appvoid commited on
Commit
0bf1ee7
·
verified ·
1 Parent(s): d0fc621

Qwen90M step 375000; 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 +241 -241
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 370000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "27d38bcb78795a7f65759cb56e8869b8f3f0b60c43b03779747aade9b5f5dfb1",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "bb0f4782fb80c8565ac6be62823151fb471a6fd6261d4d56ac9296ed0dcad246",
12
  "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
13
  "performance.json": "e3923fa86352f6d878fd8d37b6b9fc374577f52be522203b0b865badf630da15",
14
- "shard_scheduler.json": "b624032da86b57ae0825c33e90bb460d2485d24bcaf10e56c24f0be2d7391d4f",
15
- "coverage.json": "6910c159e150a2a8f8c0838a7ccd1c7e5752fb223c3f3f4a879ec42144d06cf8",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 375000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "6610966b863295dd906696faac88a04c6ff0496a879a152f0273bb1b9d396255",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "406bc235a118a90d6ce9168e0c61f5aeee892c2cd7892ea2dfe941963fb47bc3",
12
  "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
13
  "performance.json": "e3923fa86352f6d878fd8d37b6b9fc374577f52be522203b0b865badf630da15",
14
+ "shard_scheduler.json": "3446bd4676a0432db7c222c8eed67dd16a89c65bbd51ec69a9afb2a131cff95e",
15
+ "coverage.json": "eb5f67654bace9fef82be9a9d8a490ab08cd158a627ef4268f68ce6a57faf403",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 370000,
3
  "stats": {
4
- "climbmix/blocks": 13320000,
5
- "climbmix/targets": 22265879849,
6
- "climbmix/input_tokens": 22267290588,
7
- "ultra_qa/blocks": 4995000,
8
- "ultra_qa/targets": 8148669403,
9
- "ultra_qa/input_tokens": 8148685149,
10
- "cortex/blocks": 4995000,
11
- "cortex/targets": 190015600,
12
- "cortex/input_tokens": 7940414338,
13
- "ultra_style/blocks": 4995000,
14
- "ultra_style/targets": 7900652448,
15
- "ultra_style/input_tokens": 7901067026,
16
- "openmath/blocks": 3330000,
17
- "openmath/targets": 5145705012,
18
- "openmath/input_tokens": 5147419580,
19
- "rewrite6/blocks": 1665000,
20
- "rewrite6/targets": 2729490219,
21
- "rewrite6/input_tokens": 2732482833
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 1824,
26
- "row": 1985,
27
  "start": {
28
- "file": 88,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 24,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
- "documents": 7473089
38
  },
39
  "rewrite6": {
40
- "page": 37,
41
- "row": 3836,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 151552
47
  }
48
  },
49
- "epoch": 20,
50
  "offset": 0,
51
- "documents": 4263468
52
  },
53
  "openmath": {
54
- "page": 1140,
55
- "row": 244,
56
  "start": {
57
  "file": 10,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 302,
61
- "row": 940
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 4669684
67
  },
68
  "ultra_qa": {
69
- "page": 492,
70
- "row": 3817,
71
  "start": {
72
  "file": 3,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 7,
76
- "row": 4237
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
- "documents": 2019049
82
  },
83
  "ultra_style": {
84
- "page": 752,
85
- "row": 2480,
86
  "start": {
87
  "file": 4,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
- "row_group": 3,
91
- "row": 85359
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
- "documents": 3082672
97
  },
98
  "cortex": {
99
- "page": 96880,
100
- "row": 25,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 12400665
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 375000,
3
  "stats": {
4
+ "climbmix/blocks": 13500000,
5
+ "climbmix/targets": 22566571953,
6
+ "climbmix/input_tokens": 22568001890,
7
+ "ultra_qa/blocks": 5062500,
8
+ "ultra_qa/targets": 8258873649,
9
+ "ultra_qa/input_tokens": 8258889605,
10
+ "cortex/blocks": 5062500,
11
+ "cortex/targets": 192584950,
12
+ "cortex/input_tokens": 8047684352,
13
+ "ultra_style/blocks": 5062500,
14
+ "ultra_style/targets": 8007585321,
15
+ "ultra_style/input_tokens": 8008005686,
16
+ "openmath/blocks": 3375000,
17
+ "openmath/targets": 5215242560,
18
+ "openmath/input_tokens": 5216980495,
19
+ "rewrite6/blocks": 1687500,
20
+ "rewrite6/targets": 2766400210,
21
+ "rewrite6/input_tokens": 2769433064
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 1849,
26
+ "row": 879,
27
  "start": {
28
+ "file": 89,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 41,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
+ "documents": 7574383
38
  },
39
  "rewrite6": {
40
+ "page": 1,
41
+ "row": 3226,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 4096
47
  }
48
  },
49
+ "epoch": 21,
50
  "offset": 0,
51
+ "documents": 4320806
52
  },
53
  "openmath": {
54
+ "page": 1155,
55
+ "row": 2141,
56
  "start": {
57
  "file": 10,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 364,
61
+ "row": 380
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 4733021
67
  },
68
  "ultra_qa": {
69
+ "page": 499,
70
+ "row": 2492,
71
  "start": {
72
  "file": 3,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 7,
76
+ "row": 32909
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
+ "documents": 2046396
82
  },
83
  "ultra_style": {
84
+ "page": 762,
85
+ "row": 3326,
86
  "start": {
87
  "file": 4,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
+ "row_group": 4,
91
+ "row": 40896
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
+ "documents": 3124478
97
  },
98
  "cortex": {
99
+ "page": 98190,
100
+ "row": 71,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 12568391
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 370000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "27d38bcb78795a7f65759cb56e8869b8f3f0b60c43b03779747aade9b5f5dfb1", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "bb0f4782fb80c8565ac6be62823151fb471a6fd6261d4d56ac9296ed0dcad246", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "e3923fa86352f6d878fd8d37b6b9fc374577f52be522203b0b865badf630da15", "shard_scheduler.json": "b624032da86b57ae0825c33e90bb460d2485d24bcaf10e56c24f0be2d7391d4f", "coverage.json": "6910c159e150a2a8f8c0838a7ccd1c7e5752fb223c3f3f4a879ec42144d06cf8", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000370000"}
 
