/targets": 3872985034,
21
- "rewrite6/input_tokens": 3877231666
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 2588,
26
- "row": 3394,
27
  "start": {
28
- "file": 125,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 20,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
- "offset": 2048,
37
- "documents": 10603842
38
  },
39
  "rewrite6": {
40
- "page": 22,
41
- "row": 2978,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 90112
47
  }
48
  },
49
  "epoch": 29,
50
  "offset": 0,
51
- "documents": 6049806
52
  },
53
  "openmath": {
54
- "page": 1617,
55
- "row": 2871,
56
  "start": {
57
  "file": 15,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 73,
61
- "row": 482
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 6626103
67
  },
68
  "ultra_qa": {
69
- "page": 699,
70
- "row": 1991,
71
  "start": {
72
  "file": 5,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 4,
76
- "row": 6884
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
- "documents": 2865095
82
  },
83
  "ultra_style": {
84
- "page": 1067,
85
- "row": 1860,
86
  "start": {
87
  "file": 6,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 3,
91
- "row": 4718
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
- "documents": 4372292
97
  },
98
  "cortex": {
99
- "page": 137458,
100
- "row": 115,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 17594739
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 530000,
3
  "stats": {
4
+ "climbmix/blocks": 19080000,
5
+ "climbmix/targets": 31893841018,
6
+ "climbmix/input_tokens": 31895861954,
7
+ "ultra_qa/blocks": 7155000,
8
+ "ultra_qa/targets": 11672418639,
9
+ "ultra_qa/input_tokens": 11672441161,
10
+ "cortex/blocks": 7155000,
11
+ "cortex/targets": 272094364,
12
+ "cortex/input_tokens": 11373535505,
13
+ "ultra_style/blocks": 7155000,
14
+ "ultra_style/targets": 11316260253,
15
+ "ultra_style/input_tokens": 11316853356,
16
+ "openmath/blocks": 4770000,
17
+ "openmath/targets": 7370634197,
18
+ "openmath/input_tokens": 7373089745,
19
+ "rewrite6/blocks": 2385000,
20
+ "rewrite6/targets": 3909838802,
21
+ "rewrite6/input_tokens": 3914125422
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 2613,
26
+ "row": 1722,
27
  "start": {
28
+ "file": 126,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 37,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
+ "offset": 0,
37
+ "documents": 10704570
38
  },
39
  "rewrite6": {
40
+ "page": 36,
41
+ "row": 2819,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 147456
47
  }
48
  },
49
  "epoch": 29,
50
  "offset": 0,
51
+ "documents": 6106991
52
  },
53
  "openmath": {
54
+ "page": 1633,
55
+ "row": 76,
56
  "start": {
57
  "file": 15,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 139,
61
+ "row": 18
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 6688844
67
  },
68
  "ultra_qa": {
69
+ "page": 706,
70
+ "row": 539,
71
  "start": {
72
  "file": 5,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 4,
76
+ "row": 35556
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
+ "documents": 2892315
82
  },
83
  "ultra_style": {
84
+ "page": 1077,
85
+ "row": 2689,
86
  "start": {
87
  "file": 6,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 3,
91
+ "row": 45678
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
+ "documents": 4414081
97
  },
98
  "cortex": {
99
+ "page": 138765,
100
+ "row": 110,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 17762030
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 525000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "555a0b375b5d8b6df17309058cb3fac983e079e022e3453c5bb40694dc0dfd3c", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "550740c374917e84e0c390d9c1282f579154291db3c66058d82c3775128b44da", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "234744063e094dfd7b525da4e181ac493fac400c830f871c7e8c59127ed10dc9", "shard_scheduler.json": "ded67c5cf32a54e2366cba7242e168a5494026aca2d41bea5b359a7947e97553", "coverage.json": "8e20b86d8845516d961760222cf2e262fe8d2ed820d4ef25ec6904b4dc6e9d4b", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000525000"}
 
