appvoid commited on
Commit
14705e4
·
verified ·
1 Parent(s): 01a92a3

Qwen90M step 460000; 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": 455000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "4e0571b196c03b2c4a354365ed7cfb6e857d56cf683df316f0f243fc3c3c4a92",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "ed5c494fe01b44806d348a5f83f4d51c818b578b1ae84966900fdecfbb1e84ce",
12
  "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
13
  "performance.json": "58f8b6f79b191aa11b2bedb4594ca2897b6fdad2419ea7c9210688ac8022e7cb",
14
- "shard_scheduler.json": "140235f913903bf3fca7092448b8c6b8f621ac3103af86fcdc689e03b692a948",
15
- "coverage.json": "55cfee9475ccaed1654f1769d3d57fbdf4c9d032a2cd8b8c220c97dac7f99b83",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 460000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "e6df0b7486a13c4147338503b6e10d212d415f061b32918c019208df2361f792",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "f506a4eb49b59d5bfe154e01f4234e6f537368de7285d907d56b6e46d403a484",
12
  "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
13
  "performance.json": "58f8b6f79b191aa11b2bedb4594ca2897b6fdad2419ea7c9210688ac8022e7cb",
14
+ "shard_scheduler.json": "7aaac141beba6f945d5ac3dbede22069e5e37ddff308a0bbb94744199f0c5c6f",
15
+ "coverage.json": "74bb28e87fd46e4a0566d1ba75ad97ced9b1c5b7545ee80b386415b6215c402a",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 455000,
3
  "stats": {
4
- "climbmix/blocks": 16380000,
5
- "climbmix/targets": 27381108999,
6
- "climbmix/input_tokens": 27382843118,
7
- "ultra_qa/blocks": 6142500,
8
- "ultra_qa/targets": 10020265694,
9
- "ultra_qa/input_tokens": 10020284934,
10
- "cortex/blocks": 6142500,
11
- "cortex/targets": 233621534,
12
- "cortex/input_tokens": 9764332089,
13
- "ultra_style/blocks": 6142500,
14
- "ultra_style/targets": 9715613685,
15
- "ultra_style/input_tokens": 9716122899,
16
- "openmath/blocks": 4095000,
17
- "openmath/targets": 6327662730,
18
- "openmath/input_tokens": 6329771233,
19
- "rewrite6/blocks": 2047500,
20
- "rewrite6/targets": 3356578938,
21
- "rewrite6/input_tokens": 3360259205
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 2243,
26
- "row": 688,
27
  "start": {
28
- "file": 108,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 47,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
- "offset": 6144,
37
- "documents": 9188016
38
  },
39
  "rewrite6": {
40
- "page": 26,
41
- "row": 1443,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 106496
47
  }
48
  },
49
  "epoch": 25,
50
- "offset": 0,
51
- "documents": 5243039
52
  },
53
  "openmath": {
54
- "page": 1402,
55
- "row": 221,
56
  "start": {
57
  "file": 13,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 66,
61
- "row": 142
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 5742813
67
  },
68
  "ultra_qa": {
69
- "page": 606,
70
- "row": 864,
71
  "start": {
72
  "file": 4,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 6,
76
- "row": 16008
77
  }
78
  },
79
  "epoch": 0,
80
- "offset": 2048,
81
- "documents": 2483040
82
  },
83
  "ultra_style": {
