appvoid commited on
Commit
f8495e3
·
verified ·
1 Parent(s): ff53054

Qwen90M step 815000; 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 +263 -167
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 810000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "b860c45edec3c7186db36858be040ec7573c74f1e3787dd00557ac89a4199fc4",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "a24063ea8ec1a680b2cbad4348b11f3720f7b12a48fa11c042f0aa687a24bfbb",
12
  "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
13
  "performance.json": "eb45b2d9ecb051b6b7e592c18f2a0bfebf4278147b783ba131f7a9a4453a9ec7",
14
- "shard_scheduler.json": "d3fc1112c30fbcd245d56533cfa5eb69c877b6ac83ebcc2954147694a2f34a08",
15
- "coverage.json": "284e9a6216e7c4741fa1f7e01c706d28eb52bfba351b9120d3853f5e4505d59b",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"
 
1
  {
2
+ "step": 815000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "fdf8df44c98676caf03f6204fff629f95101410723ef61ce23662a551a8f427b",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "125e3addd8b07f3024e8ed1e3c3e2f705e99c3cfcfa6b9f1547db31a2081c525",
12
  "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
13
  "performance.json": "eb45b2d9ecb051b6b7e592c18f2a0bfebf4278147b783ba131f7a9a4453a9ec7",
14
+ "shard_scheduler.json": "d8c00298782e3f8a28914751297b8d31ce17e96db9eaa5de1147c633462ccb8c",
15
+ "coverage.json": "1e90d10c39719e43af2b2ea2d77129d1f5aa16ce1b06969bdac4ca8d9ac3d0a6",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 810000,
3
  "stats": {
4
- "climbmix/blocks": 29160000,
5
- "climbmix/targets": 48745844785,
6
- "climbmix/input_tokens": 48748933709,
7
- "ultra_qa/blocks": 10935000,
8
- "ultra_qa/targets": 17839350326,
9
- "ultra_qa/input_tokens": 17839384717,
10
- "cortex/blocks": 10935000,
11
- "cortex/targets": 415808421,
12
- "cortex/input_tokens": 17381926974,
13
- "ultra_style/blocks": 10935000,
14
- "ultra_style/targets": 17296108093,
15
- "ultra_style/input_tokens": 17297015211,
16
- "openmath/blocks": 7290000,
17
- "openmath/targets": 11264198535,
18
- "openmath/input_tokens": 11267951300,
19
- "rewrite6/blocks": 3645000,
20
- "rewrite6/targets": 5975471100,
21
- "rewrite6/input_tokens": 5982022644
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 3993,
26
- "row": 2480,
27
  "start": {
28
- "file": 193,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 23,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
- "offset": 2048,
37
- "documents": 16357808
38
  },
39
  "rewrite6": {
40
- "page": 22,
41
- "row": 256,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 90112
47
  }
48
  },
49
  "epoch": 45,
50
  "offset": 2048,
51
- "documents": 9333548
52
  },
53
  "openmath": {
54
- "page": 2495,
55
- "row": 3276,
56
  "start": {
57
  "file": 23,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 176,
61
- "row": 570
62
  }
63
  },
64
  "epoch": 0,
65
- "offset": 2048,
66
- "documents": 10222796
67
  },
68
  "ultra_qa": {
69
- "page": 1079,
70
- "row": 780,
71
  "start": {
72
  "file": 8,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 4,
76
- "row": 4352
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 4096,
81
- "documents": 4420364
82
  },
83
  "ultra_style": {
84
- "page": 1647,
85
- "row": 173,
86
  "start": {
87
  "file": 9,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
- "row_group": 6,
91
- "row": 72213
92
  }
93
  },
94
  "epoch": 0,
95
- "offset": 2048,
96
- "documents": 6746285
97
  },
98
  "cortex": {
99
- "page": 212069,
100
- "row": 58,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 27144890
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 815000,
3
  "stats": {
4
+ "climbmix/blocks": 29340000,
5
+ "climbmix/targets": 49047052329,
