appvoid commited on
Commit
f153b7b
·
verified ·
1 Parent(s): b248bb7

Qwen90M step 570000; complete resumable checkpoint

Browse files
Files changed (6) hide show
  1. complete.json +5 -5
  2. coverage.json +47 -47
  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": 565000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "95a5b58e2b26fc5a9ee97e43501f78f2f22c159d31689ff3bf684a9cdf3dee89",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "f2026403cd8252eb8fbef1272eb9f1d5d4b596cf0488b73b1b78201830ca2e97",
12
  "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
13
  "performance.json": "b7cb16c0913422fe54a766da97356133a598f4da7e691939f51028a8bce75667",
14
- "shard_scheduler.json": "e3f3e619a117b2827f3e52efa7e755833ae25de54c4ea382eb05fbf3b86130d8",
15
- "coverage.json": "6d935be2426b023c8381c6f1092c50faa4e567b7afaa6fdf07207b78a26ef97d",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 570000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "7d4140ba71fe97e1e82990744e5a0ec1dbc7c23b51309036d4c2230272e12e25",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "a3bc7b8df78537b2475ab2dd3ebc2aba12dbb1b97f2867ccda2914ec9a3004e2",
12
  "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
13
  "performance.json": "b7cb16c0913422fe54a766da97356133a598f4da7e691939f51028a8bce75667",
14
+ "shard_scheduler.json": "01987d1d04369b6d0c27ef5a7d5f76afa9290cfe2789b0c2414b1dca44f86adb",
15
+ "coverage.json": "ea81a0c5669ee210b85f1ffe803c00bba540b014a73ee1c73b43e81dbc6770b8",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 565000,
3
  "stats": {
4
- "climbmix/blocks": 20340000,
5
- "climbmix/targets": 34000795378,
6
- "climbmix/input_tokens": 34002950085,
7
- "ultra_qa/blocks": 7627500,
8
- "ultra_qa/targets": 12443347733,
9
- "ultra_qa/input_tokens": 12443371739,
10
- "cortex/blocks": 7627500,
11
- "cortex/targets": 290056072,
12
- "cortex/input_tokens": 12124570582,
13
- "ultra_style/blocks": 7627500,
14
- "ultra_style/targets": 12064354329,
15
- "ultra_style/input_tokens": 12064986794,
16
- "openmath/blocks": 5085000,
17
- "openmath/targets": 7857224925,
18
- "openmath/input_tokens": 7859841919,
19
- "rewrite6/blocks": 2542500,
20
- "rewrite6/targets": 4168025058,
21
- "rewrite6/input_tokens": 4172594909
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 2785,
26
- "row": 3702,
27
  "start": {
28
- "file": 134,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 64,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
- "offset": 8192,
37
- "documents": 11411062
38
  },
39
  "rewrite6": {
40
- "page": 34,
41
- "row": 3660,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 139264
47
  }
48
  },
49
  "epoch": 31,
50
  "offset": 0,
51
- "documents": 6510448
52
  },
53
  "openmath": {
54
- "page": 1740,
55
- "row": 2763,
56
  "start": {
57
  "file": 16,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 140,
61
- "row": 640
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 7129803
67
  },
68
  "ultra_qa": {
69
- "page": 752,
70
- "row": 3016,
71
  "start": {
72
  "file": 5,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 7,
76
- "row": 30558
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