84
- "page": 925,
85
- "row": 539,
86
  "start": {
87
  "file": 5,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 4,
91
- "row": 23497
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
- "documents": 3789339
97
  },
98
  "cortex": {
99
- "page": 119135,
100
- "row": 53,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 15249333
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 460000,
3
  "stats": {
4
+ "climbmix/blocks": 16560000,
5
+ "climbmix/targets": 27681620870,
6
+ "climbmix/input_tokens": 27683374095,
7
+ "ultra_qa/blocks": 6210000,
8
+ "ultra_qa/targets": 10130524498,
9
+ "ultra_qa/input_tokens": 10130543970,
10
+ "cortex/blocks": 6210000,
11
+ "cortex/targets": 236187228,
12
+ "cortex/input_tokens": 9871713742,
13
+ "ultra_style/blocks": 6210000,
14
+ "ultra_style/targets": 9822416214,
15
+ "ultra_style/input_tokens": 9822930973,
16
+ "openmath/blocks": 4140000,
17
+ "openmath/targets": 6397270462,
18
+ "openmath/input_tokens": 6399402063,
19
+ "rewrite6/blocks": 2070000,
20
+ "rewrite6/targets": 3393477095,
21
+ "rewrite6/input_tokens": 3397197651
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 2267,
26
+ "row": 3462,
27
  "start": {
28
+ "file": 109,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 60,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
+ "offset": 2048,
37
+ "documents": 9289094
38
  },
39
  "rewrite6": {
40
+ "page": 40,
41
+ "row": 1569,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 163840
47
  }
48
  },
49
  "epoch": 25,
50
+ "offset": 2048,
51
+ "documents": 5300509
52
  },
53
  "openmath": {
54
+ "page": 1417,
55
+ "row": 1939,
56
  "start": {
57
  "file": 13,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 127,
61
+ "row": 582
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 5805971
67
  },
68
  "ultra_qa": {
69
+ "page": 612,
70
+ "row": 3622,
71
  "start": {
72
  "file": 4,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 6,
76
+ "row": 40584
77
  }
78
  },
79
  "epoch": 0,
80
+ "offset": 4096,
81
+ "documents": 2510374
82
  },
83
  "ultra_style": {
84
+ "page": 935,
85
+ "row": 1084,
86
  "start": {
87
  "file": 5,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 4,
91
+ "row": 64457
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
+ "documents": 3830844
97
  },
98
  "cortex": {
99
+ "page": 120444,
100
+ "row": 91,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 15416923
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 455000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "4e0571b196c03b2c4a354365ed7cfb6e857d56cf683df316f0f243fc3c3c4a92", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "ed5c494fe01b44806d348a5f83f4d51c818b578b1ae84966900fdecfbb1e84ce", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "58f8b6f79b191aa11b2bedb4594ca2897b6fdad2419ea7c9210688ac8022e7cb", "shard_scheduler.json": "140235f913903bf3fca7092448b8c6b8f621ac3103af86fcdc689e03b692a948", "coverage.json": "55cfee9475ccaed1654f1769d3d57fbdf4c9d032a2cd8b8c220c97dac7f99b83", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000455000"}
 
