Updated max model length for text embedding models, update error handling for when errors that doesn't affect server launching show up in err logs

Author: XkunW

vec_inf/cli/_utils.py

@@ -58,12 +58,13 @@ def is_server_running(
     if isinstance(log_content, str):
         return log_content
 
+    status = None
     for line in log_content:
         if "error" in line.lower():
-            return ("FAILED", line.strip("\n"))
+            status = ("FAILED", line.strip("\n"))
         if MODEL_READY_SIGNATURE in line:
-            return "RUNNING"
-    return "LAUNCHING"
+            status = "RUNNING"
+    return "LAUNCHING" if not status else status
 
 
 def get_base_url(slurm_job_name: str, slurm_job_id: int, log_dir: Optional[str]) -> str:

vec_inf/models/models.csv

@@ -63,5 +63,5 @@ Qwen2.5-14B-Instruct,Qwen2.5,14B-Instruct,LLM,a40,m2,08:00:00,1,1,152064,32768,a
 Qwen2.5-32B-Instruct,Qwen2.5,32B-Instruct,LLM,a40,m2,08:00:00,2,1,152064,32768,auto,singularity,default,true
 Qwen2.5-72B-Instruct,Qwen2.5,72B-Instruct,LLM,a40,m2,08:00:00,4,1,152064,32768,auto,singularity,default,true
 Pixtral-12B-2409,Pixtral,12B-2409,VLM,a40,m2,08:00:00,1,1,131072,16384,auto,singularity,default,true
-bge-multilingual-gemma2,bge,multilingual-gemma2,Text Embedding,a40,m2,08:00:00,1,1,256002,8192,auto,singularity,default,true
-e5-mistral-7b-instruct,e5,mistral-7b-instruct,Text Embedding,a40,m2,08:00:00,1,1,32000,32768,auto,singularity,default,true
+bge-multilingual-gemma2,bge,multilingual-gemma2,Text Embedding,a40,m2,08:00:00,1,1,256002,4096,auto,singularity,default,true
+e5-mistral-7b-instruct,e5,mistral-7b-instruct,Text Embedding,a40,m2,08:00:00,1,1,32000,4096,auto,singularity,default,true