system instruction

docs/interface.md

@@ -44,6 +44,8 @@ This mode supports a number of command-line arguments, the details of which can
 
 - `--include_path` : Accepts a path to a folder. If passed, then all YAML files containing ` lm-eval`` compatible task configurations will be added to the task registry as available tasks. Used for when one is writing config files for their own task in a folder other than  `lm_eval/tasks/`
 
+- `--system_instruction`: Specifies a system instruction string to prepend to the prompt.
+
 - `--apply_chat_template` : If this flag is on, a chat template will be applied to the prompt. For Hugging Face models, the chat template is taken from the tokenizer, if the tokenizer does not have a chat template, a default one will be applied. For other models, a generic chat template is used.
 
 - `--predict_only`: Generates the model outputs without computing metrics. Use with `--log_samples` to retrieve decoded results.

lm_eval/__main__.py

@@ -162,6 +162,12 @@ def setup_parser() -> argparse.ArgumentParser:
         default=False,
         help="If True, write out all model outputs and documents for per-sample measurement and post-hoc analysis. Use with --output_path.",
     )
+    parser.add_argument(
+        "--system_instruction",
+        type=str,
+        default=None,
+        help="System instruction to be used in the prompt",
+    )
     parser.add_argument(
         "--apply_chat_template",
         action="store_true",
@@ -363,6 +369,7 @@ def cli_evaluate(args: Union[argparse.Namespace, None] = None) -> None:
         check_integrity=args.check_integrity,
         write_out=args.write_out,
         log_samples=args.log_samples,
+        system_instruction=args.system_instruction,
         apply_chat_template=args.apply_chat_template,
         gen_kwargs=args.gen_kwargs,
         task_manager=task_manager,

lm_eval/api/task.py

@@ -373,6 +373,7 @@ class Task(abc.ABC):
         world_size=None,
         cache_requests=False,
         rewrite_requests_cache=False,
+        system_instruction=None,
         apply_chat_template=False,
         tokenizer=None,
     ) -> None:
@@ -423,6 +424,7 @@ class Task(abc.ABC):
             fewshot_ctx = self.fewshot_context(
                 doc,
                 0 if self.config.num_fewshot is None else self.config.num_fewshot,
+                system_instruction,
                 apply_chat_template,
                 tokenizer,
             )
@@ -984,6 +986,7 @@ class ConfigurableTask(Task):
         self,
         doc: str,
         num_fewshot: int,
+        system_instruction: str = "",
         apply_chat_template: bool = False,
         tokenizer=None,
     ) -> str:
@@ -994,6 +997,8 @@ class ConfigurableTask(Task):
             The document as returned from training_docs, validation_docs, or test_docs.
         :param num_fewshot: int
             The number of fewshot examples to provide in the returned context string.
+        :param  system_instruction: str
+            System instruction to be applied to the prompt.
         :param apply_chat_template: bool
             Whether to apply the chat template to the fewshot context.
         :param tokenizer:
@@ -1001,25 +1006,43 @@ class ConfigurableTask(Task):
         :returns: str
             The fewshot context.
         """
+
+        if apply_chat_template:
+            labeled_examples = []
+        else:
+            labeled_examples = ""
+
+        # get task description
         if description := self.config.description:
             description = utils.apply_template(self.config.description, doc)
 
-        labeled_examples = []
-        if num_fewshot == 0:
-            # always prepend the (possibly empty) task description
-            if apply_chat_template:
-                labeled_examples.append({"role": "system", "content": description})
+        # create system prompt based on the provided system instruction and description
+        if system_instruction and description:
+            system_prompt = (
+                f"{system_instruction}{self.sampler.fewshot_delimiter}{description}"
+            )
+        elif system_instruction:
+            system_prompt = system_instruction
+        elif description:
+            system_prompt = description
         else:
-                labeled_examples = description
+            system_prompt = ""
+
+        # add system prompt if specified
+        if system_prompt:
+            if apply_chat_template:
+                labeled_examples.append({"role": "system", "content": system_prompt})
             else:
+                labeled_examples = system_prompt
+
+        # if few-shot - append examples after the system prompt
+        if num_fewshot > 0:
             if apply_chat_template:
                 labeled_examples = self.sampler.get_chat_context(
                     doc, num_fewshot, labeled_examples
                 )
             else:
-                labeled_examples = description + self.sampler.get_context(
-                    doc, num_fewshot
-                )
+                labeled_examples += self.sampler.get_context(doc, num_fewshot)
 
         example = self.doc_to_text(doc)
         if apply_chat_template:

lm_eval/evaluator.py

@@ -55,6 +55,7 @@ def simple_evaluate(
     check_integrity: bool = False,
     write_out: bool = False,
     log_samples: bool = True,
+    system_instruction: str = "",
     apply_chat_template: bool = False,
     gen_kwargs: Optional[str] = None,
     task_manager: Optional[TaskManager] = None,
@@ -100,6 +101,8 @@ def simple_evaluate(
         If True, write out an example document and model input for checking task integrity
     :param log_samples: bool
         If True, write out all model outputs and documents for per-sample measurement and post-hoc analysis
+    :param system_instruction: str
+        System instruction to be applied to the prompt
     :param apply_chat_template: bool
         If True, apply chat template to the prompt
     :param gen_kwargs: str
@@ -265,6 +268,7 @@ def simple_evaluate(
         bootstrap_iters=bootstrap_iters,
         write_out=write_out,
         log_samples=log_samples,
+        system_instruction=system_instruction,
         apply_chat_template=apply_chat_template,
         verbosity=verbosity,
     )
@@ -321,6 +325,7 @@ def evaluate(
     bootstrap_iters: Optional[int] = 100000,
     write_out: bool = False,
     log_samples: bool = True,
+    system_instruction: str = "",
     apply_chat_template: bool = False,
     verbosity: str = "INFO",
 ):
@@ -338,6 +343,8 @@ def evaluate(
         If True, write out an example document and model input for checking task integrity
     :param log_samples: bool
         If True, write out all model outputs and documents for per-sample measurement and post-hoc analysis
+    :param system_instruction: str
+        System instruction to be applied to the prompt
     :param apply_chat_template: bool
         If True, apply chat template to the prompt
     :return
@@ -369,6 +376,7 @@ def evaluate(
             world_size=lm.world_size,
             cache_requests=cache_requests,
             rewrite_requests_cache=rewrite_requests_cache,
+            system_instruction=system_instruction,
             apply_chat_template=apply_chat_template,
             tokenizer=lm.tokenizer if hasattr(lm, "tokenizer") else None,
         )