appvoid commited on
Commit
0692b16
·
verified ·
1 Parent(s): ee07d8a

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": 76384.46895796638,
26
- "required_pages_per_second_upper_estimate": 0.006620812174609881
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": 9548.058619745798,
40
- "required_pages_per_second_upper_estimate": 0.0038088633396145674
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": 19096.117239491596,
54
- "required_pages_per_second_upper_estimate": 0.004417550643171529
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": 28644.17585923739,
68
- "required_pages_per_second_upper_estimate": 0.0017442800254514122
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": 28644.17585923739,
82
- "required_pages_per_second_upper_estimate": 0.0022051079497977964
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": 28644.17585923739,
96
- "required_pages_per_second_upper_estimate": 0.3556647982819995
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": 0.05190601803906958,
162
- "producer_p95_seconds": 0.14367834199993013,
163
- "gpu_update_seconds": 0.9652223941043799,
164
- "producer_batches_per_second": 19.26558880412097,
165
- "gpu_batches_per_second": 1.0360306662050562,
166
- "producer_headroom": 18.59557774934418,
167
  "sustainable": true,
168
- "buffer_seconds": 60.809010828575936,
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,
 
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": 76119.66301634122,
26
+ "required_pages_per_second_upper_estimate": 0.0065978594667343345
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": 9514.957877042652,
40
+ "required_pages_per_second_upper_estimate": 0.0037956589584500766
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": 19029.915754085305,
54
+ "required_pages_per_second_upper_estimate": 0.004402236094629134
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": 28544.873631127954,
68
+ "required_pages_per_second_upper_estimate": 0.0017382330407580736
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": 28544.873631127954,
82
+ "required_pages_per_second_upper_estimate": 0.0021974633893917708
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": 28544.873631127954,
96
+ "required_pages_per_second_upper_estimate": 0.35443179695205873
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": 0.049429777875019454,
162
+ "producer_p95_seconds": 0.1342726110001422,
163
+ "gpu_update_seconds": 0.9685802206477481,
164
+ "producer_batches_per_second": 20.230720083922822,
165
+ "gpu_batches_per_second": 1.032439005755496,
166
+ "producer_headroom": 19.595075322748794,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.05197368016038,
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,
metrics/void-byte-v1/1790696019647648547-3fdea0a8/train.jsonl CHANGED
@@ -17,3 +17,4 @@
17
  {"step": 276700, "loss": 0.6008296497166157, "lr": 0.0003, "grad_norm": 0.15391543790461298, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40500433646, "tokens_per_second": 152339.59975473586, "data_wait_s": 1.401899999109446e-05, "gpu_peak_gb": 85.46427917480469}
18
  {"step": 276800, "loss": 0.6234837397933006, "lr": 0.0003, "grad_norm": 0.17697912234679625, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40515274660, "tokens_per_second": 153003.5590317808, "data_wait_s": 7.983000159583753e-06, "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}
 
 
17
  {"step": 276700, "loss": 0.6008296497166157, "lr": 0.0003, "grad_norm": 0.15391543790461298, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40500433646, "tokens_per_second": 152339.59975473586, "data_wait_s": 1.401899999109446e-05, "gpu_peak_gb": 85.46427917480469}
18
  {"step": 276800, "loss": 0.6234837397933006, "lr": 0.0003, "grad_norm": 0.17697912234679625, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40515274660, "tokens_per_second": 153003.5590317808, "data_wait_s": 7.983000159583753e-06, "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}
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 sha256:80b73f43d0996a01e99364bda274f64c12ff1dbd59064f7585e9cfc0ebf3238c
3
- size 40407
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4593a1a6c6268f8394d2d2912f85ffdf3740835a1965efbe5f3af961f01fef56
3
+ size 42525