Upload 2 files

app.py

@@ -17,17 +17,17 @@ import wikipediaapi
 wiki_wiki = wikipediaapi.Wikipedia('Organika ([email protected])', 'en')
 
 ## ctransformers disabled for now
-# from ctransformers import AutoModelForCausalLM
-# model = AutoModelForCausalLM.from_pretrained(
-#     "Colby/StarCoder-3B-WoW-JSON", 
-#     model_file="StarCoder-3B-WoW-JSON-ggml.bin",
-#     model_type="gpt_bigcode"
-#     )
+from ctransformers import AutoModelForCausalLM
+model = AutoModelForCausalLM.from_pretrained(
+    "Colby/StarCoder-3B-WoW-JSON", 
+    model_file="StarCoder-3B-WoW-JSON-ggml.bin",
+    model_type="gpt_bigcode"
+    )
 
 # Use a pipeline as a high-level helper
 from transformers import pipeline
 topic_model = pipeline("zero-shot-classification", model="valhalla/distilbart-mnli-12-9")
-model = pipeline("text-generation", model="Colby/StarCoder-3B-WoW-JSON", device=0)
+#model = pipeline("text-generation", model="Colby/StarCoder-3B-WoW-JSON", device=0)
 
 def merlin_chat(message, history):
     chat_text = ""
@@ -78,11 +78,13 @@ def merlin_chat(message, history):
     user_msg = {'role': 'user', 'content': message}
     prompt = "[" + json.dumps(system_msg) + chat_json + json.dumps(user_msg) + "{'role': 'assistant, 'content': '*recalls \""
     for attempt in range(3):
-        result = model(prompt, max_new_tokens=250, return_full_text=False, handle_long_generation="hole")
+        # result = model(prompt, max_new_tokens=250, return_full_text=False, handle_long_generation="hole")
+        result = model(prompt, stop=["]"])
         response = result[0]['generated_text']
         start = 0
         end = 0
         cleanStr = response.lstrip()
+        cleanStr = cleanStr.replace(prompt,"")
         start = cleanStr.find('{') # this should skip over whatever it recalls to what it says next
         if start<=0:
             continue

requirements.txt

@@ -1,5 +1,7 @@
+ctransformers
 transformers
 torch
 spacy
+nltk
 rake_nltk
 wikipedia-api