Update via HuggingFace Tool

app/routers/models.py

@@ -12,7 +12,7 @@ import httpx
 from app.config import settings
 
 logger = logging.getLogger(__name__)
-router = APIRouter(tags=["Models"])
+router = APIRouter(prefix="/api/models", tags=["Models"])
 
 
 class ModelInfo(BaseModel):
@@ -150,7 +150,6 @@ async def get_model_info(model_id: str):
 @router.get("/recommend/{task_type}", response_model=List[ModelRecommendation])
 async def get_model_recommendations(task_type: str):
     """Get model recommendations for a specific task."""
-    # Popular models by task type
     recommendations = {
         "causal-lm": [
             ModelRecommendation(model_id="microsoft/phi-2", reason="Compact but powerful, excellent for fine-tuning", estimated_vram="8GB", suggested_batch_size=4, supports_peft=True),
@@ -198,7 +197,6 @@ async def get_model_recommendations(task_type: str):
     result = recommendations.get(task_type, [])
     
     if not result:
-        # Default to causal-lm recommendations
         result = recommendations.get("causal-lm", [])
     
     return result[:5]
@@ -212,7 +210,6 @@ async def check_model_compatibility(
     """Check if a model is compatible for training with current setup."""
     async with httpx.AsyncClient() as client:
         try:
-            # Get model info
             response = await client.get(
                 f"{HF_API_URL}/models/{model_id}",
                 timeout=30.0
@@ -224,28 +221,21 @@ async def check_model_compatibility(
             recommendations = []
             compatible = True
             
-            # Check library
             library = data.get("library_name", "transformers")
             if library not in ["transformers", "peft", "adapter-transformers"]:
                 issues.append(f"Unsupported library: {library}")
                 compatible = False
             
-            # Check tags
             tags = data.get("tags", [])
-            model_size_tags = [t for t in tags if any(s in t for s in ['b', 'B', 'm', 'M'])]
             
-            # Check for gated models
             if "gated" in tags:
                 recommendations.append("This is a gated model. Ensure you have access and set HF_TOKEN.")
             
-            # Check private models
             if data.get("private", False):
                 recommendations.append("Private model requires authentication.")
             
-            # Estimate requirements
             pipeline_tag = data.get("pipeline_tag", "")
             
-            # Determine max sequence length based on model type
             if "bert" in model_id.lower():
                 max_seq_len = 512
             elif "llama" in model_id.lower():
@@ -257,7 +247,6 @@ async def check_model_compatibility(
             else:
                 max_seq_len = 2048
             
-            # Check task compatibility
             task_pipeline_map = {
                 "causal-lm": "text-generation",
                 "seq2seq": "text2text-generation",
@@ -271,9 +260,8 @@ async def check_model_compatibility(
             
             expected_pipeline = task_pipeline_map.get(task)
             if expected_pipeline and pipeline_tag and pipeline_tag != expected_pipeline:
-                recommendations.append(f"Model pipeline tag ({pipeline_tag}) may not match task ({task}). Consider adjusting task type.")
+                recommendations.append(f"Model pipeline tag ({pipeline_tag}) may not match task ({task}).")
             
-            # CPU training limitations
             requires_gpu = False
             model_name_lower = model_id.lower()
             if any(x in model_name_lower for x in ["7b", "13b", "30b", "70b"]):
@@ -281,7 +269,6 @@ async def check_model_compatibility(
                 requires_gpu = True
                 compatible = False
             
-            # Estimate training time (very rough)
             if "7b" in model_id.lower():
                 training_time = "Hours per epoch on GPU"
             elif "2b" in model_id.lower() or "3b" in model_id.lower():