appvoid commited on
Commit
db6e519
·
verified ·
1 Parent(s): f4b0eca

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": 77515.24693298797,
26
- "required_pages_per_second_upper_estimate": 0.005362684642976886
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": 9689.405866623496,
40
- "required_pages_per_second_upper_estimate": 0.004056751898552164
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": 19378.811733246992,
54
- "required_pages_per_second_upper_estimate": 0.004036095430403074
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": 29068.21759987049,
68
- "required_pages_per_second_upper_estimate": 0.0017425952605139125
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": 29068.21759987049,
82
- "required_pages_per_second_upper_estimate": 0.0022599260608131143
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": 29068.21759987049,
96
- "required_pages_per_second_upper_estimate": 0.37285588435077144
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": 0.05216270820316282,
162
- "producer_p95_seconds": 0.13161391600033312,
163
- "gpu_update_seconds": 0.9511419097166516,
164
- "producer_batches_per_second": 19.170783773442313,
165
- "gpu_batches_per_second": 1.0513678240693898,
166
- "producer_headroom": 18.234135889036917,
167
  "sustainable": true,
168
- "buffer_seconds": 60.8730822218657,
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": 77355.2088906546,
26
+ "required_pages_per_second_upper_estimate": 0.005351612839868844
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": 9669.401111331825,
40
+ "required_pages_per_second_upper_estimate": 0.004048376325258335
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": 19338.80222266365,
54
+ "required_pages_per_second_upper_estimate": 0.004027762504470312
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": 29008.203333995476,
68
+ "required_pages_per_second_upper_estimate": 0.0017389974969111834
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": 29008.203333995476,
82
+ "required_pages_per_second_upper_estimate": 0.0022552602156161897
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": 29008.203333995476,
96
+ "required_pages_per_second_upper_estimate": 0.37208608578642494
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": 0.046656947765640666,
162
+ "producer_p95_seconds": 0.13993321299994932,
163
+ "gpu_update_seconds": 0.9531097007858664,
164
+ "producer_batches_per_second": 21.433035118864435,
165
+ "gpu_batches_per_second": 1.0491971691983317,
166
+ "producer_headroom": 20.42803368907385,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.045911149509585,
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,
metrics/void-byte-v1/1791254470898546162-7724d9f6/train.jsonl CHANGED
@@ -23,3 +23,4 @@
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}
25
  {"step": 762500, "loss": 0.5541641116142273, "lr": 3e-05, "grad_norm": 0.2464639528453497, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111607380573, "tokens_per_second": 152839.79255499403, "data_wait_s": 1.392800004396122e-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}
25
  {"step": 762500, "loss": 0.5541641116142273, "lr": 3e-05, "grad_norm": 0.2464639528453497, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111607380573, "tokens_per_second": 152839.79255499403, "data_wait_s": 1.392800004396122e-05, "gpu_peak_gb": 89.37899398803711}
26
+ {"step": 762600, "loss": 0.5683630108833313, "lr": 3e-05, "grad_norm": 0.2581159766337038, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111621971646, "tokens_per_second": 152872.30040918954, "data_wait_s": 9.875000159809133e-06, "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:d1a813c06b1da42cdcf46b1b550a4abe7990e1131ddc48c76adfd443f0df993a
3
- size 53115
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3dfb7f10a18bcdb06ae5fc277588a0595b00aaf071a36b5e9e6325274ec93608
3
+ size 55233