appvoid commited on
Commit
5e7a908
·
verified ·
1 Parent(s): 222ad98

Qwen90M step 755000; complete resumable checkpoint

Browse files
Files changed (6) hide show
  1. complete.json +5 -5
  2. coverage.json +46 -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": 750000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "967a5dad40471a75696936f82686db770fe83a31472a4295ac8a0745743e376e",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "fa840d52fec48122d7ba48ef7ead58d0b3533e834c3dec925aa67b3f13281da5",
12
  "training_config.json": "75248af0a321738b405c3d51d1d55c7040dbc0785759d95f8dbdf6cd071a6687",
13
  "performance.json": "48e3aa7938b07d5e08810d60bc208cc2cb9133c4c39a24151758b79306675379",
14
- "shard_scheduler.json": "0b3d702d2793873e91a091708c94a6303900163fa993485826b15635a9c69670",
15
- "coverage.json": "456e31956d090f2507fe86424334753af439d32400f9d8e6814c0eb6de160a79",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "13510bfc04d8fed598fd0d09c1f778fbaebf19cefdd81f23f5ef3b909ef94a63"
 
1
  {
2
+ "step": 755000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "75cc8e7df2946f47615bba735e2412ba7d26acf35ff8d177274e2b278117ccf7",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "04389db27054d0deb18e21ea9848cc49ae3250a62872254d899b5d9fb8463946",
12
  "training_config.json": "75248af0a321738b405c3d51d1d55c7040dbc0785759d95f8dbdf6cd071a6687",
13
  "performance.json": "48e3aa7938b07d5e08810d60bc208cc2cb9133c4c39a24151758b79306675379",
14
+ "shard_scheduler.json": "6b4bb13b0a72c739a75dbb27075aa6cb85d8a6601c3ec79525308daa070d0808",
15
+ "coverage.json": "550b419d22a23b028c27e10458eb1b3ca5486992854436adfd61260ee070e83f",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "13510bfc04d8fed598fd0d09c1f778fbaebf19cefdd81f23f5ef3b909ef94a63"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 750000,
3
  "stats": {
4
- "climbmix/blocks": 27000000,
5
- "climbmix/targets": 45135657589,
6
- "climbmix/input_tokens": 45138517546,
7
- "ultra_qa/blocks": 10125000,
8
- "ultra_qa/targets": 16517745406,
9
- "ultra_qa/input_tokens": 16517777245,
10
- "cortex/blocks": 10125000,
11
- "cortex/targets": 384982808,
12
- "cortex/input_tokens": 16094233831,
13
- "ultra_style/blocks": 10125000,
14
- "ultra_style/targets": 16014758015,
15
- "ultra_style/input_tokens": 16015597822,
16
- "openmath/blocks": 6750000,
17
- "openmath/targets": 10429991701,
18
- "openmath/input_tokens": 10433466669,
19
- "rewrite6/blocks": 3375000,
20
- "rewrite6/targets": 5532858815,
21
- "rewrite6/input_tokens": 5538924694
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 3697,
26
- "row": 2703,
27
  "start": {
28
- "file": 178,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 79,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
- "offset": 16384,
37
- "documents": 15145615
38
  },
39
  "rewrite6": {
40
- "page": 3,
41
- "row": 2523,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 12288
47
  }
48
  },
49
  "epoch": 42,
50
  "offset": 0,
51
- "documents": 8641779
52
  },
53
  "openmath": {
54
- "page": 2310,
55
- "row": 3999,
56
  "start": {
57
  "file": 21,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 292,
61
- "row": 110
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 9465759
67
  },
68
  "ultra_qa": {
69
- "page": 999,
70
- "row": 1072,
71
  "start": {
72
  "file": 7,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 7,
76
- "row": 2368
