sglang 0.1.15__py3-none-any.whl → 0.1.16__py3-none-any.whl

sglang/srt/weight_utils.py

@@ -19,11 +19,12 @@ import torch
 from huggingface_hub import HfFileSystem, snapshot_download
 from safetensors.torch import load_file, safe_open, save_file
 from tqdm.auto import tqdm
-
 from vllm.config import ModelConfig
 from vllm.logger import init_logger
-from vllm.model_executor.layers.quantization import (QuantizationConfig,
-                                                     get_quantization_config)
+from vllm.model_executor.layers.quantization import (
+    QuantizationConfig,
+    get_quantization_config,
+)
 from vllm.model_executor.layers.quantization.schema import QuantParamSchema
 
 logger = init_logger(__name__)
@@ -32,17 +33,21 @@ logger = init_logger(__name__)
 # can share the same lock without error.
 # lock files in the temp directory will be automatically deleted when the
 # system reboots, so users will not complain about annoying lock files
-temp_dir = os.environ.get('TMPDIR') or os.environ.get(
-    'TEMP') or os.environ.get('TMP') or "/tmp/"
+temp_dir = (
+    os.environ.get("TMPDIR")
+    or os.environ.get("TEMP")
+    or os.environ.get("TMP")
+    or "/tmp/"
+)
 
 
 def enable_hf_transfer():
-    """automatically activates hf_transfer
-    """
+    """automatically activates hf_transfer"""
     if "HF_HUB_ENABLE_HF_TRANSFER" not in os.environ:
         try:
             # enable hf hub transfer if available
             import hf_transfer  # type: ignore # noqa
+
             huggingface_hub.constants.HF_HUB_ENABLE_HF_TRANSFER = True
         except ImportError:
             pass
@@ -65,8 +70,7 @@ def get_lock(model_name_or_path: str, cache_dir: Optional[str] = None):
     # add hash to avoid conflict with old users' lock files
     lock_file_name = hash_name + model_name + ".lock"
     # mode 0o666 is required for the filelock to be shared across users
-    lock = filelock.FileLock(os.path.join(lock_dir, lock_file_name),
-                             mode=0o666)
+    lock = filelock.FileLock(os.path.join(lock_dir, lock_file_name), mode=0o666)
     return lock
 
 
@@ -104,10 +108,12 @@ def convert_bin_to_safetensor_file(
     sf_size = os.stat(sf_filename).st_size
     pt_size = os.stat(pt_filename).st_size
     if (sf_size - pt_size) / pt_size > 0.01:
-        raise RuntimeError(f"""The file size different is more than 1%:
+        raise RuntimeError(
+            f"""The file size different is more than 1%:
          - {sf_filename}: {sf_size}
          - {pt_filename}: {pt_size}
-         """)
+         """
+        )
 
     # check if the tensors are the same
     reloaded = load_file(sf_filename)
@@ -122,8 +128,7 @@ def convert_bin_to_safetensor_file(
 def get_quant_config(model_config: ModelConfig) -> QuantizationConfig:
     quant_cls = get_quantization_config(model_config.quantization)
     # Read the quantization config from the HF model config, if available.
-    hf_quant_config = getattr(model_config.hf_config, "quantization_config",
-                              None)
+    hf_quant_config = getattr(model_config.hf_config, "quantization_config", None)
     if hf_quant_config is not None:
         return quant_cls.from_config(hf_quant_config)
     model_name_or_path = model_config.model
@@ -131,26 +136,29 @@ def get_quant_config(model_config: ModelConfig) -> QuantizationConfig:
     if not is_local:
         # Download the config files.
         with get_lock(model_name_or_path, model_config.download_dir):
-            hf_folder = snapshot_download(model_name_or_path,
-                                          revision=model_config.revision,
-                                          allow_patterns="*.json",
-                                          cache_dir=model_config.download_dir,
-                                          tqdm_class=Disabledtqdm)
+            hf_folder = snapshot_download(
+                model_name_or_path,
+                revision=model_config.revision,
+                allow_patterns="*.json",
+                cache_dir=model_config.download_dir,
+                tqdm_class=Disabledtqdm,
+            )
     else:
         hf_folder = model_name_or_path
     config_files = glob.glob(os.path.join(hf_folder, "*.json"))
 
     quant_config_files = [
-        f for f in config_files if any(
-            f.endswith(x) for x in quant_cls.get_config_filenames())
+        f
+        for f in config_files
+        if any(f.endswith(x) for x in quant_cls.get_config_filenames())
     ]
     if len(quant_config_files) == 0:
-        raise ValueError(
-            f"Cannot find the config file for {model_config.quantization}")
+        raise ValueError(f"Cannot find the config file for {model_config.quantization}")
     if len(quant_config_files) > 1:
         raise ValueError(
             f"Found multiple config files for {model_config.quantization}: "
-            f"{quant_config_files}")
+            f"{quant_config_files}"
+        )
 
