Spaces:
Running on Zero
Running on Zero
Update app.py
Browse files
app.py
CHANGED
|
@@ -31,6 +31,8 @@ MODEL_IDS: List[str] = [
|
|
| 31 |
"CodeSoft/MetaDiffusion-150M-ChatBase",
|
| 32 |
"BananaMind/BananaMind-2-Medium-Chat",
|
| 33 |
"SupraLabs/Supra2-100M-Instruct",
|
|
|
|
|
|
|
| 34 |
"HuggingFaceTB/SmolLM2-135M-Instruct",
|
| 35 |
"OpenCerebral/Boris-1.3-125M-Instruct",
|
| 36 |
"OpenCerebral/Boris-1.3-75M-Instruct",
|
|
@@ -43,6 +45,8 @@ MODEL_DISPLAY: Dict[str, str] = {
|
|
| 43 |
"CodeSoft/MetaDiffusion-150M-ChatBase": "MetaDiffusion-150M-ChatBase",
|
| 44 |
"BananaMind/BananaMind-2-Medium-Chat": "BananaMind-2-Medium-Chat",
|
| 45 |
"SupraLabs/Supra2-100M-Instruct": "Supra2-100M-Instruct",
|
|
|
|
|
|
|
| 46 |
"HuggingFaceTB/SmolLM2-135M-Instruct": "SmolLM2-135M-Instruct",
|
| 47 |
"OpenCerebral/Boris-1.3-125M-Instruct": "Boris-1.3-125M-Instruct",
|
| 48 |
"OpenCerebral/Boris-1.3-75M-Instruct": "Boris-1.3-75M-Instruct",
|
|
@@ -87,6 +91,8 @@ MODEL_PARAMS: Dict[str, float] = {
|
|
| 87 |
"CodeSoft/MetaDiffusion-150M-ChatBase": 169.5e6,
|
| 88 |
"BananaMind/BananaMind-2-Medium-Chat": 49.6e6,
|
| 89 |
"SupraLabs/Supra2-100M-Instruct": 100.0e6,
|
|
|
|
|
|
|
| 90 |
"HuggingFaceTB/SmolLM2-135M-Instruct": 135.0e6,
|
| 91 |
"OpenCerebral/Boris-1.3-125M-Instruct": 125.0e6,
|
| 92 |
"OpenCerebral/Boris-1.3-75M-Instruct": 75.0e6,
|
|
@@ -178,6 +184,8 @@ CHAT_FILE = get_chat_file()
|
|
| 178 |
GEN_DEFAULTS: Dict[str, dict] = {
|
| 179 |
"HuggingFaceTB/SmolLM2-135M-Instruct": {"max_new_tokens": 64, "temperature": 0.7, "top_p": 0.9, "repetition_penalty": 1.1, "do_sample": True},
|
| 180 |
"SupraLabs/Supra2-100M-Instruct": {"max_new_tokens": 64, "temperature": 0.7, "top_p": 0.9, "top_k": 25, "repetition_penalty": 1.1, "do_sample": True, "no_repeat_ngram_size": 3},
|
|
|
|
|
|
|
| 181 |
"BananaMind/BananaMind-2-Medium-Chat": {"max_new_tokens": 64, "temperature": 0.7, "top_p": 0.9, "repetition_penalty": 1.1, "do_sample": True},
|
| 182 |
"CodeSoft/MetaDiffusion-150M-ChatBase": {"max_new_tokens": 96, "num_steps": 128, "temperature": 0.7, "top_p": 0.9, "repetition_penalty": 1.5},
|
| 183 |
"fromziro/Zero-v0.1-150M": {"max_new_tokens": 64, "temperature": 0.8, "top_p": 0.95, "repetition_penalty": 1.1, "do_sample": True},
|
|
@@ -203,6 +211,8 @@ GEN_DEFAULTS: Dict[str, dict] = {
|
|
| 203 |
MODEL_CONTEXT: Dict[str, int] = {
|
| 204 |
"HuggingFaceTB/SmolLM2-135M-Instruct": 2048,
|
| 205 |
"SupraLabs/Supra2-100M-Instruct": 1024,
|
|
|
|
|
|
|
| 206 |
"BananaMind/BananaMind-2-Medium-Chat": 3072,
|
| 207 |
"CodeSoft/MetaDiffusion-150M-ChatBase": 5120,
|
| 208 |
"fromziro/Zero-v0.1-150M": 2048,
|
|
@@ -1383,4 +1393,4 @@ if __name__ == "__main__":
|
|
| 1383 |
logger.warning(f"Leaderboard preview failed: {e}")
|
| 1384 |
demo = create_demo()
|
| 1385 |
demo.queue(max_size=20)
|
| 1386 |
-
demo.launch(server_name="0.0.0.0", server_port=7860, show_error=True, theme=gr.themes.Base(), css=CSS)
|
|
|
|
| 31 |
"CodeSoft/MetaDiffusion-150M-ChatBase",
|
| 32 |
"BananaMind/BananaMind-2-Medium-Chat",
|
| 33 |
"SupraLabs/Supra2-100M-Instruct",
|
| 34 |
+
"SupraLabs/Supra2-Medium-Instruct",