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 4096,
81
- "documents": 4092976
82
  },
83
  "ultra_style": {
84
- "page": 1525,
85
- "row": 454,
86
  "start": {
87
  "file": 9,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
- "row_group": 0,
91
- "row": 81926
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
- "documents": 6246854
97
  },
98
  "cortex": {
99
- "page": 196363,
100
- "row": 45,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 25134509
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 755000,
3
  "stats": {
4
+ "climbmix/blocks": 27180000,
5
+ "climbmix/targets": 45436583708,
6
+ "climbmix/input_tokens": 45439462550,
7
+ "ultra_qa/blocks": 10192500,
8
+ "ultra_qa/targets": 16627829240,
9
+ "ultra_qa/input_tokens": 16627861293,
10
+ "cortex/blocks": 10192500,
11
+ "cortex/targets": 387543241,
12
+ "cortex/input_tokens": 16201531922,
13
+ "ultra_style/blocks": 10192500,
14
+ "ultra_style/targets": 16121676984,
15
+ "ultra_style/input_tokens": 16122522199,
16
+ "openmath/blocks": 6795000,
17
+ "openmath/targets": 10499419459,
18
+ "openmath/input_tokens": 10502917378,
19
+ "rewrite6/blocks": 3397500,
20
+ "rewrite6/targets": 5569717202,
21
+ "rewrite6/input_tokens": 5575823779
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 3722,
26
+ "row": 698,
27
  "start": {
28
+ "file": 180,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 14,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
+ "offset": 2048,
37
+ "documents": 15246010
38
  },
39
  "rewrite6": {
40
+ "page": 17,
41
+ "row": 3070,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 69632
47
  }
48
  },
49
  "epoch": 42,
50
  "offset": 0,
51
+ "documents": 8699670
52
  },
53
  "openmath": {
54
+ "page": 2326,
55
+ "row": 1294,
56
  "start": {
57
  "file": 21,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 357,
61
+ "row": 646
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 9528590
67
  },
68
  "ultra_qa": {
69
+ "page": 1005,
70
+ "row": 3733,
71
  "start": {
72
  "file": 7,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 7,
76
+ "row": 26944
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 4096,
81
+ "documents": 4120213
82
  },
83
  "ultra_style": {
84
+ "page": 1535,
85
+ "row": 679,
86
  "start": {
87
  "file": 9,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
+ "row_group": 1,
91
+ "row": 37522
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
+ "documents": 6288039
97
  },
98
  "cortex": {
99
+ "page": 197671,
100
+ "row": 117,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 25302005
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 750000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "967a5dad40471a75696936f82686db770fe83a31472a4295ac8a0745743e376e", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "fa840d52fec48122d7ba48ef7ead58d0b3533e834c3dec925aa67b3f13281da5", "training_config.json": "75248af0a321738b405c3d51d1d55c7040dbc0785759d95f8dbdf6cd071a6687", "performance.json": "48e3aa7938b07d5e08810d60bc208cc2cb9133c4c39a24151758b79306675379", "shard_scheduler.json": "0b3d702d2793873e91a091708c94a6303900163fa993485826b15635a9c69670", "coverage.json": "456e31956d090f2507fe86424334753af439d32400f9d8e6814c0eb6de160a79", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "13510bfc04d8fed598fd0d09c1f778fbaebf19cefdd81f23f5ef3b909ef94a63", "folder": "step-000000750000"}
 
