appvoid commited on
Commit
22969a3
·
verified ·
1 Parent(s): 6e45482

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": 77403.3418263946,
26
- "required_pages_per_second_upper_estimate": 0.005354942788047135
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": 9675.417728299324,
40
- "required_pages_per_second_upper_estimate": 0.004050895357141378
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": 19350.83545659865,
54
- "required_pages_per_second_upper_estimate": 0.004030268709761253
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": 29026.25318489797,
68
- "required_pages_per_second_upper_estimate": 0.0017400795579122616
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": 29026.25318489797,
82
- "required_pages_per_second_upper_estimate": 0.0022566635121309547
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": 29026.25318489797,
96
- "required_pages_per_second_upper_estimate": 0.3723176098933822
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": 0.052541142820327735,
162
- "producer_p95_seconds": 0.13521431299977849,
163
- "gpu_update_seconds": 0.9525170136111449,
164
- "producer_batches_per_second": 19.032703636075237,
165
- "gpu_batches_per_second": 1.0498500139213676,
166
- "producer_headroom": 18.128974028380362,
167
  "sustainable": true,
168
- "buffer_seconds": 60.00857185750213,
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": 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": 77423.90256937826,
26
+ "required_pages_per_second_upper_estimate": 0.0053563652279542415
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": 9677.987821172283,
40
+ "required_pages_per_second_upper_estimate": 0.004051971401357644
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": 19355.975642344565,
54
+ "required_pages_per_second_upper_estimate": 0.004031339274896314
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": 29033.963463516844,
68
+ "required_pages_per_second_upper_estimate": 0.0017405417773425492
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": 29033.963463516844,
82
+ "required_pages_per_second_upper_estimate": 0.0022572629523796408
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": 29033.963463516844,
96
+ "required_pages_per_second_upper_estimate": 0.37241650906885293
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": 0.046736025039074036,
162
+ "producer_p95_seconds": 0.12620448699999542,
163
+ "gpu_update_seconds": 0.9522640625604423,
164
+ "producer_batches_per_second": 21.39677046055889,
165
+ "gpu_batches_per_second": 1.0501288868459506,
166
+ "producer_headroom": 20.375375564445076,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.944900003868305,
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-v1/1791254470898546162-7724d9f6/train.jsonl CHANGED
@@ -21,3 +21,4 @@
21
  {"step": 762100, "loss": 0.5389026403427124, "lr": 3e-05, "grad_norm": 0.2438201640206466, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111549176468, "tokens_per_second": 153893.07013894522, "data_wait_s": 1.526000005469541e-05, "gpu_peak_gb": 89.37899398803711}
22
  {"step": 762200, "loss": 0.5579586625099182, "lr": 3e-05, "grad_norm": 0.24301690088266478, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111563720181, "tokens_per_second": 152356.14392541448, "data_wait_s": 1.2263999906281242e-05, "gpu_peak_gb": 89.37899398803711}
23
  {"step": 762300, "loss": 0.558387279510498, "lr": 3e-05, "grad_norm": 0.28096033875982174, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111578230490, "tokens_per_second": 152199.34418891103, "data_wait_s": 1.0274000032950426e-05, "gpu_peak_gb": 89.37899398803711}
 
 
21
  {"step": 762100, "loss": 0.5389026403427124, "lr": 3e-05, "grad_norm": 0.2438201640206466, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111549176468, "tokens_per_second": 153893.07013894522, "data_wait_s": 1.526000005469541e-05, "gpu_peak_gb": 89.37899398803711}
22
  {"step": 762200, "loss": 0.5579586625099182, "lr": 3e-05, "grad_norm": 0.24301690088266478, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111563720181, "tokens_per_second": 152356.14392541448, "data_wait_s": 1.2263999906281242e-05, "gpu_peak_gb": 89.37899398803711}
23
  {"step": 762300, "loss": 0.558387279510498, "lr": 3e-05, "grad_norm": 0.28096033875982174, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111578230490, "tokens_per_second": 152199.34418891103, "data_wait_s": 1.0274000032950426e-05, "gpu_peak_gb": 89.37899398803711}
24
+ {"step": 762400, "loss": 0.5228051543235779, "lr": 3e-05, "grad_norm": 0.24500145724663877, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111592798188, "tokens_per_second": 152837.98531226395, "data_wait_s": 1.7794000086723827e-05, "gpu_peak_gb": 89.37899398803711}
runs/void-byte-v1/events.out.tfevents.1791254470.1e0c9aa3-e42e-441e-9e17-2e25895f438d-9bhv5.366.0.1791254470898546162-7724d9f6 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:7f8be0ffeec3ddc38737b482528c33cdb7c1882464ab6e8112e93396e58e3ab0
3
- size 48879
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3d237ada5d8f78a87484f0e030114fff9a7018ffa5f2fb5fabcfd6728385666f
3
+ size 50997