- "documents": 3083208
82
  },
83
  "ultra_style": {
84
- "page": 1148,
85
- "row": 3198,
86
  "start": {
87
  "file": 6,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
- "row_group": 6,
91
- "row": 81384
92
  }
93
  },
94
  "epoch": 0,
95
- "offset": 4096,
96
- "documents": 4705406
97
  },
98
  "cortex": {
99
- "page": 147929,
100
- "row": 71,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 18934983
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 570000,
3
  "stats": {
4
+ "climbmix/blocks": 20520000,
5
+ "climbmix/targets": 34301849164,
6
+ "climbmix/input_tokens": 34304022931,
7
+ "ultra_qa/blocks": 7695000,
8
+ "ultra_qa/targets": 12553627897,
9
+ "ultra_qa/input_tokens": 12553652133,
10
+ "cortex/blocks": 7695000,
11
+ "cortex/targets": 292622737,
12
+ "cortex/input_tokens": 12231895676,
13
+ "ultra_style/blocks": 7695000,
14
+ "ultra_style/targets": 12171363052,
15
+ "ultra_style/input_tokens": 12172001166,
16
+ "openmath/blocks": 5130000,
17
+ "openmath/targets": 7926745772,
18
+ "openmath/input_tokens": 7929385871,
19
+ "rewrite6/blocks": 2565000,
20
+ "rewrite6/targets": 4205011866,
21
+ "rewrite6/input_tokens": 4209622064
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 2810,
26
+ "row": 1988,
27
  "start": {
28
+ "file": 135,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 82,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
+ "offset": 2048,
37
+ "documents": 11511748
38
  },
39
  "rewrite6": {
40
+ "page": 48,
41
+ "row": 3760,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 196608
47
  }
48
  },
49
  "epoch": 31,
50
  "offset": 0,
51
+ "documents": 6567892
52
  },
53
  "openmath": {
54
+ "page": 1756,
55
+ "row": 343,
56
  "start": {
57
  "file": 16,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 206,
61
+ "row": 176
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 7192919
67
  },
68
  "ultra_qa": {
69
+ "page": 759,
70
+ "row": 1682,
71
  "start": {
72
  "file": 5,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 7,
76
+ "row": 59230
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
+ "documents": 3110546
82
  },
83
  "ultra_style": {
84
+ "page": 1158,
85
+ "row": 3739,
86
  "start": {
87
  "file": 6,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
+ "row_group": 7,
91
+ "row": 37605
92
  }
93
  },
94
  "epoch": 0,
95
+ "offset": 0,
96
+ "documents": 4746907
97
  },
98
  "cortex": {
99
+ "page": 149240,
100
+ "row": 26,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 19102746
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 565000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "95a5b58e2b26fc5a9ee97e43501f78f2f22c159d31689ff3bf684a9cdf3dee89", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "f2026403cd8252eb8fbef1272eb9f1d5d4b596cf0488b73b1b78201830ca2e97", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "b7cb16c0913422fe54a766da97356133a598f4da7e691939f51028a8bce75667", "shard_scheduler.json": "e3f3e619a117b2827f3e52efa7e755833ae25de54c4ea382eb05fbf3b86130d8", "coverage.json": "6d935be2426b023c8381c6f1092c50faa4e567b7afaa6fdf07207b78a26ef97d", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000565000"}
 
