Update README.md
Browse files
README.md
CHANGED
|
@@ -1,19 +1,19 @@
|
|
| 1 |
---
|
| 2 |
license: mit
|
| 3 |
language:
|
| 4 |
-
- en
|
| 5 |
-
- ko
|
| 6 |
-
- code
|
| 7 |
library_name: transformers
|
| 8 |
tags:
|
| 9 |
-
- code-llama
|
| 10 |
-
- code-review
|
| 11 |
-
- fine-tuning
|
| 12 |
-
- SFT
|
| 13 |
-
- LoRA
|
| 14 |
pipeline_tag: text-generation
|
| 15 |
base_model:
|
| 16 |
-
- codellama/CodeLlama-7b-hf
|
| 17 |
---
|
| 18 |
|
| 19 |
# Model Card for codellama-7b-code-review
|
|
@@ -23,11 +23,11 @@ base_model:
|
|
| 23 |
## Model Details / λͺ¨λΈ μμΈ μ 보
|
| 24 |
|
| 25 |
<details>
|
| 26 |
-
<summary><strong
|
| 27 |
|
| 28 |
This model is fine-tuned from Meta's `codellama/CodeLlama-7b-hf` to review and provide feedback on code changes (`diffs`) from GitHub Pull Requests. It has been primarily trained on JavaScript and React code reviews, aiming to generate constructive feedback from a senior engineer's perspective on topics like code quality, architecture, performance, and conventions.
|
| 29 |
|
| 30 |
-
- **Developed by:** [
|
| 31 |
- **Model type:** Causal Language Model
|
| 32 |
- **Language(s):** English, Korean, Diff format
|
| 33 |
- **License:** apache-2.0
|
|
@@ -36,11 +36,11 @@ This model is fine-tuned from Meta's `codellama/CodeLlama-7b-hf` to review and p
|
|
| 36 |
</details>
|
| 37 |
|
| 38 |
<details>
|
| 39 |
-
<summary><strong
|
| 40 |
|
| 41 |
μ΄ λͺ¨λΈμ Metaμ `codellama/CodeLlama-7b-hf` λͺ¨λΈμ κΈ°λ°μΌλ‘, GitHub Pull Requestμ μ½λ λ³κ²½μ¬ν(`diff`)μ 리뷰νκ³ νΌλλ°±μ μ 곡νλλ‘ νμΈνλλμμ΅λλ€. μ£Όλ‘ JavaScriptμ React μ½λ 리뷰μ μ€μ μ λκ³ νμ΅λμμΌλ©°, μλμ΄ μμ§λμ΄μ κ΄μ μμ μ½λ νμ§, μν€ν
μ², μ±λ₯, 컨벀μ
λ±μ λν 건μ€μ μΈ νΌλλ°±μ μμ±νλ κ²μ λͺ©νλ‘ ν©λλ€.
|
| 42 |
|
| 43 |
-
- **κ°λ°μ:** [
|
| 44 |
- **λͺ¨λΈ μ’
λ₯:** μΈκ³Ό κ΄κ³ μΈμ΄ λͺ¨λΈ (Causal Language Model)
|
| 45 |
- **μΈμ΄:** μμ΄, νκ΅μ΄, Diff νμ
|
| 46 |
- **λΌμ΄μ μ€:** apache-2.0
|
|
@@ -55,7 +55,7 @@ This model is fine-tuned from Meta's `codellama/CodeLlama-7b-hf` to review and p
|
|
| 55 |
## Uses / μ¬μ© μ 보
|
| 56 |
|
| 57 |
<details>
|
| 58 |
-
<summary><strong
|
| 59 |
|
| 60 |
### Direct Use
|
| 61 |
|
|
@@ -74,7 +74,7 @@ This model is specialized for code review tasks. It may not perform well for oth
|
|
| 74 |
</details>
|
| 75 |
|
| 76 |
<details>
|
| 77 |
-
<summary><strong
|
| 78 |
|
| 79 |
### μ§μ μ¬μ©
|
| 80 |
|
|
@@ -95,7 +95,7 @@ This model is specialized for code review tasks. It may not perform well for oth
|
|
| 95 |
## Bias, Risks, and Limitations / νΈν₯, μν λ° νκ³
|
| 96 |
|
| 97 |
<details>
|
| 98 |
-
<summary><strong
|
| 99 |
|
| 100 |
- **Data Bias:** The model was trained on public GitHub Pull Request data, so it may be biased towards specific coding styles or patterns present in that data.
