add .isort.cfg (#378)

test/srt/model/bench_llama_low_api.py

@@ -4,6 +4,7 @@ from dataclasses import dataclass
 
 import torch
 import torch.distributed as dist
+
 from sglang.srt.managers.router.model_runner import ModelRunner
 from sglang.srt.model_config import ModelConfig
 
@@ -66,9 +67,9 @@ class BenchBatch:
             p_idx = prefix_req_idx[i // fork_num].item()
             n_idx = self.req_pool_indices[i].item()
             req_to_token[n_idx, :prefix_len] = req_to_token[p_idx, :prefix_len]
-            req_to_token[
-                n_idx, prefix_len : prefix_len + extend_len
-            ] = self.out_cache_loc[i * extend_len : (i + 1) * extend_len]
+            req_to_token[n_idx, prefix_len : prefix_len + extend_len] = (
+                self.out_cache_loc[i * extend_len : (i + 1) * extend_len]
+            )
 
     def update_decode(self, predict_ids, batch_size):
         assert predict_ids.shape[0] == batch_size
@@ -81,9 +82,9 @@ class BenchBatch:
             self.out_cache_cont_start,
             self.out_cache_cont_end,
         ) = self.token_to_kv_pool.alloc_contiguous(batch_size)
-        self.req_to_token_pool.req_to_token[
-            self.req_pool_indices, self.seq_lens
-        ] = self.out_cache_loc
+        self.req_to_token_pool.req_to_token[self.req_pool_indices, self.seq_lens] = (
+            self.out_cache_loc
+        )
         self.seq_lens.add_(1)
 
 

test/srt/model/reference_hf.py

@@ -1,5 +1,4 @@
 import argparse
-import os
 
 import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer

test/srt/model/test_llama_extend.py

@@ -1,11 +1,8 @@
 import multiprocessing
 import os
-import time
 
-import numpy as np
-import torch
-import torch.distributed as dist
 import transformers
+
 from sglang.srt.managers.router.infer_batch import Batch, ForwardMode, Req
 from sglang.srt.managers.router.model_runner import ModelRunner
 from sglang.srt.model_config import ModelConfig

test/srt/model/test_llama_low_api.py

@@ -4,6 +4,7 @@ import time
 import numpy as np
 import torch
 import torch.distributed as dist
+
 from sglang.srt.managers.router.model_runner import ModelRunner
 from sglang.srt.model_config import ModelConfig
 

test/srt/model/test_llava_low_api.py

@@ -1,12 +1,10 @@
 import multiprocessing
-import time
 
 import numpy as np
 import torch
-import torch.distributed as dist
+
 from sglang.srt.hf_transformers_utils import get_processor
-from sglang.srt.managers.router.infer_batch import ForwardMode
-from sglang.srt.managers.router.model_runner import InputMetadata, ModelRunner
+from sglang.srt.managers.router.model_runner import ModelRunner
 from sglang.srt.model_config import ModelConfig
 from sglang.srt.utils import load_image