1
+ {"step": 570000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "7d4140ba71fe97e1e82990744e5a0ec1dbc7c23b51309036d4c2230272e12e25", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "a3bc7b8df78537b2475ab2dd3ebc2aba12dbb1b97f2867ccda2914ec9a3004e2", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "b7cb16c0913422fe54a766da97356133a598f4da7e691939f51028a8bce75667", "shard_scheduler.json": "01987d1d04369b6d0c27ef5a7d5f76afa9290cfe2789b0c2414b1dca44f86adb", "coverage.json": "ea81a0c5669ee210b85f1ffe803c00bba540b014a73ee1c73b43e81dbc6770b8", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000570000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:95a5b58e2b26fc5a9ee97e43501f78f2f22c159d31689ff3bf684a9cdf3dee89
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7d4140ba71fe97e1e82990744e5a0ec1dbc7c23b51309036d4c2230272e12e25
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": 77200.01258345156,
26
- "required_pages_per_second_upper_estimate": 0.005340875999231493
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": 9650.001572931445,
40
- "required_pages_per_second_upper_estimate": 0.004040254143638525
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": 19300.00314586289,
54
- "required_pages_per_second_upper_estimate": 0.0040196816799731765
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": 28950.004718794335,
68
- "required_pages_per_second_upper_estimate": 0.001735508578794015
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": 28950.004718794335,
82
- "required_pages_per_second_upper_estimate": 0.0022507355292729556
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": 28950.004718794335,
96
- "required_pages_per_second_upper_estimate": 0.3713395764394291
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": 0.04976373940626644,
162
- "producer_p95_seconds": 0.1702074619997802,
163
- "gpu_update_seconds": 0.9550257510683902,
164
- "producer_batches_per_second": 20.094952910111818,
165
- "gpu_batches_per_second": 1.0470921845628738,
166
- "producer_headroom": 19.191197495663474,
167
  "sustainable": true,
168
- "buffer_seconds": 60.16662231730858,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 86,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 317438.5499692344,
174
- "upload_bytes_per_second": 5504378.909206007,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 68.11492323215269,
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": 0.8258607079999933,
184
- "bytes": 145464,
185
- "pages": 4,
186
- "shards": 1
187
- },
188
- {
189
- "seconds": 1.1176107269999989,
190
- "bytes": 5541224,
191
- "pages": 18,
192
- "shards": 2
193
- },
194
- {
195
- "seconds": 1.2054241769999408,
196
- "bytes": 2385611,
197
- "pages": 27,
198
- "shards": 2
199
- },
200
- {
201
- "seconds": 0.8888624019999725,
202
- "bytes": 6324210,
203
- "pages": 34,
204
- "shards": 2
205
- },
206
- {
207
- "seconds": 1.1897545489999857,
208
- "bytes": 1226596,
209
- "pages": 34,
210
  "shards": 2
211
  },
212
  {
213
- "seconds": 1.2994146069999033,
214
- "bytes": 17142614,
215
- "pages": 34,
216
- "shards": 5
217
  },
218
  {
219
- "seconds": 1.1055204159999903,
220