1
+ {"step": 375000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "6610966b863295dd906696faac88a04c6ff0496a879a152f0273bb1b9d396255", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "406bc235a118a90d6ce9168e0c61f5aeee892c2cd7892ea2dfe941963fb47bc3", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "e3923fa86352f6d878fd8d37b6b9fc374577f52be522203b0b865badf630da15", "shard_scheduler.json": "3446bd4676a0432db7c222c8eed67dd16a89c65bbd51ec69a9afb2a131cff95e", "coverage.json": "eb5f67654bace9fef82be9a9d8a490ab08cd158a627ef4268f68ce6a57faf403", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000375000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:27d38bcb78795a7f65759cb56e8869b8f3f0b60c43b03779747aade9b5f5dfb1
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6610966b863295dd906696faac88a04c6ff0496a879a152f0273bb1b9d396255
3
  size 360622744
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": 75577.40725271897,
26
- "required_pages_per_second_upper_estimate": 0.0065508581114782265
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": 9447.175906589871,
40
- "required_pages_per_second_upper_estimate": 0.003768619717005693
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": 18894.351813179743,
54
- "required_pages_per_second_upper_estimate": 0.004370875762744505
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": 28341.527719769612,
68
- "required_pages_per_second_upper_estimate": 0.001725850341629895
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": 28341.527719769612,
82
- "required_pages_per_second_upper_estimate": 0.002181809258237896
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": 28341.527719769612,
96
- "required_pages_per_second_upper_estimate": 0.35190692128797463
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,410 +158,410 @@
158
  "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 717798,
161
- "producer_batch_seconds": 0.08124377494536361,
162
- "producer_p95_seconds": 0.21642768499987142,
163
- "gpu_update_seconds": 0.9755296282322197,
164
- "producer_batches_per_second": 12.30863534680041,
165
- "gpu_batches_per_second": 1.0250841912532414,
166
- "producer_headroom": 12.00743846391016,
167
  "sustainable": true,
168
- "buffer_seconds": 60.48283695039762,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 80,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 312252.486873165,
174
- "upload_bytes_per_second": 5783967.833540881,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 67.00211481710637,
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": 0.9784717390002697,
184
- "bytes": 7343882,
185
- "pages": 29,
186
- "shards": 3
187
  },
188
  {
189
- "seconds": 0.8120560110000952,
190
- "bytes": 1154123,
191
- "pages": 33,
192
  "shards": 1
193
  },
194
  {
195
- "seconds": 1.3542424540000866,
196
- "bytes": 9851710,
197
- "pages": 34,
198
- "shards": 3
199
  },
200
  {
201
- "seconds": 1.397486900000331,
202
- "bytes": 13488820,
203
- "pages": 39,
204
  "shards": 4
205
  },
206
  {
207
- "seconds": 0.9587566239988519,
208
- "bytes": 1052447,
209
- "pages": 33,
210
  "shards": 1
211
  },
212
  {
213
- "seconds": 1.538976370000455,
214
- "bytes": 12078751,
215
- "pages": 37,
216
- "shards": 5
217
- },
218
- {
219
- "seconds": 0.9161946559997887,
220
- "bytes": 856768,
221
- "pages": 25,
222
- "shards": 1
223
  },
224
  {
225
- "seconds": 1.2508889310011,
226
- "bytes": 6008939,
227
- "pages": 36,
228
  "shards": 2
229
  },
230
  {
231
- "seconds": 1.107625275000828,
232
- "bytes": 7585534,
233
- "pages": 37,
234
  "shards": 3
235
  },
236
  {
237
- "seconds": 1.0756502530002763,
238
- "bytes": 6680691,
239
- "pages": 29,
240
  "shards": 2
241
  },
242
  {
243
- "seconds": 1.0957601480004087,
244
- "bytes": 10587203,
245
- "pages": 33,
 