1
+ {"step": 460000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "e6df0b7486a13c4147338503b6e10d212d415f061b32918c019208df2361f792", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "f506a4eb49b59d5bfe154e01f4234e6f537368de7285d907d56b6e46d403a484", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "58f8b6f79b191aa11b2bedb4594ca2897b6fdad2419ea7c9210688ac8022e7cb", "shard_scheduler.json": "7aaac141beba6f945d5ac3dbede22069e5e37ddff308a0bbb94744199f0c5c6f", "coverage.json": "74bb28e87fd46e4a0566d1ba75ad97ced9b1c5b7545ee80b386415b6215c402a", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000460000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:4e0571b196c03b2c4a354365ed7cfb6e857d56cf683df316f0f243fc3c3c4a92
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e6df0b7486a13c4147338503b6e10d212d415f061b32918c019208df2361f792
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": 76550.43485735178,
26
- "required_pages_per_second_upper_estimate": 0.005295936704911929
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": 9568.804357168972,
40
- "required_pages_per_second_upper_estimate": 0.004006258564989454
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": 19137.608714337945,
54
- "required_pages_per_second_upper_estimate": 0.003985859202515684
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": 28706.41307150692,
68
- "required_pages_per_second_upper_estimate": 0.0017209056314820437
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": 28706.41307150692,
82
- "required_pages_per_second_upper_estimate": 0.0022317973501427813
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": 28706.41307150692,
96
- "required_pages_per_second_upper_estimate": 0.36821504433635943
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": 0.2695831617967599,
162
- "producer_p95_seconds": 0.24042049499985296,
163
- "gpu_update_seconds": 0.9631297344997287,
164
- "producer_batches_per_second": 3.7094304901502158,
165
- "gpu_batches_per_second": 1.0382817227830916,
166
- "producer_headroom": 3.572662803123576,
167
  "sustainable": true,
168
- "buffer_seconds": 60.67717327348291,
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": 506639.92217322264,
174
- "upload_bytes_per_second": 5145170.27868784,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 108.71313332459935,
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": 0.9096585289998984,
184
- "bytes": 2702720,
185
- "pages": 37,
186
- "shards": 2
187
- },
188
- {
189
- "seconds": 1.212294140999802,
190
- "bytes": 5935208,
191
- "pages": 36,
192
- "shards": 2
193
- },
194
- {
195
- "seconds": 1.1118033149996336,
196
- "bytes": 6738611,
197
- "pages": 31,
198
- "shards": 2
199
- },
200
- {
201
- "seconds": 1.2066322789996775,
202
- "bytes": 7050449,
203
- "pages": 25,
204
- "shards": 3
205
- },
206
- {
207
- "seconds": 1.1100294799998665,
208
- "bytes": 10655210,
209
- "pages": 43,
210
- "shards": 3
211
  },
212
  {
213
- "seconds": 0.9483619619995807,
214
- "bytes": 1257250,
215
- "pages": 35,
216
  "shards": 1
217
  },
218
  {
219
- "seconds": 0.9075711420000516,
220
- "bytes": 7428003,
221
- "pages": 38,
222
- "shards": 3
223
  },
224
  {
225
- "seconds": 1.278834224000093,
226
- "bytes": 9834674,
227
- "pages": 30,
228
  "shards": 3
229
  },
230
  {
231
- "seconds": 1.1737758689996554,
232
- "bytes": 8190004,
233
  "pages": 33,
234
  "shards": 3
235
  },
236
  {
237
- "seconds": 1.150645027999417,
238
- "bytes": 6104360,
239
- "pages": 37,
240
  "shards": 2
241
  },
242
  {
243
- "seconds": 1.1603049650002504,
244
- "bytes": 1160533,
245
  "pages": 35,
246
- "shards": 1
247
  },
248
  {
249
- "seconds": 1.0780582049992518,
250
- "bytes": 12237678,
251
- "pages": 34,
252
- "shards": 4
253
  },
254
  {
255
- "seconds": 0.9231663379996462,
256
- "bytes": 6121395,
257
- "pages": 36,
258
  "shards": 2
259
  },
260
  {
261
- "seconds": 0.8000806340005511,
262
- "bytes": 2307263,
263
- "pages": 28,
264
  "shards": 2
265
  },
266
  {
267
- "seconds": 1.328747977000603,
268
- "bytes": 11901603,
269
- "pages": 40,
270
  "shards": 3
271
  },
272
  {
273
- "seconds": 1.0515727450001577,
274
- "bytes": 10443094,
275
- "pages": 35,
276
  "shards": 3
277
  },
278
  {
279
- "seconds": 0.9829580149998947,
280
- "bytes": 2655368,
281
  "pages": 37,
282
  "shards": 2
283
  },
284
  {
285
- "seconds": 1.023656208999455,
286
- "bytes": 5985547,
287
- "pages": 27,
288
  "shards": 2
289
  },
290
  {
291
- "seconds": 1.1010237709997455,
292
- "bytes": 2681477,
293
  "pages": 37,
294
  "shards": 2
295
  },
296
  {
297
- "seconds": 1.0707301010006631,
298
- "bytes": 10039688,
299
- "pages": 34,
300
  "shards": 3
301
  },
302
  {
303
- "seconds": 1.0178890370007139,
304
- "bytes": 11702648,
305
- "pages": 30,
306
  "shards": 3
307
  },
308
  {
309
- "seconds": 1.0858605960002023,
310
- "bytes": 2526198,
311
  "pages": 37,
312
  "shards": 2
313
  },
314
  {
315
- "seconds": 1.2877745800014964,
316
- "bytes": 5791906,
317
- "pages": 28,
318
- "shards": 2
319
- },
320
- {
321
- "seconds": 0.9569596109995473,
322
- "bytes": 5245257,
323
- "pages": 33,
324
- "shards": 2
325
  },
326
  {
327
- "seconds": 1.285435239000435,
328
- "bytes": 7473995,
329
- "pages": 31,
330
  "shards": 3
331
  },
332
  {
333
- "seconds": 1.144911826999305,
334
- "bytes": 6031558,
335
  "pages": 36,
336
  "shards": 2
337
  },
338
  {
339
- "seconds": 1.1537458799994056,
340
- "bytes": 6763666,
341
- "pages": 30,
342
  "shards": 2
343
  },
344
  {
345
- "seconds": 1.232765078999364,
346
- "bytes": 12254761,
 