|
| 35 |
+
"SupraLabs/Supra1.5-50M-Instruct-exp",
|
| 36 |
"HuggingFaceTB/SmolLM2-135M-Instruct",
|
| 37 |
"OpenCerebral/Boris-1.3-125M-Instruct",
|
| 38 |
"OpenCerebral/Boris-1.3-75M-Instruct",
|
|
|
|
| 45 |
"CodeSoft/MetaDiffusion-150M-ChatBase": "MetaDiffusion-150M-ChatBase",
|
| 46 |
"BananaMind/BananaMind-2-Medium-Chat": "BananaMind-2-Medium-Chat",
|
| 47 |
"SupraLabs/Supra2-100M-Instruct": "Supra2-100M-Instruct",
|
| 48 |
+
"SupraLabs/Supra2-Medium-Instruct": "Supra2-Medium-Instruct",
|
| 49 |
+
"SupraLabs/Supra1.5-50M-Instruct-exp": "Supra1.5-50M-Instruct-exp",
|
| 50 |
"HuggingFaceTB/SmolLM2-135M-Instruct": "SmolLM2-135M-Instruct",
|
| 51 |
"OpenCerebral/Boris-1.3-125M-Instruct": "Boris-1.3-125M-Instruct",
|
| 52 |
"OpenCerebral/Boris-1.3-75M-Instruct": "Boris-1.3-75M-Instruct",
|
|
|
|
| 91 |
"CodeSoft/MetaDiffusion-150M-ChatBase": 169.5e6,
|
| 92 |
"BananaMind/BananaMind-2-Medium-Chat": 49.6e6,
|
| 93 |
"SupraLabs/Supra2-100M-Instruct": 100.0e6,
|
| 94 |
+
"SupraLabs/Supra2-Medium-Instruct": 25.4e6,
|
| 95 |
+
"SupraLabs/Supra1.5-50M-Instruct-exp": 51.8e6,
|
| 96 |
"HuggingFaceTB/SmolLM2-135M-Instruct": 135.0e6,
|
| 97 |
"OpenCerebral/Boris-1.3-125M-Instruct": 125.0e6,
|
| 98 |
"OpenCerebral/Boris-1.3-75M-Instruct": 75.0e6,
|
|
|
|
| 184 |
GEN_DEFAULTS: Dict[str, dict] = {
|
| 185 |
"HuggingFaceTB/SmolLM2-135M-Instruct": {"max_new_tokens": 64, "temperature": 0.7, "top_p": 0.9, "repetition_penalty": 1.1, "do_sample": True},
|
| 186 |
"SupraLabs/Supra2-100M-Instruct": {"max_new_tokens": 64, "temperature": 0.7, "top_p": 0.9, "top_k": 25, "repetition_penalty": 1.1, "do_sample": True, "no_repeat_ngram_size": 3},
|
| 187 |
+
"SupraLabs/Supra2-Medium-Instruct": {"max_new_tokens": 64, "temperature": 0.2, "top_p": 0.85, "top_k": 25, "do_sample": True, "no_repeat_ngram_size": 3},
|
| 188 |
+
"SupraLabs/Supra1.5-50M-Instruct-exp": {"max_new_tokens": 64, "temperature": 0.7, "top_p": 0.9, "repetition_penalty": 1.2, "do_sample": True},
|
| 189 |
"BananaMind/BananaMind-2-Medium-Chat": {"max_new_tokens": 64, "temperature": 0.7, "top_p": 0.9, "repetition_penalty": 1.1, "do_sample": True},
|
| 190 |
"CodeSoft/MetaDiffusion-150M-ChatBase": {"max_new_tokens": 96, "num_steps": 128, "temperature": 0.7, "top_p": 0.9, "repetition_penalty": 1.5},
|
| 191 |
"fromziro/Zero-v0.1-150M": {"max_new_tokens": 64, "temperature": 0.8, "top_p": 0.95, "repetition_penalty": 1.1, "do_sample": True},
|
|
|
|
| 211 |
MODEL_CONTEXT: Dict[str, int] = {
|
| 212 |
"HuggingFaceTB/SmolLM2-135M-Instruct": 2048,
|
| 213 |
"SupraLabs/Supra2-100M-Instruct": 1024,
|
| 214 |
+
"SupraLabs/Supra2-Medium-Instruct": 1024,
|
| 215 |
+
"SupraLabs/Supra1.5-50M-Instruct-exp": 5120,
|
| 216 |
"BananaMind/BananaMind-2-Medium-Chat": 3072,
|
| 217 |
"CodeSoft/MetaDiffusion-150M-ChatBase": 5120,
|
| 218 |
"fromziro/Zero-v0.1-150M": 2048,
|
|
|
|
| 1393 |
logger.warning(f"Leaderboard preview failed: {e}")
|
| 1394 |
demo = create_demo()
|
| 1395 |
demo.queue(max_size=20)
|
| 1396 |
+
demo.launch(server_name="0.0.0.0", server_port=7860, show_error=True, theme=gr.themes.Base(), css=CSS)
|