1
+ {"step": 530000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "3345a1d3d37a56145174811805daadbe6545c69564f12badd37ca9980383c189", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "4273939d74c5226bc57df9a209306193dff7da001a42417dfdf439575063da35", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "3070ebbdf9b2b0fd6a0a237819043aa2c2a35c74a97edfa223e5c324471af4af", "shard_scheduler.json": "3e8b88313d070ffdb6572b9af938d708afe66786c34590df9f577e00e5a7ad71", "coverage.json": "1f166ed6ce6875dfc663aa1359c72063b2c84099952c88e471aa37b08f50927f", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000530000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:555a0b375b5d8b6df17309058cb3fac983e079e022e3453c5bb40694dc0dfd3c
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3345a1d3d37a56145174811805daadbe6545c69564f12badd37ca9980383c189
3
  size 360622744
performance.json CHANGED
@@ -49,11 +49,11 @@
49
  "selected": {
50
  "microbatch": 90,
51
  "mode": "compiled",
52
- "microbatch_seconds": 0.9387972919998901,
53
  "peak_gb": 89.37766551971436,
54
  "fits_budget": true,
55
- "estimated_update_seconds": 0.9387972919998901,
56
- "input_tokens_per_second": 196336.31410178967
57
  },
58
  "measurements": [
59
  {
 
