[Feature][Quant] Auto-detect quantization format from model files (#6645)

## Summary

- Add automatic quantization format detection, eliminating the need to
manually specify `--quantization` when serving quantized models.
- The detection inspects only lightweight JSON files
(`quant_model_description.json` and `config.json`) at engine
initialization time, with no `.safetensors` reads.
- User-explicit `--quantization` flags are always respected;
auto-detection only applies when the flag is omitted.

## Details

**Detection priority:**
1. `quant_model_description.json` exists → `quantization="ascend"`
(ModelSlim)
2. `config.json` contains `"quant_method": "compressed-tensors"` →
`quantization="compressed-tensors"` (LLM-Compressor)
3. Neither → default float behavior

**Technical approach:**
Hooked into `NPUPlatform.check_and_update_config()` to run detection
after `VllmConfig.__post_init__`. Since `quant_config` is already `None`
at that point, we explicitly recreate it via
`VllmConfig._get_quantization_config()` to trigger the full quantization
initialization pipeline.

## Files Changed

| File | Description |
|------|-------------|
| `vllm_ascend/quantization/utils.py` | Added
`detect_quantization_method()` and `maybe_auto_detect_quantization()` |
| `vllm_ascend/platform.py` | Integrated auto-detection in
`check_and_update_config()` |
| `vllm_ascend/quantization/modelslim_config.py` | Improved error
handling for weight loading |
- vLLM version: v0.15.0
- vLLM main:
d7e17aaacd

---------

Signed-off-by: SlightwindSec <slightwindsec@gmail.com>

tests/ut/quantization/test_modelslim_config.py

@@ -1,3 +1,6 @@
+import json
+import os
+import tempfile
 from unittest.mock import MagicMock, patch
 
 from vllm.model_executor.layers.fused_moe import FusedMoE
@@ -6,7 +9,10 @@ from vllm.model_executor.layers.linear import LinearBase
 
 from tests.ut.base import TestBase
 from vllm_ascend.ops.linear import AscendUnquantizedLinearMethod
-from vllm_ascend.quantization.modelslim_config import AscendModelSlimConfig
+from vllm_ascend.quantization.modelslim_config import (
+    MODELSLIM_CONFIG_FILENAME,
+    AscendModelSlimConfig,
+)
 from vllm_ascend.utils import ASCEND_QUANTIZATION_METHOD, vllm_version_is
 
 if vllm_version_is("v0.15.0"):
@@ -54,7 +60,7 @@ class TestAscendModelSlimConfig(TestBase):
 
     def test_get_config_filenames(self):
         filenames = AscendModelSlimConfig.get_config_filenames()
-        self.assertEqual(filenames, ["quant_model_description.json"])
+        self.assertEqual(filenames, [])
 
     def test_from_config(self):
         config = AscendModelSlimConfig.from_config(self.sample_config)
@@ -162,5 +168,90 @@ class TestAscendModelSlimConfig(TestBase):
         with self.assertRaises(ValueError):
             config.is_layer_skipped_ascend("fused_layer", fused_mapping)
 
+    def test_init_with_none_config(self):
+        config = AscendModelSlimConfig(None)
+        self.assertEqual(config.quant_description, {})
+
+    def test_init_with_default_config(self):
+        config = AscendModelSlimConfig()
+        self.assertEqual(config.quant_description, {})
+
+    def test_maybe_update_config_already_populated(self):
+        # When quant_description is already populated, should be a no-op
+        self.assertTrue(len(self.ascend_config.quant_description) > 0)
+        self.ascend_config.maybe_update_config("/some/model/path")
+        # quant_description should remain unchanged
+        self.assertEqual(self.ascend_config.quant_description,
+                         self.sample_config)
+
+    def test_maybe_update_config_loads_from_file(self):
+        config = AscendModelSlimConfig()
+        self.assertEqual(config.quant_description, {})
+
+        quant_data = {"layer1.weight": "INT8", "layer2.weight": "FLOAT"}
+        with tempfile.TemporaryDirectory() as tmpdir:
+            config_path = os.path.join(tmpdir, MODELSLIM_CONFIG_FILENAME)
+            with open(config_path, "w") as f:
+                json.dump(quant_data, f)
+
+            config.maybe_update_config(tmpdir)
+
+        self.assertEqual(config.quant_description, quant_data)
+
+    def test_maybe_update_config_raises_when_file_missing(self):
+        config = AscendModelSlimConfig()
+
+        with tempfile.TemporaryDirectory() as tmpdir:
+            with self.assertRaises(ValueError) as ctx:
+                config.maybe_update_config(tmpdir)
+
+            error_msg = str(ctx.exception)
+            self.assertIn("ModelSlim Quantization Config Not Found", error_msg)
+            self.assertIn(MODELSLIM_CONFIG_FILENAME, error_msg)
+
+    def test_maybe_update_config_raises_with_json_files_listed(self):
+        config = AscendModelSlimConfig()
+
+        with tempfile.TemporaryDirectory() as tmpdir:
+            # Create a dummy json file that is NOT the config file
+            dummy_path = os.path.join(tmpdir, "config.json")
+            with open(dummy_path, "w") as f:
+                json.dump({"dummy": True}, f)
+
+            with self.assertRaises(ValueError) as ctx:
+                config.maybe_update_config(tmpdir)
+
+            error_msg = str(ctx.exception)
+            self.assertIn("config.json", error_msg)
+
+    def test_maybe_update_config_non_directory_raises(self):
+        config = AscendModelSlimConfig()
+
+        with self.assertRaises(ValueError) as ctx:
+            config.maybe_update_config("not_a_real_directory_path")
+
+        error_msg = str(ctx.exception)
+        self.assertIn("ModelSlim Quantization Config Not Found", error_msg)
+
+    def test_apply_extra_quant_adaptations_shared_head(self):
+        config = AscendModelSlimConfig()
+        config.quant_description = {
+            "model.layers.0.shared_head.weight": "INT8",
+        }
+        config._apply_extra_quant_adaptations()
+        self.assertIn("model.layers.0.weight", config.quant_description)
+        self.assertEqual(config.quant_description["model.layers.0.weight"],
+                         "INT8")
+
+    def test_apply_extra_quant_adaptations_weight_packed(self):
+        config = AscendModelSlimConfig()
+        config.quant_description = {
+            "model.layers.0.weight_packed": "INT8",
+        }
+        config._apply_extra_quant_adaptations()
+        self.assertIn("model.layers.0.weight", config.quant_description)
+        self.assertEqual(config.quant_description["model.layers.0.weight"],
+                         "INT8")
+
     def test_get_scaled_act_names(self):
         self.assertEqual(self.ascend_config.get_scaled_act_names(), [])