1
+ {"step": 755000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "75cc8e7df2946f47615bba735e2412ba7d26acf35ff8d177274e2b278117ccf7", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "04389db27054d0deb18e21ea9848cc49ae3250a62872254d899b5d9fb8463946", "training_config.json": "75248af0a321738b405c3d51d1d55c7040dbc0785759d95f8dbdf6cd071a6687", "performance.json": "48e3aa7938b07d5e08810d60bc208cc2cb9133c4c39a24151758b79306675379", "shard_scheduler.json": "6b4bb13b0a72c739a75dbb27075aa6cb85d8a6601c3ec79525308daa070d0808", "coverage.json": "550b419d22a23b028c27e10458eb1b3ca5486992854436adfd61260ee070e83f", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "13510bfc04d8fed598fd0d09c1f778fbaebf19cefdd81f23f5ef3b909ef94a63", "folder": "step-000000755000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:967a5dad40471a75696936f82686db770fe83a31472a4295ac8a0745743e376e
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:75cc8e7df2946f47615bba735e2412ba7d26acf35ff8d177274e2b278117ccf7
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": 76589.24498588534,
26
- "required_pages_per_second_upper_estimate": 0.0052986216796557855
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": 9573.655623235667,
40
- "required_pages_per_second_upper_estimate": 0.004008289688785624
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": 19147.311246471334,
54
- "required_pages_per_second_upper_estimate": 0.00398787998408609
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": 28720.966869707,
68
- "required_pages_per_second_upper_estimate": 0.0017217781094617756
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": 28720.966869707,
82
- "required_pages_per_second_upper_estimate": 0.0022329288439374446
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": 28720.966869707,
96
- "required_pages_per_second_upper_estimate": 0.3684017248330191
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": 720750,
161
- "producer_batch_seconds": 0.10013397956231529,
162
- "producer_p95_seconds": 0.27769846199953463,
163
- "gpu_update_seconds": 0.962641687009545,
164
- "producer_batches_per_second": 9.986619970273736,
165
- "gpu_batches_per_second": 1.0388081188406757,
166
- "producer_headroom": 9.613536695707522,
167
  "sustainable": true,
168
- "buffer_seconds": 60.646426281601336,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 83,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 288485.1748269605,
174
- "upload_bytes_per_second": 5253353.013598868,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 61.90220292669879,
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": 1.3444492599992373,
184
- "bytes": 6214787,
185
- "pages": 36,
 
 
 
 
 
 
186
  "shards": 2
187
  },
188
  {
189
- "seconds": 1.2648084209977242,
190
- "bytes": 5986982,
 
 
 
 
 
 
191
  "pages": 31,
192
- "shards": 2
193
  },
194
  {
195
- "seconds": 1.4115151900005003,
196
- "bytes": 7731251,
197
  "pages": 39,
198
  "shards": 3
199
  },
200
  {
201
- "seconds": 1.1964002810018428,
202
- "bytes": 11692696,
203
- "pages": 29,
204
  "shards": 3
205
  },
206
  {
207
- "seconds": 1.0258839710004395,
208
- "bytes": 2700953,
209
- "pages": 38,
210
  "shards": 2
211
  },
212
  {
213
- "seconds": 1.2136837799989735,
214
- "bytes": 9835275,
215
- "pages": 34,
216
- "shards": 3
217
  },
218
  {
219
- "seconds": 1.2959440570011793,
220
- "bytes": 5942943,
221
- "pages": 35,
222
- "shards": 2
223
  },
224
  {
225
- "seconds": 0.9916904240017175,
226
- "bytes": 2484673,
227
- "pages": 30,
228
  "shards": 2
229
  },
230
  {
231
- "seconds": 1.0907142409996595,
232
- "bytes": 6115368,
233
- "pages": 37,
234
- "shards": 2
235
  },
236
  {
237
- "seconds": 1.3705160650024482,
238
- "bytes": 13033658,
239
- "pages": 37,
240
- "shards": 4
241
  },
242
  {
243
- "seconds": 1.2911444290002692,
244
- "bytes": 5949225,
 
 
 
 
 
 
245
  "pages": 28,
246
  "shards": 2
247
  },
248
  {
249
- "seconds": 1.0684557280001172,
250
- "bytes": 6155327,
251
- "pages": 37,
252
  "shards": 2
253
  },
254
  {
255
- "seconds": 1.2350120470000547,
256
- "bytes": 2680827,
257
- "pages": 36,
258
  "shards": 2
259
  },
260
  {
261
- "seconds": 1.5306445749993145,
262
- "bytes": 9835087,
263
  "pages": 33,
 
 
 
 
 
 
 
 
 
 
 
 
264
  "shards": 3
265
  },
266
  {
267
- "seconds": 0.9580864189993008,
268
- "bytes": 989214,
269
- "pages": 29,
270
  "shards": 1
271
  },
272
  {
273
- "seconds": 1.3733238169988908,
274
- "bytes": 13560061,
275
- "pages": 41,
276
- "shards": 4
277
  },
278
  {
279
- "seconds": 1.2763601959995867,
280
- "bytes": 5814992,
281
- "pages": 27,
 
