appvoid commited on
Commit
44f0d53
·
verified ·
1 Parent(s): 1741324

Qwen90M step 90000; complete resumable checkpoint

Browse files
Files changed (7) hide show
  1. complete.json +6 -6
  2. coverage.json +46 -46
  3. latest.json +1 -1
  4. model.safetensors +1 -1
  5. performance.json +4 -4
  6. shard_scheduler.json +29 -185
  7. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 85000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "2683c5255a46fb5734cdf83842090f23a32fa503fc084bf57f67a828bfeab926",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "b22b50c2a3f3cc11176d153ff6d2d48d52ac57f92ea3cf82eef967444d82426f",
12
  "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
13
- "performance.json": "b9785e5ac65e7e9df73669f94ffdf686e237067bf54649bd2ebf893d5327bdc9",
14
- "shard_scheduler.json": "3bcf783fa48a15540228bb1e4fd020f0a0cbc02a12b62595b447f4d19b8722cb",
15
- "coverage.json": "05c62b88d1a99622c29ec20f235ec8f3ba77c21ab09cd74eeddfb98f8759e10a",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 90000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "d1b9321795a955b99feefcdb828552461d21c50d7c0ed64202118da5f4a4eb80",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "a7a66ded6461cef76b89f27ed3606c872345dc1f83d547f29c24c93e5c270c58",
12
  "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
13
+ "performance.json": "7e827680f2cb27108bc66b58bb4d6ede7b8c07fab9033e664cac4bed5acb1a98",
14
+ "shard_scheduler.json": "1a342da44c6786263afa65805d94c6d3098e1c2f4ddb32204c37ab8faa15f028",
15
+ "coverage.json": "6ba08ee821b99cb552c6773108af7236f9e3e692b1eecc0b7f1943b6dafcefc2",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 85000,
3
  "stats": {
4
- "climbmix/blocks": 3060000,
5
- "climbmix/targets": 5113419097,
6
- "climbmix/input_tokens": 5113742843,
7
- "ultra_qa/blocks": 1147500,
8
- "ultra_qa/targets": 1871965816,
9
- "ultra_qa/input_tokens": 1871969468,
10
- "cortex/blocks": 1147500,
11
- "cortex/targets": 43652665,
12
- "cortex/input_tokens": 1823889389,
13
- "ultra_style/blocks": 1147500,
14
- "ultra_style/targets": 1814807863,
15
- "ultra_style/input_tokens": 1814902860,
16
- "openmath/blocks": 765000,
17
- "openmath/targets": 1182206521,
18
- "openmath/input_tokens": 1182600938,
19
- "rewrite6/blocks": 382500,
20
- "rewrite6/targets": 627023097,
21
- "rewrite6/input_tokens": 627710727
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 419,
26
- "row": 587,
27
  "start": {
28
- "file": 20,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 21,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
- "documents": 1716811
38
  },
39
  "rewrite6": {
40
- "page": 38,
41
- "row": 2375,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 155648
47
  }
48
  },
49
- "epoch": 4,
50
  "offset": 2048,
51
- "documents": 979639
52
  },
53
  "openmath": {
54
- "page": 262,
55
- "row": 252,
56
  "start": {
57
  "file": 2,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 199,
61
- "row": 852
62
  }
63
  },
64
  "epoch": 0,
65
- "offset": 2048,
66
- "documents": 1073404
67
  },
68
  "ultra_qa": {
69
- "page": 113,
70
- "row": 1205,
71
  "start": {
72
  "file": 0,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 7,
76
- "row": 10817
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
- "documents": 464053
82
  },
83
  "ultra_style": {
84
- "page": 172,
85
- "row": 3360,
86
  "start": {
87
  "file": 1,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 0,
91
- "row": 19573
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
- "documents": 707872
97
  },
98
  "cortex": {
99
- "page": 22250,
100
- "row": 49,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 2848049
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 90000,
3
  "stats": {
4
+ "climbmix/blocks": 3240000,
5
+ "climbmix/targets": 5413943761,
6
+ "climbmix/input_tokens": 5414286992,
7
+ "ultra_qa/blocks": 1215000,
8
+ "ultra_qa/targets": 1982074966,
9
+ "ultra_qa/input_tokens": 1982078809,
10
+ "cortex/blocks": 1215000,
11
+ "cortex/targets": 46211808,
12
+ "cortex/input_tokens": 1931141281,
13
+ "ultra_style/blocks": 1215000,
14
+ "ultra_style/targets": 1921585691,
15
+ "ultra_style/input_tokens": 1921686239,
16
+ "openmath/blocks": 810000,
17
+ "openmath/targets": 1251693248,
18
+ "openmath/input_tokens": 1252110755,
19
+ "rewrite6/blocks": 405000,
20
+ "rewrite6/targets": 663921893,
21
+ "rewrite6/input_tokens": 664649732
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 444,
26
+ "row": 130,
27
  "start": {
28
+ "file": 21,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 37,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
+ "documents": 1818754
38
  },
39
  "rewrite6": {
40
+ "page": 2,
41
+ "row": 1720,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 8192
47
  }
48
  },
49
+ "epoch": 5,
50
  "offset": 2048,
51
+ "documents": 1036932
52
  },
53
  "openmath": {
54
+ "page": 277,
55
+ "row": 1836,
56
  "start": {
57
  "file": 2,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 261,
61
+ "row": 292
62
  }
63
  },
64
  "epoch": 0,
65
+ "offset": 0,
66
+ "documents": 1136428
67
  },
68
  "ultra_qa": {
69
+ "page": 119,
70
+ "row": 3845,
71
  "start": {
72
  "file": 0,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 7,
76
+ "row": 35393
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
+ "documents": 491269
82
  },
83
  "ultra_style": {
84
+ "page": 182,
85
+ "row": 4014,
86
  "start": {
87
  "file": 1,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 0,
91
+ "row": 60533
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
+ "documents": 749486
97
  },
98
  "cortex": {
99
+ "page": 23557,
100
+ "row": 72,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 3015368
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 85000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "2683c5255a46fb5734cdf83842090f23a32fa503fc084bf57f67a828bfeab926", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "b22b50c2a3f3cc11176d153ff6d2d48d52ac57f92ea3cf82eef967444d82426f", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "b9785e5ac65e7e9df73669f94ffdf686e237067bf54649bd2ebf893d5327bdc9", "shard_scheduler.json": "3bcf783fa48a15540228bb1e4fd020f0a0cbc02a12b62595b447f4d19b8722cb", "coverage.json": "05c62b88d1a99622c29ec20f235ec8f3ba77c21ab09cd74eeddfb98f8759e10a", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000085000"}
 