- "bytes": 6016330,
221
- "pages": 35,
222
- "shards": 2
223
  },
224
  {
225
- "seconds": 0.9128079020001678,
226
- "bytes": 2402677,
227
- "pages": 27,
228
  "shards": 2
229
  },
230
  {
231
- "seconds": 1.0666157699999985,
232
- "bytes": 6099291,
233
- "pages": 35,
234
  "shards": 2
235
  },
236
  {
237
- "seconds": 1.0668149449998054,
238
- "bytes": 5027383,
239
- "pages": 32,
240
  "shards": 2
241
  },
242
  {
243
- "seconds": 1.0457134279999991,
244
- "bytes": 7722680,
245
- "pages": 38,
246
  "shards": 3
247
  },
248
  {
249
- "seconds": 1.1198342459999822,
250
- "bytes": 518226,
251
- "pages": 18,
252
- "shards": 1
253
  },
254
  {
255
- "seconds": 1.3007597939999869,
256
- "bytes": 12269409,
257
- "pages": 44,
258
  "shards": 3
259
  },
260
  {
261
- "seconds": 1.437565898999992,
262
- "bytes": 12411381,
263
- "pages": 38,
264
  "shards": 4
265
  },
266
  {
267
- "seconds": 0.765915884000151,
268
- "bytes": 1219645,
269
- "pages": 34,
270
  "shards": 1
271
  },
272
  {
273
- "seconds": 1.470795656000064,
274
- "bytes": 7604342,
275
- "pages": 30,
276
- "shards": 3
277
  },
278
  {
279
- "seconds": 0.9404902840001341,
280
- "bytes": 6153110,
281
- "pages": 37,
282
  "shards": 2
283
  },
284
  {
285
- "seconds": 1.0696395039999516,
286
- "bytes": 10812914,
287
- "pages": 37,
288
  "shards": 3
289
  },
290
  {
291
- "seconds": 1.5107661739998548,
292
- "bytes": 7696973,
293
- "pages": 34,
294
  "shards": 3
295
  },
296
  {
297
- "seconds": 0.9512165999999525,
298
- "bytes": 1125687,
299
- "pages": 30,
300
- "shards": 1
301
  },
302
  {
303
- "seconds": 1.0864234489999944,
304
- "bytes": 7369290,
305
- "pages": 38,
306
- "shards": 3
307
  },
308
  {
309
- "seconds": 1.1581361090002247,
310
- "bytes": 10067921,
311
- "pages": 37,
312
- "shards": 3
313
  },
314
  {
315
- "seconds": 0.7976953110000977,
316
- "bytes": 1131427,
317
  "pages": 35,
318
  "shards": 1
319
  },
320
  {
321
- "seconds": 1.1826321499997903,
322
- "bytes": 12996691,
323
- "pages": 33,
324
- "shards": 4
325
  },
326
  {
327
- "seconds": 0.7847848540000086,
328
- "bytes": 603387,
329
- "pages": 17,
330
- "shards": 1
331
  },
332
  {
333
- "seconds": 1.5595179319998351,
334
- "bytes": 10763416,
335
- "pages": 29,
336
  "shards": 3
337
  },
338
  {
339
- "seconds": 0.8392939139998816,
340
- "bytes": 2549967,
341
  "pages": 36,
342
- "shards": 2
343
  },
344
  {
345
- "seconds": 0.9600205120000282,
346
- "bytes": 6407325,
347
- "pages": 36,
348
- "shards": 2
349
  },
350
  {
351
- "seconds": 1.0153757429998223,
352
- "bytes": 6078141,
353
  "pages": 28,
354
  "shards": 2
355
  },
356
  {
357
- "seconds": 0.8920637669998541,
358
- "bytes": 2452797,
359
- "pages": 32,
360
- "shards": 2
361
  },
362
  {
363
- "seconds": 1.1147250890003306,
364
- "bytes": 15773840,
365
- "pages": 37,
366
  "shards": 4
367
  },
368
  {
369
- "seconds": 0.9467493070001183,
370
- "bytes": 2673204,
371
- "pages": 36,
372
- "shards": 2
373
  },
374
  {
375
- "seconds": 1.3979449759999625,
376
- "bytes": 6163668,
377
- "pages": 36,
378
- "shards": 2
379
  },
380
  {
381
- "seconds": 1.1962734589997126,
382
- "bytes": 10535589,
383
- "pages": 38,
384
  "shards": 3
