appvoid commited on
Commit
bfd79ff
·
verified ·
1 Parent(s): 5858903

Update TensorBoard training metrics

Browse files
benchmark_cache/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": 77355.21815380108,
26
- "required_pages_per_second_upper_estimate": 0.005351613480714811
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": 9669.402269225135,
40
- "required_pages_per_second_upper_estimate": 0.004048376810044085
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": 19338.80453845027,
54
- "required_pages_per_second_upper_estimate": 0.004027762986787593
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": 29008.206807675404,
68
- "required_pages_per_second_upper_estimate": 0.001738997705152991
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": 29008.206807675404,
82
- "required_pages_per_second_upper_estimate": 0.002255260485679525
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": 29008.206807675404,
96
- "required_pages_per_second_upper_estimate": 0.3720861303430613
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -155,20 +155,20 @@
155
  "shard_benchmark_scope": "Local packing/read measurements; upload capacity measured only on real data commits. Targets larger than available pages are sample-limited.",
156
  "estimated_fresh_generation_batch_seconds": 2.558590814491007,
157
  "selected": {
158
- "prefetch_batches": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
- "producer_batch_seconds": 0.05176574607812867,
162
  "producer_p95_seconds": 0.14215604600008191,
163
- "gpu_update_seconds": 0.9531095866527157,
164
- "producer_batches_per_second": 19.3177936330856,
165
- "gpu_batches_per_second": 1.0491972948377968,
166
- "producer_headroom": 18.411974304672682,
167
  "sustainable": true,
168
- "buffer_seconds": 60.04590395912109,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 68,
172
  "max_pending_pages": 4096,
173
  "generation_bytes_per_second": 0,
174
  "upload_bytes_per_second": null,
 
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": 77649.7288253755,
26
+ "required_pages_per_second_upper_estimate": 0.0053719884123332355
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": 9706.216103171937,
40
+ "required_pages_per_second_upper_estimate": 0.00406378999355734
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": 19412.432206343874,
54
+ "required_pages_per_second_upper_estimate": 0.004043097688317656
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": 29118.648309515815,
68
+ "required_pages_per_second_upper_estimate": 0.0017456185045539147
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": 29118.648309515815,
82
+ "required_pages_per_second_upper_estimate": 0.002263846826666789
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": 29118.648309515815,
96
+ "required_pages_per_second_upper_estimate": 0.3735027553458244
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
 
155
  "shard_benchmark_scope": "Local packing/read measurements; upload capacity measured only on real data commits. Targets larger than available pages are sample-limited.",
156
  "estimated_fresh_generation_batch_seconds": 2.558590814491007,
157
  "selected": {
158
+ "prefetch_batches": 64,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
+ "producer_batch_seconds": 0.05077983951563425,
162
  "producer_p95_seconds": 0.14215604600008191,
163
+ "gpu_update_seconds": 0.9494946230373196,
164
+ "producer_batches_per_second": 19.692854674976218,
165
+ "gpu_batches_per_second": 1.0531918514726495,
166
+ "producer_headroom": 18.69825962614526,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.76765587438845,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 72,
172
  "max_pending_pages": 4096,
173
  "generation_bytes_per_second": 0,
174
  "upload_bytes_per_second": null,
metrics/void-byte-v1/1791083219122319327-61ae60ac/train.jsonl CHANGED
@@ -14,3 +14,4 @@
14
  {"step": 646400, "loss": 0.507813036441803, "lr": 0.0001, "grad_norm": 0.16432517602460556, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 94616241350, "tokens_per_second": 152966.47065197848, "data_wait_s": 1.4149999969959026e-05, "gpu_peak_gb": 89.37899398803711}
15
  {"step": 646500, "loss": 0.5888087749481201, "lr": 0.0001, "grad_norm": 0.19901965227469567, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 94630940027, "tokens_per_second": 153553.19083680474, "data_wait_s": 8.095000112007256e-06, "gpu_peak_gb": 89.37899398803711}
16
  {"step": 646600, "loss": 0.5488295555114746, "lr": 0.0001, "grad_norm": 0.1862812472492112, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 94645761373, "tokens_per_second": 155561.3722798839, "data_wait_s": 1.3596999906440033e-05, "gpu_peak_gb": 89.37899398803711}
 
 
14
  {"step": 646400, "loss": 0.507813036441803, "lr": 0.0001, "grad_norm": 0.16432517602460556, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 94616241350, "tokens_per_second": 152966.47065197848, "data_wait_s": 1.4149999969959026e-05, "gpu_peak_gb": 89.37899398803711}
15
  {"step": 646500, "loss": 0.5888087749481201, "lr": 0.0001, "grad_norm": 0.19901965227469567, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 94630940027, "tokens_per_second": 153553.19083680474, "data_wait_s": 8.095000112007256e-06, "gpu_peak_gb": 89.37899398803711}
16
  {"step": 646600, "loss": 0.5488295555114746, "lr": 0.0001, "grad_norm": 0.1862812472492112, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 94645761373, "tokens_per_second": 155561.3722798839, "data_wait_s": 1.3596999906440033e-05, "gpu_peak_gb": 89.37899398803711}
17
+ {"step": 646700, "loss": 0.5115863084793091, "lr": 0.0001, "grad_norm": 0.17171189026755804, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 94660486978, "tokens_per_second": 154687.06638118604, "data_wait_s": 7.05599995853845e-06, "gpu_peak_gb": 89.37899398803711}
runs/void-byte-v1/events.out.tfevents.1791083219.1a73f132-ede3-4b51-9e4f-0cafff989449-8p4r7.435.0.1791083219122319327-61ae60ac CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:3ca87420f84ed390aaa2f4d4c5a146b591125b310c0be2bf24d06d63d710a409
3
- size 34053
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3a4401e6fddbd0a821884aad9b3163a2c38fa442c91114e59164e7d01a08b555
3
+ size 36171