appvoid commited on
Commit
194d0fb
·
verified ·
1 Parent(s): c4b3859

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": 77276.63016099708,
26
- "required_pages_per_second_upper_estimate": 0.005346176581023364
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": 9659.578770124635,
40
- "required_pages_per_second_upper_estimate": 0.0040442639161086934
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": 19319.15754024927,
54
- "required_pages_per_second_upper_estimate": 0.0040236710351885105
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": 28978.736310373904,
68
- "required_pages_per_second_upper_estimate": 0.0017372309938386101
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": 28978.736310373904,
82
- "required_pages_per_second_upper_estimate": 0.0022529692841413516
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": 28978.736310373904,
96
- "required_pages_per_second_upper_estimate": 0.3717081144466323
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": 0.05209850517968917,
162
- "producer_p95_seconds": 0.13456797300023027,
163
- "gpu_update_seconds": 0.9540788702405383,
164
- "producer_batches_per_second": 19.194408679308026,
165
- "gpu_batches_per_second": 1.0481313769666487,
166
- "producer_headroom": 18.312979747689386,
167
  "sustainable": true,
168
- "buffer_seconds": 60.10696882515391,
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": 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": 77477.60150821909,
26
+ "required_pages_per_second_upper_estimate": 0.005360080245141954
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": 9684.700188527386,
40
+ "required_pages_per_second_upper_estimate": 0.004054781729399055
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": 19369.40037705477,
54
+ "required_pages_per_second_upper_estimate": 0.004034135293102388
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": 29054.100565582157,
68
+ "required_pages_per_second_upper_estimate": 0.0017417489658637902
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": 29054.100565582157,
82
+ "required_pages_per_second_upper_estimate": 0.0022588285234914677
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": 29054.100565582157,
96
+ "required_pages_per_second_upper_estimate": 0.3726748061926111
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": 0.05401758545312063,
162
+ "producer_p95_seconds": 0.1506637600000431,
163
+ "gpu_update_seconds": 0.9516040580086709,
164
+ "producer_batches_per_second": 18.512489805155283,
165
+ "gpu_batches_per_second": 1.0508572253176416,
166
+ "producer_headroom": 17.616560422429917,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.902659712554936,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 84,
172
  "max_pending_pages": 4096,
173
  "generation_bytes_per_second": 0,
174
  "upload_bytes_per_second": null,
metrics/void-byte-v1/1791463982175382982-d4a3b372/train.jsonl CHANGED
@@ -36,3 +36,4 @@
36
  {"step": 928600, "loss": 0.5858592987060547, "lr": 3e-05, "grad_norm": 0.37578295498138514, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135920801717, "tokens_per_second": 153644.25579552943, "data_wait_s": 1.4542999906552723e-05, "gpu_peak_gb": 89.37899398803711}
37
  {"step": 928700, "loss": 0.5363717675209045, "lr": 3e-05, "grad_norm": 0.3355081071269844, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135935459183, "tokens_per_second": 153776.51483855143, "data_wait_s": 6.9359998633444775e-06, "gpu_peak_gb": 89.37899398803711}
38
  {"step": 928800, "loss": 0.5778369307518005, "lr": 3e-05, "grad_norm": 0.3463705909920343, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135949889323, "tokens_per_second": 151523.70106611072, "data_wait_s": 1.3439999747788534e-05, "gpu_peak_gb": 89.37899398803711}
 
 
36
  {"step": 928600, "loss": 0.5858592987060547, "lr": 3e-05, "grad_norm": 0.37578295498138514, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135920801717, "tokens_per_second": 153644.25579552943, "data_wait_s": 1.4542999906552723e-05, "gpu_peak_gb": 89.37899398803711}
37
  {"step": 928700, "loss": 0.5363717675209045, "lr": 3e-05, "grad_norm": 0.3355081071269844, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135935459183, "tokens_per_second": 153776.51483855143, "data_wait_s": 6.9359998633444775e-06, "gpu_peak_gb": 89.37899398803711}
38
  {"step": 928800, "loss": 0.5778369307518005, "lr": 3e-05, "grad_norm": 0.3463705909920343, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135949889323, "tokens_per_second": 151523.70106611072, "data_wait_s": 1.3439999747788534e-05, "gpu_peak_gb": 89.37899398803711}
39
+ {"step": 928900, "loss": 0.5194796323776245, "lr": 3e-05, "grad_norm": 0.3417008088336159, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135964381838, "tokens_per_second": 152211.75039808516, "data_wait_s": 7.288000233529601e-06, "gpu_peak_gb": 89.37899398803711}
runs/void-byte-v1/events.out.tfevents.1791463982.2e076f0c-3fe2-4316-8f17-4e30b2e706e2-b5972.360.0.1791463982175382982-d4a3b372 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:39f540c2ad63e5a95731be97f0c5677984bcd8c74939c301e3bbeeef30c6e1f6
3
- size 80649
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f4ad52c504a6e3aebd5e25d7eeea4a318de74ff4508439821ff8ac3365136b69
3
+ size 82767