pease

generators/model.py

@@ -126,26 +126,24 @@ class Samba():
         
         for i in range(num_comps):
             payload = {
-                "inputs": [dataclasses.asdict(message) for message in messages],
+                "inputs": messages,
                 "params": {
-                    "do_sample": {"type": "bool", "value": True},
                     "max_tokens_allowed_in_completion": {"type": "int", "value": 500},
                     "min_token_capacity_for_completion": {"type": "int", "value": 2},
-                    "temperature": {"type": "float", "value": 0.7},
-                    "top_p": {"type": "float", "value": 0.1},
-                    "top_k": {"type": "int", "value": 40},
                     "skip_special_token": {"type": "bool", "value": True},
-                    "repetition_penalty": {"type": "float", "value": 1.15},
                     "stop_sequences": {"type": "list", "value": ["[INST]", "[INST]", "[/INST]", "[/INST]"]}
                 },
-                "expert": "llama3-8b"
+                "model": "llama3-8b"
             }
-            url = 'https://kjddazcq2e2wzvzv.snova.ai/api/v1/chat/completion'
+            url = "kjddazcq2e2wzvzv.snova.ai"
+            key = "bGlnaHRuaW5nOlUyM3pMcFlHY3dmVzRzUGFy"
+            
             headers = {
-                "Authorization": "Basic bGlnaHRuaW5nOlUyM3pMcFlHY3dmVzRzUGFy",
+                "Authorization": f"Basic {key}",
                 "Content-Type": "application/json"
             }
-            post_response = requests.post(url, json=payload, headers=headers, stream=True)
+            post_response = requests.post(f'https://{url}/api/v1/chat/completion', json=payload, headers=headers, stream=True)
+
             response_text = ""
             for line in post_response.iter_lines():
                 if line.startswith(b"data: "):