 
 
 
 
 
246
  "shards": 3
247
  },
248
  {
249
- "seconds": 0.9222303849983291,
250
- "bytes": 2581513,
251
  "pages": 36,
252
  "shards": 2
253
  },
254
  {
255
- "seconds": 1.2870966139998927,
256
- "bytes": 6072045,
257
- "pages": 35,
258
- "shards": 2
259
  },
260
  {
261
- "seconds": 1.1727791039993463,
262
- "bytes": 7506500,
263
- "pages": 37,
264
  "shards": 3
265
  },
266
  {
267
- "seconds": 1.0041788279995671,
268
- "bytes": 4770921,
269
- "pages": 34,
270
  "shards": 2
271
  },
272
  {
273
- "seconds": 1.17236374300046,
274
- "bytes": 5601597,
275
- "pages": 19,
276
  "shards": 2
277
  },
278
  {
279
- "seconds": 1.1996037619992421,
280
- "bytes": 8790333,
281
- "pages": 45,
282
- "shards": 3
283
  },
284
  {
285
- "seconds": 0.831416236998848,
286
- "bytes": 5844217,
287
- "pages": 30,
288
  "shards": 2
289
  },
290
  {
291
- "seconds": 1.3119218800002272,
292
- "bytes": 12680620,
293
- "pages": 37,
294
- "shards": 4
295
  },
296
  {
297
- "seconds": 0.8689146019987675,
298
- "bytes": 1090693,
299
- "pages": 33,
300
  "shards": 1
301
  },
302
  {
303
- "seconds": 0.9899479589985276,
304
- "bytes": 6233203,
305
  "pages": 36,
306
  "shards": 2
307
  },
308
  {
309
- "seconds": 0.8259697120010969,
310
- "bytes": 2261871,
311
- "pages": 24,
312
  "shards": 2
313
  },
314
  {
315
- "seconds": 2.0209359849995963,
316
- "bytes": 15406607,
317
- "pages": 37,
318
- "shards": 4
319
  },
320
  {
321
- "seconds": 1.1222496679984033,
322
- "bytes": 5615350,
323
- "pages": 27,
324
  "shards": 2
325
  },
326
  {
327
- "seconds": 0.8820894919990678,
328
- "bytes": 2581402,
329
- "pages": 34,
330
  "shards": 2
331
  },
332
  {
333
- "seconds": 1.0837803860013082,
334
- "bytes": 6217801,
 
