fixed prompt_bench.py

pipelines/tools/batch_bill_summarizer.py

@@ -23,14 +23,10 @@ import httpx
 import typer
 from tiktoken import Encoding, get_encoding
 
+from pipelines.config import get_config_dir
 from pipelines.tools.bill_token_compression import compress_bill_text
 
-_PROMPTS_PATH = (
-    Path(__file__).resolve().parents[2]
-    / "config"
-    / "prompts"
-    / "summarization_prompts.toml"
-)
+_PROMPTS_PATH = get_config_dir() / "prompts" / "summarization_prompts.toml"
 _PROMPTS = tomllib.loads(_PROMPTS_PATH.read_text())["summarization"]
 SUMMARIZATION_SYSTEM_PROMPT: str = _PROMPTS["system_prompt"]
 SUMMARIZATION_USER_TEMPLATE: str = _PROMPTS["user_template"]

pipelines/tools/compresion_test.py

@@ -24,14 +24,10 @@ from typing import Annotated
 import httpx
 import typer
 
+from pipelines.config import get_config_dir
 from pipelines.tools.bill_token_compression import compress_bill_text
 
-_PROMPTS_PATH = (
-    Path(__file__).resolve().parents[2]
-    / "config"
-    / "prompts"
-    / "summarization_prompts.toml"
-)
+_PROMPTS_PATH = get_config_dir() / "prompts" / "summarization_prompts.toml"
 _PROMPTS = tomllib.loads(_PROMPTS_PATH.read_text())["summarization"]
 SUMMARIZATION_SYSTEM_PROMPT: str = _PROMPTS["system_prompt"]
 SUMMARIZATION_USER_TEMPLATE: str = _PROMPTS["user_template"]

pipelines/tools/finetune.py

@@ -25,6 +25,8 @@ from datasets import Dataset
 from transformers import TrainingArguments
 from trl import SFTTrainer
 
+from pipelines.config import default_config_path
+
 logger = logging.getLogger(__name__)
 
 
@@ -123,7 +125,7 @@ def main(
     config_path: Annotated[
         Path,
         typer.Option("--config", help="TOML config file"),
-    ] = Path(__file__).parent / "config.toml",
+    ] = default_config_path(),
     save_gguf: Annotated[
         bool, typer.Option("--save-gguf/--no-save-gguf", help="Also save GGUF")
     ] = False,

pipelines/tools/prompt_bench.py

@@ -11,8 +11,8 @@ from typing import Annotated
 
 import typer
 
-from pipelines.tools.containers.lib import check_gpu_free
-from pipelines.tools.containers.vllm import start_vllm, stop_vllm
+from pipelines.containers.lib import check_gpu_free
+from pipelines.containers.vllm import start_vllm, stop_vllm
 from pipelines.tools.downloader import is_model_present
 from pipelines.tools.models import BenchmarkConfig
 from pipelines.tools.vllm_client import VLLMClient