appvoid commited on
Commit
0a3f072
·
verified ·
1 Parent(s): 2d3500a

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": 76276.96952203262,
26
- "required_pages_per_second_upper_estimate": 0.006611494395957972
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": 9534.621190254078,
40
- "required_pages_per_second_upper_estimate": 0.003803502948082846
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": 19069.242380508156,
54
- "required_pages_per_second_upper_estimate": 0.0044113336175270745
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": 28603.86357076223,
68
- "required_pages_per_second_upper_estimate": 0.001741825218585507
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": 28603.86357076223,
82
- "required_pages_per_second_upper_estimate": 0.002202004598239422
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": 28603.86357076223,
96
- "required_pages_per_second_upper_estimate": 0.35516425457568856
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": 717798,
161
- "producer_batch_seconds": 0.04877243845313828,
162
- "producer_p95_seconds": 0.13781094100022528,
163
- "gpu_update_seconds": 0.9665827111642611,
164
- "producer_batches_per_second": 20.50338329835249,
165
- "gpu_batches_per_second": 1.0345726117897218,
166
- "producer_headroom": 19.818215816561576,
167
  "sustainable": true,
168
- "buffer_seconds": 60.894710803348445,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 83,
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": 75833.71175771797,
26
+ "required_pages_per_second_upper_estimate": 0.006573073936373711
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": 9479.213969714747,
40
+ "required_pages_per_second_upper_estimate": 0.0037814001793979365
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": 18958.427939429494,
54
+ "required_pages_per_second_upper_estimate": 0.004385698646851079
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": 28437.64190914424,
68
+ "required_pages_per_second_upper_estimate": 0.0017317031914906304
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": 28437.64190914424,
82
+ "required_pages_per_second_upper_estimate": 0.00218920839459706
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": 28437.64190914424,
96
+ "required_pages_per_second_upper_estimate": 0.3531003378465083
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": 717798,
161
+ "producer_batch_seconds": 0.050841793640614696,
162
+ "producer_p95_seconds": 0.14711638700009644,
163
+ "gpu_update_seconds": 0.9722325109913447,
164
+ "producer_batches_per_second": 19.668857614833545,
165
+ "gpu_batches_per_second": 1.0285605435888396,
166
+ "producer_headroom": 19.122702827200847,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.27841568146337,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 79,
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
@@ -19,3 +19,4 @@
19
  {"step": 356900, "loss": 0.6156456805765629, "lr": 0.0003, "grad_norm": 0.15816110123594962, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 52239903919, "tokens_per_second": 151397.92864944096, "data_wait_s": 1.3961000149720348e-05, "gpu_peak_gb": 85.46427917480469}
20
  {"step": 357000, "loss": 0.6048266664147377, "lr": 0.0003, "grad_norm": 0.15053552171825507, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 52254457624, "tokens_per_second": 149857.96183073396, "data_wait_s": 1.0407999980088789e-05, "gpu_peak_gb": 85.46427917480469}
21
  {"step": 357100, "loss": 0.5407655872404575, "lr": 0.0003, "grad_norm": 0.15601018103178194, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 52269172951, "tokens_per_second": 150804.18442888864, "data_wait_s": 1.0480000128154643e-05, "gpu_peak_gb": 85.46427917480469}
 
 
19
  {"step": 356900, "loss": 0.6156456805765629, "lr": 0.0003, "grad_norm": 0.15816110123594962, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 52239903919, "tokens_per_second": 151397.92864944096, "data_wait_s": 1.3961000149720348e-05, "gpu_peak_gb": 85.46427917480469}
20
  {"step": 357000, "loss": 0.6048266664147377, "lr": 0.0003, "grad_norm": 0.15053552171825507, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 52254457624, "tokens_per_second": 149857.96183073396, "data_wait_s": 1.0407999980088789e-05, "gpu_peak_gb": 85.46427917480469}
21
  {"step": 357100, "loss": 0.5407655872404575, "lr": 0.0003, "grad_norm": 0.15601018103178194, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 52269172951, "tokens_per_second": 150804.18442888864, "data_wait_s": 1.0480000128154643e-05, "gpu_peak_gb": 85.46427917480469}
22
+ {"step": 357200, "loss": 0.6043527461588383, "lr": 0.0003, "grad_norm": 0.16037414751564252, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 52283949229, "tokens_per_second": 151956.113868134, "data_wait_s": 1.650200010772096e-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 sha256:367b850723aba8e549f8d4d815cc6eaaa9224634e68acb06c0220ef5b0282bda
3
- size 44643
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0ec3707d93d11c8c0594f5c6256282fbbaec6ca83ddf68364d1001757cdbfc66
3
+ size 46761