 
 
 
 
 
335
  "pages": 36,
336
  "shards": 2
337
  },
338
  {
339
- "seconds": 1.5358513270002732,
340
- "bytes": 5360685,
341
- "pages": 28,
342
- "shards": 2
343
  },
344
  {
345
- "seconds": 1.0740509600000223,
346
- "bytes": 7543060,
347
- "pages": 37,
348
  "shards": 3
349
  },
350
  {
351
- "seconds": 1.1867137099998217,
352
- "bytes": 11783484,
353
- "pages": 32,
354
  "shards": 3
355
  },
356
  {
357
- "seconds": 1.1010340770008042,
358
- "bytes": 2510601,
 
 
 
 
 
 
359
  "pages": 35,
360
  "shards": 2
361
  },
362
  {
363
- "seconds": 1.0974370109997835,
364
- "bytes": 10722524,
365
- "pages": 40,
366
  "shards": 3
367
  },
368
  {
369
- "seconds": 1.196992418001173,
370
- "bytes": 774765,
371
- "pages": 25,
372
- "shards": 1
373
  },
374
  {
375
- "seconds": 1.1212218569999095,
376
- "bytes": 7621728,
377
- "pages": 38,
378
- "shards": 3
379
  },
380
  {
381
- "seconds": 1.0421719010009838,
382
- "bytes": 6283978,
383
- "pages": 35,
384
  "shards": 2
385
  },
386
  {
387
- "seconds": 1.2377747039990936,
388
- "bytes": 12765065,
389
- "pages": 37,
 
 
 
 
 