385
  },
386
  {
387
- "seconds": 0.9773600910002642,
388
- "bytes": 2300498,
389
- "pages": 27,
390
  "shards": 2
391
  },
392
  {
393
- "seconds": 1.3482739699993544,
394
- "bytes": 6309981,
395
- "pages": 37,
396
  "shards": 2
397
  },
398
  {
399
- "seconds": 1.1507453160002115,
400
- "bytes": 8207384,
401
- "pages": 33,
402
- "shards": 4
403
  },
404
  {
405
- "seconds": 1.092882823999389,
406
- "bytes": 10744820,
407
- "pages": 39,
408
  "shards": 3
409
  },
410
  {
411
- "seconds": 1.0739986469998257,
412
- "bytes": 6490657,
413
  "pages": 36,
414
  "shards": 2
415
  },
416
  {
417
- "seconds": 0.8932356570003321,
418
- "bytes": 2377409,
419
- "pages": 28,
420
  "shards": 2
421
  },
422
  {
423
- "seconds": 1.035186543999771,
424
- "bytes": 6290854,
425
- "pages": 36,
426
  "shards": 2
427
  },
428
  {
429
- "seconds": 1.1961185629998,
430
- "bytes": 11182870,
431
- "pages": 37,
432
  "shards": 3
433
  },
434
  {
435
- "seconds": 1.2023984399993424,
436
- "bytes": 8470279,
437
- "pages": 37,
438
  "shards": 3
439
  },
440
  {
441
- "seconds": 0.9793791169995529,
442
- "bytes": 5849196,
443
- "pages": 28,
444
  "shards": 2
445
  },
446
  {
447
- "seconds": 0.8912132279992875,
448
- "bytes": 2690813,
449
- "pages": 37,
450
- "shards": 2
451
  },
452
  {
453
- "seconds": 1.3103161509998245,
454
- "bytes": 5739919,
455
- "pages": 29,
456
  "shards": 2
457
  },
458
  {
459
- "seconds": 1.3039567450005052,
460
- "bytes": 9869427,
461
  "pages": 37,
462
- "shards": 3
463
  },
464
  {
465
- "seconds": 0.9875589590001255,
466
- "bytes": 2599753,
467
- "pages": 37,
468
  "shards": 2
469
  },
470
  {
471
- "seconds": 1.266042872000071,
472
- "bytes": 11864568,
473
- "pages": 36,
474
  "shards": 3
475
  },
476
  {
477
- "seconds": 1.21107310699972,
478
- "bytes": 11765224,
479
- "pages": 34,
480
  "shards": 4
481
  },
482
  {
483
- "seconds": 2.659455514000001,
484
- "bytes": 1230518,
485
- "pages": 35,
486
  "shards": 1
487
  },
488
  {
489
- "seconds": 1.1675947069998074,
490
- "bytes": 6176188,
491
- "pages": 36,
492
  "shards": 2
493
  },
494
  {
495
- "seconds": 0.943889709999894,
496
- "bytes": 2411720,
497
- "pages": 28,
498
  "shards": 2
499
  },
500
  {
501
- "seconds": 1.251686485999926,
502
- "bytes": 10374348,
503
- "pages": 34,
504
  "shards": 3
505
  },
506
  {
507
- "seconds": 1.5644413750005697,
508
- "bytes": 11913928,
509
  "pages": 35,
510
  "shards": 3
511
  },
512
  {
513
- "seconds": 1.3241815069995937,
514
- "bytes": 2678696,
515
  "pages": 37,
516
  "shards": 2
517
  },
518
  {
519
- "seconds": 1.1117148500006806,
520
- "bytes": 6299752,
521
  "pages": 35,
522
  "shards": 2
523
  },
524
  {
525
- "seconds": 3.5441821820004407,
526
- "bytes": 11982351,
527
- "pages": 39,
528
  "shards": 4
529
  },
530
  {
531
- "seconds": 1.0069952920002834,
532
- "bytes": 1220170,
533
- "pages": 35,
534
  "shards": 1
535
  },
536
  {
537
- "seconds": 0.9136476189996756,
538
- "bytes": 5828433,
539
- "pages": 27,
540
  "shards": 2
541
  },
542
  {
543
- "seconds": 1.4721885819999443,
544
- "bytes": 8185407,
545
- "pages": 36,
546
  "shards": 3
547
  },
548
  {
549
- "seconds": 1.3622697179998795,
550
