compat with old vllm version

Author: xusenlin

api/models.py

@@ -92,12 +92,15 @@ def create_hf_llm():
 def create_vllm_engine():
     """ get vllm generate engine for chat or completion. """
     try:
+        import vllm
         from vllm.engine.arg_utils import AsyncEngineArgs
         from vllm.engine.async_llm_engine import AsyncLLMEngine
         from api.core.vllm_engine import VllmEngine, LoRA
     except ImportError:
         raise ValueError("VLLM engine not available")
 
+    vllm_version = vllm.__version__
+
     include = {
         "tokenizer_mode",
         "trust_remote_code",
@@ -106,11 +109,14 @@ def create_vllm_engine():
         "gpu_memory_utilization",
         "max_num_seqs",
         "enforce_eager",
-        "max_seq_len_to_capture",
         "max_loras",
         "max_lora_rank",
         "lora_extra_vocab_size",
     }
+
+    if vllm_version >= "0.4.3":
+        include.add("max_seq_len_to_capture")
+
     kwargs = dictify(SETTINGS, include=include)
     engine_args = AsyncEngineArgs(
         model=SETTINGS.model_path,

api/vllm_routes/chat.py

@@ -5,6 +5,7 @@
 from typing import AsyncIterator
 
 import anyio
+import vllm
 from fastapi import APIRouter, Depends, status
 from fastapi import HTTPException, Request
 from loguru import logger
@@ -38,6 +39,7 @@
 )
 
 chat_router = APIRouter(prefix="/chat")
+vllm_version = vllm.__version__
 
 
 def get_engine():
@@ -105,17 +107,16 @@ async def create_chat_completion(
         try:
             from vllm.model_executor.guided_decoding import get_guided_decoding_logits_processor
 
-            decoding_config = await engine.model.get_decoding_config()
-
-            try:
+            if vllm_version >= "0.4.3":
+                decoding_config = await engine.model.get_decoding_config()
                 guided_decode_logits_processor = (
                     await get_guided_decoding_logits_processor(
                         request.guided_decoding_backend or decoding_config.guided_decoding_backend,
                         request,
                         engine.tokenizer,
                     )
                 )
-            except TypeError:
+            else:
                 guided_decode_logits_processor = (
                     await get_guided_decoding_logits_processor(
                         request,
@@ -128,7 +129,7 @@ async def create_chat_completion(
         except ImportError:
             pass
 
-        try:
+        if vllm_version >= "0.4.3":
             result_generator = engine.model.generate(
                 {
                     "prompt": prompt if isinstance(prompt, str) else None,
@@ -138,7 +139,7 @@ async def create_chat_completion(
                 request_id,
                 lora_request,
             )
-        except TypeError:
+        else:
             result_generator = engine.model.generate(
                 prompt if isinstance(prompt, str) else None,
                 sampling_params,

api/vllm_routes/completion.py

@@ -6,6 +6,7 @@
 from typing import AsyncIterator, Tuple
 
 import anyio
+import vllm
 from fastapi import APIRouter, Depends
 from fastapi import Request
 from loguru import logger
@@ -27,6 +28,7 @@
 )
 
 completion_router = APIRouter()
+vllm_version = vllm.__version__
 
 
 def get_engine():
@@ -144,17 +146,16 @@ async def create_completion(
         try:
             from vllm.model_executor.guided_decoding import get_guided_decoding_logits_processor
 
-            decoding_config = await engine.model.get_decoding_config()
-
-            try:
+            if vllm_version >= "0.4.3":
+                decoding_config = await engine.model.get_decoding_config()
                 guided_decode_logits_processor = (
                     await get_guided_decoding_logits_processor(
                         request.guided_decoding_backend or decoding_config.guided_decoding_backend,
                         request,
                         engine.tokenizer,
                     )
                 )
-            except TypeError:
+            else:
                 guided_decode_logits_processor = (
                     await get_guided_decoding_logits_processor(
                         request,
@@ -176,7 +177,7 @@ async def create_completion(
             else:
                 input_ids = engine.convert_to_inputs(prompt=prompt, max_tokens=request.max_tokens)
 
-            try:
+            if vllm_version >= "0.4.3":
                 generator = engine.model.generate(
                     {
                         "prompt": prompt,
@@ -186,7 +187,7 @@ async def create_completion(
                     request_id,
                     lora_request,
                 )
-            except TypeError:
+            else:
                 generator = engine.model.generate(
                     prompt,
                     sampling_params,