49
  "selected": {
50
  "microbatch": 90,
51
  "mode": "compiled",
52
+ "microbatch_seconds": 0.9447870520000379,
53
  "peak_gb": 89.37766551971436,
54
  "fits_budget": true,
55
+ "estimated_update_seconds": 0.9447870520000379,
56
+ "input_tokens_per_second": 195091.58133550777
57
  },
58
  "measurements": [
59
  {
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": 77213.6919826638,
26
- "required_pages_per_second_upper_estimate": 0.005341822371809575
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": 9651.711497832976,
40
- "required_pages_per_second_upper_estimate": 0.004040970053487503
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": 19303.42299566595,
54
- "required_pages_per_second_upper_estimate": 0.00402039394449965
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": 28955.13449349893,
68
- "required_pages_per_second_upper_estimate": 0.001735816100954843
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": 28955.13449349893,
82
- "required_pages_per_second_upper_estimate": 0.002251134346692743
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": 28955.13449349893,
96
- "required_pages_per_second_upper_estimate": 0.37140537568141674
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,410 +158,182 @@
158
  "prefetch_batches": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 704566,
161
- "producer_batch_seconds": 0.11904607814852852,
162
- "producer_p95_seconds": 0.3377903690052335,
163
- "gpu_update_seconds": 0.9548565559661825,
164
- "producer_batches_per_second": 8.400108727247144,
165
- "gpu_batches_per_second": 1.0472777232891683,
166
- "producer_headroom": 8.02089888904068,
167
  "sustainable": true,
168
- "buffer_seconds": 60.1559630258695,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 78,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 384588.48339903663,
174
- "upload_bytes_per_second": 3837530.023601385,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 82.52373577574086,
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": 1.521548144002736,
184
- "bytes": 10001036,
185
- "pages": 42,
186
- "shards": 3
187
- },
188
- {
189
- "seconds": 1.037408388001495,
190
- "bytes": 2352962,
191
- "pages": 29,
192
- "shards": 2
193
- },
194
- {
195
- "seconds": 1.4666764929934288,
196
- "bytes": 11848768,
197
- "pages": 38,
198
- "shards": 3
199
- },
200
- {
201
- "seconds": 1.1809732929978054,
202
- "bytes": 12035253,
203
- "pages": 37,
204
- "shards": 4
205
- },
206
- {
207
- "seconds": 0.976799342999584,
208
- "bytes": 895629,
209
- "pages": 27,
210
  "shards": 1
211
  },
212
  {
213
- "seconds": 1.1010828580037924,
214
- "bytes": 6038977,
215
- "pages": 38,
216
- "shards": 2
217
- },
218
- {
219
- "seconds": 1.0995818070005043,
220
- "bytes": 2583346,
221
  "pages": 34,
222
- "shards": 2
223
  },
224
  {
225
- "seconds": 0.9759389660030138,
226
- "bytes": 6031085,
227
- "pages": 33,
228
- "shards": 2
229
  },
230
  {
231
- "seconds": 1.012582071001816,
232
- "bytes": 5087118,
233
- "pages": 25,
234
- "shards": 2
235
  },
236
  {
237
- "seconds": 1.1673960980042466,
238
- "bytes": 13547167,
239
- "pages": 41,
240
  "shards": 4
241
  },
242
  {
243
- "seconds": 1.3716537860018434,
244
- "bytes": 6334955,
245
- "pages": 38,
246
- "shards": 2
247
- },
248
- {
249
- "seconds": 1.127826156996889,
250
- "bytes": 2574706,
251
- "pages": 33,
252
- "shards": 2
253
- },
254
- {
255
- "seconds": 1.1363400170012028,
256
- "bytes": 10297243,
257
- "pages": 34,
258
  "shards": 3
259
  },
260
  {
261
- "seconds": 1.2236347540019779,
262
- "bytes": 6431007,
263
- "pages": 37,
264
  "shards": 2
265
  },
266
  {
267
- "seconds": 0.9796522880060365,
268
- "bytes": 2218113,
269
- "pages": 23,
270
  "shards": 2
271
  },
272
  {
273
- "seconds": 1.3331365960038966,
274
- "bytes": 12026403,
275
- "pages": 40,
276
- "shards": 3
277
- },
278
- {
279
- "seconds": 1.336267955994117,
280
- "bytes": 7188821,
281
- "pages": 41,
282
- "shards": 3
283
- },
284
- {
285
- "seconds": 1.3482005179976113,
286
- "bytes": 6310167,
287
  "pages": 32,
288
  "shards": 2
289
  },
290
  {
291
- "seconds": 1.184783762000734,
292
- "bytes": 5973673,
293
- "pages": 38,
294
- "shards": 2
295
- },
296
- {
297
- "seconds": 0.9036034050004673,
298
- "bytes": 2269487,
299
- "pages": 25,
300
- "shards": 2
301
  },
302
  {
303
- "seconds": 1.6477199560031295,
304
- "bytes": 16059670,
305
  "pages": 45,
306
- "shards": 5
307
- },
308
- {
309
- "seconds": 1.0709953250043327,
310
- "bytes": 6125208,
311
- "pages": 31,
312
- "shards": 2
313
- },
314
- {
315
- "seconds": 0.8623744039941812,
316
- "bytes": 2649650,
317
- "pages": 39,
318
  "shards": 2
319
  },
320
  {
321
- "seconds": 1.4692237859999295,
322
- "bytes": 10546134,
323
- "pages": 31,
324
  "shards": 3
325
  },
326
  {
327
