add tokenizer logs info (#1731)

* add tokenizer logs info

* add no tokenizer case

* Update lm_eval/logging_utils.py
Co-authored-by: Hailey Schoelkopf <65563625+haileyschoelkopf@users.noreply.github.com>

* Update lm_eval/logging_utils.py
Co-authored-by: Hailey Schoelkopf <65563625+haileyschoelkopf@users.noreply.github.com>

* add updates

* fix conflict

---------
Co-authored-by: Hailey Schoelkopf <65563625+haileyschoelkopf@users.noreply.github.com>

lm_eval/evaluator.py

@@ -22,7 +22,7 @@ from lm_eval.evaluator_utils import (
     run_task_tests,
 )
 from lm_eval.loggers import EvaluationTracker
-from lm_eval.loggers.utils import add_env_info, get_git_commit_hash
+from lm_eval.logging.utils import add_env_info, add_tokenizer_info, get_git_commit_hash
 from lm_eval.tasks import TaskManager, get_task_dict
 from lm_eval.utils import (
     eval_logger,
@@ -326,6 +326,7 @@ def simple_evaluate(
         results["git_hash"] = get_git_commit_hash()
         results["date"] = start_date
         add_env_info(results)  # additional environment info to results
+        add_tokenizer_info(results, lm)  # additional info about tokenizer
         return results
     else:
         return None

lm_eval/loggers/utils.py

@@ -110,3 +110,20 @@ def add_env_info(storage: Dict[str, Any]):
         "upper_git_hash": upper_dir_commit,  # in case this repo is submodule
     }
     storage.update(added_info)
+
+
+def add_tokenizer_info(storage: Dict[str, Any], lm):
+    if getattr(lm, "tokenizer", False):
+        tokenizer_info = {
+            "tokenizer_pad_token": [lm.tokenizer.pad_token, lm.tokenizer.pad_token_id],
+            "tokenizer_eos_token": [lm.tokenizer.eos_token, lm.tokenizer.eos_token_id],
+            "tokenizer_bos_token": [lm.tokenizer.bos_token, lm.tokenizer.bos_token_id],
+            "eot_token_id": getattr(lm, "eot_token_id", None),
+            "max_length": getattr(lm, "max_length", None),
+        }
+        storage.update(tokenizer_info)
+    # seems gguf and textsynth do not have tokenizer
+    else:
+        logger.debug(
+            "LM does not have a 'tokenizer' attribute, not logging tokenizer metadata to results."
+        )