|
| 101 |
- **Inaccuracy (Hallucination):** The model may occasionally generate feedback that is factually incorrect or out of context. The generated reviews always need verification.
|
|
@@ -103,7 +103,7 @@ This model is specialized for code review tasks. It may not perform well for oth
|
|
| 103 |
</details>
|
| 104 |
|
| 105 |
<details>
|
| 106 |
-
<summary><strong
|
| 107 |
|
| 108 |
- **λ°μ΄ν° νΈν₯:** λͺ¨λΈμ 곡κ°λ GitHub Pull Request λ°μ΄ν°λ₯Ό κΈ°λ°μΌλ‘ νμ΅λμμΌλ―λ‘, ν΄λΉ λ°μ΄ν°μ μ‘΄μ¬νλ νΉμ μ½λ© μ€νμΌμ΄λ ν¨ν΄μ νΈν₯λμ΄ μμ μ μμ΅λλ€.
|
| 109 |
- **λΆμ νμ±(νκ°):** λͺ¨λΈμ λλλ‘ μ¬μ€κ³Ό λ€λ₯΄κ±°λ λ¬Έλ§₯μ λ§μ§ μλ νΌλλ°±μ μμ±ν μ μμ΅λλ€. μμ±λ 리뷰λ νμ κ²μ¦μ΄ νμν©λλ€.
|
|
@@ -113,19 +113,19 @@ This model is specialized for code review tasks. It may not perform well for oth
|
|
| 113 |
### Recommendations / κΆμ₯ μ¬ν
|
| 114 |
|
| 115 |
<details>
|
| 116 |
-
<summary><strong
|
| 117 |
Users should treat the code reviews generated by the model as a 'draft' or 'assistive tool' to help the development process, not as a final judgment. It is recommended that a human expert reviews critical changes.
|
| 118 |
</details>
|
| 119 |
|
| 120 |
<details>
|
| 121 |
-
<summary><strong
|
| 122 |
μ¬μ©μλ λͺ¨λΈμ΄ μμ±ν μ½λ 리뷰λ₯Ό μ΅μ’
μ μΈ νλ¨μ΄ μλ, κ°λ° κ³Όμ μ λλ 'μ΄μ' λλ '보쑰 λꡬ'λ‘ νμ©ν΄μΌ ν©λλ€. μ€μν λ³κ²½μ¬νμ λν΄μλ λ°λμ μΈκ° μ λ¬Έκ°μ κ²ν λ₯Ό κ±°μΉλ κ²μ κΆμ₯ν©λλ€.
|
| 123 |
</details>
|
| 124 |
|
| 125 |
## How to Get Started with the Model / λͺ¨λΈ μμνκΈ°
|
| 126 |
|
| 127 |
<details>
|
| 128 |
-
<summary><strong
|
| 129 |
|
| 130 |
**Note:** This model may be available in two versions: **Adapter** and **Merged**. Use the appropriate code for your model type.
|
| 131 |
|
|
@@ -140,15 +140,15 @@ If the model is fully merged with the base model, you can load it directly witho
|
|
| 140 |
</details>
|
| 141 |
|
| 142 |
<details>
|
| 143 |
-
<summary><strong
|
| 144 |
|
| 145 |
**μ°Έκ³ :** μ΄ λͺ¨λΈμ **μ΄λν°(Adapter)** μ **λ³ν©λ(Merged)** λ κ°μ§ λ²μ μΌλ‘ μ 곡λ μ μμ΅λλ€. μμ μ λͺ¨λΈ νμ
μ λ§λ μ½λλ₯Ό μ¬μ©νμΈμ.