 
 
 
 
 
282
  "shards": 2
283
  },
284
  {
285
- "seconds": 1.203305928996997,
286
- "bytes": 7121753,
287
  "pages": 36,
288
  "shards": 3
289
  },
290
  {
291
- "seconds": 1.1241114900003595,
292
- "bytes": 6180535,
293
- "pages": 36,
294
  "shards": 2
295
  },
296
  {
297
- "seconds": 1.1907369090004067,
298
- "bytes": 6141214,
299
- "pages": 35,
300
  "shards": 2
301
  },
302
  {
303
- "seconds": 0.9464336839992029,
304
- "bytes": 2380354,
305
- "pages": 28,
306
  "shards": 2
307
  },
308
  {
309
- "seconds": 1.6319502810001723,
310
- "bytes": 16693932,
311
- "pages": 40,
312
- "shards": 4
313
  },
314
  {
315
- "seconds": 0.91435248900234,
316
- "bytes": 2656824,
317
  "pages": 36,
318
  "shards": 2
319
  },
320
  {
321
- "seconds": 1.1157520460001251,
322
- "bytes": 5879867,
323
- "pages": 28,
324
- "shards": 2
325
- },
326
- {
327
- "seconds": 1.1945516140003747,
328
- "bytes": 6236357,
329
- "pages": 37,
330
- "shards": 2
331
  },
332
  {
333
- "seconds": 1.1350728899997193,
334
- "bytes": 701024,
335
- "pages": 21,
336
  "shards": 1
337
  },
338
  {
339
- "seconds": 1.6311224940000102,
340
- "bytes": 11525749,
341
- "pages": 46,
342
- "shards": 4
343
- },
344
- {
345
- "seconds": 1.0188665900022897,
346
- "bytes": 6807685,
347
- "pages": 33,
348
  "shards": 2
349
  },
350
  {
351
- "seconds": 1.1573900719995436,
352
- "bytes": 7406360,
353
- "pages": 30,
354
  "shards": 3
355
  },
356
  {
357
- "seconds": 1.3309355630008213,
358
- "bytes": 11158362,
359
- "pages": 38,
360
- "shards": 3
361
  },
362
  {
363
- "seconds": 0.9890253090015904,
364
- "bytes": 2755782,
365
  "pages": 37,
366
  "shards": 2
367
  },
368
  {
369
- "seconds": 1.3326452839974081,
370
- "bytes": 5962377,
 
 
 
 
 
 
371
  "pages": 28,
372
  "shards": 2
373
  },
374
  {
375
- "seconds": 1.016969195999991,
376
- "bytes": 1171347,
377
- "pages": 35,
378
- "shards": 1
379
  },
380
  {
381
- "seconds": 1.6574927980000211,
382
- "bytes": 17383397,
383
- "pages": 38,
384
- "shards": 5
385
  },
386
  {
387
- "seconds": 1.1213059549991158,
388
- "bytes": 6142168,
389
- "pages": 36,
390
  "shards": 2
391
  },
392
  {
393
- "seconds": 1.3281040260008012,
394
- "bytes": 876378,
395
- "pages": 27,
396
- "shards": 1
397
  },
398
  {
399
- "seconds": 1.2359630550017755,
400
- "bytes": 7977289,
401
- "pages": 37,
402
  "shards": 3
403
  },
404
  {
405
- "seconds": 1.4249079500004882,
406
- "bytes": 10392834,
407
- "pages": 36,
408
  "shards": 3
409
  },
410
  {
411
- "seconds": 1.2200082600029418,
412
- "bytes": 2504462,
413
  "pages": 32,
414
  "shards": 2
415
  },
416
  {
417
- "seconds": 1.3331129750004038,
418
- "bytes": 11935749,
419
  "pages": 35,
420
- "shards": 3
421
  },
422
  {
423
- "seconds": 1.076318840001477,
424
- "bytes": 1024576,
425
- "pages": 34,
426
- "shards": 1
427
  },
428
  {
429
- "seconds": 1.0973008010005287,
430
- "bytes": 11080849,
 
