scripts_full/stage6.log

##### b16v2: base p16m, 4 epochs, ex/full62.jsonl
{'train_runtime': 290.9617, 'train_samples_per_second': 147.593, 'train_steps_per_second': 4.619, 'train_loss': 0.04253844635206319, 'epoch': 4.0}
saved -> b16v2
The tokenizer you are loading from '/root/mt92/b16v2' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue.
The following generation flags are not valid and may be ignored: ['temperature', 'top_p', 'top_k']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
=== b16v2 on val_all.jsonl (4710 docs, 159s) ===
  MICRO-F1 0.9617 | P 0.9610 R 0.9624 | tp/fp/fn 8583/348/335 | malformed 4 | out tok med 34 max 256
The tokenizer you are loading from '/root/mt92/b16v2' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue.
The following generation flags are not valid and may be ignored: ['temperature', 'top_p', 'top_k']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
=== b16v2 on test_all.jsonl (5017 docs, 166s) ===
  MICRO-F1 0.9615 | P 0.9590 R 0.9641 | tp/fp/fn 8826/377/329 | malformed 2 | out tok med 33 max 256
b16v2 backed up
{ "artifact": "onnx_b16v2/final", "ttft_p50": 626, "ttft_p95": 637, "tok_s": 18.99, "cold_ms": 4599, "rss_MiB": 818, "size_MiB": 552, "requests": 68, "failures": 0, "wall_s": 65.4, "device": "DeviceProfile(machine='x86_64', system='linux', cpu_cores=20"}
staged onnx_b16v2
STAGE6_DONE