|
| 146 |
|
| 147 |
-
#### 1. μ΄λν° λͺ¨λΈ μ¬μ©λ² (`
|
| 148 |
|
| 149 |
μ΄λν° λͺ¨λΈμ μ¬μ©νλ €λ©΄, κΈ°λ° λͺ¨λΈμ λ¨Όμ λ‘λν ν `peft` λΌμ΄λΈλ¬λ¦¬λ₯Ό μ¬μ©ν΄ μ΄λν°λ₯Ό μ μ©ν΄μΌ ν©λλ€.
|
| 150 |
|
| 151 |
-
#### 2. λ³ν©λ λͺ¨λΈ μ¬μ©λ² (`
|
| 152 |
|
| 153 |
λͺ¨λΈμ΄ κΈ°λ° λͺ¨λΈκ³Ό μμ ν λ³ν©λ κ²½μ°, `peft` μμ΄ μ§μ λͺ¨λΈμ λ‘λνμ¬ μ¬μ©ν μ μμ΅λλ€.
|
| 154 |
|
|
@@ -213,39 +213,42 @@ diff_code = """
|
|
| 213 |
"""
|
| 214 |
|
| 215 |
# Prompt in Korean
|
|
|
|
|
|
|
| 216 |
prompt = f"""### μ§μ:
|
| 217 |
μ 곡λ μ½λλ pull requestμ diff λ΄μ©μ
λλ€. μ½λμ κ°μ ν μ μλ λΆλΆμ λν΄ μ΅μ 3κ°μ§ νλͺ©μΌλ‘ λλμ΄ μμΈνκ³ κ΅¬μ²΄μ μΈ νΌλλ°±μ μ 곡ν΄μ£ΌμΈμ.
|
| 218 |
|
| 219 |
### μ
λ ₯:
|
| 220 |
-
|
| 221 |
{diff_code}
|
| 222 |
-
|
| 223 |
|
| 224 |
### μλ΅:
|
| 225 |
-
|
| 226 |
1. """
|
| 227 |
|
| 228 |
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
|
| 229 |
-
outputs = model.generate(
|
| 230 |
response = tokenizer.decode(outputs[0]len(inputs.input_ids[0]):], skip_special_tokens=True)
|
| 231 |
|
| 232 |
print(response)
|
| 233 |
|
| 234 |
-
|
| 235 |
-
|
| 236 |
|
| 237 |
## Training Details / νμ΅ μμΈ μ 보
|
| 238 |
|
| 239 |
<details>
|
| 240 |
-
<summary><strong
|
| 241 |
|
| 242 |
### Training Data
|
|
|
|
| 243 |
This model was fine-tuned using the `review_dataset.json` file, which contains public Pull Request data collected from GitHub. The dataset is structured in a `instruction`, `input`(diff), `output`(review comment) format.
|
| 244 |
|
| 245 |
### Training Procedure
|
|
|
|
| 246 |
The model was fine-tuned using the QLoRA technique. It utilized the `SFTTrainer` from the `trl` library, applying 4-bit quantization and LoRA (Low-Rank Adaptation) for efficient training.
|
| 247 |
|
| 248 |
#### Training Hyperparameters
|
|
|
|
| 249 |
- **model:** `codellama/CodeLlama-7b-hf`
|
| 250 |
- **max_seq_length:** 4096
|
| 251 |
- **lora_alpha:** 128
|
|
@@ -260,15 +263,18 @@ The model was fine-tuned using the QLoRA technique. It utilized the `SFTTrainer`
|
|
| 260 |
</details>
|
| 261 |
|
| 262 |
<details>
|
| 263 |
-
<summary><strong
|
| 264 |
|
| 265 |
### νμ΅ λ°μ΄ν°
|
|
|
|
| 266 |
μ΄ λͺ¨λΈμ GitHubμμ μμ§λ κ³΅κ° Pull Request λ°μ΄ν°λ₯Ό ν¬ν¨νλ `review_dataset.json` νμΌμ μ¬μ©νμ¬ νμΈνλλμμ΅λλ€. λ°μ΄ν°μ
μ `instruction`, `input`(diff), `output`(리뷰 μ½λ©νΈ) νμμΌλ‘ ꡬμ±λμ΄ μμ΅λλ€.