6
+ "climbmix/input_tokens": 49050160293,
7
+ "ultra_qa/blocks": 11002500,
8
+ "ultra_qa/targets": 17949369763,
9
+ "ultra_qa/input_tokens": 17949404365,
10
+ "cortex/blocks": 11002500,
11
+ "cortex/targets": 418371734,
12
+ "cortex/input_tokens": 17489228968,
13
+ "ultra_style/blocks": 11002500,
14
+ "ultra_style/targets": 17402557074,
15
+ "ultra_style/input_tokens": 17403469802,
16
+ "openmath/blocks": 7335000,
17
+ "openmath/targets": 11333868401,
18
+ "openmath/input_tokens": 11337644586,
19
+ "rewrite6/blocks": 3667500,
20
+ "rewrite6/targets": 6012349385,
21
+ "rewrite6/input_tokens": 6018941143
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 4018,
26
+ "row": 620,
27
  "start": {
28
+ "file": 194,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 41,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
+ "offset": 8192,
37
+ "documents": 16458348
38
  },
39
  "rewrite6": {
40
+ "page": 36,
41
+ "row": 343,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 147456
47
  }
48
  },
49
  "epoch": 45,
50
  "offset": 2048,
51
+ "documents": 9390979
52
  },
53
  "openmath": {
54
+ "page": 2511,
55
+ "row": 1078,
56
  "start": {
57
  "file": 23,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 242,
61
+ "row": 106
62
  }
63
  },
64
  "epoch": 0,
65
+ "offset": 0,
66
+ "documents": 10286134
67
  },
68
  "ultra_qa": {
69
+ "page": 1085,
70
+ "row": 3490,
71
  "start": {
72
  "file": 8,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 4,
76
+ "row": 28928
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 4096,
81
+ "documents": 4447650
82
  },
83
  "ultra_style": {
84
+ "page": 1657,
85
+ "row": 1167,
86
  "start": {
87
  "file": 9,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
+ "row_group": 7,
91
+ "row": 27375
92
  }
93
  },
94
  "epoch": 0,
95
+ "offset": 4096,
96
+ "documents": 6788239
97
  },
98
  "cortex": {
99
+ "page": 213380,
100
+ "row": 6,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 27312646
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 810000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "b860c45edec3c7186db36858be040ec7573c74f1e3787dd00557ac89a4199fc4", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "a24063ea8ec1a680b2cbad4348b11f3720f7b12a48fa11c042f0aa687a24bfbb", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "eb45b2d9ecb051b6b7e592c18f2a0bfebf4278147b783ba131f7a9a4453a9ec7", "shard_scheduler.json": "d3fc1112c30fbcd245d56533cfa5eb69c877b6ac83ebcc2954147694a2f34a08", "coverage.json": "284e9a6216e7c4741fa1f7e01c706d28eb52bfba351b9120d3853f5e4505d59b", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000810000"}
 
1
+ {"step": 815000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "fdf8df44c98676caf03f6204fff629f95101410723ef61ce23662a551a8f427b", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "125e3addd8b07f3024e8ed1e3c3e2f705e99c3cfcfa6b9f1547db31a2081c525", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "eb45b2d9ecb051b6b7e592c18f2a0bfebf4278147b783ba131f7a9a4453a9ec7", "shard_scheduler.json": "d8c00298782e3f8a28914751297b8d31ce17e96db9eaa5de1147c633462ccb8c", "coverage.json": "1e90d10c39719e43af2b2ea2d77129d1f5aa16ce1b06969bdac4ca8d9ac3d0a6", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000815000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b860c45edec3c7186db36858be040ec7573c74f1e3787dd00557ac89a4199fc4
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fdf8df44c98676caf03f6204fff629f95101410723ef61ce23662a551a8f427b
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": 76742.70439241717,
26
- "required_pages_per_second_upper_estimate": 0.0053092383574745395
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": 9592.838049052147,
40