     quant_config_file = quant_config_files[0]
     with open(quant_config_file, "r") as f:
@@ -166,8 +174,7 @@ def prepare_hf_model_weights(
     revision: Optional[str] = None,
 ) -> Tuple[str, List[str], bool]:
     # Download model weights from huggingface.
-    is_local = os.path.isdir(model_name_or_path) \
-               and load_format != "tensorizer"
+    is_local = os.path.isdir(model_name_or_path) and load_format != "tensorizer"
     use_safetensors = False
     # Some quantized models use .pt files for storing the weights.
     if load_format == "auto":
@@ -203,11 +210,13 @@ def prepare_hf_model_weights(
         # Use file lock to prevent multiple processes from
         # downloading the same model weights at the same time.
         with get_lock(model_name_or_path, cache_dir):
-            hf_folder = snapshot_download(model_name_or_path,
-                                          allow_patterns=allow_patterns,
-                                          cache_dir=cache_dir,
-                                          tqdm_class=Disabledtqdm,
-                                          revision=revision)
+            hf_folder = snapshot_download(
+                model_name_or_path,
+                allow_patterns=allow_patterns,
+                cache_dir=cache_dir,
+                tqdm_class=Disabledtqdm,
+                revision=revision,
+            )
     else:
         hf_folder = model_name_or_path
     hf_weights_files: List[str] = []
@@ -228,16 +237,14 @@ def prepare_hf_model_weights(
             "scaler.pt",
         ]
         hf_weights_files = [
-            f for f in hf_weights_files
-            if not any(f.endswith(x) for x in blacklist)
+            f for f in hf_weights_files if not any(f.endswith(x) for x in blacklist)
         ]
 
     if load_format == "tensorizer":
         return hf_folder, hf_weights_files, use_safetensors
 
     if len(hf_weights_files) == 0:
-        raise RuntimeError(
-            f"Cannot find any model weights with `{model_name_or_path}`")
+        raise RuntimeError(f"Cannot find any model weights with `{model_name_or_path}`")
 
     return hf_folder, hf_weights_files, use_safetensors
 
@@ -254,7 +261,8 @@ def hf_model_weights_iterator(
         cache_dir=cache_dir,
         load_format=load_format,
         fall_back_to_pt=fall_back_to_pt,
-        revision=revision)
+        revision=revision,
+    )
 
     if load_format == "npcache":
         # Currently np_cache only support *.bin checkpoints
@@ -289,22 +297,25 @@ def hf_model_weights_iterator(
                 param = np.load(f)
             yield name, torch.from_numpy(param)
     elif load_format == "tensorizer":
-        from vllm.model_executor.tensorizer_loader import (TensorDeserializer,
-                                                           open_stream,
-                                                           tensorizer_warning)
+        from vllm.model_executor.tensorizer_loader import (
+            TensorDeserializer,
+            open_stream,
+            tensorizer_warning,
+        )
+
         tensorizer_args = load_format.params
         tensorizer_warning(
             "Deserializing HuggingFace models is not optimized for "
             "loading on vLLM, as tensorizer is forced to load to CPU. "
             "Consider deserializing a vLLM model instead for faster "
             "load times. See the examples/tensorize_vllm_model.py example "
-            "script for serializing vLLM models.")
+            "script for serializing vLLM models."
+        )
 
         deserializer_args = tensorizer_args.deserializer_params
         stream_params = tensorizer_args.stream_params
         stream = open_stream(tensorizer_args.tensorizer_uri, **stream_params)
-        with TensorDeserializer(stream, **deserializer_args,
-                                device="cpu") as state:
+        with TensorDeserializer(stream, **deserializer_args, device="cpu") as state:
             for name, param in state.items():
                 yield name, param
         del state
@@ -324,8 +335,12 @@ def hf_model_weights_iterator(
 
 
 def kv_cache_scales_loader(
-        filename: str, tp_rank: int, tp_size: int, num_hidden_layers: int,
-        model_type: Optional[str]) -> Iterable[Tuple[int, float]]:
+    filename: str,
+    tp_rank: int,
+    tp_size: int,
+    num_hidden_layers: int,
+    model_type: Optional[str],
+) -> Iterable[Tuple[int, float]]:
     """
     A simple utility to read in KV cache scaling factors that have been
     previously serialized to disk. Used by the model to populate the appropriate
@@ -343,8 +358,7 @@ def kv_cache_scales_loader(
                 "tp_size": tp_size,
             }
             schema_dct = json.load(f)
-            schema = QuantParamSchema.model_validate(schema_dct,
-                                                     context=context)
+            schema = QuantParamSchema.model_validate(schema_dct, context=context)
             layer_scales_map = schema.kv_cache.scaling_factor[tp_rank]
             return layer_scales_map.items()
 
