Maia Pelletier commited on
Commit ·
b484ab6
1
Parent(s): 85532f2
fix deprecated models + timeout error fallback
Browse files
app.py
CHANGED
|
@@ -93,11 +93,11 @@ class RAGChatbot:
|
|
| 93 |
|
| 94 |
# Default and fallback models (try in order until one is supported by your Inference API providers)
|
| 95 |
# Check current availability at https://huggingface.co/inference/models
|
| 96 |
-
|
|
|
|
| 97 |
FALLBACK_CHAT_MODELS = [
|
| 98 |
-
"meta-llama/Llama-3.3-70B-Instruct",
|
| 99 |
"google/gemma-4-31B-it",
|
| 100 |
-
"
|
| 101 |
]
|
| 102 |
|
| 103 |
# Shown to users when the HF account behind HF_TOKEN has run out of Inference Providers credits (HTTP 402)
|
|
@@ -201,6 +201,11 @@ class RAGChatbot:
|
|
| 201 |
):
|
| 202 |
print(f"[DEBUG] Model {model} not available, trying next fallback.")
|
| 203 |
continue
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 204 |
print(f"[DEBUG] _generate_with_chat exception for {model}: {type(e).__name__}: {e}")
|
| 205 |
import traceback
|
| 206 |
traceback.print_exc()
|
|
|
|
| 93 |
|
| 94 |
# Default and fallback models (try in order until one is supported by your Inference API providers)
|
| 95 |
# Check current availability at https://huggingface.co/inference/models
|
| 96 |
+
# All ungated (no licence acceptance needed) and non-reasoning
|
| 97 |
+
DEFAULT_CHAT_MODEL = "Qwen/Qwen3-235B-A22B-Instruct-2507"
|
| 98 |
FALLBACK_CHAT_MODELS = [
|
|
|
|
| 99 |
"google/gemma-4-31B-it",
|
| 100 |
+
"deepseek-ai/DeepSeek-V3-0324",
|
| 101 |
]
|
| 102 |
|
| 103 |
# Shown to users when the HF account behind HF_TOKEN has run out of Inference Providers credits (HTTP 402)
|
|
|
|
| 201 |
):
|
| 202 |
print(f"[DEBUG] Model {model} not available, trying next fallback.")
|
| 203 |
continue
|
| 204 |
+
# Provider too slow or temporarily down (e.g. 504 Gateway Time-out): try the next model
|
| 205 |
+
status = getattr(getattr(e, "response", None), "status_code", None)
|
| 206 |
+
if status in (502, 503, 504) or "timed out" in err_str or "time-out" in err_str:
|
| 207 |
+
print(f"[DEBUG] Model {model} timed out or provider unavailable ({status}), trying next fallback.")
|
| 208 |
+
continue
|
| 209 |
print(f"[DEBUG] _generate_with_chat exception for {model}: {type(e).__name__}: {e}")
|
| 210 |
import traceback
|
| 211 |
traceback.print_exc()
|