- "required_pages_per_second_upper_estimate": 0.0040163209699003825
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": 19185.676098104293,
54
- "required_pages_per_second_upper_estimate": 0.0039958703709320606
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": 28778.514147156442,
68
- "required_pages_per_second_upper_estimate": 0.0017252279808752647
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": 28778.514147156442,
82
- "required_pages_per_second_upper_estimate": 0.002237402891635416
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": 28778.514147156442,
96
- "required_pages_per_second_upper_estimate": 0.3691398795186881
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,171 +158,27 @@
158
  "prefetch_batches": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
- "producer_batch_seconds": 0.1648581179296542,
162
- "producer_p95_seconds": 0.19809514199914702,
163
- "gpu_update_seconds": 0.9607167298014186,
164
- "producer_batches_per_second": 6.065822008393333,
165
- "gpu_batches_per_second": 1.0408895452530542,
166
- "producer_headroom": 5.8275366834611155,
167
  "sustainable": true,
168
- "buffer_seconds": 60.525153977489374,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 81,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 375296.30858700874,
174
- "upload_bytes_per_second": 4130574.7776954207,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 80.52985137183853,
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": 1.6723634379995929,
184
- "bytes": 50251,
185
- "pages": 1,
186
- "shards": 1
187
- },
188
- {
189
- "seconds": 2.1410375759996896,
190
- "bytes": 10006036,
191
- "pages": 40,
192
- "shards": 3
193
- },
194
- {
195
- "seconds": 1.5198384699997405,
196
- "bytes": 2430008,
197
- "pages": 30,
198
- "shards": 2
199
- },
200
- {
201
- "seconds": 1.2644712469991646,
202
- "bytes": 6316580,
203
- "pages": 36,
204
- "shards": 2
205
- },
206
- {
207
- "seconds": 1.6103071739998995,
208
- "bytes": 7727369,
209
- "pages": 38,
210
- "shards": 3
211
- },
212
- {
213
- "seconds": 1.5575974499997756,
214
- "bytes": 11152630,
215
- "pages": 33,
216
- "shards": 3
217
- },
218
- {
219
- "seconds": 1.9014979050007241,
220
- "bytes": 5893171,
221
- "pages": 27,
222
- "shards": 2
223
- },
224
- {
225
- "seconds": 1.375044964999688,
226
- "bytes": 2500711,
227
- "pages": 36,
228
- "shards": 2
229
- },
230
- {
231
- "seconds": 1.54583506400013,
232
- "bytes": 5956633,
233
- "pages": 28,
234
- "shards": 2
235
- },
236
- {
237
- "seconds": 1.6070335669992346,
238
- "bytes": 5667674,
239
- "pages": 32,
240
- "shards": 2
241
- },
242
- {
243
- "seconds": 1.649759572000221,
244
- "bytes": 7616382,
245
- "pages": 37,
246
- "shards": 3
247
- },
248
- {
249
- "seconds": 1.8874110320002728,
250
- "bytes": 11916705,
251
- "pages": 38,
252
- "shards": 3
253
- },
254
- {
255
- "seconds": 1.604149986000266,
256
- "bytes": 2375336,
257
- "pages": 28,
258
- "shards": 2
259
- },
260
- {
261
- "seconds": 1.4722619869999107,
262
- "bytes": 10022494,
263
- "pages": 34,
264
- "shards": 3
265
- },
266
- {
267
- "seconds": 1.460268009000174,
268
- "bytes": 1130080,
269
- "pages": 34,
270
- "shards": 1
271
- },
272
- {
273
- "seconds": 1.6416546800001015,
274
- "bytes": 7689453,
275
- "pages": 38,
276
- "shards": 4
277
- },
278
- {
279
- "seconds": 1.5363667999999961,
280
- "bytes": 6120972,
281
- "pages": 36,
282
- "shards": 2
283
- },
284
- {
285
- "seconds": 1.693875957000273,
286
- "bytes": 10707952,
287
- "pages": 29,
288
- "shards": 3
289
- },
290
- {
291
- "seconds": 1.540582020000329,
292
- "bytes": 7589375,
293
- "pages": 37,
294
- "shards": 3
295
- },
296
- {
297
- "seconds": 1.573228470999311,
298
- "bytes": 6089737,
299