 
 
 
 
 
347
  "pages": 37,
348
  "shards": 4
349
  },
350
  {
351
- "seconds": 1.0962526939983945,
352
- "bytes": 1063767,
353
- "pages": 31,
354
  "shards": 1
355
  },
356
  {
357
- "seconds": 1.2301429309991363,
358
- "bytes": 7644122,
359
- "pages": 35,
360
- "shards": 4
361
  },
362
  {
363
- "seconds": 1.0025934540008166,
364
- "bytes": 6106895,
365
- "pages": 36,
366
- "shards": 2
367
  },
368
  {
369
- "seconds": 1.0866103590014973,
370
- "bytes": 4773621,
371
- "pages": 30,
372
- "shards": 2
373
  },
374
  {
375
- "seconds": 1.0814739849993202,
376
- "bytes": 2089687,
377
- "pages": 19,
378
  "shards": 2
379
  },
380
  {
381
- "seconds": 1.1024076169996988,
382
- "bytes": 12461960,
383
- "pages": 47,
384
- "shards": 3
385
  },
386
  {
387
- "seconds": 1.0951313269997627,
388
- "bytes": 6163611,
389
  "pages": 36,
390
  "shards": 2
391
  },
392
  {
393
- "seconds": 1.1898623689994565,
394
- "bytes": 6904260,
395
- "pages": 25,
396
- "shards": 3
397
  },
398
  {
399
- "seconds": 1.1612172679997457,
400
- "bytes": 6026166,
401
- "pages": 34,
402
- "shards": 2
403
  },
404
  {
405
- "seconds": 1.3405283600004623,
406
- "bytes": 7840338,
 
 
 
 
 
