workspace/sprint_1_2/CODEBASE/infra/implementation/triton_run/run.sh

@@ -1 +1,2 @@
+cd ../PILOT_PROJECT/workspace/sprint_1_2/CODEBASE/infra/implementation/triton_run
 modal deploy PILOT_PROJECT/workspace/sprint_1_2/codebase/infra/implementation/triton_run/modal_triton.py

workspace/sprint_1_2/CODEBASE/infra/tests/test.py

@@ -0,0 +1,260 @@
+import os
+import sys
+import json
+import base64
+import io
+import time
+from pathlib import Path
+import numpy as np
+from PIL import Image
+import requests
+
+
+TRITON_URL = "https://dtj-tran--triton-s3-service-unified-triton-server.modal.run"
+MODEL_NAME = "msk_vision_pipeline_ensemble"
+
+BASE_DIR = Path(__file__).resolve().parent
+TEST_IMAGE_DIR = BASE_DIR / "test_images"
+
+ANGLE_CLASSES = ["med-lat", "post-trans", "sup-trans-flex", "sup-up-long"]
+
+ANGLE_OUTPUT_NAMES = [
+    "angle_classify_convnext_tiny_logits",
+    "angle_classify_resnet50_logits",
+    "angle_classify_swin_v2_s_logits",
+    "angle_classify_densenet_logits",
+    "angle_classify_efficientnet_logits",
+]
+
+SUP_SEG_OUTPUT_NAMES = [
+    "segmentation_model_unet_resnet101_logits",
+    "segmentation_model_unet3plus_att_logits",
+]
+
+POST_SEG_OUTPUT_NAMES = [
+    "segmentation_model_post_deeplabv3_resnet101_logits",
+    "segmentation_model_post_deeplabv3_logits",
+    "segmentation_model_post_efficientfeedback_logits",
+]
+
+
+def load_image(path: Path) -> Image.Image:
+    return Image.open(path).convert("RGB")
+
+
+def preprocess_224(img: Image.Image) -> np.ndarray:
+    img_resized = img.resize((224, 224), Image.Resampling.BILINEAR)
+    arr = np.asarray(img_resized).astype(np.float32) / 255.0
+    mean = np.array([0.485, 0.456, 0.406], dtype=np.float32)
+    std = np.array([0.229, 0.224, 0.225], dtype=np.float32)
+    arr = (arr - mean) / std
+    arr = arr.transpose(2, 0, 1)  # HWC -> CHW
+    arr = np.expand_dims(arr, axis=0)  # NCHW
+    return arr
+
+
+def preprocess_512(img: Image.Image) -> np.ndarray:
+    img_resized = img.resize((512, 512), Image.Resampling.BILINEAR)
+    arr = np.asarray(img_resized).astype(np.float32) / 255.0
+    arr = arr.transpose(2, 0, 1)  # HWC -> CHW
+    arr = np.expand_dims(arr, axis=0)  # NCHW
+    return arr
+
+def build_ensemble_request(input_224: np.ndarray, input_512: np.ndarray, model_name: str) -> tuple[bytes, dict]:
+    inputs = [
+        {
+            "name": "input_224",
+            "shape": list(input_224.shape),
+            "datatype": "FP32",
+            "parameters": {"binary_data_size": input_224.nbytes},
+        },
+        {
+            "name": "input_512",
+            "shape": list(input_512.shape),
+            "datatype": "FP32",
+            "parameters": {"binary_data_size": input_512.nbytes},
+        },
+    ]
+    outputs = [
+        {"name": name}
+        for name in (
+            ANGLE_OUTPUT_NAMES
+            + ["inflammation_model_efficientnet_b0_ultrasound_2_cls_logits"]
+            + SUP_SEG_OUTPUT_NAMES
+            + POST_SEG_OUTPUT_NAMES
+        )
+    ]
+
+    metadata = {"model_name": model_name, "model_version": "", "inputs": inputs, "outputs": outputs}
+    metadata_bytes = json.dumps(metadata).encode("utf-8")
+