|
| 267 |
|
| 268 |
### νμ΅ μ μ°¨
|
|
|
|
| 269 |
λͺ¨λΈμ QLoRA κΈ°λ²μ μ¬μ©νμ¬ νμΈνλλμμ΅λλ€. `trl` λΌμ΄λΈλ¬λ¦¬μ `SFTTrainer`λ₯Ό μ¬μ©νμΌλ©°, 4-bit μμνμ LoRA(Low-Rank Adaptation)λ₯Ό μ μ©νμ¬ ν¨μ¨μ μΈ νμ΅μ μ§ννμ΅λλ€.
|
| 270 |
|
| 271 |
#### νμ΅ νμ΄νΌνλΌλ―Έν°
|
|
|
|
| 272 |
- **λͺ¨λΈ:** `codellama/CodeLlama-7b-hf`
|
| 273 |
- **μ΅λ μνμ€ κΈΈμ΄:** 4096
|
| 274 |
- **LoRA Alpha:** 128
|
|
@@ -285,16 +291,19 @@ The model was fine-tuned using the QLoRA technique. It utilized the `SFTTrainer`
|
|
| 285 |
## Compute Infrastructure / μ»΄ν¨ν
μΈνλΌ
|
| 286 |
|
| 287 |
<details>
|
| 288 |
-
<summary><strong
|
| 289 |
|
| 290 |
- **Hardware Type:** RunPod Cloud GPU
|
| 291 |
- **Cloud Provider:** RunPod
|
| 292 |
</details>
|
| 293 |
|
| 294 |
<details>
|
| 295 |
-
<summary><strong
|
| 296 |
|
| 297 |
- **νλμ¨μ΄ μ’
λ₯:** RunPod ν΄λΌμ°λ GPU
|
| 298 |
- **ν΄λΌμ°λ μ 곡μ
체:** RunPod
|
| 299 |
</details>
|
| 300 |
-
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: mit
|
| 3 |
language:
|
| 4 |
+
- en
|
| 5 |
+
- ko
|
| 6 |
+
- code
|
| 7 |
library_name: transformers
|
| 8 |
tags:
|
| 9 |
+
- code-llama
|
| 10 |
+
- code-review
|
| 11 |
+
- fine-tuning
|
| 12 |
+
- SFT
|
| 13 |
+
- LoRA
|
| 14 |
pipeline_tag: text-generation
|
| 15 |
base_model:
|
| 16 |
+
- codellama/CodeLlama-7b-hf
|
| 17 |
---
|
| 18 |
|
| 19 |
# Model Card for codellama-7b-code-review
|
|
|
|
| 23 |
## Model Details / λͺ¨λΈ μμΈ μ 보
|
| 24 |
|
| 25 |
<details>
|
| 26 |
+
<summary><strong>πΊπΈ English</strong></summary>
|
| 27 |
|
| 28 |
This model is fine-tuned from Meta's `codellama/CodeLlama-7b-hf` to review and provide feedback on code changes (`diffs`) from GitHub Pull Requests. It has been primarily trained on JavaScript and React code reviews, aiming to generate constructive feedback from a senior engineer's perspective on topics like code quality, architecture, performance, and conventions.
|
| 29 |
|
| 30 |
+
- **Developed by:** [ken12377](https://huggingface.co/ken12377)
|
| 31 |
- **Model type:** Causal Language Model
|
| 32 |
- **Language(s):** English, Korean, Diff format
|
| 33 |
- **License:** apache-2.0
|
|
|
|
| 36 |
</details>
|
| 37 |
|
| 38 |
<details>
|
| 39 |
+
<summary><strong>π°π· νκ΅μ΄</strong></summary>
|
| 40 |
|
| 41 |
μ΄ λͺ¨λΈμ Metaμ `codellama/CodeLlama-7b-hf` λͺ¨λΈμ κΈ°λ°μΌλ‘, GitHub Pull Requestμ μ½λ λ³κ²½μ¬ν(`diff`)μ 리뷰νκ³ νΌλλ°±μ μ 곡νλλ‘ νμΈνλλμμ΅λλ€. μ£Όλ‘ JavaScriptμ React μ½λ 리뷰μ μ€μ μ λκ³ νμ΅λμμΌλ©°, μλμ΄ μμ§λμ΄μ κ΄μ μμ μ½λ νμ§, μν€ν
μ², μ±λ₯, 컨벀μ
λ±μ λν 건μ€μ μΈ νΌλλ°±μ μμ±νλ κ²μ λͺ©νλ‘ ν©λλ€.