 
407
  "pages": 38,
408
  "shards": 3
409
  },
410
  {
411
- "seconds": 0.9687580539994087,
412
- "bytes": 835437,
413
- "pages": 25,
414
- "shards": 1
415
  },
416
  {
417
- "seconds": 14.438800792999245,
418
- "bytes": 11778517,
419
- "pages": 32,
420
- "shards": 3
421
  },
422
  {
423
- "seconds": 1.142588306000107,
424
- "bytes": 12693549,
425
- "pages": 49,
426
- "shards": 4
427
  },
428
  {
429
- "seconds": 0.9721095380009501,
430
- "bytes": 837498,
431
- "pages": 26,
432
- "shards": 1
433
  },
434
  {
435
- "seconds": 0.9322610279996297,
436
- "bytes": 6251371,
437
- "pages": 36,
438
  "shards": 2
439
  },
440
  {
441
- "seconds": 1.1527972080002655,
442
- "bytes": 6634416,
443
- "pages": 39,
444
  "shards": 3
445
  },
446
  {
447
- "seconds": 1.1540112700004101,
448
- "bytes": 11807839,
449
- "pages": 33,
450
  "shards": 3
451
  },
452
  {
453
- "seconds": 0.8518780800004606,
454
- "bytes": 2360936,
455
- "pages": 30,
456
  "shards": 2
457
  },
458
  {
459
- "seconds": 1.2458887120010331,
460
- "bytes": 6052831,
461
- "pages": 35,
462
  "shards": 2
463
  },
464
  {
465
- "seconds": 1.0937011059995712,
466
- "bytes": 11191807,
467
- "pages": 40,
468
  "shards": 3
469
  },
470
  {
471
- "seconds": 1.0167957749999914,
472
- "bytes": 2251564,
473
  "pages": 28,
474
  "shards": 2
475
  },
476
  {
477
- "seconds": 0.9138518070012651,
478
- "bytes": 6438645,
479
- "pages": 35,
480
  "shards": 2
481
  },
482
  {
483
- "seconds": 1.2295570830010547,
484
- "bytes": 11846540,
485
- "pages": 32,
486
  "shards": 3
487
  },
488
  {
489
- "seconds": 0.9956274190008116,
490
- "bytes": 6486340,
491
- "pages": 41,
492
  "shards": 3
493
  },
494
  {
495
- "seconds": 1.1072240289995534,
496
- "bytes": 6495721,
497
- "pages": 37,
498
- "shards": 2
499
  },
500
  {
501
- "seconds": 0.8532074099985039,
502
- "bytes": 2480365,
503
- "pages": 29,
504
- "shards": 2
505
  },
506
  {
507
- "seconds": 0.959426125000391,
508
- "bytes": 5985197,
509
- "pages": 37,
 
 
 
 
 
 
510
  "shards": 2
511
  },
512
  {
513
- "seconds": 1.0452366619992972,
514
- "bytes": 10510253,
515
  "pages": 33,
516
  "shards": 3
517
  },
518
  {
519
- "seconds": 1.1982196040007693,
520
- "bytes": 8227634,
521
- "pages": 32,
522
- "shards": 3
523
  },
524
  {
525
- "seconds": 1.881202160999237,
526
- "bytes": 6307877,
527
- "pages": 37,
528
  "shards": 2
529
  },
530
  {
531
- "seconds": 1.1178816689989617,
532
- "bytes": 7521323,
 
 
 
 
 
 
533
  "pages": 39,
534
  "shards": 3
535
  },
536
  {
537
- "seconds": 1.323959568000646,
538
- "bytes": 5092035,
 
 
 
 
 
 
539
  "pages": 32,
540
- "shards": 3
541
  },
542
  {
543
- "seconds": 1.185996062000413,
544
- "bytes": 6230664,
 
 
 
 
 