1
+ {"step": 90000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "d1b9321795a955b99feefcdb828552461d21c50d7c0ed64202118da5f4a4eb80", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "a7a66ded6461cef76b89f27ed3606c872345dc1f83d547f29c24c93e5c270c58", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "7e827680f2cb27108bc66b58bb4d6ede7b8c07fab9033e664cac4bed5acb1a98", "shard_scheduler.json": "1a342da44c6786263afa65805d94c6d3098e1c2f4ddb32204c37ab8faa15f028", "coverage.json": "6ba08ee821b99cb552c6773108af7236f9e3e692b1eecc0b7f1943b6dafcefc2", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000090000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:2683c5255a46fb5734cdf83842090f23a32fa503fc084bf57f67a828bfeab926
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d1b9321795a955b99feefcdb828552461d21c50d7c0ed64202118da5f4a4eb80
3
  size 360622744
performance.json CHANGED
@@ -49,11 +49,11 @@
49
  "selected": {
50
  "microbatch": 86,
51
  "mode": "compiled",
52
- "microbatch_seconds": 0.8982481310000026,
53
- "peak_gb": 85.39913082122803,
54
  "fits_budget": true,
55
- "estimated_update_seconds": 0.9539577840000106,
56
- "input_tokens_per_second": 196079.45056776243
57
  },
58
  "measurements": [
59
  {
 