+    body = metadata_bytes + input_224.tobytes() + input_512.tobytes()
+    headers = {"Inference-Header-Content-Length": str(len(metadata_bytes)), "Content-Type": "application/octet-stream"}
+    return body, headers
+
+def parse_kserve_v2_response(resp: requests.Response) -> dict:
+    # 1. Pull the header length sent by Triton
+    header_length = int(resp.headers.get("Inference-Header-Content-Length", "0"))
+    response_bytes = resp.content
+    
+    # 2. Extract the JSON metadata section
+    metadata = json.loads(response_bytes[:header_length].decode("utf-8"))
+    binary_data = response_bytes[header_length:]
+
+    # Dictionary to map Triton datatypes to numpy types & byte sizes
+    dtype_map = {
+        "FP32": (np.float32, 4),
+        "INT32": (np.int32, 4),
+        "INT64": (np.int64, 8),
+        "FP16": (np.float16, 2),
+        "UINT8": (np.uint8, 1)
+    }
+
+    result = {"metadata": metadata, "outputs": {}}
+    offset = 0
+    
+    for desc in metadata.get("outputs", []):
+        name = desc["name"]
+        shape = desc["shape"]
+        datatype = desc["datatype"]
+        
+        # Determine the data type and total size safely
+        np_type, element_size = dtype_map.get(datatype, (np.float32, 4))
+        total_elements = int(np.prod(shape))
+        
+        # SAFE EXTRACT: Look for 'binary_data_size'. 
+        # If it's missing (common in ensemble nodes), calculate it manually!
+        params = desc.get("parameters", {})
+        size = params.get("binary_data_size", total_elements * element_size)
+        
+        # Extract the slice of raw bytes
+        raw = binary_data[offset : offset + size]
+        
+        # Convert raw buffer back into a structured NumPy array
+        arr = np.frombuffer(raw, dtype=np_type).reshape(shape)
+        result["outputs"][name] = arr
+        
+        # Move our index cursor forward
+        offset += size
+        
+    return result
+
+
+def softmax(x: np.ndarray) -> np.ndarray:
+    e = np.exp(x - np.max(x))
+    return e / np.sum(e)
+
+
+def decode_angle_from_ensemble(outputs: dict) -> tuple[str, float]:
+    logits_list = [outputs[name] for name in ANGLE_OUTPUT_NAMES if name in outputs]
+    if not logits_list:
+        raise ValueError("No angle logits found in ensemble outputs")
+    
+    # stack and average across models
+    avg_logits = np.mean(np.stack(logits_list), axis=0)
+    # remove batch dimension if present
+    if avg_logits.ndim > 1 and avg_logits.shape[0] == 1:
+        avg_logits = avg_logits[0]
+    probs = softmax(avg_logits)
+    idx = int(np.argmax(probs))
+    return ANGLE_CLASSES[idx], float(probs[idx])
+
+
+def decode_inflammation(outputs: dict) -> tuple[bool, float]:
+    key = "inflammation_model_efficientnet_b0_ultrasound_2_cls_logits"
+    logits = outputs[key]
+    probs = softmax(logits)
+    prob_inflam = float(probs[1])
+    return bool(prob_inflam >= 0.5), prob_inflam
+
+
+def infer_ensemble(image: Image.Image) -> dict:
+    input_224 = preprocess_224(image)
+    input_512 = preprocess_512(image)
+    body, headers = build_ensemble_request(input_224, input_512, MODEL_NAME)
+    resp = requests.post(
+        f"{TRITON_URL}/v2/models/{MODEL_NAME}/infer",