 
 
 
 
 
431
  "pages": 33,
432
  "shards": 4
433
  },
434
  {
435
- "seconds": 1.1580557140005112,
436
- "bytes": 6155625,
437
- "pages": 38,
438
  "shards": 2
439
  },
440
  {
441
- "seconds": 0.8461053749997518,
442
- "bytes": 1155308,
443
- "pages": 32,
444
- "shards": 1
445
  },
446
  {
447
- "seconds": 1.2314064579986734,
448
- "bytes": 7429916,
449
- "pages": 31,
450
  "shards": 3
451
  },
452
  {
453
- "seconds": 1.4280758529966988,
454
- "bytes": 11440670,
455
- "pages": 39,
 
 
 
 
 
 
456
  "shards": 3
457
  },
458
  {
459
- "seconds": 1.364633949000563,
460
- "bytes": 7290084,
461
- "pages": 30,
462
  "shards": 3
463
  },
464
  {
465
- "seconds": 1.0805251250021684,
466
- "bytes": 6054722,
467
- "pages": 36,
468
  "shards": 2
469
  },
470
  {
471
- "seconds": 2.882830210000975,
472
- "bytes": 1047147,
473
- "pages": 31,
474
  "shards": 1
475
  },
476
  {
477
- "seconds": 1.6690692930023943,
478
- "bytes": 11069831,
479
- "pages": 40,
480
  "shards": 4
481
  },
482
  {
483
- "seconds": 2.282538455998292,
484
- "bytes": 6276829,
485
- "pages": 36,
486
  "shards": 2
487
  },
488
  {
489
- "seconds": 1.4166184950008756,
490
- "bytes": 8285024,
491
- "pages": 35,
492
- "shards": 3
493
  },
494
  {
495
- "seconds": 1.271756786001788,
496
- "bytes": 10114603,
497
- "pages": 34,
498
  "shards": 3
499
  },
500
  {
501
- "seconds": 1.2068859079990943,
502
- "bytes": 6031296,
503
  "pages": 36,
504
  "shards": 3
505
  },
506
  {
507
- "seconds": 1.0317238650022773,
508
- "bytes": 2334152,
509
- "pages": 28,
510
- "shards": 2
511
- },
512
- {
513
- "seconds": 1.198212211002101,
514
- "bytes": 5956669,
515
- "pages": 36,
516
- "shards": 2
517
- },
518
- {
519
- "seconds": 1.182239486999606,
520
- "bytes": 5356316,
521
- "pages": 28,
522
  "shards": 2
523
  },
524
  {
525
- "seconds": 1.588657971999055,
526
- "bytes": 13335657,
527
  "pages": 33,
528
- "shards": 4
529
- },
530
- {
531
- "seconds": 0.8740047409992258,
532
- "bytes": 947216,
533
- "pages": 27,
534
- "shards": 1
535
- },
536
- {
537
- "seconds": 1.4317332190003071,
538
- "bytes": 7796978,
539
- "pages": 38,
540
- "shards": 3
541
- },
542
- {
543
- "seconds": 1.0188206910024746,
544
- "bytes": 812811,
545
- "pages": 23,
546
- "shards": 1
547
- },
548
- {
549
- "seconds": 1.1825939109985484,
550
- "bytes": 10557366,
551
- "pages": 46,
552
- "shards": 3
553
- },
554
- {
555
- "seconds": 1.6627466680001817,
556
- "bytes": 7279837,
557
- "pages": 30,
558
- "shards": 3
559
  },
560
  {
561
- "seconds": 1.578546805998485,
562
- "bytes": 6985484,
563
- "pages": 38,
564
  "shards": 2
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": 76927.36162945612,
26
+ "required_pages_per_second_upper_estimate": 0.005322013373596714
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": 9615.920203682015,
40
+ "required_pages_per_second_upper_estimate": 0.0040259849860337085
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": 19231.84040736403,
54
+ "required_pages_per_second_upper_estimate": 0.00400548517911616
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": 28847.760611046047,
68
+ "required_pages_per_second_upper_estimate": 0.0017293792006522182
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": 28847.760611046047,
82
+ "required_pages_per_second_upper_estimate": 0.002242786499619828
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": 28847.760611046047,
96
+ "required_pages_per_second_upper_estimate": 0.3700280988064038
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": 0.11758859633584962,
162
+ "producer_p95_seconds": 0.2860190520004835,
163
+ "gpu_update_seconds": 0.9584106153949902,
164
+ "producer_batches_per_second": 8.504226014773227,
165
+ "gpu_batches_per_second": 1.0433941193231353,
166