- "bytes": 10825600,
551
- "pages": 37,
552
- "shards": 3
553
  },
554
  {
555
- "seconds": 1.2178643509996618,
556
- "bytes": 7619131,
557
- "pages": 36,
558
  "shards": 3
559
  },
560
  {
561
- "seconds": 0.8663683109998601,
562
- "bytes": 939323,
563
- "pages": 28,
564
  "shards": 1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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": 76299.61100662185,
26
+ "required_pages_per_second_upper_estimate": 0.005278584128926917
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": 9537.451375827732,
40
+ "required_pages_per_second_upper_estimate": 0.003993131726426578
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": 19074.902751655463,
54
+ "required_pages_per_second_upper_estimate": 0.003972799204156364
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": 28612.354127483195,
68
+ "required_pages_per_second_upper_estimate": 0.0017152669414075132
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": 28612.354127483195,
82
+ "required_pages_per_second_upper_estimate": 0.002224484680966508
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": 28612.354127483195,
96
+ "required_pages_per_second_upper_estimate": 0.36700855719504877
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": 0.11649422678902965,
162
+ "producer_p95_seconds": 0.21004571299999952,
163
+ "gpu_update_seconds": 0.9662958831284386,
164
+ "producer_batches_per_second": 8.5841163769514,
165
+ "gpu_batches_per_second": 1.03487970657853,
166
+ "producer_headroom": 8.294796315343547,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.87664063709163,
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": 322997.3148886,
174
+ "upload_bytes_per_second": 5468898.378129647,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 69.30770478242398,
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": 1.035186543999771,
184
+ "bytes": 6290854,
185
+ "pages": 36,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
186
  "shards": 2
187
  },
188
  {
189
+ "seconds": 1.1961185629998,
190
+ "bytes": 11182870,
191
+ "pages": 37,
192
+ "shards": 3
193
  },
194
  {
195
+ "seconds": 1.2023984399993424,
196
+ "bytes": 8470279,
197
+ "pages": 37,
198
+ "shards": 3
199
  },
200
  {
201
+ "seconds": 0.9793791169995529,
202
+ "bytes": 5849196,
203
+ "pages": 28,
204
  "shards": 2
205
  },
206
  {
207
+ "seconds": 0.8912132279992875,
208
+ "bytes": 2690813,
209
+ "pages": 37,
210
  "shards": 2
211
  },
212
  {
213
+ "seconds": 1.3103161509998245,
214
+ "bytes": 5739919,
215
+ "pages": 29,
216
  "shards": 2
217
  },
218
  {
219
+ "seconds": 1.3039567450005052,
220
+ "bytes": 9869427,
221
+ "pages": 37,
222
  "shards": 3
223
  },
224
  {
225
+ "seconds": 0.9875589590001255,
226
+ "bytes": 2599753,
227
+ "pages": 37,
228
+ "shards": 2
229
  },
230
  {
231
+ "seconds": 1.266042872000071,
232
+ "bytes": 11864568,
233
+ "pages": 36,
234
  "shards": 3
235
  },
236
  {
237
+ "seconds": 1.21107310699972,
238
+ "bytes": 11765224,
239
+ "pages": 34,
240
  "shards": 4
241
  },
242
  {
243
+ "seconds": 2.659455514000001,
244
+ "bytes": 1230518,
245
+ "pages": 35,
246
  "shards": 1
247
  },
248
  {
249