 
390
  "shards": 4
391
  },
392
  {
393
- "seconds": 1.0026179279993812,
394
- "bytes": 5878114,
395
- "pages": 28,
396
- "shards": 2
397
  },
398
  {
399
- "seconds": 0.8248095879989705,
400
- "bytes": 1051965,
401
- "pages": 33,
402
  "shards": 1
403
  },
404
  {
405
- "seconds": 1.2737176890004775,
406
- "bytes": 7770049,
407
- "pages": 38,
408
  "shards": 3
409
  },
410
  {
411
- "seconds": 1.1648699769993982,
412
- "bytes": 6106287,
413
- "pages": 25,
414
  "shards": 2
415
  },
416
  {
417
- "seconds": 1.0237944010004867,
418
- "bytes": 5653726,
419
- "pages": 43,
420
- "shards": 2
421
  },
422
  {
423
- "seconds": 2.0509548539994285,
424
- "bytes": 12950750,
425
- "pages": 26,
426
- "shards": 4
427
  },
428
  {
429
- "seconds": 0.8878694479990372,
430
- "bytes": 1157611,
431
  "pages": 34,
432
  "shards": 1
433
  },
434
  {
435
- "seconds": 1.0454248690002714,
436
- "bytes": 6130575,
437
- "pages": 33,
438
- "shards": 2
439
  },
440
  {
441
- "seconds": 1.3716071430008014,
442
- "bytes": 11837179,
443
  "pages": 36,
444
- "shards": 4
445
  },
446
  {
447
- "seconds": 0.8982555259990477,
448
- "bytes": 1117404,
449
- "pages": 34,
450
  "shards": 1
451
  },
452
  {
453
- "seconds": 1.2563196339997376,
454
- "bytes": 7398482,
455
- "pages": 29,
456
  "shards": 3
457
  },
458
  {
459
- "seconds": 1.2963822889996663,
460
- "bytes": 6635646,
461
- "pages": 28,
462
- "shards": 2
463
- },
464
- {
465
- "seconds": 1.2522562880003534,
466
- "bytes": 9539459,
467
- "pages": 41,
468
  "shards": 3
469
  },
470
  {
471
- "seconds": 0.8906719839997095,
472
- "bytes": 2274847,
473
- "pages": 27,
474
- "shards": 2
475
- },
476
- {
477
- "seconds": 1.014783542999794,
478
- "bytes": 6341777,
479
  "pages": 35,
480
- "shards": 2
481
  },
482
  {
483
- "seconds": 1.1726115760011453,
484
- "bytes": 10149882,
485
  "pages": 35,
486
- "shards": 3
487
  },
488
  {
489
- "seconds": 1.0351456819989835,
490
- "bytes": 2694003,
491
- "pages": 36,
492
- "shards": 2
493
  },
494
  {
495
- "seconds": 1.6696938850000151,
496
- "bytes": 11635327,
497
- "pages": 30,
498
  "shards": 3
499
  },
500
  {
501
- "seconds": 1.0197001609994913,
502
- "bytes": 7722133,
503
- "pages": 38,
504
  "shards": 3
505
  },
506
  {
507
- "seconds": 1.0303949059998558,
508
- "bytes": 5043636,
509
- "pages": 33,
510
- "shards": 2
511
  },
512
  {
513
- "seconds": 1.2701704230003088,
514
- "bytes": 6145157,
515
- "pages": 35,
516
  "shards": 2
517
  },
518
  {
519
- "seconds": 1.1758933869987231,
520
- "bytes": 2380578,
521
- "pages": 31,
522
  "shards": 2
523
  },
524
  {
525
- "seconds": 1.1564406300003611,
526
- "bytes": 5972956,
527
- "pages": 32,
528
- "shards": 2
529
  },
530
  {
531
- "seconds": 1.3616979350008478,
532
- "bytes": 18065144,
533
- "pages": 42,
534
- "shards": 5
535
  },
536
  {
537
- "seconds": 0.8625212450006075,
538
- "bytes": 879863,
539
- "pages": 25,
540
  "shards": 1
541
  },
542
  {
543
- "seconds": 1.04706289600108,
544
- "bytes": 6208354,
545
- "pages": 36,
546
- "shards": 2
547
  },
548
  {
549
- "seconds": 0.9189691079991462,
550
- "bytes": 2568805,
551
- "pages": 36,
552
  "shards": 2
553
  },
554
  {
555
- "seconds": 1.252516081000067,
556
- "bytes": 9866864,
557
- "pages": 30,
558
  "shards": 3
559
- },
560
- {
561
- "seconds": 1.0568891879993316,
562
- "bytes": 6084196,
563
- "pages": 35,
564
- "shards": 2
565
  }
566
  ],
567
  "cache_key": {
 