+ "producer_headroom": 8.150540488276892,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.37986876988438,
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": 451012.4654282901,
174
+ "upload_bytes_per_second": 5097647.062415216,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 96.77677604805496,
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": 1.0973008010005287,
184
+ "bytes": 11080849,
185
+ "pages": 33,
186
+ "shards": 4
187
+ },
188
+ {
189
+ "seconds": 1.1580557140005112,
190
+ "bytes": 6155625,
191
+ "pages": 38,
192
  "shards": 2
193
  },
194
  {
195
+ "seconds": 0.8461053749997518,
196
+ "bytes": 1155308,
197
+ "pages": 32,
198
+ "shards": 1
199
+ },
200
+ {
201
+ "seconds": 1.2314064579986734,
202
+ "bytes": 7429916,
203
  "pages": 31,
204
+ "shards": 3
205
  },
206
  {
207
+ "seconds": 1.4280758529966988,
208
+ "bytes": 11440670,
209
  "pages": 39,
210
  "shards": 3
211
  },
212
  {
213
+ "seconds": 1.364633949000563,
214
+ "bytes": 7290084,
215
+ "pages": 30,
216
  "shards": 3
217
  },
218
  {
219
+ "seconds": 1.0805251250021684,
220
+ "bytes": 6054722,
221
+ "pages": 36,
222
  "shards": 2
223
  },
224
  {
225
+ "seconds": 2.882830210000975,
226
+ "bytes": 1047147,
227
+ "pages": 31,
228
+ "shards": 1
229
  },
230
  {
231
+ "seconds": 1.6690692930023943,
232
+ "bytes": 11069831,
233
+ "pages": 40,
234
+ "shards": 4
235
  },
236
  {
237
+ "seconds": 2.282538455998292,
238
+ "bytes": 6276829,
239
+ "pages": 36,
240
  "shards": 2
241
  },
242
  {
243
+ "seconds": 1.4166184950008756,
244
+ "bytes": 8285024,
245
+ "pages": 35,
246
+ "shards": 3
247
  },
248
  {
249
+ "seconds": 1.271756786001788,
250
+ "bytes": 10114603,
251
+ "pages": 34,
252
+ "shards": 3
253
  },
254
  {
255
+ "seconds": 1.2068859079990943,
256
+ "bytes": 6031296,
257
+ "pages": 36,
258
+ "shards": 3
259
+ },
260
+ {
261
+ "seconds": 1.0317238650022773,
262
+ "bytes": 2334152,
263
  "pages": 28,
264
  "shards": 2
265
  },
266
  {
267
+ "seconds": 1.198212211002101,
268
+ "bytes": 5956669,
269
+ "pages": 36,
270
  "shards": 2
271
  },
272
  {
273
+ "seconds": 1.182239486999606,
274
+ "bytes": 5356316,
275
+ "pages": 28,
276
  "shards": 2
277
  },
278
  {
279
+ "seconds": 1.588657971999055,
280
+ "bytes": 13335657,
281
  "pages": 33,
282
+ "shards": 4
283
+ },
284
+ {
285
+ "seconds": 0.8740047409992258,
286
+ "bytes": 947216,
287
+ "pages": 27,
288
+ "shards": 1
289
+ },
290
+ {
291
+ "seconds": 1.4317332190003071,
292
+ "bytes": 7796978,
293
+ "pages": 38,
294
  "shards": 3
295
  },
296
  {
297
+ "seconds": 1.0188206910024746,
298
+ "bytes": 812811,
299
+ "pages": 23,
300
  "shards": 1
301
  },
302
  {
303
+ "seconds": 1.1825939109985484,
304
+ "bytes": 10557366,
305
+ "pages": 46,
306
+ "shards": 3
307
  },
308
  {
309
+ "seconds": 1.6627466680001817,
310
+ "bytes": 7279837,
311
+ "pages": 30,
312
+ "shards": 3
313
+ },
314
+ {
315
+ "seconds": 1.578546805998485,
316
+ "bytes": 6985484,
317
+ "pages": 38,
318
  "shards": 2
319
  },
320
  {
321
+ "seconds": 1.4114699799974915,
322
+ "bytes": 10209990,
323
  "pages": 36,
324
  "shards": 3
325
  },
326
  {
327
+ "seconds": 1.0779238770010124,
328
+ "bytes": 1827152,
329
+ "pages": 11,
330
  "shards": 2
331
  },
332
  {
333
+ "seconds": 0.9918526799992833,
334
+ "bytes": 6095255,
335
+ "pages": 37,
336
  "shards": 2
337
  },
338
  {
339
+ "seconds": 1.0435619140007475,
340
+ "bytes": 6434451,
341
+ "pages": 37,
342
  "shards": 2
343
  },
344
  {
345
+ "seconds": 1.1101431559982302,
346
+ "bytes": 7100100,
347
+ "pages": 31,
348
+ "shards": 3
349
  },
350
  {
351
+ "seconds": 1.0310441840010753,
352
+ "bytes": 5933626,
353
  "pages": 36,
354
  "shards": 2
355
  },
356
  {
357
+ "seconds": 1.339776704000542,
358
+ "bytes": 13443293,
359
+ "pages": 35,
360
+ "shards": 4
 
