feat: Update allowed models to be used with Prompt Node (#4018)

* Update allowed models to be used with Prompt Node

* Added try except block around the config to skip over OpenAI models.

* Fixing tests

* Adding warning message

* Adding test for different HF models that could be used in prompt node

haystack/nodes/prompt/prompt_node.py

@@ -11,12 +11,13 @@ import requests
 import torch
 from transformers import (
     pipeline,
-    AutoModelForSeq2SeqLM,
+    AutoConfig,
     StoppingCriteria,
     StoppingCriteriaList,
-    PreTrainedTokenizerFast,
     PreTrainedTokenizer,
+    PreTrainedTokenizerFast,
 )
+from transformers.models.auto.modeling_auto import MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING_NAMES
 
 from haystack import MultiLabel
 from haystack.environment import HAYSTACK_REMOTE_API_BACKOFF_SEC, HAYSTACK_REMOTE_API_MAX_RETRIES
@@ -265,8 +266,6 @@ class HFLocalInvocationLayer(PromptModelInvocationLayer):
         includes: trust_remote_code, revision, feature_extractor, tokenizer, config, use_fast, torch_dtype, device_map.
         For more details about these kwargs, see
         Hugging Face [documentation](https://huggingface.co/docs/transformers/en/main_classes/pipelines#transformers.pipeline).
-
-
         """
         super().__init__(model_name_or_path, max_length)
         self.use_auth_token = use_auth_token
@@ -365,15 +364,22 @@ class HFLocalInvocationLayer(PromptModelInvocationLayer):
 
     @classmethod
     def supports(cls, model_name_or_path: str) -> bool:
-        if not all(m in model_name_or_path for m in ["google", "flan", "t5"]):
+        try:
+            config = AutoConfig.from_pretrained(model_name_or_path)
+        except OSError:
+            # This is needed so OpenAI models are skipped over
             return False
 
-        try:
-            # if it is google flan t5, load it, we'll use it anyway and also check if model loads correctly
-            AutoModelForSeq2SeqLM.from_pretrained(model_name_or_path)
-        except EnvironmentError:
-            return False
-        return True
+        if not all(m in model_name_or_path for m in ["flan", "t5"]):
+            logger.warning(
+                "PromptNode has been potentially initialized with a language model not fine-tuned on instruction following tasks. "
+                "Many of the default prompts and PromptTemplates will likely not work as intended. "
+                "Please use custom prompts and PromptTemplates specific to the %s model",
+                model_name_or_path,
+            )
+
+        supported_models = list(MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING_NAMES.values())
+        return config.architectures[0] in supported_models
 
 
 class OpenAIInvocationLayer(PromptModelInvocationLayer):
@@ -578,8 +584,8 @@ class PromptModel(BaseComponent):
                 )
         raise ValueError(
             f"Model {self.model_name_or_path} is not supported - no invocation layer found."
-            f"Currently supported models are: {self.invocation_layers}"
-            f"Register new invocation layer for {self.model_name_or_path} using the register method."
+            f" Currently supported models are: {self.invocation_layers}"
+            f" Register a new invocation layer for {self.model_name_or_path} using the register method."
         )
 
     def register(self, invocation_layer: Type[PromptModelInvocationLayer]):
@@ -694,7 +700,6 @@ class PromptNode(BaseComponent):
     LLM does not "follow" prompt instructions well. This is why we recommend using T5 flan or OpenAI InstructGPT models.
 
     For more details, see the PromptNode [documentation](https://docs.haystack.deepset.ai/docs/prompt_node).
-
     """
 
     outgoing_edges: int = 1

test/nodes/test_prompt_node.py

@@ -8,6 +8,7 @@ import torch
 from haystack import Document, Pipeline, BaseComponent, MultiLabel
 from haystack.errors import OpenAIError
 from haystack.nodes.prompt import PromptTemplate, PromptNode, PromptModel
+from haystack.nodes.prompt.prompt_node import HFLocalInvocationLayer
 
 
 def is_openai_api_key_set(api_key: str):
@@ -109,16 +110,6 @@ def test_create_prompt_node():
     assert prompt_node.model_name_or_path == "text-davinci-003"
     assert prompt_node.prompt_model is not None
 
-    with pytest.raises(ValueError, match="Model vblagoje/bart_lfqa is not supported"):
-        # yes vblagoje/bart_lfqa is AutoModelForSeq2SeqLM, can be downloaded, however it is useless for prompting
-        # currently support only T5-Flan models
-        prompt_node = PromptNode("vblagoje/bart_lfqa")
-
-    with pytest.raises(ValueError, match="Model valhalla/t5-base-e2e-qg is not supported"):
-        # yes valhalla/t5-base-e2e-qg is AutoModelForSeq2SeqLM, can be downloaded, however it is useless for prompting
-        # currently support only T5-Flan models
-        prompt_node = PromptNode("valhalla/t5-base-e2e-qg")
-
     with pytest.raises(ValueError, match="Model some-random-model is not supported"):
         PromptNode("some-random-model")
 
@@ -713,3 +704,8 @@ def test_complex_pipeline_with_multiple_same_prompt_node_components_yaml(tmp_pat
         )
     pipeline = Pipeline.load_from_yaml(path=tmp_path / "tmp_config.yml")
     assert pipeline is not None
+
+
+def test_HFLocalInvocationLayer_supports():
+    assert HFLocalInvocationLayer.supports("philschmid/flan-t5-base-samsum")
+    assert HFLocalInvocationLayer.supports("bigscience/T0_3B")