Upload folder using huggingface_hub

config.json

@@ -1,10 +1,11 @@
 {
+  "_name_or_path": "shhossain/all-MiniLM-L6-v2-sentiment-classifier",
   "architectures": [
     "SententenceTransformerSentimentModel"
   ],
   "auto_map": {
-    "AutoConfig": "config.SentimentConfig",
-    "AutoModelForSequenceClassification": "model.SententenceTransformerSentimentModel"
+    "AutoConfig": "shhossain/all-MiniLM-L6-v2-sentiment-classifier--config.SentimentConfig",
+    "AutoModelForSequenceClassification": "shhossain/all-MiniLM-L6-v2-sentiment-classifier--model.SententenceTransformerSentimentModel"
   },
   "class_map": {
     "0": "sad",
@@ -14,6 +15,16 @@
     "4": "fear",
     "5": "surprise"
   },
+  "custom_pipelines": {
+    "text-classification": {
+      "impl": "smodelpipeline.SentimentModelPipe",
+      "pt": [
+        "AutoModelForSequenceClassification"
+      ],
+      "tf": [],
+      "type": "text"
+    }
+  },
   "embedding_model": "all-MiniLM-L6-v2",
   "h1": 44,
   "h2": 46,

smodelpipeline.py

@@ -0,0 +1,23 @@
+from transformers import Pipeline
+from sentence_transformers import SentenceTransformer
+import torch
+
+class SentimentModelPipe(Pipeline):
+
+    def __init__(self, **kwargs):
+        Pipeline.__init__(self, **kwargs)
+        self.smodel = SentenceTransformer(kwargs.get("embedding_model", "sentence-transformers/all-MiniLM-L6-v2"))
+
+    def _sanitize_parameters(self, **kw):
+        return {}, {}, {}
+
+    def preprocess(self, inputs):
+        return self.smodel.encode(inputs, convert_to_tensor=True)
+    
+    def postprocess(self, outputs):
+        return outputs.argmax(1).item()
+    
+    def _forward(self, tensor):
+        with torch.no_grad():
+            out = self.model(tensor)
+        return out