49
  "selected": {
50
  "microbatch": 86,
51
  "mode": "compiled",
52
+ "microbatch_seconds": 0.899068480999631,
53
+ "peak_gb": 85.39889430999756,
54
  "fits_budget": true,
55
+ "estimated_update_seconds": 0.9546496269995259,
56
+ "input_tokens_per_second": 195900.5389713715
57
  },
58
  "measurements": [
59
  {
shard_scheduler.json CHANGED
@@ -22,8 +22,8 @@
22
  "tokens_per_second": 1062967.6168812835,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
- "required_token_positions_per_second": 75341.36127254891,
26
- "required_pages_per_second_upper_estimate": 0.006530398244170153
27
  },
28
  {
29
  "source": "rewrite6",
@@ -36,8 +36,8 @@
36
  "tokens_per_second": 1206672.263377472,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
- "required_token_positions_per_second": 9417.670159068613,
40
- "required_pages_per_second_upper_estimate": 0.003756849433169225
41
  },
42
  {
43
  "source": "openmath",
@@ -50,8 +50,8 @@
50
  "tokens_per_second": 410042.7581134458,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
- "required_token_positions_per_second": 18835.340318137227,
54
- "required_pages_per_second_upper_estimate": 0.0043572244919332604
55
  },
56
  {
57
  "source": "ultra_qa",
@@ -64,8 +64,8 @@
64
  "tokens_per_second": 2163486.5624068812,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
- "required_token_positions_per_second": 28253.010477205844,
68
- "required_pages_per_second_upper_estimate": 0.0017204601059718416
69
  },
70
  {
71
  "source": "ultra_style",
@@ -78,8 +78,8 @@
78
  "tokens_per_second": 1546694.2737633225,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
- "required_token_positions_per_second": 28253.010477205844,
82
- "required_pages_per_second_upper_estimate": 0.0021749949558739273
83
  },
84
  {
85
  "source": "cortex",
@@ -92,8 +92,8 @@
92
  "tokens_per_second": 1015177.8395898147,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
  "estimated_pages_for_buffer": 22,
95
- "required_token_positions_per_second": 28253.010477205844,
96
- "required_pages_per_second_upper_estimate": 0.3508078333834864
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,188 +158,32 @@
158
  "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 688946,
161
- "producer_batch_seconds": 0.04785197319529999,
162
- "producer_p95_seconds": 0.1573215110001911,
163
- "gpu_update_seconds": 0.9785859819188487,
164
- "producer_batches_per_second": 20.89777982443198,
165
- "gpu_batches_per_second": 1.0218826127461604,
166
- "producer_headroom": 20.450274389415675,
167
  "sustainable": true,
168
- "buffer_seconds": 60.67233087896862,
169
  "buffer_drain_seconds": null,
170
- "shard_target_mib": 32,
171
- "upload_pages_per_commit": 164,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 611305.9510963782,
174
- "upload_bytes_per_second": 5634180.224215531,
175
- "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 65.58601331552748,
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": 1.104392732000008,
184
- "bytes": 36786,
185
  "pages": 1,
186
  "shards": 1
187
- },
188
- {
189
- "seconds": 0.71818233099998,
190
- "bytes": 253911,
191
- "pages": 7,
192
- "shards": 1
193
- },
194
- {
195
- "seconds": 1.0387228129999357,
196
- "bytes": 25849,
197
- "pages": 1,
198
- "shards": 1
199
- },
200
- {
201
- "seconds": 1.3780422429999817,
202
- "bytes": 2340308,
203
- "pages": 30,
204
- "shards": 2
205
- },
206
- {
207
- "seconds": 1.0020929300000034,
208
- "bytes": 10862030,
209
- "pages": 33,
210
- "shards": 3
211
- },
212
- {
213
- "seconds": 0.8503723880000962,
214
- "bytes": 6155947,
215
- "pages": 35,
216
- "shards": 2
217
- },
218
- {
219
- "seconds": 0.8228095520000807,
220
- "bytes": 2489087,
221
- "pages": 34,
222
- "shards": 2
223
- },
224
- {
225
- "seconds": 1.2741382629999407,
226
- "bytes": 11572130,
227
- "pages": 33,
228
- "shards": 3
229
- },
230
- {
231
- "seconds": 1.0936447959998077,
232
- "bytes": 6527222,
233
- "pages": 37,
234
- "shards": 3
235
- },
236
- {
237
- "seconds": 0.8425330919999396,
238
- "bytes": 6207711,
239
- "pages": 34,
240
- "shards": 2
241
- },
242
- {
243
- "seconds": 1.0241649789998064,
244
- "bytes": 5879797,
245
- "pages": 26,
246
- "shards": 2
247
- },
248
- {
249
- "seconds": 0.9143005570001606,
250
- "bytes": 2623386,
251
- "pages": 36,
252
- "shards": 2
253
- },
254
- {
255
- "seconds": 1.1859510890001275,
256
- "bytes": 10202108,
257
- "pages": 38,
258
- "shards": 3
259
- },
260
- {
261
- "seconds": 1.2170790539998961,
262
- "bytes": 6598891,
263
- "pages": 27,
264
- "shards": 2
265
- },
266
- {
267
- "seconds": 0.9556911770000625,
268
- "bytes": 7849875,
269
- "pages": 36,
270
- "shards": 3
271
- },
272
- {
273
- "seconds": 0.9569393540000419,
274
- "bytes": 10947040,
275
- "pages": 39,
276
- "shards": 3
277
- },
278
- {
279
- "seconds": 0.8917274570003428,
280
- "bytes": 2243299,
281
- "pages": 26,
282
- "shards": 2
283
- },
284
- {
285
- "seconds": 1.2289755599999808,
286
- "bytes": 6161182,
287
- "pages": 34,
288
- "shards": 2
289
- },
290
- {
291
- "seconds": 0.8229349630000797,
292
- "bytes": 1172160,
293
- "pages": 33,
294
- "shards": 1
295
- },
296
- {
297
- "seconds": 1.011977818000105,
298
- "bytes": 13444268,
299
- "pages": 40,
300
- "shards": 5
301
- },
302
- {
303
- "seconds": 1.0623180309999043,
304
- "bytes": 9531595,
305
- "pages": 31,
306
- "shards": 3
307
- },
308
- {
309
- "seconds": 0.9543692199999896,
310
- "bytes": 1054926,
311
- "pages": 32,
312
- "shards": 1
313
- },
314
- {
315
- "seconds": 1.207006858999648,
316
- "bytes": 7769728,
317
- "pages": 37,
318
- "shards": 3
319
- },
320
- {
321
- "seconds": 0.9489059839997935,
322
- "bytes": 5950611,
323
- "pages": 30,
324
- "shards": 2
325
- },
326
- {
327
- "seconds": 1.4311316419998548,
328
- "bytes": 7407679,
329
- "pages": 37,
330
- "shards": 3
331
- },
332
- {
333
- "seconds": 1.307497912999679,
334
- "bytes": 11505681,
335
- "pages": 31,
336
- "shards": 3
337
- },
338
- {
339
- "seconds": 0.7801257850001093,
340
- "bytes": 1090489,
341
- "pages": 33,
342
- "shards": 1
343
  }
344
  ],
345
  "cache_key": {
@@ -384,7 +228,7 @@
384
  },
385
  "cache_reused": true,
386
  "resume_probe_seconds": [
387
- 2.5095109330000014,
388
- 0.10111166000001504
389
  ]
390
  }
 