 
545
  "pages": 36,
546
  "shards": 2
547
  },
548
  {
549
- "seconds": 0.9238252899995132,
550
- "bytes": 2424834,
551
- "pages": 29,
552
  "shards": 2
553
  },
554
  {
555
- "seconds": 1.1364503580007295,
556
- "bytes": 11698392,
557
  "pages": 34,
558
  "shards": 3
559
- },
560
- {
561
- "seconds": 1.0955314789989643,
562
- "bytes": 11552086,
563
- "pages": 37,
564
- "shards": 4
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": 77107.71522175097,
26
+ "required_pages_per_second_upper_estimate": 0.0053344906535896615
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": 9638.464402718871,
40
+ "required_pages_per_second_upper_estimate": 0.00403542377139403
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": 19276.928805437743,
54
+ "required_pages_per_second_upper_estimate": 0.004014875903373896
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": 28915.393208156613,
68
+ "required_pages_per_second_upper_estimate": 0.0017334336715800003
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": 28915.393208156613,
82
+ "required_pages_per_second_upper_estimate": 0.0022480446365608186
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": 28915.393208156613,
96
+ "required_pages_per_second_upper_estimate": 0.37089561714391317
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": 0.40328230902341033,
162
+ "producer_p95_seconds": 0.21793054399677203,
163
+ "gpu_update_seconds": 0.956168909790267,
164
+ "producer_batches_per_second": 2.479652535271391,
165
+ "gpu_batches_per_second": 1.0458403214755718,
166
+ "producer_headroom": 2.3709666613091174,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.238641316786826,
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": 290697.8459376837,
174
+ "upload_bytes_per_second": 5108075.817065287,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 62.376990638712726,
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": 1.142588306000107,
184
+ "bytes": 12693549,
185
+ "pages": 49,
186
+ "shards": 4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
187
  },
188
  {
189
+ "seconds": 0.9721095380009501,
190
+ "bytes": 837498,
191
+ "pages": 26,
192
  "shards": 1
193
  },
194
  {
195
+ "seconds": 0.9322610279996297,
196
+ "bytes": 6251371,
197
+ "pages": 36,
198
+ "shards": 2
199
  },
200
  {
201
+ "seconds": 1.1527972080002655,
202
+ "bytes": 6634416,
203
+ "pages": 39,
204
  "shards": 3
205
  },
206
  {
207
+ "seconds": 1.1540112700004101,
208
+ "bytes": 11807839,
209
  "pages": 33,
210
  "shards": 3
211
  },
212
  {
213
+ "seconds": 0.8518780800004606,
214
+ "bytes": 2360936,
215
+ "pages": 30,
216
  "shards": 2
217
  },
218
  {
219
+ "seconds": 1.2458887120010331,
220
+ "bytes": 6052831,
221
  "pages": 35,
222
+ "shards": 2
223
  },
224
  {
225
+ "seconds": 1.0937011059995712,
226
+ "bytes": 11191807,
227
+ "pages": 40,
228
+ "shards": 3
229
  },
230
  {
231
+ "seconds": 1.0167957749999914,
232
+ "bytes": 2251564,
233
+ "pages": 28,
234
  "shards": 2
235
  },
236
  {
237
+ "seconds": 0.9138518070012651,
238
+ "bytes": 6438645,
239
+ "pages": 35,
240
  "shards": 2
241
  },
242
  {
243
+ "seconds": 1.2295570830010547,
244
+ "bytes": 11846540,
245
+ "pages": 32,
246
  "shards": 3
247
  },
248
  {
249
+ "seconds": 0.9956274190008116,
250
+ "bytes": 6486340,
251
+ "pages": 41,
252
  "shards": 3
253
  },
254
  {
255
+ "seconds": 1.1072240289995534,
256
+ "bytes": 6495721,
257
  "pages": 37,
258
  "shards": 2
259
  },
260
  {
261
+ "seconds": 0.8532074099985039,
262
+ "bytes": 2480365,
263
+ "pages": 29,
264
  "shards": 2
265
  },
266
  {
267
+ "seconds": 0.959426125000391,
268
+ "bytes": 5985197,
269
  "pages": 37,
270
  "shards": 2
271
  },
272
  {
273
+ "seconds": 1.0452366619992972,
274
+ "bytes": 10510253,
275
+ "pages": 33,
276
  "shards": 3
277
  },
278
  {
279
+ "seconds": 1.1982196040007693,
280
+ "bytes": 8227634,
281
+ "pages": 32,
282
  "shards": 3
283
  },
284
  {
285
+ "seconds": 1.881202160999237,
286
+ "bytes": 6307877,
287
  "pages": 37,
288
  "shards": 2
289
  },
290
  {
291
+ "seconds": 1.1178816689989617,
292
+ "bytes": 7521323,
293
+ "pages": 39,
294
+ "shards": 3
 
 
 
 
 
