[PYPROJECT TOML] [FIXES]

pyproject.toml

@@ -39,6 +39,7 @@ faiss-cpu = "*"
 backoff = "*"
 marshmallow = "*"
 datasets = "*"
+optimum = "*"
 diffusers = "*"
 PyPDF2 = "*"
 accelerate = "*"
@@ -57,7 +58,6 @@ pydantic = "1.10.12"
 tenacity = "*"
 Pillow = "*"
 chromadb = "*"
-opencv-python-headless = "*"
 tabulate = "*"
 termcolor = "*"
 black = "*"

requirements.txt

@@ -25,6 +25,7 @@ sentencepiece
 PyPDF2
 accelerate
 chromadb
+optimum
 tiktoken
 tabulate 
 colored

sequential_workflow_example.py

@@ -16,7 +16,7 @@ llm = OpenAIChat(
     temperature=0.5,
     model_name="gpt-4",
     openai_api_key=api_key,
-    max_tokens=4000
+    max_tokens=4000,
 )
 
 

swarms/models/__init__.py

@@ -8,6 +8,7 @@ from swarms.models.openai_models import (
     AzureOpenAI,
     OpenAIChat,
 )  # noqa: E402
+
 # from swarms.models.zephyr import Zephyr  # noqa: E402
 from swarms.models.biogpt import BioGPT  # noqa: E402
 from swarms.models.huggingface import HuggingfaceLLM  # noqa: E402

swarms/models/huggingface.py

@@ -270,9 +270,7 @@ class HuggingfaceLLM:
         self.print_dashboard(task)
 
         try:
-            inputs = self.tokenizer.encode(
-                task, return_tensors="pt"
-            )
+            inputs = self.tokenizer.encode(task, return_tensors="pt")
 
             # self.log.start()
 

swarms/models/huggingface_pipeline.py

@@ -0,0 +1,55 @@
+from abc import abstractmethod
+from termcolor import colored
+import torch
+
+from swarms.models.base_llm import AbstractLLM
+
+if torch.cuda.is_available():
+    try:
+        from optimum.nvidia.pipelines import pipeline
+    except ImportError:
+        from transformers.pipelines import pipeline
+
+
+class HuggingfacePipeline(AbstractLLM):
+    """HuggingfacePipeline
+
+
+
+    Args:
+        AbstractLLM (AbstractLLM): [description]
+        task (str, optional): [description]. Defaults to "text-generation".
+        model_name (str, optional): [description]. Defaults to None.
+        use_fp8 (bool, optional): [description]. Defaults to False.
+        *args: [description]
+        **kwargs: [description]
+
+    Raises:
+
+    """
+
+    def __init__(
+        self,
+        task: str = "text-generation",
+        model_name: str = None,
+        use_fp8: bool = False,
+        *args,
+        **kwargs,
+    ):
+        super().__init__(*args, **kwargs)
+        self.pipe = pipeline(
+            task, model_name, use_fp8=use_fp8 * args, **kwargs
+        )
+
+    @abstractmethod
+    def run(self, task: str, *args, **kwargs):
+        try:
+            out = self.pipeline(task, *args, **kwargs)
+            return out
+        except Exception as e:
+            print(
+                colored(
+                    f"Error in {self.__class__.__name__} pipeline",
+                    "red",
+                )
+            )

tests/models/test_gpt4_vision_api.py

@@ -35,9 +35,7 @@ def test_encode_image(vision_api):
 
 
 def test_run_success(vision_api):
-    expected_response = {
-        "This is the model's response."
-    }
+    expected_response = {"This is the model's response."}
     with patch(
         "requests.post",
         return_value=Mock(json=lambda: expected_response),

tests/models/test_huggingface.py

@@ -18,7 +18,10 @@ def llm_instance():
 
 # Test for instantiation and attributes
 def test_llm_initialization(llm_instance):
-    assert llm_instance.model_id == "NousResearch/Nous-Hermes-2-Vision-Alpha"
+    assert (
+        llm_instance.model_id
+        == "NousResearch/Nous-Hermes-2-Vision-Alpha"
+    )
     assert llm_instance.max_length == 500
     # ... add more assertions for all default attributes
 
@@ -77,7 +80,10 @@ def test_llm_memory_consumption(llm_instance):
     [
         ("NousResearch/Nous-Hermes-2-Vision-Alpha", 100),
         ("microsoft/Orca-2-13b", 200),
-        ("berkeley-nest/Starling-LM-7B-alpha", None),  # None to check default behavior
+        (
+            "berkeley-nest/Starling-LM-7B-alpha",
+            None,
+        ),  # None to check default behavior
     ],
 )
 def test_llm_initialization_params(model_id, max_length):
@@ -99,7 +105,6 @@ def test_llm_set_invalid_device(llm_instance):
         llm_instance.set_device("quantum_processor")
 
 
-
 # Mocking external API call to test run method without network
 @patch("swarms.models.huggingface.HuggingfaceLLM.run")
 def test_llm_run_without_network(mock_run, llm_instance):
@@ -210,9 +215,6 @@ def test_llm_cleanup(mock_model, mock_tokenizer, llm_instance):
     mock_tokenizer.delete.assert_called_once()
 
 
-
-
-
 # Test model's ability to handle multilingual input
 @patch("swarms.models.huggingface.HuggingfaceLLM.run")
 def test_llm_multilingual_input(mock_run, llm_instance):
@@ -236,6 +238,5 @@ def test_llm_caching_mechanism(mock_run, llm_instance):
     assert first_run_result == second_run_result
 
 
-
 # These tests are provided as examples. In real-world scenarios, you will need to adapt these tests to the actual logic of your `HuggingfaceLLM` class.
 # For instance, "mock_model.delete.assert_called_once()" and similar lines are based on hypothetical methods and behaviors that you need to replace with actual implementations.