+        data=body,
+        headers=headers,
+        timeout=120,
+    )
+    resp.raise_for_status()
+    if resp.status_code != 200:
+        raise RuntimeError(f"Failed to get inference response: {resp.status_code} {resp.text}")
+    else:
+        print(f"Received inference response: {resp.status_code}")
+    return parse_kserve_v2_response(resp)
+
+
+def analyze_image_flow(image: Image.Image) -> dict:
+    t0 = time.time()
+    parsed = infer_ensemble(image)
+    outputs = parsed["outputs"]
+
+    angle, angle_conf = decode_angle_from_ensemble(outputs)
+    print(f"Angle: {angle} ({angle_conf*100:.2f}%)")
+
+    result = {
+        "angle": {"class": angle, "confidence": round(angle_conf * 100, 2)},
+        "inflammation": None,
+        "segmentation": None,
+        "measurement": None,
+        "severity": None,
+    }
+
+    if "post-trans" in angle.lower():
+        has_inflammation, inflam_conf = decode_inflammation(outputs)
+        result["inflammation"] = {"detected": has_inflammation, "confidence": round(inflam_conf * 100, 2)}
+        print(f"Inflammation POST: {has_inflammation} ({result['inflammation']['confidence']}%)")
+    elif "sup-up-long" in angle.lower():
+        has_inflammation, inflam_conf = decode_inflammation(outputs)
+        result["inflammation"] = {"detected": has_inflammation, "confidence": round(inflam_conf * 100, 2)}
+        print(f"Inflammation SUP: {has_inflammation} ({result['inflammation']['confidence']}%)")
+
+    elapsed = time.time() - t0
+    print(f"Elapsed: {elapsed:.2f}s")
+    return result
+
+
+def main():
+    print(f"TRITON_URL={TRITON_URL}")
+    print(f"MODEL_NAME={MODEL_NAME}")
+    print(f"TEST_IMAGE_DIR={TEST_IMAGE_DIR}")
+    print("=" * 60)
+
+    folders = {
+        # "sup-up-long_positive": TEST_IMAGE_DIR / "sup-up-long_positive",
+        # "sup-up-long_negative": TEST_IMAGE_DIR / "sup-up-long_negative",
+        # "post_trans_positive": TEST_IMAGE_DIR / "post_trans_positive",
+        # "post_trans_negative": TEST_IMAGE_DIR / "post_trans_negative",
+        "other_angle": TEST_IMAGE_DIR / "other_angle",
+    }
+
+    summary = []
+    for category, folder in folders.items():
+        if not folder.exists():
+            continue
+        for img_path in sorted(folder.iterdir()):
+            if img_path.suffix.lower() not in {".jpg", ".jpeg", ".png"}:
+                continue
+            print(f"\n>>> {category}: {img_path.name}")
+            try:
+                img = load_image(img_path)
+                res = analyze_image_flow(img)
+                res["file"] = str(img_path.relative_to(BASE_DIR))
+                res["category"] = category
+                summary.append(res)
+            except Exception as e:
+                print(f"Error: {e}")
+                import traceback
+                traceback.print_exc()
+                summary.append({"file": str(img_path.relative_to(BASE_DIR)), "category": category, "error": str(e)})
+
+    out_path = BASE_DIR / "result.json"
+    with open(out_path, "w", encoding="utf-8") as f:
+        json.dump(summary, f, indent=2, ensure_ascii=False)
+    print(f"\nSaved results to {out_path}")
+
+
+if __name__ == "__main__":
+    main()