 
295
  },
296
  {
297
+ "seconds": 1.323959568000646,
298
+ "bytes": 5092035,
299
+ "pages": 32,
300
  "shards": 3
301
  },
302
  {
303
+ "seconds": 1.185996062000413,
304
+ "bytes": 6230664,
305
  "pages": 36,
306
  "shards": 2
307
  },
308
  {
309
+ "seconds": 0.9238252899995132,
310
+ "bytes": 2424834,
311
+ "pages": 29,
312
  "shards": 2
313
  },
314
  {
315
+ "seconds": 1.1364503580007295,
316
+ "bytes": 11698392,
317
+ "pages": 34,
318
+ "shards": 3
319
+ },
320
+ {
321
+ "seconds": 1.0955314789989643,
322
+ "bytes": 11552086,
323
  "pages": 37,
324
  "shards": 4
325
  },
326
  {
327
+ "seconds": 0.7776427819990204,
328
+ "bytes": 236866,
329
+ "pages": 8,
330
  "shards": 1
331
  },
332
  {
333
+ "seconds": 0.9514035880001757,
334
+ "bytes": 6117975,
335
+ "pages": 37,
336
+ "shards": 2
337
  },
338
  {
339
+ "seconds": 0.8122407450009632,
340
+ "bytes": 1177705,
341
+ "pages": 35,
342
+ "shards": 1
343
  },
344
  {
345
+ "seconds": 0.9919227340014913,
346
+ "bytes": 7505854,
347
+ "pages": 31,
348
+ "shards": 3
349
  },
350
  {
351
+ "seconds": 1.4834884449992387,
352
+ "bytes": 5324170,
353
+ "pages": 33,
354
  "shards": 2
355
  },
356
  {
357
+ "seconds": 1.2237625110010413,
358
+ "bytes": 13342349,
359
+ "pages": 36,
360
+ "shards": 4
361
  },
362
  {
363
+ "seconds": 1.0075590680007736,
364
+ "bytes": 6216793,
365
  "pages": 36,
366
  "shards": 2
367
  },
368
  {
369
+ "seconds": 0.8775070329993468,
370
+ "bytes": 1128371,
371
+ "pages": 32,
372
+ "shards": 1
373
  },
374
  {
375
+ "seconds": 1.7969425399987813,
376
+ "bytes": 12329538,
377
+ "pages": 40,
378
+ "shards": 4
379
  },
380
  {
381
+ "seconds": 0.9383213960009016,
382
+ "bytes": 933737,
383
+ "pages": 26,
384
+ "shards": 1
385
+ },
386
+ {
387
+ "seconds": 3.437801178999507,
388
+ "bytes": 7738259,
389
  "pages": 38,
390
  "shards": 3
391
  },
392
  {
393
+ "seconds": 1.1072633740004676,
394
+ "bytes": 11735446,
395
+ "pages": 36,
396
+ "shards": 3
397
  },
398
  {
399
+ "seconds": 1.0690352890014765,
400
+ "bytes": 5156190,
401
+ "pages": 34,
402
+ "shards": 2
403
  },
404
  {
405
+ "seconds": 1.2706896670024435,
406
+ "bytes": 7639609,
407
+ "pages": 39,
408
+ "shards": 3
409
  },
410
  {
411
+ "seconds": 1.257235208999191,
412
+ "bytes": 5936765,
413
+ "pages": 28,
414
+ "shards": 2
415
  },
416
  {
417
+ "seconds": 1.1808424570008356,
418
+ "bytes": 2620310,
419
+ "pages": 37,
420
  "shards": 2
421
  },
422
  {
423
+ "seconds": 1.1430494310006907,
424
+ "bytes": 10900083,
425
+ "pages": 37,
426
  "shards": 3
427
  },
428
  {
429
+ "seconds": 1.591609420000168,
430
+ "bytes": 11822162,
431
+ "pages": 35,
432
  "shards": 3
433
  },
434
  {
435
+ "seconds": 0.9329451240009803,
436
+ "bytes": 2485686,
437
+ "pages": 29,
438
