Requirements: Update Exllamav2

v0.0.15

Signed-off-by: kingbri <bdashore3@proton.me>

backends/exllamav2/model.py

@@ -10,6 +10,7 @@ from exllamav2 import (
     ExLlamaV2Config,
     ExLlamaV2Cache,
     ExLlamaV2Cache_8bit,
+    ExLlamaV2Cache_Q4,
     ExLlamaV2Tokenizer,
     ExLlamaV2Lora,
 )
@@ -27,14 +28,6 @@ from common.templating import (
 from common.utils import coalesce, unwrap
 from common.logger import init_logger
 
-# Optional imports for dependencies
-try:
-    from exllamav2 import ExLlamaV2Cache_Q4
-
-    _exllamav2_has_int4 = True
-except ImportError:
-    _exllamav2_has_int4 = False
-
 logger = init_logger(__name__)
 
 
@@ -116,16 +109,7 @@ class ExllamaV2Container:
         """
 
         self.quiet = quiet
-
+        self.cache_mode = unwrap(kwargs.get("cache_mode"), "FP16")
-        cache_mode = unwrap(kwargs.get("cache_mode"), "FP16")
-        if cache_mode == "Q4" and not _exllamav2_has_int4:
-            logger.warning(
-                "Q4 cache is not available "
-                "in the currently installed ExllamaV2 version. Using FP16."
-            )
-            cache_mode = "FP16"
-
-        self.cache_mode = cache_mode
 
         # Turn off GPU split if the user is using 1 GPU
         gpu_count = torch.cuda.device_count()
@@ -415,7 +399,7 @@ class ExllamaV2Container:
 
         batch_size = 2 if self.use_cfg else 1
 
-        if self.cache_mode == "Q4" and _exllamav2_has_int4:
+        if self.cache_mode == "Q4":
             self.cache = ExLlamaV2Cache_Q4(
                 self.model, lazy=self.gpu_split_auto, batch_size=batch_size
             )

backends/exllamav2/utils.py

@@ -9,7 +9,7 @@ logger = init_logger(__name__)
 def check_exllama_version():
     """Verifies the exllama version"""
 
-    required_version = version.parse("0.0.14")
+    required_version = version.parse("0.0.15")
     current_version = version.parse(package_version("exllamav2").split("+")[0])
 
     if current_version < required_version:

requirements-amd.txt

@@ -3,8 +3,8 @@
 torch ~= 2.2
 
 # Exllamav2
-https://github.com/turboderp/exllamav2/releases/download/v0.0.14/exllamav2-0.0.14+rocm5.6-cp311-cp311-linux_x86_64.whl; python_version == "3.11"
+https://github.com/turboderp/exllamav2/releases/download/v0.0.15/exllamav2-0.0.15+rocm5.6-cp311-cp311-linux_x86_64.whl; python_version == "3.11"
-https://github.com/turboderp/exllamav2/releases/download/v0.0.14/exllamav2-0.0.14+rocm5.6-cp310-cp310-linux_x86_64.whl; python_version == "3.10"
+https://github.com/turboderp/exllamav2/releases/download/v0.0.15/exllamav2-0.0.15+rocm5.6-cp310-cp310-linux_x86_64.whl; python_version == "3.10"
 
 # Pip dependencies
 fastapi

requirements-cu118.txt

@@ -5,12 +5,12 @@ torch ~= 2.2
 # Exllamav2
 
 # Windows
-https://github.com/turboderp/exllamav2/releases/download/v0.0.14/exllamav2-0.0.14+cu118-cp311-cp311-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
+https://github.com/turboderp/exllamav2/releases/download/v0.0.15/exllamav2-0.0.15+cu121-cp311-cp311-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
-https://github.com/turboderp/exllamav2/releases/download/v0.0.14/exllamav2-0.0.14+cu118-cp310-cp310-win_amd64.whl; platform_system == "Windows" and python_version == "3.10"
+https://github.com/turboderp/exllamav2/releases/download/v0.0.15/exllamav2-0.0.15+cu118-cp310-cp310-win_amd64.whl; platform_system == "Windows" and python_version == "3.10"
 
 # Linux
-https://github.com/turboderp/exllamav2/releases/download/v0.0.14/exllamav2-0.0.14+cu118-cp311-cp311-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
+https://github.com/turboderp/exllamav2/releases/download/v0.0.15/exllamav2-0.0.15+cu118-cp311-cp311-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
-https://github.com/turboderp/exllamav2/releases/download/v0.0.14/exllamav2-0.0.14+cu118-cp310-cp310-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.10"
+https://github.com/turboderp/exllamav2/releases/download/v0.0.15/exllamav2-0.0.15+cu118-cp310-cp310-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.10"
 
 # Pip dependencies
 fastapi

requirements.txt

@@ -5,12 +5,12 @@ torch ~= 2.2
 # Exllamav2
 
 # Windows
-https://github.com/turboderp/exllamav2/releases/download/v0.0.14/exllamav2-0.0.14+cu121-cp311-cp311-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
+https://github.com/turboderp/exllamav2/releases/download/v0.0.15/exllamav2-0.0.15+cu121-cp311-cp311-win_amd64.whl; platform_system == "Windows" and python_version == "3.11"
-https://github.com/turboderp/exllamav2/releases/download/v0.0.14/exllamav2-0.0.14+cu121-cp310-cp310-win_amd64.whl; platform_system == "Windows" and python_version == "3.10"
+https://github.com/turboderp/exllamav2/releases/download/v0.0.15/exllamav2-0.0.15+cu121-cp310-cp310-win_amd64.whl; platform_system == "Windows" and python_version == "3.10"
 
 # Linux
-https://github.com/turboderp/exllamav2/releases/download/v0.0.14/exllamav2-0.0.14+cu121-cp311-cp311-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
+https://github.com/turboderp/exllamav2/releases/download/v0.0.15/exllamav2-0.0.15+cu121-cp311-cp311-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.11"
-https://github.com/turboderp/exllamav2/releases/download/v0.0.14/exllamav2-0.0.14+cu121-cp310-cp310-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.10"
+https://github.com/turboderp/exllamav2/releases/download/v0.0.15/exllamav2-0.0.15+cu121-cp310-cp310-linux_x86_64.whl; platform_system == "Linux" and platform_machine == "x86_64" and python_version == "3.10"
 
 # Pip dependencies
 fastapi