appvoid commited on
Commit
0688d1f
·
verified ·
1 Parent(s): e99c7e1

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": 75865.33291868574,
26
- "required_pages_per_second_upper_estimate": 0.0065758147784632974
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": 9483.166614835718,
40
- "required_pages_per_second_upper_estimate": 0.003782976948634003
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": 18966.333229671436,
54
- "required_pages_per_second_upper_estimate": 0.0043875273966201955
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": 28449.499844507154,
68
- "required_pages_per_second_upper_estimate": 0.0017324252775404532
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": 28449.499844507154,
82
- "required_pages_per_second_upper_estimate": 0.002190121251286163
83
  },
84
  {
85
  "source": "cortex",
@@ -92,8 +92,8 @@
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": 28449.499844507154,
96
- "required_pages_per_second_upper_estimate": 0.35324757371775894
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": 717798,
161
- "producer_batch_seconds": 0.04355434992973173,
162
- "producer_p95_seconds": 0.13079734600023585,
163
- "gpu_update_seconds": 0.9718272781986393,
164
- "producer_batches_per_second": 22.95981920550638,
165
- "gpu_batches_per_second": 1.0289894330333895,
166
- "producer_headroom": 22.31297860642011,
167
  "sustainable": true,
168
- "buffer_seconds": 60.25329124831564,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 72,
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": 76209.30229895536,
26
+ "required_pages_per_second_upper_estimate": 0.006605629172562643
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": 9526.16278736942,
40
+ "required_pages_per_second_upper_estimate": 0.0038001287647077627
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": 19052.32557473884,
54
+ "required_pages_per_second_upper_estimate": 0.004407420212237956
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": 28578.48836210826,
68
+ "required_pages_per_second_upper_estimate": 0.0017402800015119107
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": 28578.48836210826,
82
+ "required_pages_per_second_upper_estimate": 0.0022000511444341658
83
  },
84
  {
85
  "source": "cortex",
 
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": 28578.48836210826,
96
+ "required_pages_per_second_upper_estimate": 0.3548491794095665
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": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 717798,
161
+ "producer_batch_seconds": 0.053835043320304976,
162
+ "producer_p95_seconds": 0.13447191699970062,
163
+ "gpu_update_seconds": 0.9674409524283314,
164
+ "producer_batches_per_second": 18.57526135997052,
165
+ "gpu_batches_per_second": 1.0336548163378276,
166
+ "producer_headroom": 17.970468541695062,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.948780002984876,
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-v1/1790782192342215557-281f6532/train.jsonl CHANGED
@@ -17,3 +17,4 @@
17
  {"step": 356700, "loss": 0.6328282952308655, "lr": 0.0003, "grad_norm": 0.19027880083121018, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 52210646056, "tokens_per_second": 150503.8607225417, "data_wait_s": 7.793999884597724e-06, "gpu_peak_gb": 85.46427917480469}
18
  {"step": 356800, "loss": 0.5971382074058056, "lr": 0.0003, "grad_norm": 0.17656761639722304, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 52225208874, "tokens_per_second": 149479.63006219224, "data_wait_s": 1.575900023453869e-05, "gpu_peak_gb": 85.46427917480469}
19
  {"step": 356900, "loss": 0.6156456805765629, "lr": 0.0003, "grad_norm": 0.15816110123594962, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 52239903919, "tokens_per_second": 151397.92864944096, "data_wait_s": 1.3961000149720348e-05, "gpu_peak_gb": 85.46427917480469}
 
 
17
  {"step": 356700, "loss": 0.6328282952308655, "lr": 0.0003, "grad_norm": 0.19027880083121018, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 52210646056, "tokens_per_second": 150503.8607225417, "data_wait_s": 7.793999884597724e-06, "gpu_peak_gb": 85.46427917480469}
18
  {"step": 356800, "loss": 0.5971382074058056, "lr": 0.0003, "grad_norm": 0.17656761639722304, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 52225208874, "tokens_per_second": 149479.63006219224, "data_wait_s": 1.575900023453869e-05, "gpu_peak_gb": 85.46427917480469}
19
  {"step": 356900, "loss": 0.6156456805765629, "lr": 0.0003, "grad_norm": 0.15816110123594962, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 52239903919, "tokens_per_second": 151397.92864944096, "data_wait_s": 1.3961000149720348e-05, "gpu_peak_gb": 85.46427917480469}
20
+ {"step": 357000, "loss": 0.6048266664147377, "lr": 0.0003, "grad_norm": 0.15053552171825507, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 52254457624, "tokens_per_second": 149857.96183073396, "data_wait_s": 1.0407999980088789e-05, "gpu_peak_gb": 85.46427917480469}
runs/void-byte-v1/events.out.tfevents.1790782192.062ccfe6-664f-43fe-9e0e-62f43f0c4870-64tpq.429.0.1790782192342215557-281f6532 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:77e44cb6a1c469a1dca4bafa02f51fdfa27475e386b3c121d09d6f03b66f59bf
3
- size 40407
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aac8645296a7807df3f106e02dc67df32f965a4514e0e0526b103d138538084d
3
+ size 42525