|
| 42 |
|
| 43 |
+
- **κ°λ°μ:** [ken12377](https://huggingface.co/ken12377)
|
| 44 |
- **λͺ¨λΈ μ’
λ₯:** μΈκ³Ό κ΄κ³ μΈμ΄ λͺ¨λΈ (Causal Language Model)
|
| 45 |
- **μΈμ΄:** μμ΄, νκ΅μ΄, Diff νμ
|
| 46 |
- **λΌμ΄μ μ€:** apache-2.0
|
|
|
|
| 55 |
## Uses / μ¬μ© μ 보
|
| 56 |
|
| 57 |
<details>
|
| 58 |
+
<summary><strong>πΊπΈ English</strong></summary>
|
| 59 |
|
| 60 |
### Direct Use
|
| 61 |
|
|
|
|
| 74 |
</details>
|
| 75 |
|
| 76 |
<details>
|
| 77 |
+
<summary><strong>π°π· νκ΅μ΄</strong></summary>
|
| 78 |
|
| 79 |
### μ§μ μ¬μ©
|
| 80 |
|
|
|
|
| 95 |
## Bias, Risks, and Limitations / νΈν₯, μν λ° νκ³
|
| 96 |
|
| 97 |
<details>
|
| 98 |
+
<summary><strong>πΊπΈ English</strong></summary>
|
| 99 |
|
| 100 |
- **Data Bias:** The model was trained on public GitHub Pull Request data, so it may be biased towards specific coding styles or patterns present in that data.
|
| 101 |
- **Inaccuracy (Hallucination):** The model may occasionally generate feedback that is factually incorrect or out of context. The generated reviews always need verification.
|
|
|
|
| 103 |
</details>
|
| 104 |
|
| 105 |
<details>
|
| 106 |
+
<summary><strong>π°π· νκ΅μ΄</strong></summary>
|
| 107 |
|
| 108 |
- **λ°μ΄ν° νΈν₯:** λͺ¨λΈμ 곡κ°λ GitHub Pull Request λ°μ΄ν°λ₯Ό κΈ°λ°μΌλ‘ νμ΅λμμΌλ―λ‘, ν΄λΉ λ°μ΄ν°μ μ‘΄μ¬νλ νΉμ μ½λ© μ€νμΌμ΄λ ν¨ν΄μ νΈν₯λμ΄ μμ μ μμ΅λλ€.
|
| 109 |
- **λΆμ νμ±(νκ°):** λͺ¨λΈμ λλλ‘ μ¬μ€κ³Ό λ€λ₯΄κ±°λ λ¬Έλ§₯μ λ§μ§ μλ νΌλλ°±μ μμ±ν μ μμ΅λλ€. μμ±λ 리뷰λ νμ κ²μ¦μ΄ νμν©λλ€.
|
|
|
|
| 113 |
### Recommendations / κΆμ₯ μ¬ν
|
| 114 |
|
| 115 |
<details>
|
| 116 |
+
<summary><strong>πΊπΈ English</strong></summary>
|
| 117 |
Users should treat the code reviews generated by the model as a 'draft' or 'assistive tool' to help the development process, not as a final judgment. It is recommended that a human expert reviews critical changes.
|
| 118 |
</details>
|
| 119 |
|
| 120 |
<details>
|
| 121 |
+
<summary><strong>π°π· νκ΅μ΄</strong></summary>
|
| 122 |
μ¬μ©μλ λͺ¨λΈμ΄ μμ±ν μ½λ 리뷰λ₯Ό μ΅μ’
μ μΈ νλ¨μ΄ μλ, κ°λ° κ³Όμ μ λλ 'μ΄μ' λλ '보쑰 λꡬ'λ‘ νμ©ν΄μΌ ν©λλ€. μ€μν λ³κ²½μ¬νμ λν΄μλ λ°λμ μΈκ° μ λ¬Έκ°μ κ²ν λ₯Ό κ±°μΉλ κ²μ κΆμ₯ν©λλ€.
