fix(model): Fix load some quantization model error

pilot/model/model_adapter.py

@@ -51,6 +51,12 @@ _OLD_MODELS = [
 class LLMModelAdaper:
     """New Adapter for DB-GPT LLM models"""
 
+    def use_fast_tokenizer(self) -> bool:
+        """Whether use a [fast Rust-based tokenizer](https://huggingface.co/docs/tokenizers/index) if it is supported
+        for a given model.
+        """
+        return False
+
     def model_type(self) -> str:
         return ModelType.HF
 
@@ -169,6 +175,9 @@ class OldLLMModelAdaperWrapper(LLMModelAdaper):
         self._adapter = adapter
         self._chat_adapter = chat_adapter
 
+    def use_fast_tokenizer(self) -> bool:
+        return self._adapter.use_fast_tokenizer()
+
     def model_type(self) -> str:
         return self._adapter.model_type()
 
@@ -200,6 +209,9 @@ class FastChatLLMModelAdaperWrapper(LLMModelAdaper):
     def __init__(self, adapter: "BaseModelAdapter") -> None:
         self._adapter = adapter
 
+    def use_fast_tokenizer(self) -> bool:
+        return self._adapter.use_fast_tokenizer
+
     def load(self, model_path: str, from_pretrained_kwargs: dict):
         return self._adapter.load_model(model_path, from_pretrained_kwargs)
 

pilot/utils/tracer/span_storage.py

@@ -34,7 +34,9 @@ class FileSpanStorage(SpanStorage):
         self.flush_signal_queue = queue.Queue()
 
         if not os.path.exists(filename):
-            with open(filename, "w") as _:
+            # New file if not exist
+            os.makedirs(os.path.dirname(filename), exist_ok=True)
+            with open(filename, "a"):
                 pass
         self.flush_thread = threading.Thread(target=self._flush_to_file, daemon=True)
         self.flush_thread.start()

pilot/utils/tracer/tests/test_base.py

@@ -1,7 +1,7 @@
 from typing import Dict
 from pilot.component import SystemApp
 
-from pilot.utils.tracer import Span, SpanStorage, Tracer
+from pilot.utils.tracer import Span, SpanType, SpanStorage, Tracer
 
 
 # Mock implementations
@@ -31,7 +31,9 @@ class MockTracer(Tracer):
             self._new_uuid() if parent_span_id is None else parent_span_id.split(":")[0]
         )
         span_id = f"{trace_id}:{self._new_uuid()}"
-        span = Span(trace_id, span_id, parent_span_id, operation_name, metadata)
+        span = Span(
+            trace_id, span_id, SpanType.BASE, parent_span_id, operation_name, metadata
+        )
         self.current_span = span
         return span
 
@@ -50,7 +52,14 @@ class MockTracer(Tracer):
 
 
 def test_span_creation():
-    span = Span("trace_id", "span_id", "parent_span_id", "operation", {"key": "value"})
+    span = Span(
+        "trace_id",
+        "span_id",
+        SpanType.BASE,
+        "parent_span_id",
+        "operation",
+        {"key": "value"},
+    )
     assert span.trace_id == "trace_id"
     assert span.span_id == "span_id"
     assert span.parent_span_id == "parent_span_id"

pilot/utils/tracer/tests/test_span_storage.py

@@ -5,7 +5,7 @@ import json
 import tempfile
 import time
 
-from pilot.utils.tracer import SpanStorage, FileSpanStorage, Span
+from pilot.utils.tracer import SpanStorage, FileSpanStorage, Span, SpanType
 
 
 @pytest.fixture
@@ -44,7 +44,7 @@ def read_spans_from_file(filename):
     "storage", [{"batch_size": 1, "flush_interval": 5}], indirect=True
 )
 def test_write_span(storage: SpanStorage):
-    span = Span("1", "a", "b", "op1")
+    span = Span("1", "a", SpanType.BASE, "b", "op1")
     storage.append_span(span)
     time.sleep(0.1)
 
@@ -57,8 +57,8 @@ def test_write_span(storage: SpanStorage):
     "storage", [{"batch_size": 1, "flush_interval": 5}], indirect=True
 )
 def test_incremental_write(storage: SpanStorage):
-    span1 = Span("1", "a", "b", "op1")
-    span2 = Span("2", "c", "d", "op2")
+    span1 = Span("1", "a", SpanType.BASE, "b", "op1")
+    span2 = Span("2", "c", SpanType.BASE, "d", "op2")
 
     storage.append_span(span1)
     storage.append_span(span2)
@@ -72,7 +72,7 @@ def test_incremental_write(storage: SpanStorage):
     "storage", [{"batch_size": 2, "flush_interval": 5}], indirect=True
 )
 def test_sync_and_async_append(storage: SpanStorage):
-    span = Span("1", "a", "b", "op1")
+    span = Span("1", "a", SpanType.BASE, "b", "op1")
 
     storage.append_span(span)
 
@@ -88,7 +88,7 @@ def test_sync_and_async_append(storage: SpanStorage):
 
 @pytest.mark.asyncio
 async def test_flush_policy(storage: SpanStorage):
-    span = Span("1", "a", "b", "op1")
+    span = Span("1", "a", SpanType.BASE, "b", "op1")
 
     for _ in range(storage.batch_size - 1):
         storage.append_span(span)
@@ -108,8 +108,8 @@ async def test_flush_policy(storage: SpanStorage):
     "storage", [{"batch_size": 2, "file_does_not_exist": True}], indirect=True
 )
 def test_non_existent_file(storage: SpanStorage):
-    span = Span("1", "a", "b", "op1")
-    span2 = Span("2", "c", "d", "op2")
+    span = Span("1", "a", SpanType.BASE, "b", "op1")
+    span2 = Span("2", "c", SpanType.BASE, "d", "op2")
     storage.append_span(span)
     time.sleep(0.1)