save logs

app.py

@@ -9,6 +9,8 @@ import torch
 import spaces
 import gradio as gr
 from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig, TextIteratorStreamer
+from huggingface_hub import HfApi
+from datetime import datetime
 
 subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
 
@@ -21,6 +23,9 @@ EMOJI = os.environ.get("EMOJI")
 DESCRIPTION = os.environ.get("DESCRIPTION")
 
 DISCORD_WEBHOOK = os.environ.get("DISCORD_WEBHOOK")
+TOKEN = os.environ.get("TOKEN")
+
+
 
 def send_discord(i,o):
     url = DISCORD_WEBHOOK
@@ -111,6 +116,31 @@ def predict(message, history, system_prompt, temperature, max_new_tokens, top_k,
     send_discord(instruction, "".join(outputs))
 
 
+    api = HfApi()
+    day=datetime.now().strftime("%Y-%m-%d")
+    timestamp=datetime.now().timestamp()
+    dd={
+        'message': message, 
+        'history': history, 
+        'system_prompt':system_prompt, 
+        'temperature':temperature, 
+        'max_new_tokens':max_new_tokens, 
+        'top_k':top_k, 
+        'repetition_penalty':repetition_penalty, 
+        'top_p':top_p, 
+        'instruction':instruction, 
+        'output':"".join(outputs)
+    }
+    api.upload_file(
+        path_or_fileobj=json.dumps(dd, indent=2, ensure_ascii=False).encode('utf-8'),
+        path_in_repo=f"{day}/{timestamp}.json",
+        repo_id="speakleash/bielik-logs",
+        repo_type="dataset",
+        commit_message=f"X",
+        token=TOKEN,
+        run_as_future=True
+    )
+
 # Load model
 device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
 quantization_config = BitsAndBytesConfig(