appvoid commited on
Commit
f61c488
·
verified ·
1 Parent(s): ab9f72e

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": 76165.46955987094,
26
- "required_pages_per_second_upper_estimate": 0.006601829861831867
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": 9520.683694983867,
40
- "required_pages_per_second_upper_estimate": 0.0037979430728354345
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": 19041.367389967734,
54
- "required_pages_per_second_upper_estimate": 0.004404885229048625
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": 28562.051084951603,
68
- "required_pages_per_second_upper_estimate": 0.0017392790575727986
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": 28562.051084951603,
82
- "required_pages_per_second_upper_estimate": 0.0021987857573373512
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": 28562.051084951603,
96
- "required_pages_per_second_upper_estimate": 0.3546450834393087
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": 0.051125705460950854,
162
  "producer_p95_seconds": 0.1299934950000079,
163
- "gpu_update_seconds": 0.9679977084897384,
164
- "producer_batches_per_second": 19.55963230206744,
165
- "gpu_batches_per_second": 1.0330602967647426,
166
- "producer_headroom": 18.933679247303147,
167
  "sustainable": true,
168
- "buffer_seconds": 60.01585792636378,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 60,
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": 76418.88780981897,
26
+ "required_pages_per_second_upper_estimate": 0.006623795513454681
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": 9552.360976227372,
40
+ "required_pages_per_second_upper_estimate": 0.0038105796139410292
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": 19104.721952454744,
54
+ "required_pages_per_second_upper_estimate": 0.004419541192077777
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": 28657.082928682117,
68
+ "required_pages_per_second_upper_estimate": 0.0017450659982624363
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": 28657.082928682117,
82
+ "required_pages_per_second_upper_estimate": 0.0022061015717327066
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": 28657.082928682117,
96
+ "required_pages_per_second_upper_estimate": 0.35582506088732035
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": 0.05022175149999519,
162
  "producer_p95_seconds": 0.1299934950000079,
163
+ "gpu_update_seconds": 0.9647876606564113,
164
+ "producer_batches_per_second": 19.911691052831873,
165
+ "gpu_batches_per_second": 1.0364975017607827,
166
+ "producer_headroom": 19.21055383057486,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.78162262135391,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 58,
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
@@ -7,3 +7,4 @@
7
  {"step": 355700, "loss": 0.5800438262522221, "lr": 0.0003, "grad_norm": 0.14151087830719344, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 52064543552, "tokens_per_second": 150241.3835507587, "data_wait_s": 1.281000004382804e-05, "gpu_peak_gb": 85.46427917480469}
8
  {"step": 355800, "loss": 0.6531054731458426, "lr": 0.0003, "grad_norm": 0.17719566197741718, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 52079015267, "tokens_per_second": 148980.5927603168, "data_wait_s": 1.1954000001423992e-05, "gpu_peak_gb": 85.46427917480469}
9
  {"step": 355900, "loss": 0.5476674549281597, "lr": 0.0003, "grad_norm": 0.1451582331959875, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 52093511172, "tokens_per_second": 149512.22858483213, "data_wait_s": 1.1758000027839444e-05, "gpu_peak_gb": 85.46427917480469}
 
 
7
  {"step": 355700, "loss": 0.5800438262522221, "lr": 0.0003, "grad_norm": 0.14151087830719344, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 52064543552, "tokens_per_second": 150241.3835507587, "data_wait_s": 1.281000004382804e-05, "gpu_peak_gb": 85.46427917480469}
8
  {"step": 355800, "loss": 0.6531054731458426, "lr": 0.0003, "grad_norm": 0.17719566197741718, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 52079015267, "tokens_per_second": 148980.5927603168, "data_wait_s": 1.1954000001423992e-05, "gpu_peak_gb": 85.46427917480469}
9
  {"step": 355900, "loss": 0.5476674549281597, "lr": 0.0003, "grad_norm": 0.1451582331959875, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 52093511172, "tokens_per_second": 149512.22858483213, "data_wait_s": 1.1758000027839444e-05, "gpu_peak_gb": 85.46427917480469}
10
+ {"step": 356000, "loss": 0.5913226958364248, "lr": 0.0003, "grad_norm": 0.16621256028653167, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 52108078318, "tokens_per_second": 149870.95399072525, "data_wait_s": 1.1811000149464235e-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 sha256:8a77aa0379dd7bc301eabc7a0255d555d1e3a2f67df061ca2eb3231d7e187817
3
- size 19227
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8954edc84b2cde0d0969acd08c27ce4276f77c21376f3c6642717e9f944a06ab
3
+ size 21345