Spaces:
Running on Zero
Running on Zero
Update app.py
Browse files
app.py
CHANGED
|
@@ -69,6 +69,9 @@ BASE_MODEL_IDS: List[str] = [
|
|
| 69 |
"BananaMind/BananaMind-2-Medium",
|
| 70 |
"BananaMind/BananaMind-2-Mini",
|
| 71 |
"BananaMind/BananaMind-2-Nano",
|
|
|
|
|
|
|
|
|
|
| 72 |
]
|
| 73 |
|
| 74 |
BASE_MODEL_DISPLAY: Dict[str, str] = {
|
|
@@ -85,6 +88,9 @@ BASE_MODEL_DISPLAY: Dict[str, str] = {
|
|
| 85 |
"BananaMind/BananaMind-2-Medium": "BananaMind-2-Medium",
|
| 86 |
"BananaMind/BananaMind-2-Mini": "BananaMind-2-Mini",
|
| 87 |
"BananaMind/BananaMind-2-Nano": "BananaMind-2-Nano",
|
|
|
|
|
|
|
|
|
|
| 88 |
}
|
| 89 |
|
| 90 |
MODEL_PARAMS: Dict[str, float] = {
|
|
@@ -112,6 +118,9 @@ MODEL_PARAMS: Dict[str, float] = {
|
|
| 112 |
"BananaMind/BananaMind-2-Medium": 55.85e6,
|
| 113 |
"BananaMind/BananaMind-2-Mini": 28.32e6,
|
| 114 |
"BananaMind/BananaMind-2-Nano": 12.07e6,
|
|
|
|
|
|
|
|
|
|
| 115 |
}
|
| 116 |
|
| 117 |
FALLBACK_IDS: Dict[str, str] = {}
|
|
@@ -206,6 +215,9 @@ GEN_DEFAULTS: Dict[str, dict] = {
|
|
| 206 |
"BananaMind/BananaMind-2-Medium": {"max_new_tokens": 64, "temperature": 0.8, "top_p": 0.95, "repetition_penalty": 1.1, "do_sample": True},
|
| 207 |
"BananaMind/BananaMind-2-Mini": {"max_new_tokens": 64, "temperature": 0.8, "top_p": 0.95, "repetition_penalty": 1.1, "do_sample": True},
|
| 208 |
"BananaMind/BananaMind-2-Nano": {"max_new_tokens": 64, "temperature": 0.8, "top_p": 0.95, "repetition_penalty": 1.1, "do_sample": True},
|
|
|
|
|
|
|
|
|
|
| 209 |
}
|
| 210 |
|
| 211 |
MODEL_CONTEXT: Dict[str, int] = {
|
|
@@ -233,6 +245,9 @@ MODEL_CONTEXT: Dict[str, int] = {
|
|
| 233 |
"BananaMind/BananaMind-2-Medium": 3072,
|
| 234 |
"BananaMind/BananaMind-2-Mini": 4096,
|
| 235 |
"BananaMind/BananaMind-2-Nano": 4096,
|
|
|
|
|
|
|
|
|
|
| 236 |
}
|
| 237 |
|
| 238 |
|
|
|
|
| 69 |
"BananaMind/BananaMind-2-Medium",
|
| 70 |
"BananaMind/BananaMind-2-Mini",
|
| 71 |
"BananaMind/BananaMind-2-Nano",
|
| 72 |
+
"DedeProGames/Kiyo-230M-Preview",
|
| 73 |
+
"DedeProGames/Kiyo-135M",
|
| 74 |
+
"DedeProGames/Kiyo-65M",
|
| 75 |
]
|
| 76 |
|
| 77 |
BASE_MODEL_DISPLAY: Dict[str, str] = {
|
|
|
|
| 88 |
"BananaMind/BananaMind-2-Medium": "BananaMind-2-Medium",
|
| 89 |
"BananaMind/BananaMind-2-Mini": "BananaMind-2-Mini",
|
| 90 |
"BananaMind/BananaMind-2-Nano": "BananaMind-2-Nano",
|
| 91 |
+
"DedeProGames/Kiyo-230M-Preview": "Kiyo-230M-Preview",
|
| 92 |
+
"DedeProGames/Kiyo-135M": "Kiyo-135M",
|
| 93 |
+
"DedeProGames/Kiyo-65M": "Kiyo-65M",
|
| 94 |
}
|
| 95 |
|
| 96 |
MODEL_PARAMS: Dict[str, float] = {
|
|
|
|
| 118 |
"BananaMind/BananaMind-2-Medium": 55.85e6,
|
| 119 |
"BananaMind/BananaMind-2-Mini": 28.32e6,
|
| 120 |
"BananaMind/BananaMind-2-Nano": 12.07e6,
|
| 121 |
+
"DedeProGames/Kiyo-230M-Preview": 229.7e6,
|
| 122 |
+
"DedeProGames/Kiyo-135M": 134.5e6,
|
| 123 |
+
"DedeProGames/Kiyo-65M": 65.0e6,
|
| 124 |
}
|
| 125 |
|
| 126 |
FALLBACK_IDS: Dict[str, str] = {}
|
|
|
|
| 215 |
"BananaMind/BananaMind-2-Medium": {"max_new_tokens": 64, "temperature": 0.8, "top_p": 0.95, "repetition_penalty": 1.1, "do_sample": True},
|
| 216 |
"BananaMind/BananaMind-2-Mini": {"max_new_tokens": 64, "temperature": 0.8, "top_p": 0.95, "repetition_penalty": 1.1, "do_sample": True},
|
| 217 |
"BananaMind/BananaMind-2-Nano": {"max_new_tokens": 64, "temperature": 0.8, "top_p": 0.95, "repetition_penalty": 1.1, "do_sample": True},
|
| 218 |
+
"DedeProGames/Kiyo-230M-Preview": {"max_new_tokens": 64, "temperature": 0.8, "top_p": 0.95, "repetition_penalty": 1.1, "do_sample": True},
|
| 219 |
+
"DedeProGames/Kiyo-135M": {"max_new_tokens": 64, "temperature": 0.8, "top_p": 0.95, "repetition_penalty": 1.1, "do_sample": True},
|
| 220 |
+
"DedeProGames/Kiyo-65M": {"max_new_tokens": 64, "temperature": 0.8, "top_p": 0.95, "repetition_penalty": 1.1, "do_sample": True},
|
| 221 |
}
|
| 222 |
|
| 223 |
MODEL_CONTEXT: Dict[str, int] = {
|
|
|
|
| 245 |
"BananaMind/BananaMind-2-Medium": 3072,
|
| 246 |
"BananaMind/BananaMind-2-Mini": 4096,
|
| 247 |
"BananaMind/BananaMind-2-Nano": 4096,
|
| 248 |
+
"DedeProGames/Kiyo-230M-Preview": 2048,
|
| 249 |
+
"DedeProGames/Kiyo-135M": 8192,
|
| 250 |
+
"DedeProGames/Kiyo-65M": 2048,
|
| 251 |
}
|
| 252 |
|
| 253 |
|