@@ -357,9 +371,11 @@ def kv_cache_scales_loader(
     # This section is reached if and only if any of the excepts are hit
     # Return an empty iterable (list) => no KV cache scales are loaded
     # which ultimately defaults to 1.0 scales
-    logger.warning("Defaulting to KV cache scaling factors = 1.0 "
-                   f"for all layers in TP rank {tp_rank} "
-                   "as an error occurred during loading.")
+    logger.warning(
+        "Defaulting to KV cache scaling factors = 1.0 "
+        f"for all layers in TP rank {tp_rank} "
+        "as an error occurred during loading."
+    )
     return []
 
 
@@ -378,8 +394,7 @@ def convert_pyslice_to_tensor(x: Any) -> torch.Tensor:
     return x
 
 
-def default_weight_loader(param: torch.Tensor,
-                          loaded_weight: torch.Tensor) -> None:
+def default_weight_loader(param: torch.Tensor, loaded_weight: torch.Tensor) -> None:
     """Default weight loader."""
     assert param.size() == loaded_weight.size()
     param.data.copy_(loaded_weight)
@@ -399,4 +414,4 @@ def initialize_dummy_weights(
     """
     for param in model.state_dict().values():
         if torch.is_floating_point(param):
-            param.data.uniform_(low, high)
+            param.data.uniform_(low, high)

sglang/utils.py

@@ -2,40 +2,23 @@
 
 import base64
 import json
+import os
+import sys
 import threading
+import traceback
 import urllib.request
+from concurrent.futures import ThreadPoolExecutor
 from io import BytesIO
 from json import dumps
 
+import numpy as np
 import requests
 
 
-def get_available_gpu_memory(gpu_id, distributed=True):
-    """
-    Get available memory for cuda:gpu_id device.
-    When distributed is True, the available memory is the minimum available memory of all GPUs.
-    """
-    import torch
-
-    num_gpus = torch.cuda.device_count()
-    assert gpu_id < num_gpus
-
-    if torch.cuda.current_device() != gpu_id:
-        print(
-            f"WARNING: current device is not {gpu_id}, but {torch.cuda.current_device()}, ",
-            "which may cause useless memory allocation for torch CUDA context.",
-        )
-
-    free_gpu_memory, _ = torch.cuda.mem_get_info(gpu_id)
-
-    if distributed:
-        tensor = torch.tensor(free_gpu_memory, dtype=torch.float32).to(
-            torch.device("cuda", gpu_id)
-        )
-        torch.distributed.all_reduce(tensor, op=torch.distributed.ReduceOp.MIN)
-        free_gpu_memory = tensor.item()
-
-    return free_gpu_memory / (1 << 30)
+def get_exception_traceback():
+    etype, value, tb = sys.exc_info()
+    err_str = "".join(traceback.format_exception(etype, value, tb))
+    return err_str
 
 
 def is_same_type(values):
@@ -130,6 +113,74 @@ def encode_image_base64(image_path):
         return base64.b64encode(buffered.getvalue()).decode("utf-8")
 
 
+def encode_frame(frame):
+    import cv2  # pip install opencv-python-headless
+    from PIL import Image
+
+    # Convert the frame to RGB (OpenCV uses BGR by default)
+    frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
+
+    # Convert the frame to PIL Image to easily convert to bytes
+    im_pil = Image.fromarray(frame)
+
+    # Convert to bytes
+    buffered = BytesIO()
+
+    # frame_format = str(os.getenv('FRAME_FORMAT', "JPEG"))
+
+    im_pil.save(buffered, format="PNG")
+
+    frame_bytes = buffered.getvalue()
+
+    # Return the bytes of the frame
+    return frame_bytes
+
+
+def encode_video_base64(video_path, num_frames=16):
+    import cv2
+    cap = cv2.VideoCapture(video_path)
+    if not cap.isOpened():
+        raise IOError(f"Could not open video file:{video_path}")
+
+    total_frames = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
+    print(f"target_frames: {num_frames}")
+
+    frame_indices = np.linspace(0, total_frames - 1, num_frames, dtype=int)
+
+    frames = []
+    for i in range(total_frames):
+        ret, frame = cap.read()
+        if ret:
+            frames.append(frame)
+        else:
+            # Handle the case where the frame could not be read
+            # print(f"Warning: Could not read frame at index {i}.")
+            pass
+
+    cap.release()
+
+    # Safely select frames based on frame_indices, avoiding IndexError
+    frames = [frames[i] for i in frame_indices if i < len(frames)]
+
+    # If there are not enough frames, duplicate the last frame until we reach the target
+    while len(frames) < num_frames:
+        frames.append(frames[-1])
+
+    # Use ThreadPoolExecutor to process and encode frames in parallel
+    with ThreadPoolExecutor() as executor:
+        encoded_frames = list(executor.map(encode_frame, frames))
+
+    # encoded_frames = list(map(encode_frame, frames))
+
+    # Concatenate all frames bytes
+    video_bytes = b"".join(encoded_frames)
+
+    # Encode the concatenated bytes to base64
+    video_base64 = "video:" + base64.b64encode(video_bytes).decode("utf-8")
+
+    return video_base64
+
+
 def _is_chinese_char(cp):
     """Checks whether CP is the codepoint of a CJK character."""
     # This defines a "chinese character" as anything in the CJK Unicode block:

{sglang-0.1.15.dist-info → sglang-0.1.16.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: sglang
-Version: 0.1.15
+Version: 0.1.16
 Summary: A structured generation langauge for LLMs.
 License: Apache License
                                    Version 2.0, January 2004
@@ -237,8 +237,10 @@ Requires-Dist: vllm >=0.4.2 ; extra == 'srt'
 Requires-Dist: interegular ; extra == 'srt'
 Requires-Dist: pydantic ; extra == 'srt'
 Requires-Dist: pillow ; extra == 'srt'
-Requires-Dist: outlines >=0.0.27 ; extra == 'srt'
 Requires-Dist: packaging ; extra == 'srt'
+Requires-Dist: huggingface-hub ; extra == 'srt'
+Requires-Dist: hf-transfer ; extra == 'srt'
+Requires-Dist: outlines >=0.0.34 ; extra == 'srt'
 
 <div align="center">
 <img src="assets/logo.png" alt="logo" width="400"></img>
@@ -568,15 +570,17 @@ response = client.chat.completions.create(
 print(response)
 ```
 
-In above example, the server uses the chat template specified in the model tokenizer.
-You can override the chat template if needed when launching the server:
+
+By default, the server uses the chat template specified in the model tokenizer from Hugging Face. It should just work for most official models such as Llama-2/Llama-3.
+
+If needed, you can also override the chat template when launching the server:
 
 ```
 python -m sglang.launch_server --model-path meta-llama/Llama-2-7b-chat-hf --port 30000 --chat-template llama-2
 ```
 
 If the chat template you are looking for is missing, you are welcome to contribute it.
-Meanwhile, you can also temporary register your chat template as follows:
+Meanwhile, you can also temporarily register your chat template as follows:
 