 
 
 
 
 
361
  },
362
  {
363
+ "seconds": 1.7319635569983802,
364
+ "bytes": 934829,
365
+ "pages": 27,
366
  "shards": 1
367
  },
368
  {
369
+ "seconds": 1.0800791440015018,
370
+ "bytes": 6107947,
371
+ "pages": 36,
 
 
 
 
 
 
372
  "shards": 2
373
  },
374
  {
375
+ "seconds": 1.2829358429989952,
376
+ "bytes": 6479810,
377
+ "pages": 33,
378
  "shards": 3
379
  },
380
  {
381
+ "seconds": 1.002878891002183,
382
+ "bytes": 6148473,
383
+ "pages": 37,
384
+ "shards": 2
385
  },
386
  {
387
+ "seconds": 1.581660068997735,
388
+ "bytes": 6199462,
389
  "pages": 37,
390
  "shards": 2
391
  },
392
  {
393
+ "seconds": 1.3033671679986583,
394
+ "bytes": 13020043,
395
+ "pages": 37,
396
+ "shards": 4
397
+ },
398
+ {
399
+ "seconds": 1.1211934040002234,
400
+ "bytes": 5810769,
401
  "pages": 28,
402
  "shards": 2
403
  },
404
  {
405
+ "seconds": 1.2197048900015943,
406
+ "bytes": 2772365,
407
+ "pages": 38,
408
+ "shards": 2
409
  },
410
  {
411
+ "seconds": 1.9406323449984484,
412
+ "bytes": 5788082,
413
+ "pages": 28,
414
+ "shards": 2
415
  },
416
  {
417
+ "seconds": 1.5139052369995625,
418
+ "bytes": 6128565,
419
+ "pages": 37,
420
  "shards": 2
421
  },
422
  {
423
+ "seconds": 0.9614488920014992,
424
+ "bytes": 6537093,
425
+ "pages": 33,
426
+ "shards": 3
427
  },
428
  {
429
+ "seconds": 1.2357967999996617,
430
+ "bytes": 11997860,
431
+ "pages": 35,
432
  "shards": 3
433
  },
434
  {
435
+ "seconds": 1.7381740189994161,
436
+ "bytes": 7938212,
437
+ "pages": 40,
438
  "shards": 3
439
  },
440
  {
441
+ "seconds": 1.2239695549978933,
442
+ "bytes": 5214335,
443
  "pages": 32,
444
  "shards": 2
445
  },
446
  {
447
+ "seconds": 1.2570640209996782,
448
+ "bytes": 6323177,
449
  "pages": 35,
450
+ "shards": 2
451
  },
452
  {
453
+ "seconds": 1.1513970990017697,
454
+ "bytes": 2421895,
455
+ "pages": 31,
456
+ "shards": 2
457
  },
458
  {
459
+ "seconds": 0.9733352420007577,
460
+ "bytes": 6126927,
461
+ "pages": 37,
462
+ "shards": 2
463
+ },
464
+ {
465
+ "seconds": 1.3508422129998507,
466
+ "bytes": 16469825,
467
  "pages": 33,
468
  "shards": 4
469
  },
470
  {
471
+ "seconds": 1.220427556996583,
472
