appvoid commited on
Commit
9c5d5e0
·
verified ·
1 Parent(s): 6ca9163

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": 75854.16197691974,
26
- "required_pages_per_second_upper_estimate": 0.006574846509543513
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": 9481.770247114968,
40
- "required_pages_per_second_upper_estimate": 0.0037824199166726375
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": 18963.540494229936,
54
- "required_pages_per_second_upper_estimate": 0.004386881346426269
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": 28445.310741344903,
68
- "required_pages_per_second_upper_estimate": 0.0017321701831363979
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": 28445.310741344903,
82
- "required_pages_per_second_upper_estimate": 0.002189798762528553
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": 28445.310741344903,
96
- "required_pages_per_second_upper_estimate": 0.35319555907651023
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": 690670,
161
- "producer_batch_seconds": 0.05652020725776907,
162
- "producer_p95_seconds": 0.15567023499988863,
163
- "gpu_update_seconds": 0.9719703979121584,
164
- "producer_batches_per_second": 17.69278720864109,
165
- "gpu_batches_per_second": 1.0288379174386901,
166
- "producer_headroom": 17.196865423358027,
167
  "sustainable": true,
168
- "buffer_seconds": 60.26216467055382,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 86,
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": 76388.50942967138,
26
+ "required_pages_per_second_upper_estimate": 0.0066211623924567874
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": 9548.563678708922,
40
+ "required_pages_per_second_upper_estimate": 0.0038090648151862623
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": 19097.127357417845,
54
+ "required_pages_per_second_upper_estimate": 0.004417784316176299
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": 28645.691036126766,
68
+ "required_pages_per_second_upper_estimate": 0.0017443722917744565
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": 28645.691036126766,
82
+ "required_pages_per_second_upper_estimate": 0.0022052245923090187
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": 28645.691036126766,
96
+ "required_pages_per_second_upper_estimate": 0.3556836117079947
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": 690670,
161
+ "producer_batch_seconds": 0.06634447944529143,
162
+ "producer_p95_seconds": 0.17216890799954854,
163
+ "gpu_update_seconds": 0.9651713399104767,
164
+ "producer_batches_per_second": 15.072844166704387,
165
+ "gpu_batches_per_second": 1.036085468609909,
166
+ "producer_headroom": 14.547877200639885,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.80579441436003,
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
@@ -39,3 +39,4 @@
39
  {"step": 278900, "loss": 0.5618883538991213, "lr": 0.0003, "grad_norm": 0.1863066547620779, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40823246291, "tokens_per_second": 150619.76547717387, "data_wait_s": 1.0720999853219837e-05, "gpu_peak_gb": 85.46427917480469}
40
  {"step": 279000, "loss": 0.6319178156554699, "lr": 0.0003, "grad_norm": 0.17287498253791164, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40838042164, "tokens_per_second": 152059.73354522864, "data_wait_s": 1.053100004355656e-05, "gpu_peak_gb": 85.46427917480469}
41
  {"step": 279100, "loss": 0.5977139621973038, "lr": 0.0003, "grad_norm": 0.15597813372850852, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40852708011, "tokens_per_second": 151328.98478430815, "data_wait_s": 1.2958999832335394e-05, "gpu_peak_gb": 85.46427917480469}
 
 
39
  {"step": 278900, "loss": 0.5618883538991213, "lr": 0.0003, "grad_norm": 0.1863066547620779, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40823246291, "tokens_per_second": 150619.76547717387, "data_wait_s": 1.0720999853219837e-05, "gpu_peak_gb": 85.46427917480469}
40
  {"step": 279000, "loss": 0.6319178156554699, "lr": 0.0003, "grad_norm": 0.17287498253791164, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40838042164, "tokens_per_second": 152059.73354522864, "data_wait_s": 1.053100004355656e-05, "gpu_peak_gb": 85.46427917480469}
41
  {"step": 279100, "loss": 0.5977139621973038, "lr": 0.0003, "grad_norm": 0.15597813372850852, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40852708011, "tokens_per_second": 151328.98478430815, "data_wait_s": 1.2958999832335394e-05, "gpu_peak_gb": 85.46427917480469}
42
+ {"step": 279200, "loss": 0.6283826157450676, "lr": 0.0003, "grad_norm": 0.1496751405195431, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40867476963, "tokens_per_second": 152577.3529347542, "data_wait_s": 1.2758000593748875e-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 sha256:277974a277ec6bded38983d7288fd06bf7be0b2a5fd5ac2a3188fcc0938ead81
3
- size 87003
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:21422c67723d0ad77663b2e05bdacfbd4a98c743a7c0eeb396a906c132602dd6
3
+ size 89121