 ```json
 {

sglang-0.1.16.dist-info/RECORD

@@ -0,0 +1,72 @@
+sglang/__init__.py,sha256=lKabCNZM2OhtymVLUuW4bpt-Jdxwk81wP1TkhVqIJEg,1058
+sglang/api.py,sha256=hnVPt_p2ALLrraAKpVbkGocVtgb0MqgOH5NUQKOA6sY,4548
+sglang/global_config.py,sha256=LxoF7VGCYszeEafC8zBbzUQ5PPFdv2rPzw2zEGPLgfg,961
+sglang/launch_server.py,sha256=jKPZRDN5bUe8Wgz5eoDkqeePhmKa8DLD4DpXQLT5auo,294
+sglang/launch_server_llavavid.py,sha256=UWo_qUCJ9yknp1TVPzrz4B_aZtEuQpLQq0l96FMgynI,1058
+sglang/utils.py,sha256=Xp5mmhLoXNLB5U0NmCg-WMkfV0Ov4KVqzOvGZa3XKmc,7610
+sglang/backend/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+sglang/backend/anthropic.py,sha256=gpxYWNRKDiRs1-dUUA53tuBH6TT2mSVgi-J9iOKuNNo,2075
+sglang/backend/base_backend.py,sha256=APiMht4WYECLCOGRPCEUF6lX-an1vjVe2dWoMSgymWY,1831
+sglang/backend/openai.py,sha256=QQS09WHqMpgg70r-uB1LocqxUZ7vhv4R3FHlt7NNaKg,9583
+sglang/backend/runtime_endpoint.py,sha256=ZnQ4DtbNIUr_Me5F6iYwMYsYhom8ZCs6A5kRjWwAANA,8695
+sglang/backend/vertexai.py,sha256=XNkbUzOdLIz-1qP_BBieYIfUXZf6gsfdghlaulNpBM8,4714
+sglang/lang/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+sglang/lang/chat_template.py,sha256=ogIT8iMlDcSEgcNBTh5pRLoCkdQI_ec5Hc27wFUFDIg,11532
+sglang/lang/compiler.py,sha256=wNn_UqV6Sxl22mv-PpzFUtRgiFFV-Y4OYpO4LshEoRM,7527
+sglang/lang/interpreter.py,sha256=GSIbO9N6ThfchdURb7XzQMZ9U6p1xirKHgXGmqLxKtg,28434
+sglang/lang/ir.py,sha256=NxvIWlUidvtpQpPG4GAXZEN64Y2vLOBjN2Z2JkZVG1U,13350
+sglang/lang/tracer.py,sha256=QcslAObEjepk8XmiqCobwzWaDpihofEQXjeRs_3B8NQ,8282
+sglang/srt/backend_config.py,sha256=UIV6kIU2j-Xh0eoezn1aXcYIy0miftHsWFeAZwqpbGE,227
+sglang/srt/conversation.py,sha256=NwTVuQXd3NqPq5WCllaYUgPLG2w2pMMbzIKDQfJMMO0,15491
+sglang/srt/flush_cache.py,sha256=JOXLH4pmweVbuEWDPu3SEDrLYFG82nR2SpzbslW4b-A,381
+sglang/srt/hf_transformers_utils.py,sha256=UneOMsw3w7taH9EKIi6uHZ-GNUZG0vbZIWN-ZoQZ5gM,5417
+sglang/srt/memory_pool.py,sha256=5bqI8d5_JURbKwIhv1BwlcIO2IDHewHvIqezPG-b_5M,3284
+sglang/srt/mm_utils.py,sha256=OptgAHDX-73Bk4jAdr2BOAJtiEXJNzPrMhaM-dy275c,8889
+sglang/srt/model_config.py,sha256=843L1KxEPZcEk1uwQH10BwSX9L5DYJ3OGUUBo8wMdZg,1695
+sglang/srt/openai_api_adapter.py,sha256=w3zvahyzvCnQd2pphQ6ViRBgHJmyI-TyIul6Q-CBY5Q,13214
+sglang/srt/openai_protocol.py,sha256=87pLM0hxocd5LUvhYopnL61cEKz3iu8TKdJtHbk3C5o,5211
+sglang/srt/sampling_params.py,sha256=dQbVr7JmTJ9JEn_sy3clB56yT9kyr9ldWFZ-GaNXOy0,3023
+sglang/srt/server.py,sha256=YAUiniJs9ebNrJ0Lweg2TnUL_yZ0P3PtWoT0Z_3d8vk,10371
+sglang/srt/server_args.py,sha256=TQxIEdF0crqtY6WfZ6q7SKOQcCSomBEVjJ5K4HyTSvQ,9539
+sglang/srt/utils.py,sha256=cr2uZmEB-Exq-wi3Y8B3yQu7kFUiyV4PAvzouvKYkWg,13090
+sglang/srt/weight_utils.py,sha256=bFNh9-T8gseB0zKeu1qsMww8FpyrGFxbPcOFSeJtL5Q,15505
+sglang/srt/constrained/__init__.py,sha256=BPRNDJnWtzYJ13X4urRS5aE6wFuwAVNBA9qeWIHF8rE,1236
+sglang/srt/constrained/base_cache.py,sha256=QQjmFEiT8jlOskJoZobhrDl2TKB-B4b1LPQo9JQCP_w,1405
+sglang/srt/constrained/fsm_cache.py,sha256=B9FPtpqzm4jKqciXTbfgNJL44hV2-rUG6-omDECN7iA,902
+sglang/srt/constrained/jump_forward.py,sha256=fUa4AlnGX40gYiWTLuICTJfq4b7wA3AL5dydTqT3jz4,2483
+sglang/srt/layers/context_flashattention_nopad.py,sha256=bENdVltDozccR5mLY_CcYDjqLob28tHA9f2s03D8UFQ,5210
