Spaces:
Running
on
Zero
Running
on
Zero
save logs
Browse files
app.py
CHANGED
@@ -9,6 +9,8 @@ import torch
|
|
9 |
import spaces
|
10 |
import gradio as gr
|
11 |
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig, TextIteratorStreamer
|
|
|
|
|
12 |
|
13 |
subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
|
14 |
|
@@ -21,6 +23,9 @@ EMOJI = os.environ.get("EMOJI")
|
|
21 |
DESCRIPTION = os.environ.get("DESCRIPTION")
|
22 |
|
23 |
DISCORD_WEBHOOK = os.environ.get("DISCORD_WEBHOOK")
|
|
|
|
|
|
|
24 |
|
25 |
def send_discord(i,o):
|
26 |
url = DISCORD_WEBHOOK
|
@@ -111,6 +116,31 @@ def predict(message, history, system_prompt, temperature, max_new_tokens, top_k,
|
|
111 |
send_discord(instruction, "".join(outputs))
|
112 |
|
113 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
114 |
# Load model
|
115 |
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
|
116 |
quantization_config = BitsAndBytesConfig(
|
|
|
9 |
import spaces
|
10 |
import gradio as gr
|
11 |
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig, TextIteratorStreamer
|
12 |
+
from huggingface_hub import HfApi
|
13 |
+
from datetime import datetime
|
14 |
|
15 |
subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
|
16 |
|
|
|
23 |
DESCRIPTION = os.environ.get("DESCRIPTION")
|
24 |
|
25 |
DISCORD_WEBHOOK = os.environ.get("DISCORD_WEBHOOK")
|
26 |
+
TOKEN = os.environ.get("TOKEN")
|
27 |
+
|
28 |
+
|
29 |
|
30 |
def send_discord(i,o):
|
31 |
url = DISCORD_WEBHOOK
|
|
|
116 |
send_discord(instruction, "".join(outputs))
|
117 |
|
118 |
|
119 |
+
api = HfApi()
|
120 |
+
day=datetime.now().strftime("%Y-%m-%d")
|
121 |
+
timestamp=datetime.now().timestamp()
|
122 |
+
dd={
|
123 |
+
'message': message,
|
124 |
+
'history': history,
|
125 |
+
'system_prompt':system_prompt,
|
126 |
+
'temperature':temperature,
|
127 |
+
'max_new_tokens':max_new_tokens,
|
128 |
+
'top_k':top_k,
|
129 |
+
'repetition_penalty':repetition_penalty,
|
130 |
+
'top_p':top_p,
|
131 |
+
'instruction':instruction,
|
132 |
+
'output':"".join(outputs)
|
133 |
+
}
|
134 |
+
api.upload_file(
|
135 |
+
path_or_fileobj=json.dumps(dd, indent=2, ensure_ascii=False).encode('utf-8'),
|
136 |
+
path_in_repo=f"{day}/{timestamp}.json",
|
137 |
+
repo_id="speakleash/bielik-logs",
|
138 |
+
repo_type="dataset",
|
139 |
+
commit_message=f"X",
|
140 |
+
token=TOKEN,
|
141 |
+
run_as_future=True
|
142 |
+
)
|
143 |
+
|
144 |
# Load model
|
145 |
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
|
146 |
quantization_config = BitsAndBytesConfig(
|