appvoid commited on
Commit
c5a2ec0
·
verified ·
1 Parent(s): 0dcf3e9

Update TensorBoard training metrics

Browse files
benchmark_cache/shard_scheduler.json CHANGED
@@ -22,8 +22,8 @@
22
  "tokens_per_second": 1062967.6168812835,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
- "required_token_positions_per_second": 75624.81359048597,
26
- "required_pages_per_second_upper_estimate": 0.006554967172685605
27
  },
28
  {
29
  "source": "rewrite6",
@@ -36,8 +36,8 @@
36
  "tokens_per_second": 1206672.263377472,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
- "required_token_positions_per_second": 9453.101698810746,
40
- "required_pages_per_second_upper_estimate": 0.003770983604121089
41
  },
42
  {
43
  "source": "openmath",
@@ -50,8 +50,8 @@
50
  "tokens_per_second": 410042.7581134458,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
- "required_token_positions_per_second": 18906.20339762149,
54
- "required_pages_per_second_upper_estimate": 0.004373617418224341
55
  },
56
  {
57
  "source": "ultra_qa",
@@ -64,8 +64,8 @@
64
  "tokens_per_second": 2163486.5624068812,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
- "required_token_positions_per_second": 28359.305096432236,
68
- "required_pages_per_second_upper_estimate": 0.001726932890597962
69
  },
70
  {
71
  "source": "ultra_style",
@@ -78,8 +78,8 @@
78
  "tokens_per_second": 1546694.2737633225,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
- "required_token_positions_per_second": 28359.305096432236,
82
- "required_pages_per_second_upper_estimate": 0.002183177809904313
83
  },
84
  {
85
  "source": "cortex",
@@ -91,9 +91,9 @@
91
  "read_seconds": 0.042780299000014566,
92
  "tokens_per_second": 1015177.8395898147,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
- "estimated_pages_for_buffer": 22,
95
- "required_token_positions_per_second": 28359.305096432236,
96
- "required_pages_per_second_upper_estimate": 0.35212765680907204
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": 0.1434574083060457,
157
  "selected": {
158
- "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 688946,
161
- "producer_batch_seconds": 0.05614438421092771,
162
- "producer_p95_seconds": 0.1496411060002174,
163
- "gpu_update_seconds": 0.9749181055736897,
164
- "producer_batches_per_second": 17.811220375008837,
165
- "gpu_batches_per_second": 1.025727180860541,
166
- "producer_headroom": 17.364481225959118,
167
  "sustainable": true,
168
- "buffer_seconds": 60.444922545568765,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 78,
172
  "max_pending_pages": 4096,
173
  "generation_bytes_per_second": 0,
174
  "upload_bytes_per_second": null,
 
22
  "tokens_per_second": 1062967.6168812835,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
+ "required_token_positions_per_second": 74872.98955274138,
26
+ "required_pages_per_second_upper_estimate": 0.006489800970574507
27
  },
28
  {
29
  "source": "rewrite6",
 
36
  "tokens_per_second": 1206672.263377472,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
+ "required_token_positions_per_second": 9359.123694092672,
40
+ "required_pages_per_second_upper_estimate": 0.0037334943729426646
41
  },
42
  {
43
  "source": "openmath",
 
50
  "tokens_per_second": 410042.7581134458,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
+ "required_token_positions_per_second": 18718.247388185344,
54
+ "required_pages_per_second_upper_estimate": 0.004330137103354076
55
  },
56
  {
57
  "source": "ultra_qa",
 
64
  "tokens_per_second": 2163486.5624068812,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
+ "required_token_positions_per_second": 28077.371082278016,
68
+ "required_pages_per_second_upper_estimate": 0.0017097645883294241
69
  },
70
  {
71
  "source": "ultra_style",
 
78
  "tokens_per_second": 1546694.2737633225,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
+ "required_token_positions_per_second": 28077.371082278016,
82
+ "required_pages_per_second_upper_estimate": 0.0021614737490514195
83
  },
84
  {
85
  "source": "cortex",
 
91
  "read_seconds": 0.042780299000014566,
92
  "tokens_per_second": 1015177.8395898147,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
+ "estimated_pages_for_buffer": 21,
95
+ "required_token_positions_per_second": 28077.371082278016,
96
+ "required_pages_per_second_upper_estimate": 0.3486269799257238
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": 0.1434574083060457,
157
  "selected": {
158
+ "prefetch_batches": 61,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 688946,
161
+ "producer_batch_seconds": 0.055460317054674135,
162
+ "producer_p95_seconds": 0.1378342259999954,
163
+ "gpu_update_seconds": 0.984707575327484,
164
+ "producer_batches_per_second": 18.030910263534476,
165
+ "gpu_batches_per_second": 1.0155299147235972,
166
+ "producer_headroom": 17.75517392655248,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.067162094976524,
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,
metrics/void-byte-v1/1790606918076636751-3952f1c5/train.jsonl CHANGED
@@ -26,3 +26,4 @@
26
  {"step": 197600, "loss": 0.5868651624768972, "lr": 0.0003, "grad_norm": 0.17262299639540224, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28920839583, "tokens_per_second": 149402.8234699391, "data_wait_s": 1.622199988560169e-05, "gpu_peak_gb": 85.46427917480469}
27
  {"step": 197700, "loss": 0.6026234813034534, "lr": 0.0003, "grad_norm": 0.15512142850393992, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28935300702, "tokens_per_second": 147646.94076719318, "data_wait_s": 1.7273000139539363e-05, "gpu_peak_gb": 85.46427917480469}
28
  {"step": 197800, "loss": 0.5235539898276329, "lr": 0.0003, "grad_norm": 0.149242034860401, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28949997669, "tokens_per_second": 150145.07525097489, "data_wait_s": 1.1808000181190437e-05, "gpu_peak_gb": 85.46427917480469}
 
 
26
  {"step": 197600, "loss": 0.5868651624768972, "lr": 0.0003, "grad_norm": 0.17262299639540224, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28920839583, "tokens_per_second": 149402.8234699391, "data_wait_s": 1.622199988560169e-05, "gpu_peak_gb": 85.46427917480469}
27
  {"step": 197700, "loss": 0.6026234813034534, "lr": 0.0003, "grad_norm": 0.15512142850393992, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28935300702, "tokens_per_second": 147646.94076719318, "data_wait_s": 1.7273000139539363e-05, "gpu_peak_gb": 85.46427917480469}
28
  {"step": 197800, "loss": 0.5235539898276329, "lr": 0.0003, "grad_norm": 0.149242034860401, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28949997669, "tokens_per_second": 150145.07525097489, "data_wait_s": 1.1808000181190437e-05, "gpu_peak_gb": 85.46427917480469}
29
+ {"step": 197900, "loss": 0.5999762620776892, "lr": 0.0003, "grad_norm": 0.14391978348168138, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 28964748234, "tokens_per_second": 150920.02659706326, "data_wait_s": 1.72149998434179e-05, "gpu_peak_gb": 85.46427917480469}
runs/void-byte-v1/events.out.tfevents.1790606918.0a7f8bdc-23d3-4ecb-bd0c-4a4bf460026e-rl2cx.1192.0.1790606918076636751-3952f1c5 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:815a2dd026143da7cfc2b865c5a23487f23a08efa20bd6ee1426d9221c35e08d
3
- size 59469
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5d5b4eb032fa81593a8d5f975f3680ed51840e38531073919667eadfc0cf756f
3
+ size 61587