some fixes in generating kwargs

app.py

@@ -1,40 +1,23 @@
 import gradio as gr
 import torch
-from transformers import AutoModelForCausalLM, AutoTokenizer, StoppingCriteria, StoppingCriteriaList, TextIteratorStreamer
+from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
 from threading import Thread
+from functools import partial
 
 
 tokenizer = AutoTokenizer.from_pretrained("IlyaGusev/saiga_llama3_8b")
 model = AutoModelForCausalLM.from_pretrained("IlyaGusev/saiga_llama3_8b", torch_dtype=torch.bfloat16)
-model = model  #.to('cuda')
-
-
-class StopOnTokens(StoppingCriteria):
-    def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor, **kwargs) -> bool:
-        stop_ids = [29, 0]
-        for stop_id in stop_ids:
-            if input_ids[0][-1] == stop_id:
-                return True
-        return False
+model = model
 
 
 def predict(message, history):
     print(history)
     history_transformer_format = history + [{"role": "user", "content": message}, 
                                             {"role": "assistant", "content": ""}]
-    stop = StopOnTokens()
 
-    # messages = "".join(["".join(["<|start_header_id|>user<|end_header_id|>\n"+item[0], 
-    #                              "<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n"+item[1]])
-    #             for item in history_transformer_format])
-    # messages = [{"role": "user", item[0], "content": item[1]} for item in history_transformer_format]
-    #print(messages)
-
-    # model_inputs = tokenizer([messages], return_tensors="pt") # .to("cuda")
     model_inputs = tokenizer.apply_chat_template(history_transformer_format, return_tensors="pt")
     streamer = TextIteratorStreamer(tokenizer, timeout=10., skip_prompt=True, skip_special_tokens=True)
     generate_kwargs = dict(
-        model_inputs,
         streamer=streamer,
         max_new_tokens=1024,
         do_sample=True,
@@ -42,9 +25,9 @@ def predict(message, history):
         top_k=1000,
         temperature=1.0,
         num_beams=1,
-        stopping_criteria=StoppingCriteriaList([stop])
-        )
-    t = Thread(target=model.generate, kwargs=generate_kwargs)
+    )
+    generating_func = partial(model.generate, model_inputs)
+    t = Thread(target=generating_func, kwargs=generate_kwargs)
     t.start()
 
     partial_message = ""
@@ -53,4 +36,5 @@ def predict(message, history):
             partial_message += new_token
             yield partial_message
 
+
 gr.ChatInterface(predict).launch(share=True)