appvoid commited on
Commit
ec907e2
·
verified ·
1 Parent(s): 66a9c08

Update TensorBoard training metrics

Browse files
benchmark_cache/shard_scheduler.json CHANGED
@@ -22,8 +22,8 @@
22
  "tokens_per_second": 598315.063954628,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
- "required_token_positions_per_second": 76829.40532140016,
26
- "required_pages_per_second_upper_estimate": 0.005315236528915454
27
  },
28
  {
29
  "source": "rewrite6",
@@ -36,8 +36,8 @@
36
  "tokens_per_second": 8533.957213384612,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
- "required_token_positions_per_second": 9603.67566517502,
40
- "required_pages_per_second_upper_estimate": 0.00402085845345587
41
  },
42
  {
43
  "source": "openmath",
@@ -50,8 +50,8 @@
50
  "tokens_per_second": 380144.35761751566,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
- "required_token_positions_per_second": 19207.35133035004,
54
- "required_pages_per_second_upper_estimate": 0.004000384750194536
55
  },
56
  {
57
  "source": "ultra_qa",
@@ -64,8 +64,8 @@
64
  "tokens_per_second": 43742.16556775435,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
- "required_token_positions_per_second": 28811.026995525062,
68
- "required_pages_per_second_upper_estimate": 0.0017271770764907184
69
  },
70
  {
71
  "source": "ultra_style",
@@ -78,8 +78,8 @@
78
  "tokens_per_second": 22018.86276523721,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
- "required_token_positions_per_second": 28811.026995525062,
82
- "required_pages_per_second_upper_estimate": 0.00223993062258717
83
  },
84
  {
85
  "source": "cortex",
@@ -91,9 +91,9 @@
91
  "read_seconds": 0.04061180899998362,
92
  "tokens_per_second": 910315.184299418,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
- "estimated_pages_for_buffer": 23,
95
- "required_token_positions_per_second": 28811.026995525062,
96
- "required_pages_per_second_upper_estimate": 0.369556919427984
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": 2.558590814491007,
157
  "selected": {
158
- "prefetch_batches": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
- "producer_batch_seconds": 0.05445190348437379,
162
- "producer_p95_seconds": 0.12841596800035404,
163
- "gpu_update_seconds": 0.9596325741631597,
164
- "producer_batches_per_second": 18.364830905993447,
165
- "gpu_batches_per_second": 1.042065501863609,
166
- "producer_headroom": 17.623489956389644,
167
  "sustainable": true,
168
- "buffer_seconds": 60.45685217227906,
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,
 
22
  "tokens_per_second": 598315.063954628,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
+ "required_token_positions_per_second": 75467.24499801552,
26
+ "required_pages_per_second_upper_estimate": 0.005220999117096302
27
  },
28
  {
29
  "source": "rewrite6",
 
36
  "tokens_per_second": 8533.957213384612,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
+ "required_token_positions_per_second": 9433.40562475194,
40
+ "required_pages_per_second_upper_estimate": 0.003949569943173496
41
  },
42
  {
43
  "source": "openmath",
 
50
  "tokens_per_second": 380144.35761751566,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
+ "required_token_positions_per_second": 18866.81124950388,
54
+ "required_pages_per_second_upper_estimate": 0.003929459232000135
55
  },
56
  {
57
  "source": "ultra_qa",
 
64
  "tokens_per_second": 43742.16556775435,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
+ "required_token_positions_per_second": 28300.21687425582,
68
+ "required_pages_per_second_upper_estimate": 0.0016965547896825213
69
  },
70
  {
71
  "source": "ultra_style",
 
78
  "tokens_per_second": 22018.86276523721,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
+ "required_token_positions_per_second": 28300.21687425582,
82
+ "required_pages_per_second_upper_estimate": 0.0022002173824747593
83
  },
84
  {
85
  "source": "cortex",
 
91
  "read_seconds": 0.04061180899998362,
92
  "tokens_per_second": 910315.184299418,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
+ "estimated_pages_for_buffer": 22,
95
+ "required_token_positions_per_second": 28300.21687425582,
96
+ "required_pages_per_second_upper_estimate": 0.36300479565751875
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": 2.558590814491007,
157
  "selected": {
158
+ "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
+ "producer_batch_seconds": 0.05197926053122259,
162
+ "producer_p95_seconds": 0.13209945799962952,
163
+ "gpu_update_seconds": 0.9769536439542578,
164
+ "producer_batches_per_second": 19.2384422129154,
165
+ "gpu_batches_per_second": 1.0235900200468686,
166
+ "producer_headroom": 18.795066223911114,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.57112592516398,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 80,
172
  "max_pending_pages": 4096,
173
  "generation_bytes_per_second": 0,
174
  "upload_bytes_per_second": null,
metrics/void-byte-v1/1791209746770022608-f3fe5fb7/train.jsonl CHANGED
@@ -18,3 +18,4 @@
18
  {"step": 721800, "loss": 0.597263753414154, "lr": 5e-05, "grad_norm": 0.21663511348799927, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105651647851, "tokens_per_second": 153471.2782438447, "data_wait_s": 1.55860002450936e-05, "gpu_peak_gb": 89.37899398803711}
19
  {"step": 721900, "loss": 0.5127010941505432, "lr": 5e-05, "grad_norm": 0.2314292797174232, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105666361451, "tokens_per_second": 153174.7397481044, "data_wait_s": 7.755999831715599e-06, "gpu_peak_gb": 89.37899398803711}
20
  {"step": 722000, "loss": 0.5648269057273865, "lr": 5e-05, "grad_norm": 0.2197077038734712, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105681078705, "tokens_per_second": 152364.29076591463, "data_wait_s": 1.9053999949392164e-05, "gpu_peak_gb": 89.37899398803711}
 
 
18
  {"step": 721800, "loss": 0.597263753414154, "lr": 5e-05, "grad_norm": 0.21663511348799927, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105651647851, "tokens_per_second": 153471.2782438447, "data_wait_s": 1.55860002450936e-05, "gpu_peak_gb": 89.37899398803711}
19
  {"step": 721900, "loss": 0.5127010941505432, "lr": 5e-05, "grad_norm": 0.2314292797174232, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105666361451, "tokens_per_second": 153174.7397481044, "data_wait_s": 7.755999831715599e-06, "gpu_peak_gb": 89.37899398803711}
20
  {"step": 722000, "loss": 0.5648269057273865, "lr": 5e-05, "grad_norm": 0.2197077038734712, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105681078705, "tokens_per_second": 152364.29076591463, "data_wait_s": 1.9053999949392164e-05, "gpu_peak_gb": 89.37899398803711}
21
+ {"step": 722100, "loss": 0.5536484122276306, "lr": 5e-05, "grad_norm": 0.21087483701386997, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105695631662, "tokens_per_second": 151489.05163858933, "data_wait_s": 8.944000001065433e-06, "gpu_peak_gb": 89.37899398803711}
runs/void-byte-v1/events.out.tfevents.1791209746.ebd20fd8-0561-42d8-bf2b-56d08e7106ef-6qqk2.329.0.1791209746770022608-f3fe5fb7 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:f5588aa3e84735b250a9a7786e6cfd19c01f13c01352c9a6e306466cb6453238
3
- size 42525
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3505bf62df65d20f162f6d171bdfc7a0ac08c5d2605888203d0fc0fe7f9742ac
3
+ size 44643