- "seconds": 1.1281011890023365,
328
- "bytes": 2729862,
329
- "pages": 38,
330
  "shards": 2
331
  },
332
  {
333
- "seconds": 2.671323365997523,
334
- "bytes": 5596441,
335
- "pages": 20,
336
- "shards": 2
337
- },
338
- {
339
- "seconds": 1.0342092059945571,
340
- "bytes": 930671,
341
- "pages": 29,
342
- "shards": 1
343
- },
344
- {
345
- "seconds": 1.692445526998199,
346
- "bytes": 13078461,
347
- "pages": 32,
348
- "shards": 4
349
- },
350
- {
351
- "seconds": 1.3595879119966412,
352
- "bytes": 10348219,
353
- "pages": 44,
354
- "shards": 3
355
- },
356
- {
357
- "seconds": 1.0515126419995795,
358
- "bytes": 1003171,
359
- "pages": 30,
360
- "shards": 1
361
- },
362
- {
363
- "seconds": 1.000422951001383,
364
- "bytes": 7649877,
365
- "pages": 39,
366
- "shards": 3
367
- },
368
- {
369
- "seconds": 1.2835066039988305,
370
- "bytes": 10437972,
371
- "pages": 31,
372
- "shards": 3
373
- },
374
- {
375
- "seconds": 0.8616337460043724,
376
- "bytes": 2430044,
377
  "pages": 28,
378
  "shards": 2
379
  },
380
  {
381
- "seconds": 1.3565980959974695,
382
- "bytes": 12199549,
383
- "pages": 43,
384
  "shards": 3
385
  },
386
  {
387
- "seconds": 1.012810416999855,
388
- "bytes": 6201884,
389
- "pages": 37,
390
- "shards": 2
391
- },
392
- {
393
- "seconds": 0.8832714039963321,
394
- "bytes": 2126873,
395
- "pages": 22,
396
  "shards": 2
397
  },
398
  {
399
- "seconds": 1.4999150360017666,
400
- "bytes": 10166456,
401
- "pages": 43,
402
  "shards": 3
403
  },
404
  {
405
- "seconds": 0.9999620900052832,
406
- "bytes": 2535870,
407
- "pages": 34,
408
- "shards": 2
409
- },
410
- {
411
- "seconds": 1.0642618760030018,
412
- "bytes": 5883611,
413
- "pages": 24,
414
- "shards": 2
415
- },
416
- {
417
- "seconds": 2.4567707870010054,
418
- "bytes": 16775004,
419
- "pages": 44,
420
- "shards": 4
421
- },
422
- {
423
- "seconds": 1.3269825160023174,
424
- "bytes": 2735395,
425
- "pages": 38,
426
  "shards": 2
427
  },
428
  {
429
- "seconds": 2.3559383290048572,
430
- "bytes": 6286319,
431
- "pages": 38,
432
  "shards": 2
433
  },
434
  {
435
- "seconds": 0.9432465139980195,
436
- "bytes": 858963,
437
  "pages": 27,
438
- "shards": 1
439
- },
440
- {
441
- "seconds": 1.6026845509986742,
442
- "bytes": 11607409,
443
- "pages": 29,
444
- "shards": 4
445
- },
446
- {
447
- "seconds": 1.04873497600056,
448
- "bytes": 6766779,
449
- "pages": 32,
450
  "shards": 2
451
  },
452
  {
453
- "seconds": 1.4542191169966827,
454
- "bytes": 7891732,
455
- "pages": 46,
456
- "shards": 3
457
- },
458
- {
459
- "seconds": 1.0971116009968682,
460
- "bytes": 6168592,
461
- "pages": 38,
462
- "shards": 2
463
- },
464
- {
465
- "seconds": 1.321997156002908,
466
- "bytes": 9700555,
467
- "pages": 34,
468
  "shards": 3
469
  },
470
  {
471
- "seconds": 1.1995469100002083,
472
- "bytes": 2400407,
473
- "pages": 29,
474
- "shards": 2
475
- },
476
- {
477
- "seconds": 1.5711508849999518,
478
- "bytes": 6147230,
479
- "pages": 38,
480
- "shards": 2
481
- },
482
- {
483
- "seconds": 1.4379745080004795,
484
- "bytes": 8148657,
485
- "pages": 31,
486
  "shards": 3
487
  },
488
  {
489
- "seconds": 1.2602131539970287,
490
- "bytes": 10952650,
491
- "pages": 39,
492
- "shards": 4
493
- },
494
- {
495
- "seconds": 1.072258756001247,
496
- "bytes": 6000592,
497
- "pages": 29,
498
- "shards": 2
499
- },
500
- {
501
- "seconds": 0.8994364020036301,
502
- "bytes": 2733894,
503
  "pages": 37,
504
  "shards": 2
505
  },
506
  {
507
- "seconds": 3.8195672939982614,
508
- "bytes": 6121969,
509
- "pages": 34,
510
  "shards": 2
511
  },
512
  {
513
- "seconds": 1.2331268449997879,
514
- "bytes": 7295036,
515
- "pages": 32,
516
- "shards": 3
517
- },
518
- {
519
- "seconds": 31.867704994001542,
520
- "bytes": 12094685,
521
- "pages": 36,
522
- "shards": 3
523
- },
524
- {
525
- "seconds": 1.475563729996793,
526
- "bytes": 6268574,
527
  "pages": 37,
528
- "shards": 2
529
- },
530
- {
531
- "seconds": 1.4652358960011043,
532
- "bytes": 7414584,
533
- "pages": 40,
534
  "shards": 3
535
  },
536
  {
537
- "seconds": 2.4988657550056814,
538
- "bytes": 5294252,
539
- "pages": 30,
540
  "shards": 2
541
- },
542
- {
543
- "seconds": 1.1802646279975306,
544
- "bytes": 7575689,
545
- "pages": 41,
546
- "shards": 3
547
- },
548
- {
549
- "seconds": 0.8871877959973062,
550
- "bytes": 955502,
551
- "pages": 27,
552
- "shards": 1
553
- },
554
- {
555
- "seconds": 1.3968777099944418,
556
- "bytes": 12101140,
557
- "pages": 39,
558
- "shards": 3
559
- },
560
- {
561
- "seconds": 1.1570743730044342,
562
- "bytes": 12058095,
563
- "pages": 40,
564
- "shards": 4
565
  }
566
  ],
567
  "cache_key": {
@@ -606,7 +378,7 @@
606
  },
607
  "cache_reused": true,
608
  "resume_probe_seconds": [
609
- 1.0198370660000364,
610
- 0.013382547000219347
611
  ]
612
  }
 