workspace/sprint_1_2/CODEBASE/infra/tests/test_1_model.py

@@ -0,0 +1,134 @@
+import json
+import time
+from pathlib import Path
+import numpy as np
+from PIL import Image
+import requests
+
+# Server configuration
+TRITON_URL = "https://dtj-tran--triton-s3-service-unified-triton-server.modal.run"
+TARGET_MODEL = "angle_classify_resnet50"  # Target a single sub-model directly
+ANGLE_CLASSES = ["med-lat", "post-trans", "sup-trans-flex", "sup-up-long"]
+
+BASE_DIR = Path(__file__).resolve().parent
+
+def load_image(path: Path) -> Image.Image:
+    if not path.exists():
+        raise FileNotFoundError(f"Test image not found at: {path}")
+    return Image.open(path).convert("RGB")
+
+def preprocess_224(img: Image.Image) -> np.ndarray:
+    """Preprocesses image to NCHW FP32 [1, 3, 224, 224] matching ResNet50 input requirements"""
+    img_resized = img.resize((224, 224), Image.Resampling.BILINEAR)
+    arr = np.asarray(img_resized).astype(np.float32) / 255.0
+    mean = np.array([0.485, 0.456, 0.406], dtype=np.float32)
+    std = np.array([0.229, 0.224, 0.225], dtype=np.float32)
+    arr = (arr - mean) / std
+    arr = arr.transpose(2, 0, 1)  # HWC -> CHW
+    arr = np.expand_dims(arr, axis=0)  # Add batch dim -> NCHW
+    return arr
+
+def softmax(x: np.ndarray) -> np.ndarray:
+    e = np.exp(x - np.max(x))
+    return e / np.sum(e)
+
+def build_binary_request(input_tensor: np.ndarray, model_name: str) -> tuple[bytes, dict]:
+    """Constructs a strict KServe v2 binary payload for a single input tensor"""
+    # NOTE: Check your individual model's config.pbtxt to verify if the 
+    # expected input name is 'input_image', 'input_0', etc.
+    input_name = "input_image" 
+
+    inputs = [
+        {
+            "name": input_name,
+            "shape": list(input_tensor.shape),
+            "datatype": "FP32",
+            "parameters": {"binary_data_size": input_tensor.nbytes},
+        }
+    ]
+    
+    # We request the standard 'logits' output tensor from this model
+    outputs = [{"name": "logits"}]
+
+    metadata = {
+        "model_name": model_name, 
+        "model_version": "", 
+        "inputs": inputs, 
+        "outputs": outputs
+    }
+    
+    metadata_bytes = json.dumps(metadata).encode("utf-8")
+    body = metadata_bytes + input_tensor.tobytes()
+    
+    headers = {
+        "Inference-Header-Content-Length": str(len(metadata_bytes)), 
+        "Content-Type": "application/octet-stream"
+    }
+    return body, headers
+
+def parse_binary_response(resp: requests.Response) -> np.ndarray:
+    """Parses Triton's binary response stream for a single output tensor"""
+    header_length = int(resp.headers.get("Inference-Header-Content-Length", "0"))
+    response_bytes = resp.content
+    
+    metadata = json.loads(response_bytes[:header_length].decode("utf-8"))
+    binary_data = response_bytes[header_length:]
+
+    # Extract the first available output tensor
+    output_desc = metadata["outputs"][0]
+    shape = output_desc["shape"]
+    
+    # Unpack raw bytes back into numpy array
+    arr = np.frombuffer(binary_data, dtype=np.float32).reshape(shape)
+    return arr
+
+def main():
+    # Define a path to a real local image to test with
+    image_path = BASE_DIR / "test_images" / "other_angle" / "med_lat_2.png"
+    
+    print(f"Targeting Individual Model: {TARGET_MODEL}")
+    print(f"Loading image from: {image_path}")
+    
+    try:
+        # 1. Prepare image
+        img = load_image(image_path)
+        input_data = preprocess_224(img)
+        
+        # 2. Build the payload
+        body, headers = build_binary_request(input_data, TARGET_MODEL)
+        
+        # 3. Fire request to the targeted model endpoint
+        url = f"{TRITON_URL}/v2/models/{TARGET_MODEL}/infer"
+        print("Sending request to Triton server...")
+        
+        t0 = time.time()
+        resp = requests.post(url, data=body, headers=headers, timeout=30)
+        resp.raise_for_status()
+        latency = time.time() - t0
+        
+        # 4. Parse output logits
+        logits = parse_binary_response(resp)
+        logits = np.squeeze(logits)  # Drop batch dimension
+        
+        # 5. Decode probabilities
+        probs = softmax(logits)
+        predicted_idx = int(np.argmax(probs))
+        
+        print("\n" + "="*40)
+        print("🎉 SUCCESSFUL INFERENCE")
+        print(f"Network Roundtrip Time: {latency:.4f}s")
+        print("-" * 40)
+        print("Class Probabilities:")
+        for cls_name, prob in zip(ANGLE_CLASSES, probs):
+            print(f"  {cls_name:<15}: {prob * 100:.2f}%")
+        print("-" * 40)
+        print(f"Prediction: {ANGLE_CLASSES[predicted_idx]} ({probs[predicted_idx]*100:.2f}%)")
+        print("="*40)
+
+    except Exception as e:
+        print(f"\n❌ Inference Failed: {e}")
+        if 'resp' in locals() and hasattr(resp, 'text'):
+            print(f"Server Error Message: {resp.text}")
+
+if __name__ == "__main__":
+    main()

workspace/sprint_1_2/CODEBASE/infra/tests/test_config.py

@@ -0,0 +1,2 @@
+TRITON_URL = "https://dtj-tran--triton-s3-service-unified-triton-server.modal.run"
+MODEL_NAME = "msk_vision_pipeline_ensemble"