|
| 123 |
</details>
|
| 124 |
|
| 125 |
## How to Get Started with the Model / λͺ¨λΈ μμνκΈ°
|
| 126 |
|
| 127 |
<details>
|
| 128 |
+
<summary><strong>πΊπΈ English</strong></summary>
|
| 129 |
|
| 130 |
**Note:** This model may be available in two versions: **Adapter** and **Merged**. Use the appropriate code for your model type.
|
| 131 |
|
|
|
|
| 140 |
</details>
|
| 141 |
|
| 142 |
<details>
|
| 143 |
+
<summary><strong>π°π· νκ΅μ΄</strong></summary>
|
| 144 |
|
| 145 |
**μ°Έκ³ :** μ΄ λͺ¨λΈμ **μ΄λν°(Adapter)** μ **λ³ν©λ(Merged)** λ κ°μ§ λ²μ μΌλ‘ μ 곡λ μ μμ΅λλ€. μμ μ λͺ¨λΈ νμ
μ λ§λ μ½λλ₯Ό μ¬μ©νμΈμ.
|
| 146 |
|
| 147 |
+
#### 1. μ΄λν° λͺ¨λΈ μ¬μ©λ² (`ken12377/codellama-7b-code-review-adapter`)
|
| 148 |
|
| 149 |
μ΄λν° λͺ¨λΈμ μ¬μ©νλ €λ©΄, κΈ°λ° λͺ¨λΈμ λ¨Όμ λ‘λν ν `peft` λΌμ΄λΈλ¬λ¦¬λ₯Ό μ¬μ©ν΄ μ΄λν°λ₯Ό μ μ©ν΄μΌ ν©λλ€.
|
| 150 |
|
| 151 |
+
#### 2. λ³ν©λ λͺ¨λΈ μ¬μ©λ² (`ken12377/codellama-7b-code-review`)
|
| 152 |
|
| 153 |
λͺ¨λΈμ΄ κΈ°λ° λͺ¨λΈκ³Ό μμ ν λ³ν©λ κ²½μ°, `peft` μμ΄ μ§μ λͺ¨λΈμ λ‘λνμ¬ μ¬μ©ν μ μμ΅λλ€.
|
| 154 |
|
|
|
|
| 213 |
"""
|
| 214 |
|
| 215 |
# Prompt in Korean
|
| 216 |
+
# λ§ν¬λ€μ΄ νμμ νΌλμ νΌνκΈ° μν΄ μ½λ λΈλ‘ ꡬλΆμλ₯Ό λ³μλ‘ λ§λ€μ΄ μ¬μ©ν©λλ€.
|
| 217 |
+
diff_block_delimiter = "```"
|
| 218 |
prompt = f"""### μ§μ:
|
| 219 |
μ 곡λ μ½λλ pull requestμ diff λ΄μ©μ
λλ€. μ½λμ κ°μ ν μ μλ λΆλΆμ λν΄ μ΅μ 3κ°μ§ νλͺ©μΌλ‘ λλμ΄ μμΈνκ³ κ΅¬μ²΄μ μΈ νΌλλ°±μ μ 곡ν΄μ£ΌμΈμ.