- "pages": 35,
300
- "shards": 2
301
- },
302
- {
303
- "seconds": 1.5443801999999778,
304
- "bytes": 7496481,
305
- "pages": 38,
306
- "shards": 3
307
- },
308
- {
309
- "seconds": 1.4706297409993567,
310
- "bytes": 5704544,
311
- "pages": 26,
312
- "shards": 2
313
- },
314
- {
315
- "seconds": 1.637408425999638,
316
- "bytes": 1096275,
317
- "pages": 34,
318
- "shards": 1
319
- },
320
- {
321
- "seconds": 1.9842258310000034,
322
- "bytes": 13303551,
323
- "pages": 43,
324
- "shards": 4
325
- },
326
  {
327
  "seconds": 1.4946416839993617,
328
  "bytes": 5892242,
@@ -466,6 +322,246 @@
466
  "bytes": 12131753,
467
  "pages": 31,
468
  "shards": 3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
469
  }
470
  ],
471
  "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": 76422.60943566434,
26
+ "required_pages_per_second_upper_estimate": 0.005287093445643752
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": 9552.826179458043,
40
+ "required_pages_per_second_upper_estimate": 0.003999568835644181
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": 19105.652358916086,
54
+ "required_pages_per_second_upper_estimate": 0.003979203536427075
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": 28658.47853837413,
68
+ "required_pages_per_second_upper_estimate": 0.001718032029412527
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": 28658.47853837413,
82
+ "required_pages_per_second_upper_estimate": 0.0022280706510334355
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": 28658.47853837413,
96
+ "required_pages_per_second_upper_estimate": 0.3676001916134238
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
 
158
  "prefetch_batches": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
+ "producer_batch_seconds": 0.2541760841795906,
162
+ "producer_p95_seconds": 0.2388652240006195,
163
+ "gpu_update_seconds": 0.9647406774570715,
164
+ "producer_batches_per_second": 3.9342804545428445,
165
+ "gpu_batches_per_second": 1.0365479795418884,
166
+ "producer_headroom": 3.795560391021779,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.7786626797955,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 82,
172
  "max_pending_pages": 4096,
173
+ "generation_bytes_per_second": 448277.4506157306,
174
+ "upload_bytes_per_second": 3886909.442531961,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 96.18990553716601,
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": 1.4946416839993617,
184
  "bytes": 5892242,
 
322
  "bytes": 12131753,
323
  "pages": 31,
324
  "shards": 3
325
+ },
326
+ {
327
+ "seconds": 1.6632540899991,
328
+ "bytes": 5726975,
329
+ "pages": 37,
330
+ "shards": 2
331
+ },
332
+ {
333
+ "seconds": 1.7110617899998033,
334
+ "bytes": 7430334,
335
+ "pages": 30,
336
+ "shards": 3
337
+ },
338
+ {
339
+ "seconds": 1.6944532489997073,
340
+ "bytes": 1160058,
341
+ "pages": 35,
342
+ "shards": 1
343
+ },
344
+ {
345
+ "seconds": 2.4445176089993765,
346
+ "bytes": 7341476,
347
+ "pages": 34,
348
+ "shards": 3
349
+ },
350
+ {
351
+ "seconds": 1.587468388999696,
352
+ "bytes": 5869829,
353
+ "pages": 27,
354
+ "shards": 2
355
+ },
356
+ {
357
+ "seconds": 1.9299586070010264,
358
+ "bytes": 10891963,
359
+ "pages": 40,
360
+ "shards": 3
361
+ },
362
+ {
363
+ "seconds": 1.62565908400029,
364
+ "bytes": 7724112,
365
+ "pages": 39,
366
+ "shards": 3
367
+ },
368
+ {
369
+ "seconds": 1.4196104830007243,
370
+ "bytes": 988295,
371
+ "pages": 27,
372
+ "shards": 1
373
+ },
374
+ {
375
+ "seconds": 2.0304724819998228,
376
+ "bytes": 10476032,
377
+ "pages": 30,
378
+ "shards": 3
379
+ },
380
+ {
381
+ "seconds": 1.5808353909997095,
382
+ "bytes": 7239609,
383
+ "pages": 35,
384
+ "shards": 3
385
+ },
386
+ {
387
+ "seconds": 1.859183167000083,
