Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -28,8 +28,8 @@ LLM-судьёй). `pass@1` при greedy-декодировании.
|
|
| 28 |
|
| 29 |
| Бенчмарк | Задач | pass@1 |
|
| 30 |
|---|---|---|
|
| 31 |
-
| HumanEval | 164 |
|
| 32 |
-
| MBPP (sanitized) | 257 |
|
| 33 |
|
| 34 |
## Сравнение с другими моделями по кодингу
|
| 35 |
|
|
@@ -38,7 +38,7 @@ LLM-судьёй). `pass@1` при greedy-декодировании.
|
|
| 38 |
|
| 39 |
| Модель | Параметров | HumanEval | MBPP |
|
| 40 |
|---|---|---|---|
|
| 41 |
-
| **TupoyCoder (ours)** | **308M** | **
|
| 42 |
| Salesforce CodeGen-350M-mono | 350M | 12.8 | — |
|
| 43 |
| Qwen2.5-Coder-0.5B-Instruct | 0.5B | ~30 | ~40 |
|
| 44 |
| TinyLlama-1.1B | 1.1B | ~10 | ~15 |
|
|
|
|
| 28 |
|
| 29 |
| Бенчмарк | Задач | pass@1 |
|
| 30 |
|---|---|---|
|
| 31 |
+
| HumanEval | 164 | 5.5% |
|
| 32 |
+
| MBPP (sanitized) | 257 | 7.4% |
|
| 33 |
|
| 34 |
## Сравнение с другими моделями по кодингу
|
| 35 |
|
|
|
|
| 38 |
|
| 39 |
| Модель | Параметров | HumanEval | MBPP |
|
| 40 |
|---|---|---|---|
|
| 41 |
+
| **TupoyCoder (ours)** | **308M** | **5.5** | **7.4** |
|
| 42 |
| Salesforce CodeGen-350M-mono | 350M | 12.8 | — |
|
| 43 |
| Qwen2.5-Coder-0.5B-Instruct | 0.5B | ~30 | ~40 |
|
| 44 |
| TinyLlama-1.1B | 1.1B | ~10 | ~15 |
|