+ "seconds": 1.1675947069998074,
250
+ "bytes": 6176188,
251
+ "pages": 36,
252
+ "shards": 2
253
  },
254
  {
255
+ "seconds": 0.943889709999894,
256
+ "bytes": 2411720,
257
+ "pages": 28,
258
  "shards": 2
259
  },
260
  {
261
+ "seconds": 1.251686485999926,
262
+ "bytes": 10374348,
263
+ "pages": 34,
264
  "shards": 3
265
  },
266
  {
267
+ "seconds": 1.5644413750005697,
268
+ "bytes": 11913928,
269
+ "pages": 35,
270
  "shards": 3
271
  },
272
  {
273
+ "seconds": 1.3241815069995937,
274
+ "bytes": 2678696,
275
+ "pages": 37,
276
+ "shards": 2
277
  },
278
  {
279
+ "seconds": 1.1117148500006806,
280
+ "bytes": 6299752,
281
+ "pages": 35,
282
+ "shards": 2
283
  },
284
  {
285
+ "seconds": 3.5441821820004407,
286
+ "bytes": 11982351,
287
+ "pages": 39,
288
+ "shards": 4
289
  },
290
  {
291
+ "seconds": 1.0069952920002834,
292
+ "bytes": 1220170,
293
  "pages": 35,
294
  "shards": 1
295
  },
296
  {
297
+ "seconds": 0.9136476189996756,
298
+ "bytes": 5828433,
299
+ "pages": 27,
300
+ "shards": 2
301
  },
302
  {
303
+ "seconds": 1.4721885819999443,
304
+ "bytes": 8185407,
305
+ "pages": 36,
306
+ "shards": 3
307
  },
308
  {
309
+ "seconds": 1.3622697179998795,
310
+ "bytes": 10825600,
311
+ "pages": 37,
312
  "shards": 3
313
  },
314
  {
315
+ "seconds": 1.2178643509996618,
316
+ "bytes": 7619131,
317
  "pages": 36,
318
+ "shards": 3
319
  },
320
  {
321
+ "seconds": 0.8663683109998601,
322
+ "bytes": 939323,
323
+ "pages": 28,
324
+ "shards": 1
325
  },
326
  {
327
+ "seconds": 0.8838954740003828,
328
+ "bytes": 5887982,
329
  "pages": 28,
330
  "shards": 2
331
  },
332
  {
333
+ "seconds": 1.2055649489993812,
334
+ "bytes": 6199136,
335
+ "pages": 30,
336
+ "shards": 3
337
  },
338
  {
339
+ "seconds": 1.1954915420010366,
340
+ "bytes": 12042262,
341
+ "pages": 38,
342
  "shards": 4
343
  },
344
  {
345
+ "seconds": 1.2374154409990297,
346
+ "bytes": 7885146,
347
+ "pages": 38,
348
+ "shards": 3
349
  },
350
  {
351
+ "seconds": 0.751855661999798,
352
+ "bytes": 918929,
353
+ "pages": 26,
354
+ "shards": 1
355
  },
356
  {
357
+ "seconds": 1.077764785999534,
358
+ "bytes": 11149283,
359
+ "pages": 40,
360
  "shards": 3
361
  },
362
  {
363
+ "seconds": 1.0994711280000047,
364
+ "bytes": 2636989,
365
+ "pages": 36,
366
  "shards": 2
367
  },
368
  {
369
+ "seconds": 0.9980228669992357,
370
+ "bytes": 5628518,
371
+ "pages": 28,
372
  "shards": 2
373
  },
374
  {
375
+ "seconds": 1.4070167600002605,
376
+ "bytes": 12110556,
377
+ "pages": 40,
378
+ "shards": 3
379
  },
380
  {
381
+ "seconds": 1.1031485080002312,
382
+ "bytes": 6510962,
383
+ "pages": 32,
384
  "shards": 3
385
  },
386
  {
387
+ "seconds": 1.2347332800000004,
388
+ "bytes": 6132473,
389
  "pages": 36,
390
  "shards": 2
391
  },
392
  {
393
+ "seconds": 0.8308523640007479,
394
+ "bytes": 2627309,
395
+ "pages": 37,
396
  "shards": 2
397
  },
398
  {
399
+ "seconds": 2.325367698998889,
400
+ "bytes": 6081267,
401
+ "pages": 28,
402
  "shards": 2
403
  },
404
  {
405
+ "seconds": 1.1514169719994243,
406
+ "bytes": 10850299,
407
+ "pages": 40,
408
  "shards": 3
409
  },
410
  {
411
+ "seconds": 1.1272267000003922,
412
+ "bytes": 8136270,
413