|
| 220 |
|
| 221 |
### μ
λ ₯:
|
| 222 |
+
{diff_block_delimiter}diff
|
| 223 |
{diff_code}
|
| 224 |
+
{diff_block_delimiter}
|
| 225 |
|
| 226 |
### μλ΅:
|
|
|
|
| 227 |
1. """
|
| 228 |
|
| 229 |
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
|
| 230 |
+
outputs = model.generate(**inputs, max_new_tokens=512, temperature=0.7, repetition_penalty=1.2)
|
| 231 |
response = tokenizer.decode(outputs[0]len(inputs.input_ids[0]):], skip_special_tokens=True)
|
| 232 |
|
| 233 |
print(response)
|
| 234 |
|
| 235 |
+
````
|
|
|
|
| 236 |
|
| 237 |
## Training Details / νμ΅ μμΈ μ 보
|
| 238 |
|
| 239 |
<details>
|
| 240 |
+
<summary><strong>πΊπΈ English</strong></summary>
|
| 241 |
|
| 242 |
### Training Data
|
| 243 |
+
|
| 244 |
This model was fine-tuned using the `review_dataset.json` file, which contains public Pull Request data collected from GitHub. The dataset is structured in a `instruction`, `input`(diff), `output`(review comment) format.
|
| 245 |
|
| 246 |
### Training Procedure
|
| 247 |
+
|
| 248 |
The model was fine-tuned using the QLoRA technique. It utilized the `SFTTrainer` from the `trl` library, applying 4-bit quantization and LoRA (Low-Rank Adaptation) for efficient training.
|
| 249 |
|
| 250 |
#### Training Hyperparameters
|
| 251 |
+
|
| 252 |
- **model:** `codellama/CodeLlama-7b-hf`
|
| 253 |
- **max_seq_length:** 4096
|
| 254 |
- **lora_alpha:** 128
|
|
|
|
| 263 |
</details>
|
| 264 |
|
| 265 |
<details>
|
| 266 |
+
<summary><strong>π°π· νκ΅μ΄</strong></summary>
|
| 267 |
|
| 268 |
### νμ΅ λ°μ΄ν°
|
| 269 |
+
|
| 270 |
μ΄ λͺ¨λΈμ GitHubμμ μμ§λ κ³΅κ° Pull Request λ°μ΄ν°λ₯Ό ν¬ν¨νλ `review_dataset.json` νμΌμ μ¬μ©νμ¬ νμΈνλλμμ΅λλ€. λ°μ΄ν°μ
μ `instruction`, `input`(diff), `output`(리뷰 μ½λ©νΈ) νμμΌλ‘ ꡬμ±λμ΄ μμ΅λλ€.
|
| 271 |
|
| 272 |
### νμ΅ μ μ°¨
|
| 273 |
+
|
| 274 |
λͺ¨λΈμ QLoRA κΈ°λ²μ μ¬μ©νμ¬ νμΈνλλμμ΅λλ€. `trl` λΌμ΄λΈλ¬λ¦¬μ `SFTTrainer`λ₯Ό μ¬μ©νμΌλ©°, 4-bit μμνμ LoRA(Low-Rank Adaptation)λ₯Ό μ μ©νμ¬ ν¨μ¨μ μΈ νμ΅μ μ§ννμ΅λλ€.
|
| 275 |
|
| 276 |
#### νμ΅ νμ΄νΌνλΌλ―Έν°
|
| 277 |
+
|
| 278 |
- **λͺ¨λΈ:** `codellama/CodeLlama-7b-hf`
|
| 279 |
- **μ΅λ μνμ€ κΈΈμ΄:** 4096
|
| 280 |
- **LoRA Alpha:** 128
|
|
|
|
| 291 |
## Compute Infrastructure / μ»΄ν¨ν
μΈνλΌ
|
| 292 |
|
| 293 |
<details>
|
| 294 |
+
<summary><strong>πΊπΈ English</strong></summary>
|
| 295 |
|
| 296 |
- **Hardware Type:** RunPod Cloud GPU
|
| 297 |
- **Cloud Provider:** RunPod
|
| 298 |
</details>
|
| 299 |
|
| 300 |
<details>
|
| 301 |
+
<summary><strong>π°π· νκ΅μ΄</strong></summary>
|
| 302 |
|
| 303 |
- **νλμ¨μ΄ μ’
λ₯:** RunPod ν΄λΌμ°λ GPU
|
| 304 |
- **ν΄λΌμ°λ μ 곡μ
체:** RunPod
|
| 305 |
</details>
|
| 306 |
+
|
| 307 |
+
```
|
| 308 |
+
|
| 309 |
+
```
|