+sglang/srt/layers/extend_attention.py,sha256=5gvRggy6qPLrLvjctoMMsYh1w70mOGxiPjxstHqjqsY,12623
+sglang/srt/layers/logits_processor.py,sha256=Vbkr6ANNfiBGkkNobqjNm1KQTqtuYQWZvmPjhhIWnS8,7267
+sglang/srt/layers/radix_attention.py,sha256=PBucvAdGI27Z1qQOUxUi-YJp-tKGm6LX3L2kp99pOV4,5598
+sglang/srt/layers/token_attention.py,sha256=Wm-Gj0VdmFE8krZeHjDWic9dmVxRvg1WRAIHbbA3M34,8517
+sglang/srt/managers/detokenizer_manager.py,sha256=-zuI2ZLyLD3wf21u8xWZm91JkcZZ57DwUFbFxnP2vFI,3462
+sglang/srt/managers/io_struct.py,sha256=fFfUQtC-D31xGYdCAfuNVuX3QyaNDgGpfzC8qnKt0YA,4294
+sglang/srt/managers/tokenizer_manager.py,sha256=TlGyFhWz1b24vkeUVvCwKFBERffi-esxGRhoukBnET8,13116
+sglang/srt/managers/router/infer_batch.py,sha256=a1F3EjSBdER5pbgZFifuTdrE2Xom8Mt4aT9rmB8n35M,20204
+sglang/srt/managers/router/manager.py,sha256=tdvYmwGHMeG2MMYZ4ZThdAJ_b4fp94UpemISFWOddno,2697
+sglang/srt/managers/router/model_rpc.py,sha256=FJFgf1KAJ0Z8Yq4EPyczxZkCmZBjwNwCwXcjwyhU0k4,29775
+sglang/srt/managers/router/model_runner.py,sha256=fp9wPh4sQY6Q-5PVtv_e9p5GgkkixSDUIqfFt7lVlV8,16527
+sglang/srt/managers/router/radix_cache.py,sha256=GE6oY8bppRJCIxZWiDKO4P6al58zcqLQe605Y1d2bdo,7924
+sglang/srt/managers/router/scheduler.py,sha256=pvlKSyCyIXmu14eyy1mvP9-QdG78eLUqMlr4cnfes2Y,2259
+sglang/srt/models/commandr.py,sha256=DVdUF5C5etm82RoXJTNjYqlS2W2_9awzxzXNMubRoVg,13579
+sglang/srt/models/dbrx.py,sha256=NIhlJp2B_y_L1ltK_Y7SEenAiHTUUp3p1rf8LIydC0o,14173
+sglang/srt/models/dbrx_config.py,sha256=6EKMCAP1kS4pkQ9Ycr39PeEeTCPG4JhKRm2rtA4jS2s,11071
+sglang/srt/models/gemma.py,sha256=Wk25zFkqkdG62xVVJEzeIjDES1LnoO0EY2W2p9XMvbA,11637
+sglang/srt/models/llama2.py,sha256=Y2XwS5XXG77OfPAvbju7zp53CP5izzee_4-laVqu5ZM,11655
+sglang/srt/models/llava.py,sha256=HtR7lUnAYW39vWw6xmDZkbG7AueswZDJxXeu6rQfpSU,14921
+sglang/srt/models/llavavid.py,sha256=ueImEwOR4ZlNFUoBvXbwZPNRcrYWg54sPNK7pmGnrp0,13219
+sglang/srt/models/mistral.py,sha256=XSn7fiZqspyWVTYrpVAacAnWdwAybBtyn9-Sh9AvMTM,254
+sglang/srt/models/mixtral.py,sha256=1aggGw0P0MVQu5C5D3pMaZpRpY_PmrK_nwBOygOlPEM,13839
+sglang/srt/models/qwen.py,sha256=cakvxjghKdGg5iGq9TJ_nGlVQaJ4-9V91EyyZnV4rmc,9390
+sglang/srt/models/qwen2.py,sha256=PyOA8-RA_frRVLXfh8d1Ui1hUd1YmM3GfsPw2q5rCDI,11351
+sglang/srt/models/stablelm.py,sha256=TCfQumj0acu2lCGujJj_PuzHFp3kFIwENQEfT-hnHUA,10867
+sglang/srt/models/yivl.py,sha256=q8MUvIFIWpKCQ4pSZBoFpw-pnbdjkfr-M8jBJfGFu7E,4393
+sglang/test/test_conversation.py,sha256=1zIrXcXiwEliPHgDAsqsQUA7JKzZ5fnQEU-U6L887FU,1592
+sglang/test/test_openai_protocol.py,sha256=eePzoskYR3PqfWczSVZvg8ja63qbT8TFUNEMyzDZpa8,1657
+sglang/test/test_programs.py,sha256=-2AoddzOOmXoj3muVUKX6Uih63UNTm3MFg2fcNnsy7Y,11498
+sglang/test/test_utils.py,sha256=9VFNGUMW0LBvmtDEHZ7ponakv5ZVF7B2Lg3xX353DXw,10083
+sglang-0.1.16.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
+sglang-0.1.16.dist-info/METADATA,sha256=yiziPDpVr6NPPhX58sA0GaLYKCut4FnBKD7TE50HH6k,28911
+sglang-0.1.16.dist-info/WHEEL,sha256=GJ7t_kWBFywbagK5eo9IoUwLW6oyOeTKmQ-9iHFVNxQ,92
+sglang-0.1.16.dist-info/top_level.txt,sha256=yxhh3pYQkcnA7v3Bg889C2jZhvtJdEincysO7PEB09M,7
+sglang-0.1.16.dist-info/RECORD,,

