appvoid commited on
Commit
de7c20e
·
verified ·
1 Parent(s): a59f32a

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": 76084.04838430988,
26
- "required_pages_per_second_upper_estimate": 0.006594772480694317
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": 9510.506048038735,
40
- "required_pages_per_second_upper_estimate": 0.003793883057299639
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": 19021.01209607747,
54
- "required_pages_per_second_upper_estimate": 0.00440017639004805
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": 28531.518144116206,
68
- "required_pages_per_second_upper_estimate": 0.001737419761669883
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": 28531.518144116206,
82
- "required_pages_per_second_upper_estimate": 0.0021964352470312523
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": 28531.518144116206,
96
- "required_pages_per_second_upper_estimate": 0.35426596650131253
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": 0.05652759409375108,
162
- "producer_p95_seconds": 0.13663131199996315,
163
- "gpu_update_seconds": 0.9690336090896584,
164
- "producer_batches_per_second": 17.690475174681925,
165
- "gpu_batches_per_second": 1.031955951393092,
166
- "producer_headroom": 17.14266500503303,
167
  "sustainable": true,
168
- "buffer_seconds": 60.08008376355882,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 67,
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": 76299.06374361439,
26
+ "required_pages_per_second_upper_estimate": 0.006613409467087387
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": 9537.382967951798,
40
+ "required_pages_per_second_upper_estimate": 0.003804604662498723
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": 19074.765935903597,
54
+ "required_pages_per_second_upper_estimate": 0.004412611394856555
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": 28612.148903855395,
68
+ "required_pages_per_second_upper_estimate": 0.0017423297519010973
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": 28612.148903855395,
82
+ "required_pages_per_second_upper_estimate": 0.0022026424261162034
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": 28612.148903855395,
96
+ "required_pages_per_second_upper_estimate": 0.3552671306834796
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": 688946,
161
+ "producer_batch_seconds": 0.04763045408594735,
162
+ "producer_p95_seconds": 0.13063349799995194,
163
+ "gpu_update_seconds": 0.9663028139866321,
164
+ "producer_batches_per_second": 20.994970952734104,
165
+ "gpu_batches_per_second": 1.0348722838489366,
166
+ "producer_headroom": 20.287499511194568,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.877077281157824,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 68,
172
  "max_pending_pages": 4096,
173
  "generation_bytes_per_second": 0,
174
  "upload_bytes_per_second": null,
metrics/void-byte-v1/1790519810984249790-cb9c9513/train.jsonl CHANGED
@@ -12,3 +12,4 @@
12
  {"step": 116200, "loss": 0.6335533894598484, "lr": 0.0003, "grad_norm": 0.14737405640119294, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17004101407, "tokens_per_second": 149923.8179464614, "data_wait_s": 1.2765999827024643e-05, "gpu_peak_gb": 85.46427917480469}
13
  {"step": 116300, "loss": 0.6132048312574625, "lr": 0.0003, "grad_norm": 0.1570570949934492, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17018720502, "tokens_per_second": 150144.02422067648, "data_wait_s": 1.3435000028039212e-05, "gpu_peak_gb": 85.46427917480469}
14
  {"step": 116400, "loss": 0.5653194822371006, "lr": 0.0003, "grad_norm": 0.13992974011068288, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17033279015, "tokens_per_second": 150077.5783915966, "data_wait_s": 9.144000159722054e-06, "gpu_peak_gb": 85.46427917480469}
 
 
12
  {"step": 116200, "loss": 0.6335533894598484, "lr": 0.0003, "grad_norm": 0.14737405640119294, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17004101407, "tokens_per_second": 149923.8179464614, "data_wait_s": 1.2765999827024643e-05, "gpu_peak_gb": 85.46427917480469}
13
  {"step": 116300, "loss": 0.6132048312574625, "lr": 0.0003, "grad_norm": 0.1570570949934492, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17018720502, "tokens_per_second": 150144.02422067648, "data_wait_s": 1.3435000028039212e-05, "gpu_peak_gb": 85.46427917480469}
14
  {"step": 116400, "loss": 0.5653194822371006, "lr": 0.0003, "grad_norm": 0.13992974011068288, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17033279015, "tokens_per_second": 150077.5783915966, "data_wait_s": 9.144000159722054e-06, "gpu_peak_gb": 85.46427917480469}
15
+ {"step": 116500, "loss": 0.6138330940157175, "lr": 0.0003, "grad_norm": 0.13738930956550519, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17047979953, "tokens_per_second": 151677.26381402943, "data_wait_s": 1.4385999747901224e-05, "gpu_peak_gb": 85.46427917480469}
runs/void-byte-v1/events.out.tfevents.1790519810.cad0b19a-59f2-43f0-8fb4-2a94097843e2-tjh7c.1336.0.1790519810984249790-cb9c9513 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:c0aa190927aee46a317c12fa0ca8e1efc324d12cbc38808bab3e2732b5af82f0
3
- size 29817
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:db3670f7b80004dc1f3b158000b2fac82ab6f6d71ef2da9489987295e66262c7
3
+ size 31935