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": 77015.98464267456,
26
+ "required_pages_per_second_upper_estimate": 0.005328144519284888
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": 9626.99808033432,
40
+ "required_pages_per_second_upper_estimate": 0.0040306230616556585
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": 19253.99616066864,
54
+ "required_pages_per_second_upper_estimate": 0.004010099638243004
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": 28880.99424100296,
68
+ "required_pages_per_second_upper_estimate": 0.0017313715060233205
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": 28880.99424100296,
82
+ "required_pages_per_second_upper_estimate": 0.0022453702681696883
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": 28880.99424100296,
96
+ "required_pages_per_second_upper_estimate": 0.3704543841279994
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": 0.320096299234379,
162
+ "producer_p95_seconds": 0.15855196899974544,
163
+ "gpu_update_seconds": 0.95730776334381,
164
+ "producer_batches_per_second": 3.1240598607101857,
165
+ "gpu_batches_per_second": 1.0445961458696094,
166
+ "producer_headroom": 2.9906867578086427,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.310389090660024,
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": 386229.8712076355,
174
+ "upload_bytes_per_second": 3476176.657096584,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 82.87593938991354,
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": 1.0629590909998115,
184
+ "bytes": 25106,
185
+ "pages": 1,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
186
  "shards": 1
187
  },
188
  {
189
+ "seconds": 1.218460234000304,
190
+ "bytes": 6273746,
 