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": 75451.54704015724,
26
+ "required_pages_per_second_upper_estimate": 0.006539948867243167
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": 9431.443380019655,
40
+ "required_pages_per_second_upper_estimate": 0.0037623437769346
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": 18862.88676003931,
54
+ "required_pages_per_second_upper_estimate": 0.004363596876466488
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": 28294.330140058966,
68
+ "required_pages_per_second_upper_estimate": 0.00172297625665208
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": 28294.330140058966,
82
+ "required_pages_per_second_upper_estimate": 0.002178175857900508
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": 28294.330140058966,
96
+ "required_pages_per_second_upper_estimate": 0.3513208853081064
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": 717798,
161
+ "producer_batch_seconds": 0.11261634734350423,
162
+ "producer_p95_seconds": 0.27807193199987523,
163
+ "gpu_update_seconds": 0.9771569025716607,
164
+ "producer_batches_per_second": 8.879705509802974,
165
+ "gpu_batches_per_second": 1.023377102866716,
166
+ "producer_headroom": 8.676865531707582,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.58372795944296,
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": 413636.13457235816,
174
+ "upload_bytes_per_second": 5862199.890527139,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 88.75668552282389,
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": 2.0509548539994285,
184
+ "bytes": 12950750,
185
+ "pages": 26,
186
+ "shards": 4
187
  },
188
  {
189
+ "seconds": 0.8878694479990372,
190
+ "bytes": 1157611,
191
+ "pages": 34,
192
  "shards": 1
193
  },
194
  {
195
+ "seconds": 1.0454248690002714,
196
+ "bytes": 6130575,
197
+ "pages": 33,
198
+ "shards": 2
199
  },
200
  {
201
+ "seconds": 1.3716071430008014,
202
+ "bytes": 11837179,
203
+ "pages": 36,
204
  "shards": 4
205
  },
206
  {
207
+ "seconds": 0.8982555259990477,
208
+ "bytes": 1117404,
209
+ "pages": 34,
210
  "shards": 1
211
  },
212
  {
213
+ "seconds": 1.2563196339997376,
214
+ "bytes": 7398482,
215
+ "pages": 29,
216
+ "shards": 3
 
 
 
 
 