sglang-0.1.15.dist-info/RECORD

@@ -1,69 +0,0 @@
-sglang/__init__.py,sha256=Ef_3iE98hM5y45k97dcEXLqlRcSovIvGxEbTebnOre4,1034
-sglang/api.py,sha256=c2MIXPgtkmsgDY7BvXPOYkRaaJJRkCSBjGjvUz2zkkM,4455
-sglang/global_config.py,sha256=TLmmeWsk4mrjNr-ryj0w7irSr8HRekXYrYZON2sABdk,854
-sglang/launch_server.py,sha256=FteIWF2C73RN1qSPkh7cfIURV5rFvfHyKLHGDRUYJIA,294
-sglang/utils.py,sha256=2dUXLMPz9VhhzbIRQABmfZnVW5yz61F3UVtb6yKyevM,6237
-sglang/backend/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-sglang/backend/anthropic.py,sha256=gpxYWNRKDiRs1-dUUA53tuBH6TT2mSVgi-J9iOKuNNo,2075
-sglang/backend/base_backend.py,sha256=APiMht4WYECLCOGRPCEUF6lX-an1vjVe2dWoMSgymWY,1831
-sglang/backend/openai.py,sha256=QQS09WHqMpgg70r-uB1LocqxUZ7vhv4R3FHlt7NNaKg,9583
-sglang/backend/runtime_endpoint.py,sha256=ZnQ4DtbNIUr_Me5F6iYwMYsYhom8ZCs6A5kRjWwAANA,8695
-sglang/backend/vertexai.py,sha256=XNkbUzOdLIz-1qP_BBieYIfUXZf6gsfdghlaulNpBM8,4714
-sglang/lang/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-sglang/lang/chat_template.py,sha256=-pmALN5tV2upD5kb4RKP7DOvQY1s4nuvRdOcXKclXnw,11260
-sglang/lang/compiler.py,sha256=wNn_UqV6Sxl22mv-PpzFUtRgiFFV-Y4OYpO4LshEoRM,7527
-sglang/lang/interpreter.py,sha256=W1uwgTJqeHXrkG3K7mZfH8JX9Oc9poYIwtCWRIH7lhI,27251
-sglang/lang/ir.py,sha256=8Ap-uEUz6K9eNQTOKtMixePuLwRFHFKcN0Z5Yn44nKk,13320
-sglang/lang/tracer.py,sha256=vArGy7RNUP0qzE26kohsIHWRIfB0d88Ph2aiLq_P_fU,8284
-sglang/srt/backend_config.py,sha256=UIV6kIU2j-Xh0eoezn1aXcYIy0miftHsWFeAZwqpbGE,227
-sglang/srt/conversation.py,sha256=NwTVuQXd3NqPq5WCllaYUgPLG2w2pMMbzIKDQfJMMO0,15491
-sglang/srt/hf_transformers_utils.py,sha256=mwDuBMZcp66U6hZWpiO1KeOmjXXXG9fbX_ZwEqjzzn0,5286
-sglang/srt/memory_pool.py,sha256=5bqI8d5_JURbKwIhv1BwlcIO2IDHewHvIqezPG-b_5M,3284
-sglang/srt/mm_utils.py,sha256=OptgAHDX-73Bk4jAdr2BOAJtiEXJNzPrMhaM-dy275c,8889
-sglang/srt/model_config.py,sha256=Tw13FKY7qA4hJOskl3gmdb_W5gTEoB2m0PEArUiINQU,1546
-sglang/srt/openai_api_adapter.py,sha256=w3zvahyzvCnQd2pphQ6ViRBgHJmyI-TyIul6Q-CBY5Q,13214
-sglang/srt/openai_protocol.py,sha256=87pLM0hxocd5LUvhYopnL61cEKz3iu8TKdJtHbk3C5o,5211
-sglang/srt/sampling_params.py,sha256=dQbVr7JmTJ9JEn_sy3clB56yT9kyr9ldWFZ-GaNXOy0,3023
-sglang/srt/server.py,sha256=SQXIo9XLG0fuW123-UF4VA0Os75I73upQoAzZ_U2su8,9923
-sglang/srt/server_args.py,sha256=ySWe8RA4ukJQTnN4rs4_42XoYcVz1XPfeT8Ps551MlY,9510
-sglang/srt/utils.py,sha256=n8OLrrbdNbA6ow1s2wbJU7a35fHGQmnFfewcgzTBecE,9201
-sglang/srt/weight_utils.py,sha256=TBNP9jWb32gohPLj4-qWRn_Yn64gqWk1ZGLWrv967uU,15930
-sglang/srt/constrained/__init__.py,sha256=BPRNDJnWtzYJ13X4urRS5aE6wFuwAVNBA9qeWIHF8rE,1236
-sglang/srt/constrained/base_cache.py,sha256=QQjmFEiT8jlOskJoZobhrDl2TKB-B4b1LPQo9JQCP_w,1405
-sglang/srt/constrained/fsm_cache.py,sha256=B9FPtpqzm4jKqciXTbfgNJL44hV2-rUG6-omDECN7iA,902
-sglang/srt/constrained/jump_forward.py,sha256=fUa4AlnGX40gYiWTLuICTJfq4b7wA3AL5dydTqT3jz4,2483
-sglang/srt/layers/context_flashattention_nopad.py,sha256=bENdVltDozccR5mLY_CcYDjqLob28tHA9f2s03D8UFQ,5210