388
+ "bytes": 1137645,
389
+ "pages": 35,
390
+ "shards": 1
391
+ },
392
+ {
393
+ "seconds": 4.019235966999986,
394
+ "bytes": 13232606,
395
+ "pages": 36,
396
+ "shards": 4
397
+ },
398
+ {
399
+ "seconds": 1.720672621000631,
400
+ "bytes": 611156,
401
+ "pages": 21,
402
+ "shards": 1
403
+ },
404
+ {
405
+ "seconds": 1.6194256800008588,
406
+ "bytes": 9820718,
407
+ "pages": 47,
408
+ "shards": 3
409
+ },
410
+ {
411
+ "seconds": 1.6360844549999456,
412
+ "bytes": 7454978,
413
+ "pages": 29,
414
+ "shards": 3
415
+ },
416
+ {
417
+ "seconds": 1.568186391999916,
418
+ "bytes": 1164036,
419
+ "pages": 35,
420
+ "shards": 1
421
+ },
422
+ {
423
+ "seconds": 2.5616059220010357,
424
+ "bytes": 11807815,
425
+ "pages": 42,
426
+ "shards": 4
427
+ },
428
+ {
429
+ "seconds": 1.6904037299991614,
430
+ "bytes": 12097177,
431
+ "pages": 34,
432
+ "shards": 3
433
+ },
434
+ {
435
+ "seconds": 1.2079874939990987,
436
+ "bytes": 1225243,
437
+ "pages": 35,
438
+ "shards": 1
439
+ },
440
+ {
441
+ "seconds": 1.7503577689985832,
442
+ "bytes": 11722749,
443
+ "pages": 33,
444
+ "shards": 4
445
+ },
446
+ {
447
+ "seconds": 1.717468271999678,
448
+ "bytes": 6201366,
449
+ "pages": 37,
450
+ "shards": 2
451
+ },
452
+ {
453
+ "seconds": 1.4180366079999658,
454
+ "bytes": 818969,
455
+ "pages": 25,
456
+ "shards": 1
457
+ },
458
+ {
459
+ "seconds": 1.935665021999739,
460
+ "bytes": 7829235,
461
+ "pages": 39,
462
+ "shards": 3
463
+ },
464
+ {
465
+ "seconds": 1.7852341269990575,
466
+ "bytes": 11660999,
467
+ "pages": 37,
468
+ "shards": 3
469
+ },
470
+ {
471
+ "seconds": 1.8856627909990493,
472
+ "bytes": 7118698,
473
+ "pages": 30,
474
+ "shards": 3
475
+ },
476
+ {
477
+ "seconds": 1.6447518009990745,
478
+ "bytes": 6129393,
479
+ "pages": 37,
480
+ "shards": 2
481
+ },
482
+ {
483
+ "seconds": 1.743864690999544,
484
+ "bytes": 1054003,
485
+ "pages": 35,
486
+ "shards": 1
487
+ },
488
+ {
489
+ "seconds": 1.5393589430004795,
490
+ "bytes": 11424070,
491
+ "pages": 35,
492
+ "shards": 4
493
+ },
494
+ {
495
+ "seconds": 1.9194490070003667,
496
+ "bytes": 6288888,
497
+ "pages": 35,
498
+ "shards": 2
499
+ },
500
+ {
501
+ "seconds": 1.4686280049991183,
502
+ "bytes": 8470485,
503
+ "pages": 37,
504
+ "shards": 4
505
+ },
506
+ {
507
+ "seconds": 1.9930366190001223,
508
+ "bytes": 5866239,
509
+ "pages": 26,
510
+ "shards": 2
511
+ },
512
+ {
513
+ "seconds": 2.011558195001271,
514
+ "bytes": 5842602,
515
+ "pages": 41,
516
+ "shards": 2
517
+ },
518
+ {
519
+ "seconds": 1.6693870859999151,
520
+ "bytes": 7612861,
521
+ "pages": 29,
522
+ "shards": 3
523
+ },
524
+ {
525
+ "seconds": 1.3870315900003334,
526
+ "bytes": 1261477,
527
+ "pages": 35,
528
+ "shards": 1
529
+ },
530
+ {
531
+ "seconds": 1.6420892379992438,
532
+ "bytes": 7607765,
533
+ "pages": 39,
534
+ "shards": 3
535
+ },
536
+ {
537
+ "seconds": 1.5253431410001212,
538
+ "bytes": 5688723,
539
+ "pages": 19,
540
+ "shards": 2
541
+ },
542
+ {
543
+ "seconds": 1.593651556999248,
544
+ "bytes": 11614889,
545
+ "pages": 46,
546
+ "shards": 3
547
+ },
548
+ {
549
+ "seconds": 1.818591990000641,
550
+ "bytes": 7461958,
551
+ "pages": 30,
552
+ "shards": 3
553
+ },
554
+ {
555
+ "seconds": 1.4790617950002343,
556
+ "bytes": 1048131,
557
+ "pages": 34,
558
+ "shards": 1
559
+ },
560
+ {
561
+ "seconds": 1.3689376410002296,
562
+ "bytes": 5834070,
563
+ "pages": 28,
564
+ "shards": 2
565
  }
566
  ],
567
  "cache_key": {
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:a24063ea8ec1a680b2cbad4348b11f3720f7b12a48fa11c042f0aa687a24bfbb
3
- size 721768959
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:125e3addd8b07f3024e8ed1e3c3e2f705e99c3cfcfa6b9f1547db31a2081c525
3
+ size 721763839