brianbellx commited on
Commit
35e291c
·
verified ·
1 Parent(s): 161cb23

Fix refiner prompt and use validated Qwen 1.5B model

Browse files

100 tests passed; frontend build and hygiene gate passed. Five-demo greedy CPU check: kept improvements increased from 0/5 to 3/5. Scorer pin retained.

Files changed (3) hide show
  1. Dockerfile +2 -2
  2. README.md +2 -2
  3. server/prompt.py +1 -1
Dockerfile CHANGED
@@ -13,8 +13,8 @@ RUN cd web && npm run build
13
 
14
  FROM python:3.12-slim
15
  ARG CODEHYGIENE_REF
16
- ARG REFINER_REPO=Qwen/Qwen2.5-Coder-0.5B-Instruct-GGUF
17
- ARG REFINER_FILE=qwen2.5-coder-0.5b-instruct-q4_k_m.gguf
18
 
19
  RUN apt-get update \
20
  && apt-get install -y --no-install-recommends libgomp1 git \
 
13
 
14
  FROM python:3.12-slim
15
  ARG CODEHYGIENE_REF
16
+ ARG REFINER_REPO=Qwen/Qwen2.5-Coder-1.5B-Instruct-GGUF
17
+ ARG REFINER_FILE=qwen2.5-coder-1.5b-instruct-q4_k_m.gguf
18
 
19
  RUN apt-get update \
20
  && apt-get install -y --no-install-recommends libgomp1 git \
README.md CHANGED
@@ -21,8 +21,8 @@ The scorer runs in the browser. The refiner runs here on the CPU. A patch is kep
21
  | Variable | Default | What it does |
22
  |---|---|---|
23
  | `REFINER_BACKEND` | `llama` | `llama` runs the GGUF model. `mock` replays recorded patches. |
24
- | `REFINER_REPO` | `Qwen/Qwen2.5-Coder-0.5B-Instruct-GGUF` | Build arg. Hub repo of the GGUF. |
25
- | `REFINER_FILE` | `qwen2.5-coder-0.5b-instruct-q4_k_m.gguf` | Build arg. GGUF file name. |
26
  | `REFINER_MODEL_PATH` | set by the Dockerfile | Path of the GGUF inside the image. |
27
  | `CODEHYGIENE_REF` | pinned commit | Build arg. Version of the scorer and web assets. |
28
  | `MAX_FILE_CHARS` | `12000` | Longer files are not sent to the model. |
 
21
  | Variable | Default | What it does |
22
  |---|---|---|
23
  | `REFINER_BACKEND` | `llama` | `llama` runs the GGUF model. `mock` replays recorded patches. |
24
+ | `REFINER_REPO` | `Qwen/Qwen2.5-Coder-1.5B-Instruct-GGUF` | Build arg. Hub repo of the GGUF. |
25
+ | `REFINER_FILE` | `qwen2.5-coder-1.5b-instruct-q4_k_m.gguf` | Build arg. GGUF file name. |
26
  | `REFINER_MODEL_PATH` | set by the Dockerfile | Path of the GGUF inside the image. |
27
  | `CODEHYGIENE_REF` | pinned commit | Build arg. Version of the scorer and web assets. |
28
  | `MAX_FILE_CHARS` | `12000` | Longer files are not sent to the model. |
server/prompt.py CHANGED
@@ -11,7 +11,7 @@ def render_chatml(messages: list[dict[str, str]]) -> str:
11
  return (
12
  f"<|im_start|>system\n{system['content']}<|im_end|>\n"
13
  f"<|im_start|>user\n{user['content']}<|im_end|>\n"
14
- "<|im_start|>assistant\n<think>\n\n</think>\n\n"
15
  )
16
 
17
 
 
11
  return (
12
  f"<|im_start|>system\n{system['content']}<|im_end|>\n"
13
  f"<|im_start|>user\n{user['content']}<|im_end|>\n"
14
+ "<|im_start|>assistant\n"
15
  )
16
 
17