-sglang/srt/layers/extend_attention.py,sha256=5gvRggy6qPLrLvjctoMMsYh1w70mOGxiPjxstHqjqsY,12623
-sglang/srt/layers/logits_processor.py,sha256=Vbkr6ANNfiBGkkNobqjNm1KQTqtuYQWZvmPjhhIWnS8,7267
-sglang/srt/layers/radix_attention.py,sha256=PBucvAdGI27Z1qQOUxUi-YJp-tKGm6LX3L2kp99pOV4,5598
-sglang/srt/layers/token_attention.py,sha256=Wm-Gj0VdmFE8krZeHjDWic9dmVxRvg1WRAIHbbA3M34,8517
-sglang/srt/managers/detokenizer_manager.py,sha256=-zuI2ZLyLD3wf21u8xWZm91JkcZZ57DwUFbFxnP2vFI,3462
-sglang/srt/managers/io_struct.py,sha256=hdCHrBMoZ_4vc2l6mgbGGOW5b8STd4GSlQm-J_BCmw0,3716
-sglang/srt/managers/tokenizer_manager.py,sha256=hgLGkZYWs5enyeJzDjht6hOjSjTEBJSvUrFHNmjszbQ,11900
-sglang/srt/managers/router/infer_batch.py,sha256=CsNErbPt2XxoUxA3MkQeP4Tr3ipNK7eF0_K7IxdEpeY,19920
-sglang/srt/managers/router/manager.py,sha256=iNmLd-0V0aTU-B3FH6YutmcKJVtuhRcTU28EqbU8PII,2683
-sglang/srt/managers/router/model_rpc.py,sha256=8fDGBsqyo8lAFhr4_N6rB3D3we7zTfyjeV36IR1M7Ds,28325
-sglang/srt/managers/router/model_runner.py,sha256=k7YMEvqU3GSIGpaBde2rCoGlWDpVjTOJgO-3xrsz0uI,16545
-sglang/srt/managers/router/radix_cache.py,sha256=ZXSYyUb2e_xHwXDi_c9U6g2-0zmX3c_wX9UWs33F6u4,6685
-sglang/srt/managers/router/scheduler.py,sha256=V-LAnVSzgD2ddy2eXW3jWURCeq9Lv7YxCGk4kHyytfM,2818
-sglang/srt/models/commandr.py,sha256=GHcgyksXAnp4Nlnij1qULpFk0D1iA_lV3SzhLBD6Yus,13599
-sglang/srt/models/dbrx.py,sha256=OK9xmb9f1m-nrO3yFB7bvy7u6ofyobaKU2fsa0oIteQ,14158
-sglang/srt/models/dbrx_config.py,sha256=6EKMCAP1kS4pkQ9Ycr39PeEeTCPG4JhKRm2rtA4jS2s,11071
-sglang/srt/models/gemma.py,sha256=Y4iLdmH4U_oySEk2-UrxqXsW3tsT_vnY0bJFywxdRyU,11630
-sglang/srt/models/llama2.py,sha256=lAYVI5bE1oy_jY0tvSvRSI9wxfalidNtIZc8VXEsaNQ,11648
-sglang/srt/models/llava.py,sha256=ocaWPocml74UoUHaAKE0oWF7Je5Dw_3fXw1c7b53zKk,14941
-sglang/srt/models/mistral.py,sha256=XSn7fiZqspyWVTYrpVAacAnWdwAybBtyn9-Sh9AvMTM,254
-sglang/srt/models/mixtral.py,sha256=jC7LR9NWjeQE9I28TfNeNGy65GdzcH3kxdWfIocpvho,13892
-sglang/srt/models/qwen.py,sha256=eGkWsgYAhXVNkcS9iR8T3pk65UnIdTRjzSnRveYdigQ,9320
-sglang/srt/models/qwen2.py,sha256=nXF5UJlgVFuY5TjDL2nqOy4_R1xn73EYpzHj2mL5odU,11344
-sglang/srt/models/stablelm.py,sha256=d1pP5e-6CtOppWRzUtQar_0ULhGIHDZlXTh9lKMWbv4,10828
-sglang/srt/models/yivl.py,sha256=Aoo_AlGu9PYMDvj6bQj9PX7Ui7-oIe9MArLe5N6FAno,4406
-sglang/test/test_conversation.py,sha256=1zIrXcXiwEliPHgDAsqsQUA7JKzZ5fnQEU-U6L887FU,1592
-sglang/test/test_openai_protocol.py,sha256=eePzoskYR3PqfWczSVZvg8ja63qbT8TFUNEMyzDZpa8,1657
-sglang/test/test_programs.py,sha256=-2AoddzOOmXoj3muVUKX6Uih63UNTm3MFg2fcNnsy7Y,11498
-sglang/test/test_utils.py,sha256=9VFNGUMW0LBvmtDEHZ7ponakv5ZVF7B2Lg3xX353DXw,10083
-sglang-0.1.15.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
-sglang-0.1.15.dist-info/METADATA,sha256=9pKA1HIo0OFpZz-peDJCVekVjaZvqj55sK3n5Dchd4A,28727
-sglang-0.1.15.dist-info/WHEEL,sha256=GJ7t_kWBFywbagK5eo9IoUwLW6oyOeTKmQ-9iHFVNxQ,92
-sglang-0.1.15.dist-info/top_level.txt,sha256=yxhh3pYQkcnA7v3Bg889C2jZhvtJdEincysO7PEB09M,7
-sglang-0.1.15.dist-info/RECORD,,