 
217
  },
218
  {
219
+ "seconds": 1.2963822889996663,
220
+ "bytes": 6635646,
221
+ "pages": 28,
222
  "shards": 2
223
  },
224
  {
225
+ "seconds": 1.2522562880003534,
226
+ "bytes": 9539459,
227
+ "pages": 41,
228
  "shards": 3
229
  },
230
  {
231
+ "seconds": 0.8906719839997095,
232
+ "bytes": 2274847,
233
+ "pages": 27,
234
  "shards": 2
235
  },
236
  {
237
+ "seconds": 1.014783542999794,
238
+ "bytes": 6341777,
239
+ "pages": 35,
240
+ "shards": 2
241
+ },
242
+ {
243
+ "seconds": 1.1726115760011453,
244
+ "bytes": 10149882,
245
+ "pages": 35,
246
  "shards": 3
247
  },
248
  {
249
+ "seconds": 1.0351456819989835,
250
+ "bytes": 2694003,
251
  "pages": 36,
252
  "shards": 2
253
  },
254
  {
255
+ "seconds": 1.6696938850000151,
256
+ "bytes": 11635327,
257
+ "pages": 30,
258
+ "shards": 3
259
  },
260
  {
261
+ "seconds": 1.0197001609994913,
262
+ "bytes": 7722133,
263
+ "pages": 38,
264
  "shards": 3
265
  },
266
  {
267
+ "seconds": 1.0303949059998558,
268
+ "bytes": 5043636,
269
+ "pages": 33,
270
  "shards": 2
271
  },
272
  {
273
+ "seconds": 1.2701704230003088,
274
+ "bytes": 6145157,
275
+ "pages": 35,
276
  "shards": 2
277
  },
278
  {
279
+ "seconds": 1.1758933869987231,
280
+ "bytes": 2380578,
281
+ "pages": 31,
282
+ "shards": 2
283
  },
284
  {
285
+ "seconds": 1.1564406300003611,
286
+ "bytes": 5972956,
287
+ "pages": 32,
288
  "shards": 2
289
  },
290
  {
291
+ "seconds": 1.3616979350008478,
292
+ "bytes": 18065144,
293
+ "pages": 42,
294
+ "shards": 5
295
  },
296
  {
297
+ "seconds": 0.8625212450006075,
298
+ "bytes": 879863,
299
+ "pages": 25,
300
  "shards": 1
301
  },
302
  {
303
+ "seconds": 1.04706289600108,
304
+ "bytes": 6208354,
305
  "pages": 36,
306
  "shards": 2
307
  },
308
  {
309
+ "seconds": 0.9189691079991462,
310
+ "bytes": 2568805,
311
+ "pages": 36,
312
  "shards": 2
313
  },
314
  {
315
+ "seconds": 1.252516081000067,
316
+ "bytes": 9866864,
317
+ "pages": 30,
318
+ "shards": 3
319
  },
320
  {
321
+ "seconds": 1.0568891879993316,
322
+ "bytes": 6084196,
323
+ "pages": 35,
324
  "shards": 2
325
  },
326
  {
327
+ "seconds": 0.8473804010009189,
328
+ "bytes": 2318684,
329
+ "pages": 28,
330
  "shards": 2
331
  },
332
  {
333
+ "seconds": 1.1035010350005905,
334
+ "bytes": 11692213,
335
+ "pages": 31,
336
+ "shards": 3
337
+ },
338
+ {
339
+ "seconds": 1.0510043949998362,
340
+ "bytes": 4790558,
341
  "pages": 36,
342
  "shards": 2
343
  },
344
  {
345
+ "seconds": 1.1652991789997031,
346
+ "bytes": 7620124,
347
+ "pages": 34,
348
+ "shards": 3
349
  },
350
  {
351
+ "seconds": 1.0473349340009008,
352
+ "bytes": 5879474,
353
+ "pages": 30,
354
  "shards": 3
355
  },
356
  {
357
+ "seconds": 0.929622351000944,
358
+ "bytes": 6601933,
359
+ "pages": 39,
360
  "shards": 3
361
  },
362
  {
363
+ "seconds": 2.1250439309987996,
364
+ "bytes": 6028843,
365
+ "pages": 31,
366
+ "shards": 2
367
+ },
368
+ {
369
+ "seconds": 0.9982244909988367,
370
+ "bytes": 6934101,
371
  "pages": 35,
372
  "shards": 2
373
  },
374
  {
375
+ "seconds": 1.4433788420028577,
376
+ "bytes": 7611195,
377
+ "pages": 35,
378
  "shards": 3
379
  },
380
  {
381
+ "seconds": 1.260750511999504,
382
+ "bytes": 10895855,
383
+ "pages": 34,
384
+ "shards": 3
385
  },
386
  {
387
+ "seconds": 1.124983618999977,
388
+ "bytes": 2583658,
389
+ "pages": 36,
390
+ "shards": 2
391
  },
392
  {
393
+ "seconds": 1.0313133240015304,
394
+ "bytes": 5497040,
395
+ "pages": 26,
396
  "shards": 2
397
  },
398
  {
399
+ "seconds": 1.0232497649994912,
400
+ "bytes": 1108747,
401
+ "pages": 34,
402
+ "shards": 1
403
+ },
404
+ {
405
+ "seconds": 1.292045906000567,
406
+ "bytes": 13335425,
407
+ "pages": 35,
408
  "shards": 4
409
  },
410
  {
411
+ "seconds": 1.1053190409984381,
412
+ "bytes": 10451885,
413
+ "pages": 40,
414
+ "shards": 3
415
  },
416
  {
417
+ "seconds": 0.8453881149980589,
418
+ "bytes": 832969,
419
+ "pages": 26,
420
  "shards": 1
421
  },
422
  {
423
+ "seconds": 1.4130192459997488,
424
+ "bytes": 7383726,
425
+ "pages": 37,
426
  "shards": 3
427
  },
428
  {
429
+ "seconds": 1.007008101001702,
430
+ "bytes": 6024431,
431
+ "pages": 35,
432
  "shards": 2
433
  },
434
  {
435
+ "seconds": 1.0379702050013293,
436
+ "bytes": 6386142,
437
+ "pages": 34,
438
+ "shards": 3
439
  },
440
  {
441
+ "seconds": 1.1697385600018606,
442
+ "bytes": 11844982,
443
+ "pages": 34,
444
+ "shards": 3
445
  },
446
  {
447
+ "seconds": 1.079595678998885,
448
+ "bytes": 1118845,
449
  "pages": 34,
450
  "shards": 1
451
  },
452
  {
453
+ "seconds": 1.1773780209987308,
454
+ "bytes": 12342839,
455
+ "pages": 35,
456
+ "shards": 4
457
  },
458
  {
459
+ "seconds": 1.146244535997539,
460
+ "bytes": 6228835,
461
  "pages": 36,
462
+ "shards": 2
463
  },
464
  {
465
+ "seconds": 1.0997196959979192,
466
+ "bytes": 878203,
467
+ "pages": 25,
468
  "shards": 1
469
  },
470
  {
471
+ "seconds": 1.3262504000012996,
472
+ "bytes": 7612878,
473
+ "pages": 38,
474
  "shards": 3
475
  },
476
  {
477
+ "seconds": 1.397075862998463,
478
+ "bytes": 11930697,
479
+ "pages": 36,
 