+ "pages": 29,
414
  "shards": 3
415
  },
416
  {
417
+ "seconds": 1.007810305000021,
418
+ "bytes": 6126457,
419
+ "pages": 37,
420
  "shards": 2
421
  },
422
  {
423
+ "seconds": 1.4154219219999504,
424
+ "bytes": 7407761,
425
+ "pages": 38,
426
+ "shards": 3
427
  },
428
  {
429
+ "seconds": 1.1865251550007088,
430
+ "bytes": 5346992,
431
+ "pages": 31,
432
  "shards": 2
433
  },
434
  {
435
+ "seconds": 1.0492023780007003,
436
+ "bytes": 6288876,
437
  "pages": 37,
438
+ "shards": 2
439
  },
440
  {
441
+ "seconds": 1.7196641009995801,
442
+ "bytes": 2291853,
443
+ "pages": 28,
444
  "shards": 2
445
  },
446
  {
447
+ "seconds": 1.0487870670003758,
448
+ "bytes": 11950097,
449
+ "pages": 41,
450
  "shards": 3
451
  },
452
  {
453
+ "seconds": 1.2208401020016026,
454
+ "bytes": 11777262,
455
+ "pages": 36,
456
  "shards": 4
457
  },
458
  {
459
+ "seconds": 0.9635194839993346,
460
+ "bytes": 1029009,
461
+ "pages": 33,
462
  "shards": 1
463
  },
464
  {
465
+ "seconds": 1.0835146409990557,
466
+ "bytes": 5946250,
467
+ "pages": 30,
468
  "shards": 2
469
  },
470
  {
471
+ "seconds": 0.9391229799985013,
472
+ "bytes": 2632765,
473
+ "pages": 34,
474
  "shards": 2
475
  },
476
  {
477
+ "seconds": 1.0898778940008924,
478
+ "bytes": 10697456,
479
+ "pages": 35,
480
  "shards": 3
481
  },
482
  {
483
+ "seconds": 1.3169437420001486,
484
+ "bytes": 11722554,
485
  "pages": 35,
486
  "shards": 3
487
  },
488
  {
489
+ "seconds": 1.2027081380001619,
490
+ "bytes": 2601682,
491
  "pages": 37,
492
  "shards": 2
493
  },
494
  {
495
+ "seconds": 1.0705139929996221,
496
+ "bytes": 6406862,
497
  "pages": 35,
498
  "shards": 2
499
  },
500
  {
501
+ "seconds": 3.478892785999051,
502
+ "bytes": 11471341,
503
+ "pages": 38,
504
  "shards": 4
505
  },
506
  {
507
+ "seconds": 0.9223496029990201,
508
+ "bytes": 897417,
509
+ "pages": 26,
510
  "shards": 1
511
  },
512
  {
513
+ "seconds": 1.0840730929994606,
514
+ "bytes": 6199927,
515
+ "pages": 34,
516
  "shards": 2
517
  },
518
  {
519
+ "seconds": 1.112097929999436,
520
+ "bytes": 8142857,
521
+ "pages": 30,
522
  "shards": 3
523
  },
524
  {
525
+ "seconds": 1.17854872099997,
526
+ "bytes": 10442242,
527
+ "pages": 43,
528
+ "shards": 4
529
  },
530
  {
531
+ "seconds": 1.163782216999607,
532
+ "bytes": 7671664,
533
+ "pages": 39,
534
  "shards": 3
535
  },
536
  {
537
+ "seconds": 1.1290235629985546,
538
+ "bytes": 1186173,
539
+ "pages": 34,
540
  "shards": 1
541
+ },
542
+ {
543
+ "seconds": 1.2037071409995406,
544
+ "bytes": 10277758,
545
+ "pages": 28,
546
+ "shards": 3
547
+ },
548
+ {
549
+ "seconds": 0.9222849990001123,
550
+ "bytes": 2428641,
551
+ "pages": 31,
552
+ "shards": 2
553
+ },
554
+ {
555
+ "seconds": 1.1589227389995358,
556
+ "bytes": 11976819,
557
+ "pages": 37,
558
+ "shards": 3
559
+ },
560
+ {
561
+ "seconds": 1.0021623419997923,
562
+ "bytes": 6273488,
563
+ "pages": 37,
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:f2026403cd8252eb8fbef1272eb9f1d5d4b596cf0488b73b1b78201830ca2e97
3
- size 721761919
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a3bc7b8df78537b2475ab2dd3ebc2aba12dbb1b97f2867ccda2914ec9a3004e2
3
+ size 721777983