 
 
 
 
 
191
  "pages": 34,
192
+ "shards": 3
193
  },
194
  {
195
+ "seconds": 1.1453777450001326,
196
+ "bytes": 12048299,
197
+ "pages": 34,
198
+ "shards": 3
199
  },
200
  {
201
+ "seconds": 1.011093388000063,
202
+ "bytes": 1199201,
203
+ "pages": 34,
204
+ "shards": 1
205
  },
206
  {
207
+ "seconds": 1.0240878049999083,
208
+ "bytes": 7205273,
209
+ "pages": 28,
210
  "shards": 4
211
  },
212
  {
213
+ "seconds": 1.1896957620001558,
214
+ "bytes": 10758587,
215
+ "pages": 42,
 
 
 
 
 
 
 
 
 
 
 
 
216
  "shards": 3
217
  },
218
  {
219
+ "seconds": 0.9081728179999118,
220
+ "bytes": 2358450,
221
+ "pages": 28,
222
  "shards": 2
223
  },
224
  {
225
+ "seconds": 1.152849502999743,
226
+ "bytes": 6119990,
227
+ "pages": 36,
228
  "shards": 2
229
  },
230
  {
231
+ "seconds": 1.0355856849996599,
232
+ "bytes": 6832540,
 
 
 
 
 
 
 
 
 
 
 
 
233
  "pages": 32,
234
  "shards": 2
235
  },
236
  {
237
+ "seconds": 20.55054765299974,
238
+ "bytes": 11612692,
239
+ "pages": 35,
240
+ "shards": 4
 
 
 
 
 
 
241
  },
242
  {
243
+ "seconds": 1.2162191410006926,
244
+ "bytes": 6401722,
245
  "pages": 45,
 
 
 
 
 
 
 
 
 
 
 
 
246
  "shards": 2
247
  },
248
  {
249
+ "seconds": 1.3737986060004914,
250
+ "bytes": 7819084,
251
+ "pages": 35,
252
  "shards": 3
253
  },
254
  {
255
+ "seconds": 1.079906239000593,
256
+ "bytes": 5671187,
257
+ "pages": 35,
258
  "shards": 2
259
  },
260
  {
261
+ "seconds": 1.1921040119996178,
262
+ "bytes": 5850261,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
263
  "pages": 28,
264
  "shards": 2
265
  },
266
  {
267
+ "seconds": 0.9688384539995241,
268
+ "bytes": 8315017,
269
+ "pages": 35,
270
  "shards": 3
271
  },
272
  {
273
+ "seconds": 1.0787890209994657,
274
+ "bytes": 6174644,
275
+ "pages": 35,
 
 
 
 
 
 
276
  "shards": 2
277
  },
278
  {
279
+ "seconds": 1.431994874999873,
280
+ "bytes": 7632897,
281
+ "pages": 39,
282
  "shards": 3
283
  },
284
  {
285
+ "seconds": 1.237401497999599,
286
+ "bytes": 5052734,
287
+ "pages": 33,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
288
  "shards": 2
289
  },
290
  {
291
+ "seconds": 1.0236900689997128,
292
+ "bytes": 6421100,
293
+ "pages": 37,
294
  "shards": 2
295
  },
296
  {
297
+ "seconds": 0.9477791999997862,
298
+ "bytes": 2414897,
299
  "pages": 27,
 
 
 
 
 
 
 
 
 
 
 
 
300
  "shards": 2
301
  },
302
  {
303
+ "seconds": 1.0987587250001525,
304
+ "bytes": 11899150,
305
+ "pages": 30,
 
 
 
 
 
 
 
 
 
 
 
 
306
  "shards": 3
307
  },
308
  {
309
+ "seconds": 1.4687352659993849,
310
+ "bytes": 10700369,
311
+ "pages": 42,
 
 
 
 
 
 
 
 
 
 
 
 
312
  "shards": 3
313
  },
314
  {
315
+ "seconds": 1.1858354729993152,
316
+ "bytes": 2742758,
 
 
 
 
 
 
 
 
 
 
 
 
317
  "pages": 37,
318
  "shards": 2
319
  },
320
  {
321
+ "seconds": 1.046531567000784,
322
+ "bytes": 5904437,
323
+ "pages": 28,
324
  "shards": 2
325
  },
326
  {
327
+ "seconds": 1.4689589510007863,
328
+ "bytes": 7230449,
 
 
 
 
 
 
 
 
 
 
 
 
329
  "pages": 37,
 
 
 
 
 
 
330
  "shards": 3
331
  },
332
  {
333
+ "seconds": 1.0110029770003166,
334
+ "bytes": 6117097,
335
+ "pages": 37,
336
  "shards": 2
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
337
  }
338
  ],
339
  "cache_key": {
 
378
  },
379
  "cache_reused": true,
380
  "resume_probe_seconds": [
381
+ 1.0657202160000452,
382
+ 0.013942628999984663
383
  ]
384
  }
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:550740c374917e84e0c390d9c1282f579154291db3c66058d82c3775128b44da
3
- size 721783935
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4273939d74c5226bc57df9a209306193dff7da001a42417dfdf439575063da35
3
+ size 721771007