  "shards": 2
439
  },
440
  {
441
+ "seconds": 4.417064211000252,
442
+ "bytes": 6221614,
443
+ "pages": 37,
444
  "shards": 2
445
  },
446
  {
447
+ "seconds": 1.0521548180004174,
448
+ "bytes": 6267595,
449
+ "pages": 39,
450
  "shards": 3
451
  },
452
  {
453
+ "seconds": 0.9623302439977124,
454
+ "bytes": 6111937,
455
  "pages": 28,
456
  "shards": 2
457
  },
458
  {
459
+ "seconds": 1.1251186529989354,
460
+ "bytes": 6276648,
461
+ "pages": 36,
462
  "shards": 2
463
  },
464
  {
465
+ "seconds": 1.272974397001235,
466
+ "bytes": 8064738,
467
+ "pages": 28,
468
  "shards": 3
469
  },
470
  {
471
+ "seconds": 1.1465204630003427,
472
+ "bytes": 10636216,
473
+ "pages": 42,
474
  "shards": 3
475
  },
476
  {
477
+ "seconds": 0.9501123180016293,
478
+ "bytes": 1078327,
479
+ "pages": 34,
480
+ "shards": 1
481
  },
482
  {
483
+ "seconds": 3.8270818399978452,
484
+ "bytes": 7309721,
485
+ "pages": 35,
486
+ "shards": 4
487
  },
488
  {
489
+ "seconds": 1.1971988570003305,
490
+ "bytes": 10569382,
491
+ "pages": 35,
492
+ "shards": 3
493
+ },
494
+ {
495
+ "seconds": 0.8849704640015261,
496
+ "bytes": 2587633,
497
+ "pages": 35,
498
  "shards": 2
499
  },
500
  {
501
+ "seconds": 2.1376686900002824,
502
+ "bytes": 11886724,
503
  "pages": 33,
504
  "shards": 3
505
  },
506
  {
507
+ "seconds": 1.1407382979996328,
508
+ "bytes": 5980776,
509
+ "pages": 35,
510
+ "shards": 2
511
  },
512
  {
513
+ "seconds": 0.9373651969981438,
514
+ "bytes": 2246840,
515
+ "pages": 25,
516
  "shards": 2
517
  },
518
  {
519
+ "seconds": 2.0274646729994856,
520
+ "bytes": 10144494,
521
+ "pages": 44,
522
+ "shards": 3
523
+ },
524
+ {
525
+ "seconds": 2.5764055170002393,
526
+ "bytes": 7638759,
527
  "pages": 39,
528
  "shards": 3
529
  },
530
  {
531
+ "seconds": 0.8547688260005089,
532
+ "bytes": 858247,
533
+ "pages": 26,
534
+ "shards": 1
535
+ },
536
+ {
537
+ "seconds": 1.4054743929991673,
538
+ "bytes": 16596361,
539
  "pages": 32,
540
+ "shards": 4
541
  },
542
  {
543
+ "seconds": 3.107019715000206,
544
+ "bytes": 2583987,
545
+ "pages": 34,
546
+ "shards": 2
547
+ },
548
+ {
549
+ "seconds": 1.155031209000299,
550
+ "bytes": 6268449,
551
  "pages": 36,
552
  "shards": 2
553
  },
554
  {
555
+ "seconds": 0.8907613930023217,
556
+ "bytes": 6205142,
557
+ "pages": 37,
558
  "shards": 2
559
  },
560
  {
561
+ "seconds": 1.1056236590011395,
562
+ "bytes": 6648792,
563
  "pages": 34,
564
  "shards": 3
 
 
 
 
 
 
565
  }
566
  ],
567
  "cache_key": {
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:ed5c494fe01b44806d348a5f83f4d51c818b578b1ae84966900fdecfbb1e84ce
3
- size 721756031
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f506a4eb49b59d5bfe154e01f4234e6f537368de7285d907d56b6e46d403a484
3
+ size 721757439