 
 
 
 
 
480
  "shards": 3
481
  },
482
  {
483
+ "seconds": 1.03103892800209,
484
+ "bytes": 6422195,
 
 
 
 
 
 
485
  "pages": 35,
486
+ "shards": 3
487
  },
488
  {
489
+ "seconds": 1.064075487000082,
490
+ "bytes": 6211136,
491
  "pages": 35,
492
+ "shards": 2
493
  },
494
  {
495
+ "seconds": 0.7981215300023905,
496
+ "bytes": 910382,
497
+ "pages": 26,
498
+ "shards": 1
499
  },
500
  {
501
+ "seconds": 1.0455480339987844,
502
+ "bytes": 7415331,
503
+ "pages": 34,
504
  "shards": 3
505
  },
506
  {
507
+ "seconds": 1.2304811680005514,
508
+ "bytes": 10294757,
509
+ "pages": 37,
510
  "shards": 3
511
  },
512
  {
513
+ "seconds": 1.5211265639991325,
514
+ "bytes": 8178789,
515
+ "pages": 36,
516
+ "shards": 3
517
  },
518
  {
519
+ "seconds": 0.9424325320032949,
520
+ "bytes": 6190430,
521
+ "pages": 36,
522
  "shards": 2
523
  },
524
  {
525
+ "seconds": 0.982011802996567,
526
+ "bytes": 5926739,
527
+ "pages": 33,
528
  "shards": 2
529
  },
530
  {
531
+ "seconds": 1.048674845998903,
532
+ "bytes": 7634002,
533
+ "pages": 29,
534
+ "shards": 3
535
  },
536
  {
537
+ "seconds": 1.2676819320004142,
538
+ "bytes": 6003365,
539
+ "pages": 36,
540
+ "shards": 3
541
  },
542
  {
543
+ "seconds": 0.7857203480016324,
544
+ "bytes": 1089066,
545
+ "pages": 33,
546
  "shards": 1
547
  },
548
  {
549
+ "seconds": 1.0898243610026839,
550
+ "bytes": 13191576,
551
+ "pages": 31,
552
+ "shards": 4
553
  },
554
  {
555
+ "seconds": 0.8497731899988139,
556
+ "bytes": 5387577,
557
+ "pages": 34,
558
  "shards": 2
559
  },
560
  {
561
+ "seconds": 1.2923507409977901,
562
+ "bytes": 7469632,
563
+ "pages": 37,
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:bb0f4782fb80c8565ac6be62823151fb471a6fd6261d4d56ac9296ed0dcad246
3
- size 721784447
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:406bc235a118a90d6ce9168e0c61f5aeee892c2cd7892ea2dfe941963fb47bc3
3
+ size 721764351