== training exited 20:22
step 123/124  loss 0.0204  191.7s/step  ~0.1 h left
step 124/124  loss 0.0152  141.9s/step  ~0.0 h left
Final validation loss: 0.0173
Adapter saved to /Users/umer/Documents/6 - mix/endpoint-audit/training/models/auditor-lora/final
== base model 20:22
/Users/umer/Documents/6 - mix/endpoint-audit/training/.venv/lib/python3.11/site-packages/transformers/generation/configuration_utils.py:590: UserWarning: `do_sample` is set to `False`. However, `temperature` is set to `0.7` -- this flag is only used in sample-based generation modes. You should set `do_sample=True` or unset `temperature`.
  warnings.warn(
/Users/umer/Documents/6 - mix/endpoint-audit/training/.venv/lib/python3.11/site-packages/transformers/generation/configuration_utils.py:595: UserWarning: `do_sample` is set to `False`. However, `top_p` is set to `0.8` -- this flag is only used in sample-based generation modes. You should set `do_sample=True` or unset `top_p`.
  warnings.warn(
/Users/umer/Documents/6 - mix/endpoint-audit/training/.venv/lib/python3.11/site-packages/transformers/generation/configuration_utils.py:612: UserWarning: `do_sample` is set to `False`. However, `top_k` is set to `20` -- this flag is only used in sample-based generation modes. You should set `do_sample=True` or unset `top_k`.
  warnings.warn(
The attention mask is not set and cannot be inferred from input because pad token is same as eos token. As a consequence, you may observe unexpected behavior. Please pass your input's `attention_mask` to obtain reliable results.
[1/30] ai_classification      26s
[2/30] finding                80s
[3/30] finding               136s
[4/30] finding               194s
[5/30] finding               207s
[6/30] finding               219s
[7/30] summary               354s
[8/30] ai_classification     368s
[9/30] finding               395s
[10/30] finding               427s
[11/30] finding               446s
[12/30] finding               491s
[13/30] finding               544s
[14/30] summary               716s
[15/30] ai_classification     853s
[16/30] finding               879s
[17/30] finding               913s
[18/30] finding               943s
[19/30] finding               975s
[20/30] finding               989s
[21/30] summary              1163s
[22/30] ai_classification    1301s
[23/30] finding              1332s
[24/30] finding              1362s
[25/30] finding              1381s
[26/30] finding              1393s
[27/30] finding              1438s
[28/30] summary              1599s
[29/30] ai_classification    1666s
[30/30] finding              1695s

Results for base model on 30 test examples:
task                 n  valid JSON  grounded  fields match
ai_classification    5         0%        0%            0%
finding             21         0%        0%            0%
summary              4         0%        0%            0%
== fine-tuned v0.1 pilot 20:51
/Users/umer/Documents/6 - mix/endpoint-audit/training/.venv/lib/python3.11/site-packages/transformers/generation/configuration_utils.py:590: UserWarning: `do_sample` is set to `False`. However, `temperature` is set to `0.7` -- this flag is only used in sample-based generation modes. You should set `do_sample=True` or unset `temperature`.
  warnings.warn(
/Users/umer/Documents/6 - mix/endpoint-audit/training/.venv/lib/python3.11/site-packages/transformers/generation/configuration_utils.py:595: UserWarning: `do_sample` is set to `False`. However, `top_p` is set to `0.8` -- this flag is only used in sample-based generation modes. You should set `do_sample=True` or unset `top_p`.
  warnings.warn(
/Users/umer/Documents/6 - mix/endpoint-audit/training/.venv/lib/python3.11/site-packages/transformers/generation/configuration_utils.py:612: UserWarning: `do_sample` is set to `False`. However, `top_k` is set to `20` -- this flag is only used in sample-based generation modes. You should set `do_sample=True` or unset `top_k`.
  warnings.warn(
The attention mask is not set and cannot be inferred from input because pad token is same as eos token. As a consequence, you may observe unexpected behavior. Please pass your input's `attention_mask` to obtain reliable results.
[1/30] ai_classification      41s
[2/30] finding                84s
[3/30] finding               133s
[4/30] finding               167s
[5/30] finding               188s
[6/30] finding               214s
[7/30] summary               325s
[8/30] ai_classification     386s
[9/30] finding               434s
[10/30] finding               462s
[11/30] finding               492s
[12/30] finding               517s
[13/30] finding               551s
[14/30] summary               659s
[15/30] ai_classification     762s
[16/30] finding               784s
[17/30] finding               814s
[18/30] finding               843s
[19/30] finding               862s
[20/30] finding               887s
[21/30] summary              1009s
[22/30] ai_classification    1104s
[23/30] finding              1123s
[24/30] finding              1148s
[25/30] finding              1172s
[26/30] finding              1189s
[27/30] finding              1212s
[28/30] summary              1338s
[29/30] ai_classification    1370s
[30/30] finding              1387s

Results for adapter training/models/auditor-lora/final on 30 test examples:
task                 n  valid JSON  grounded  fields match
ai_classification    5       100%      100%          100%
finding             21       100%      100%          100%
summary              4       100%      100%            0%
== done 21:14