+ "bytes": 2558602,
473
+ "pages": 34,
474
  "shards": 2
475
  },
476
  {
477
+ "seconds": 1.1153945629994269,
478
+ "bytes": 6138686,
479
+ "pages": 33,
480
+ "shards": 2
481
  },
482
  {
483
+ "seconds": 1.4000027559959562,
484
+ "bytes": 7623395,
485
+ "pages": 39,
486
  "shards": 3
487
  },
488
  {
489
+ "seconds": 1.0437422140021226,
490
+ "bytes": 887331,
491
+ "pages": 27,
492
+ "shards": 1
493
+ },
494
+ {
495
+ "seconds": 1.3929757290025009,
496
+ "bytes": 11170736,
497
+ "pages": 35,
498
  "shards": 3
499
  },
500
  {
501
+ "seconds": 1.5524120170011884,
502
+ "bytes": 8373554,
503
+ "pages": 38,
504
  "shards": 3
505
  },
506
  {
507
+ "seconds": 1.806914691995189,
508
+ "bytes": 5833325,
509
+ "pages": 30,
510
  "shards": 2
511
  },
512
  {
513
+ "seconds": 1.1171102930020425,
514
+ "bytes": 1254155,
515
+ "pages": 34,
516
  "shards": 1
517
  },
518
  {
519
+ "seconds": 1.3684268549986882,
520
+ "bytes": 11656046,
521
+ "pages": 37,
522
  "shards": 4
523
  },
524
  {
525
+ "seconds": 1.130367224002839,
526
+ "bytes": 6468621,
527
+ "pages": 37,
528
  "shards": 2
529
  },
530
  {
531
+ "seconds": 1.1125211710022995,
532
+ "bytes": 2380554,
533
+ "pages": 28,
534
+ "shards": 2
535
  },
536
  {
537
+ "seconds": 1.3700068609978189,
538
+ "bytes": 12013278,
539
+ "pages": 37,
540
  "shards": 3
541
  },
542
  {
543
+ "seconds": 1.232138457999099,
544
+ "bytes": 9991874,
545
  "pages": 36,
546
  "shards": 3
547
  },
548
  {
549
+ "seconds": 1.27796818799834,
550
+ "bytes": 2573475,
551
+ "pages": 32,
 
 
 
 
 
 
 
 
 
 
 
 
552
  "shards": 2
553
  },
554
  {
555
+ "seconds": 1.1166499749961076,
556
+ "bytes": 6085922,
557
  "pages": 33,
558
+ "shards": 2
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
559
  },
560
  {
561
+ "seconds": 0.8856082540005445,
562
+ "bytes": 2405776,
563
+ "pages": 29,
564
  "shards": 2
565
  }
566
  ],
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:fa840d52fec48122d7ba48ef7ead58d0b3533e834c3dec925aa67b3f13281da5
3
- size 721765951
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:04389db27054d0deb18e21ea9848cc49ae3250a62872254d899b5d9fb8463946
3
+ size 721760703