Use absolute imports (#90)

- Switch to absolute imports to ensure `vptq.ops` is correctly located
on a local setup.
- Allow running `vptq` as a script without requiring the `-m` flag.
- Resolve Issue #89 where CUDA kernel is not found in local setups.

---------

Co-authored-by: Yang Wang <wyatuestc@gmail.com>

vptq/layers/__init__.py

@@ -3,4 +3,4 @@
 # Licensed under the MIT License.
 # --------------------------------------------------------------------------
 
-from vptq.model_base import AutoModelForCausalLM as AutoModelForCausalLM
+from vptq.layers.model_base import AutoModelForCausalLM as AutoModelForCausalLM

vptq/layers/model_base.py

@@ -14,7 +14,7 @@
 import transformers
 from tqdm import tqdm
 
-from vptq.vqlinear import VQuantLinear
+from vptq.layers.vqlinear import VQuantLinear
 
 
 def set_op_by_name(layer, name, new_module):