37 improve dolly in order to accept parameters as falcon does

cht-dolly-v2/models.py

@@ -1,15 +1,19 @@
 import os
+from abc import ABC, abstractmethod
+from typing import List
 
 import torch
-from transformers import pipeline
+from transformers import logging, pipeline
 
+logging.set_verbosity_error()
 
-class ChatModel:
-    @classmethod
+
+class ChatModel(ABC):
+    @abstractmethod
     def get_model(cls):
         pass
 
-    @classmethod
+    @abstractmethod
     def generate(
         cls,
         messages: list,
@@ -23,13 +27,14 @@ def generate(
     ):
         pass
 
-    @classmethod
+    @abstractmethod
     def embeddings(cls, text):
         pass
 
 
 class DollyBasedModel(ChatModel):
     model = None
+    tokenizer = None
 
     @classmethod
     def generate(
@@ -42,9 +47,20 @@ def generate(
         max_tokens: int = 128,
         stop: str = "",
         **kwargs,
-    ):
+    ) -> List:
         message = messages[-1]["content"]
-        return [cls.model(message)[0]["generated_text"]]
+        return [
+            cls.model(
+                message,
+                max_length=max_tokens,
+                temperature=temperature,
+                top_p=top_p,
+                num_return_sequences=n,
+                return_full_text=kwargs.get("return_full_text", False),
+                do_sample=kwargs.get("do_sample", True),
+                stop_sequence=stop[0] if stop else None,
+            )[0]["generated_text"]
+        ]
 
     @classmethod
     def get_model(cls):

scripts/cht_dolly_v2.sh

@@ -2,7 +2,7 @@
 
 set -e
 
-export VERSION=1.0.2
+export VERSION=1.0.3
 
 docker buildx build --push \
     --cache-from ghcr.io/premai-io/chat-dolly-v2-12b-gpu:latest \