22
  "tokens_per_second": 1062967.6168812835,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
+ "required_token_positions_per_second": 75156.2396899193,
26
+ "required_pages_per_second_upper_estimate": 0.006514352374574179
27
  },
28
  {
29
  "source": "rewrite6",
 
36
  "tokens_per_second": 1206672.263377472,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
+ "required_token_positions_per_second": 9394.529961239912,
40
+ "required_pages_per_second_upper_estimate": 0.0037476184622785673
41
  },
42
  {
43
  "source": "openmath",
 
50
  "tokens_per_second": 410042.7581134458,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
+ "required_token_positions_per_second": 18789.059922479824,
54
+ "required_pages_per_second_upper_estimate": 0.004346518336905065
55
  },
56
  {
57
  "source": "ultra_qa",
 
64
  "tokens_per_second": 2163486.5624068812,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
+ "required_token_positions_per_second": 28183.589883719735,
68
+ "required_pages_per_second_upper_estimate": 0.0017162327560502429
69
  },
70
  {
71
  "source": "ultra_style",
 
78
  "tokens_per_second": 1546694.2737633225,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
+ "required_token_positions_per_second": 28183.589883719735,
82
+ "required_pages_per_second_upper_estimate": 0.0021696507664188645
83
  },
84
  {
85
  "source": "cortex",
 
92
  "tokens_per_second": 1015177.8395898147,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
  "estimated_pages_for_buffer": 22,
95
+ "required_token_positions_per_second": 28183.589883719735,
96
+ "required_pages_per_second_upper_estimate": 0.34994586194816957
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
 
158
  "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 688946,
161
+ "producer_batch_seconds": 0.08091448022655356,
162
+ "producer_p95_seconds": 0.18719405599949823,
163
+ "gpu_update_seconds": 0.9809963923712529,
164
+ "producer_batches_per_second": 12.35872735263313,
165
+ "gpu_batches_per_second": 1.019371740585928,
166
+ "producer_headroom": 12.123866947233024,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.82177632701768,
169
  "buffer_drain_seconds": null,
170
+ "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 87,
172
  "max_pending_pages": 4096,
173
+ "generation_bytes_per_second": 397083.56006381963,
174
+ "upload_bytes_per_second": 43958.880398765425,
175
+ "upload_sustainable": false,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 85.20488835750524,
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": 0.9955667570011428,
184
+ "bytes": 43764,
185
  "pages": 1,
186
  "shards": 1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
187
  }
188
  ],
189
  "cache_key": {
 
228
  },
229
  "cache_reused": true,
230
  "resume_probe_seconds": [
231
+ 0.5039531120000902,
232
+ 0.014973258001191425
233
  ]
234
  }
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b22b50c2a3f3cc11176d153ff6d2d48d52ac57f92ea3cf82eef967444d82426f
3
- size 721760447
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a7a66ded6461cef76b89f27ed3606c872345dc1f83d547f29c24c93e5c270c58
3
+ size 721778879