appvoid commited on
Commit
7dce055
·
verified ·
1 Parent(s): 9e1987c

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": 76197.96714523742,
26
- "required_pages_per_second_upper_estimate": 0.006604646670167057
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": 9524.745893154677,
40
- "required_pages_per_second_upper_estimate": 0.0037995635444210457
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": 19049.491786309354,
54
- "required_pages_per_second_upper_estimate": 0.004406764665157767
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": 28574.23767946403,
68
- "required_pages_per_second_upper_estimate": 0.0017400211572404854
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": 28574.23767946403,
82
- "required_pages_per_second_upper_estimate": 0.0021997239151175705
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": 28574.23767946403,
96
- "required_pages_per_second_upper_estimate": 0.35479640015724484
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": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 690670,
161
- "producer_batch_seconds": 0.05407736080466563,
162
- "producer_p95_seconds": 0.1416256800002884,
163
- "gpu_update_seconds": 0.9675848682350079,
164
- "producer_batches_per_second": 18.492026702488836,
165
- "gpu_batches_per_second": 1.0335010734759849,
166
- "producer_headroom": 17.89260522032591,
167
  "sustainable": true,
168
- "buffer_seconds": 60.9578466988055,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 82,
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": 75487.6494258315,
26
+ "required_pages_per_second_upper_estimate": 0.006543078130532753
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": 9435.956178228938,
40
+ "required_pages_per_second_upper_estimate": 0.003764143999612629
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": 18871.912356457877,
54
+ "required_pages_per_second_upper_estimate": 0.004365684789352851
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": 28307.868534686815,
68
+ "required_pages_per_second_upper_estimate": 0.0017238006738544487
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": 28307.868534686815,
82
+ "required_pages_per_second_upper_estimate": 0.002179218080995635
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": 28307.868534686815,
96
+ "required_pages_per_second_upper_estimate": 0.3514889868592922
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": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 690670,
161
+ "producer_batch_seconds": 0.0579027166952919,
162
+ "producer_p95_seconds": 0.13588079900000594,
163
+ "gpu_update_seconds": 0.9766895718807563,
164
+ "producer_batches_per_second": 17.270346834716143,
165
+ "gpu_batches_per_second": 1.0238667728112998,
166
+ "producer_headroom": 16.867767656231084,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.554753456606896,
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,
metrics/void-byte-v1/1790696019647648547-3fdea0a8/train.jsonl CHANGED
@@ -19,3 +19,4 @@
19
  {"step": 276900, "loss": 0.6361541375517845, "lr": 0.0003, "grad_norm": 0.14672465720243766, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40530044420, "tokens_per_second": 152390.50278827036, "data_wait_s": 1.304199986407184e-05, "gpu_peak_gb": 85.46427917480469}
20
  {"step": 277000, "loss": 0.6080015897750854, "lr": 0.0003, "grad_norm": 0.1453749490976164, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40544820615, "tokens_per_second": 152366.88545407142, "data_wait_s": 1.2525999864010373e-05, "gpu_peak_gb": 85.46427917480469}
21
  {"step": 277100, "loss": 0.561845388263464, "lr": 0.0003, "grad_norm": 0.13909139895340553, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40559593939, "tokens_per_second": 151182.7518431791, "data_wait_s": 1.3313000181369716e-05, "gpu_peak_gb": 85.46427917480469}
 
 
19
  {"step": 276900, "loss": 0.6361541375517845, "lr": 0.0003, "grad_norm": 0.14672465720243766, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40530044420, "tokens_per_second": 152390.50278827036, "data_wait_s": 1.304199986407184e-05, "gpu_peak_gb": 85.46427917480469}
20
  {"step": 277000, "loss": 0.6080015897750854, "lr": 0.0003, "grad_norm": 0.1453749490976164, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40544820615, "tokens_per_second": 152366.88545407142, "data_wait_s": 1.2525999864010373e-05, "gpu_peak_gb": 85.46427917480469}
21
  {"step": 277100, "loss": 0.561845388263464, "lr": 0.0003, "grad_norm": 0.13909139895340553, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40559593939, "tokens_per_second": 151182.7518431791, "data_wait_s": 1.3313000181369716e-05, "gpu_peak_gb": 85.46427917480469}
22
+ {"step": 277200, "loss": 0.6728517189621925, "lr": 0.0003, "grad_norm": 0.1942734196814653, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40574057787, "tokens_per_second": 149074.59816718844, "data_wait_s": 9.401000170328189e-06, "gpu_peak_gb": 85.46427917480469}
runs/void-byte-v1/events.out.tfevents.1790696019.3e6d470a-8cc7-4d7e-a224-51a0f9a2ce55-85btc.22407.0.1790696019647648547-3fdea0a8 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b08001cd975307098ea8ec9c8b58816c12e68bcc145badd3ad3feabf6b0216b9
3
- size 44643
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:94bc1ff179e96d2e561631200db1101ddf2cfe43524fc0c63c1fccca6565aa71
3
+ size 46761