appvoid commited on
Commit
836cfdf
·
verified ·
1 Parent(s): b865a9b

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": 77454.1153760207,
26
- "required_pages_per_second_upper_estimate": 0.005358455420021135
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": 9681.764422002587,
40
- "required_pages_per_second_upper_estimate": 0.004053552585260899
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": 19363.528844005174,
54
- "required_pages_per_second_upper_estimate": 0.004032912407610896
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": 29045.29326600776,
68
- "required_pages_per_second_upper_estimate": 0.0017412209817022728
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": 29045.29326600776,
82
- "required_pages_per_second_upper_estimate": 0.002258143794688775
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": 29045.29326600776,
96
- "required_pages_per_second_upper_estimate": 0.3725618356102123
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": 64,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
- "producer_batch_seconds": 0.05481918577343947,
162
- "producer_p95_seconds": 0.19849216200009323,
163
- "gpu_update_seconds": 0.9518926094768325,
164
- "producer_batches_per_second": 18.241788634600837,
165
- "gpu_batches_per_second": 1.0505386742624334,
166
- "producer_headroom": 17.364223784915016,
167
  "sustainable": true,
168
- "buffer_seconds": 60.92112700651728,
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": 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": 77084.74304473559,
26
+ "required_pages_per_second_upper_estimate": 0.005332901385080942
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": 9635.592880591948,
40
+ "required_pages_per_second_upper_estimate": 0.004034221525043688
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": 19271.185761183897,
54
+ "required_pages_per_second_upper_estimate": 0.004013679778710082
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": 28906.778641775847,
68
+ "required_pages_per_second_upper_estimate": 0.0017329172414791546
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": 28906.778641775847,
82
+ "required_pages_per_second_upper_estimate": 0.002247374891916187
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": 28906.778641775847,
96
+ "required_pages_per_second_upper_estimate": 0.3707851187359814
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": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
+ "producer_batch_seconds": 0.05218987320311541,
162
+ "producer_p95_seconds": 0.13707047500020053,
163
+ "gpu_update_seconds": 0.9564538595816875,
164
+ "producer_batches_per_second": 19.160805317693438,
165
+ "gpu_batches_per_second": 1.0455287413836751,
166
+ "producer_headroom": 18.32642619880121,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.256593153646314,
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
@@ -31,3 +31,4 @@
31
  {"step": 928100, "loss": 0.5575608015060425, "lr": 3e-05, "grad_norm": 0.35268365170265237, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135847590762, "tokens_per_second": 152781.83261242192, "data_wait_s": 6.800999926781515e-06, "gpu_peak_gb": 89.37899398803711}
32
  {"step": 928200, "loss": 0.5369636416435242, "lr": 3e-05, "grad_norm": 0.32165364746021524, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135862068058, "tokens_per_second": 152026.8280368941, "data_wait_s": 6.794999990233919e-06, "gpu_peak_gb": 89.37899398803711}
33
  {"step": 928300, "loss": 0.5516103506088257, "lr": 3e-05, "grad_norm": 0.328012424760905, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135876730454, "tokens_per_second": 153856.77058871454, "data_wait_s": 1.1735000043699984e-05, "gpu_peak_gb": 89.37899398803711}
 
 
31
  {"step": 928100, "loss": 0.5575608015060425, "lr": 3e-05, "grad_norm": 0.35268365170265237, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135847590762, "tokens_per_second": 152781.83261242192, "data_wait_s": 6.800999926781515e-06, "gpu_peak_gb": 89.37899398803711}
32
  {"step": 928200, "loss": 0.5369636416435242, "lr": 3e-05, "grad_norm": 0.32165364746021524, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135862068058, "tokens_per_second": 152026.8280368941, "data_wait_s": 6.794999990233919e-06, "gpu_peak_gb": 89.37899398803711}
33
  {"step": 928300, "loss": 0.5516103506088257, "lr": 3e-05, "grad_norm": 0.328012424760905, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135876730454, "tokens_per_second": 153856.77058871454, "data_wait_s": 1.1735000043699984e-05, "gpu_peak_gb": 89.37899398803711}
34
+ {"step": 928400, "loss": 0.584817111492157, "lr": 3e-05, "grad_norm": 0.34219711359203886, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135891380783, "tokens_per_second": 153575.66740318137, "data_wait_s": 1.710800006549107e-05, "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 sha256:18873e73d6fad262097f185301734dc015284d16643899fda26c2efd3c347023
3
- size 70059
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f71badfe28d365adb00fe7271fe68a46867b6e225cad875503b3fa0470413bc3
3
+ size 72177