do the loop

app.py

@@ -1,64 +1,19 @@
 import gradio as gr
-from huggingface_hub import InferenceClient
+import loop
 
-"""
-For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-"""
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
+MODEL_ID = "TinyLlama/TinyLlama_v1.1_math_code"
 
+handler = loop.EndpointHandler(MODEL_ID)
 
-def respond(
-    message,
-    history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-):
-    messages = [{"role": "system", "content": system_message}]
+def respond(prompt, grammar):
+    args = { "inputs": prompt, "grammar": grammar }
+    return handler(prompt)
 
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
-
-    messages.append({"role": "user", "content": message})
-
-    response = ""
-
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
-
-        response += token
-        yield response
-
-
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
-demo = gr.ChatInterface(
+demo = gr.Interface(
     respond,
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
-    ],
+    inputs=["text", "text"],
+    outputs=["text"]
 )
 
-
 if __name__ == "__main__":
     demo.launch()

loop.py

@@ -0,0 +1,51 @@
+from typing import Dict, List, Any
+
+from transformers import AutoTokenizer, AutoModelForCausalLM
+from transformers.generation.logits_process import LogitsProcessorList, InfNanRemoveLogitsProcessor
+from transformers_gad.grammar_utils import IncrementalGrammarConstraint
+from transformers_gad.generation.logits_process import GrammarAlignedOracleLogitsProcessor
+
+class EndpointHandler():
+    def __init__(self, path=""):
+        # Preload
+        self.tokenizer = AutoTokenizer.from_pretrained(path)
+        self.model = AutoModelForCausalLM.from_pretrained(path)
+
+    def __call__(self, data: Dict[str, Any]) -> List[Dict[str, Any]]:
+        # do it!
+        inputs = data.get("inputs",data)
+        grammar_str = data.get("grammar", "")
+        MAX_NEW_TOKENS=512
+        MAX_TIME=30
+        print(grammar_str)
+        grammar = IncrementalGrammarConstraint(grammar_str, "root", self.tokenizer)
+
+        # Initialize logits processor for the grammar
+        gad_oracle_processor = GrammarAlignedOracleLogitsProcessor(grammar)
+        inf_nan_remove_processor = InfNanRemoveLogitsProcessor()
+        logits_processors = LogitsProcessorList([
+            inf_nan_remove_processor,
+            gad_oracle_processor,
+        ])
+
+        input_ids = self.tokenizer([inputs], add_special_tokens=False, return_tensors="pt")["input_ids"]
+
+        output = self.model.generate(
+                    input_ids,
+                    do_sample=True,
+                    max_time=MAX_TIME,
+                    max_new_tokens=MAX_NEW_TOKENS,
+                    logits_processor=logits_processors
+                )
+        
+        gad_oracle_processor.reset()
+
+        # Detokenize generated output
+        input_length = 1 if self.model.config.is_encoder_decoder else input_ids.shape[1]
+        if (hasattr(output, "sequences")):
+            generated_tokens = output.sequences[:, input_length:]
+        else:
+            generated_tokens = output[:, input_length:]
+
+        generations = self.tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)
+        return generations