appvoid commited on
Commit
eb821d9
·
verified ·
1 Parent(s): 1396d9b

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": 76309.04021336343,
26
- "required_pages_per_second_upper_estimate": 0.006614274202200102
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": 9538.63002667043,
40
- "required_pages_per_second_upper_estimate": 0.003805102132866774
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": 19077.26005334086,
54
- "required_pages_per_second_upper_estimate": 0.004413188365030698
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": 28615.89008001129,
68
- "required_pages_per_second_upper_estimate": 0.001742557569900557
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": 28615.89008001129,
82
- "required_pages_per_second_upper_estimate": 0.0022029304322129226
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": 28615.89008001129,
96
- "required_pages_per_second_upper_estimate": 0.35531358357042464
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": 0.04529537349217705,
162
- "producer_p95_seconds": 0.12967073999971035,
163
- "gpu_update_seconds": 0.9661764817622298,
164
- "producer_batches_per_second": 22.077309952476927,
165
- "gpu_batches_per_second": 1.035007598380038,
166
- "producer_headroom": 21.33057765665842,
167
  "sustainable": true,
168
- "buffer_seconds": 60.869118351020475,
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": 75727.503948841,
26
+ "required_pages_per_second_upper_estimate": 0.006563868112681485
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": 9465.937993605125,
40
+ "required_pages_per_second_upper_estimate": 0.0037761041940342767
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": 18931.87598721025,
54
+ "required_pages_per_second_upper_estimate": 0.0043795563200035555
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": 28397.813980815372,
68
+ "required_pages_per_second_upper_estimate": 0.001729277879616393
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": 28397.813980815372,
82
+ "required_pages_per_second_upper_estimate": 0.0021861423304235417
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": 28397.813980815372,
96
+ "required_pages_per_second_upper_estimate": 0.35260580827216526
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": 0.0559557438515661,
162
+ "producer_p95_seconds": 0.1513643759999468,
163
+ "gpu_update_seconds": 0.9735960668901514,
164
+ "producer_batches_per_second": 17.871266310974292,
165
+ "gpu_batches_per_second": 1.0271200079866671,
166
+ "producer_headroom": 17.399394590711037,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.36295614718939,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 85,
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
@@ -29,3 +29,4 @@
29
  {"step": 277900, "loss": 0.5786179415881634, "lr": 0.0003, "grad_norm": 0.15734794308123032, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40676732679, "tokens_per_second": 151542.69303228127, "data_wait_s": 1.229699955729302e-05, "gpu_peak_gb": 85.46427917480469}
30
  {"step": 278000, "loss": 0.5908274631947279, "lr": 0.0003, "grad_norm": 0.14033135135848648, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40691426002, "tokens_per_second": 151521.56213670966, "data_wait_s": 1.3419999959296547e-05, "gpu_peak_gb": 85.46427917480469}
31
  {"step": 278100, "loss": 0.5920380316674709, "lr": 0.0003, "grad_norm": 0.1494044021047503, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40706113567, "tokens_per_second": 150710.19329023614, "data_wait_s": 1.1837999863928417e-05, "gpu_peak_gb": 85.46427917480469}
 
 
29
  {"step": 277900, "loss": 0.5786179415881634, "lr": 0.0003, "grad_norm": 0.15734794308123032, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40676732679, "tokens_per_second": 151542.69303228127, "data_wait_s": 1.229699955729302e-05, "gpu_peak_gb": 85.46427917480469}
30
  {"step": 278000, "loss": 0.5908274631947279, "lr": 0.0003, "grad_norm": 0.14033135135848648, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40691426002, "tokens_per_second": 151521.56213670966, "data_wait_s": 1.3419999959296547e-05, "gpu_peak_gb": 85.46427917480469}
31
  {"step": 278100, "loss": 0.5920380316674709, "lr": 0.0003, "grad_norm": 0.1494044021047503, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40706113567, "tokens_per_second": 150710.19329023614, "data_wait_s": 1.1837999863928417e-05, "gpu_peak_gb": 85.46427917480469}
32
+ {"step": 278200, "loss": 0.6077223289757967, "lr": 0.0003, "grad_norm": 0.1378774455084411, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40720756769, "tokens_per_second": 151397.99093341763, "data_wait_s": 1.0745000054157572e-05, "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:e8ffdaa7178b8b414775cafad880488ced1f924a8370976ffdc5d015f0bbf665
3
- size 65823
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7ad3cd8c9c632c609c712e79cd2521b9f297c68ee5027dd8f3c938b3322c2da2
3
+ size 67941