raca-workspace-v1/autotrainer-v0
收藏Hugging Face2026-04-08 更新2026-04-12 收录
下载链接：
https://hf-mirror.com/datasets/raca-workspace-v1/autotrainer-v0
下载链接
链接失效反馈官方服务：
资源简介：
---
license: mit
tags:
- autotrainer
- rlvr
- countdown
- agent-controlled-training
configs:
- config_name: eval_traces
data_files:
- split: train
path: eval_traces/train-*
- config_name: rounds
data_files:
- split: train
path: rounds/train-*
- config_name: state
data_files:
- split: train
path: state/train-*
- config_name: train_traces
data_files:
- split: train
path: train_traces/train-*
dataset_info:
- config_name: eval_traces
features:
- name: question_id
dtype: string
- name: input
dtype: string
- name: target
dtype: int64
- name: numbers
dtype: string
- name: expected_answer
dtype: string
- name: model_response
dtype: string
- name: correct
dtype: bool
- name: score
dtype: float64
- name: error_type
dtype: string
- name: num_tokens
dtype: int64
- name: round_id
dtype: string
splits:
- name: train
num_bytes: 2511714
num_examples: 668
download_size: 433635
dataset_size: 2511714
- config_name: rounds
features:
- name: round_id
dtype: string
- name: config
dtype: string
- name: reward_function_code
dtype: string
- name: prompt_template
dtype: string
- name: agent_reasoning
dtype: string
- name: agent_trace
dtype: string
- name: training_metrics_per_step
dtype: string
- name: eval_accuracy
dtype: float64
- name: eval_error_breakdown
dtype: string
- name: crash_log
dtype: string
- name: status
dtype: string
splits:
- name: train
num_bytes: 34564
num_examples: 5
download_size: 24843
dataset_size: 34564
- config_name: state
features:
- name: state_json
dtype: string
splits:
- name: train
num_bytes: 2441
num_examples: 1
download_size: 11693
dataset_size: 2441
- config_name: train_traces
features:
- name: step
dtype: int64
- name: question_id
dtype: string
- name: input
dtype: string
- name: model_response
dtype: string
- name: reward
dtype: float64
- name: target
dtype: int64
- name: numbers
dtype: string
- name: round_id
dtype: string
splits:
- name: train
num_bytes: 3880910
num_examples: 1280
download_size: 1564967
dataset_size: 3880910
---
# autotrainer-v0
AutoTrainer-v0: LLM-agent-controlled GRPO training on Countdown
## Dataset Info
- **Rows**: 1
- **Columns**: 1
## Columns
| Column | Type | Description |
|--------|------|-------------|
| state_json | Value('string') | Full autotrainer state as JSON string |
## Generation Parameters
```json
{
"script_name": "run_round.py",
"model": "Qwen/Qwen2.5-1.5B-Instruct",
"description": "AutoTrainer-v0: LLM-agent-controlled GRPO training on Countdown",
"experiment_id": "autotrainer-v0",
"hyperparameters": {},
"input_datasets": []
}
```
## Usage
```python
from datasets import load_dataset
dataset = load_dataset("raca-workspace-v1/autotrainer-v0", split="train")
print(f"Loaded {len(dataset)} rows")
```
---
*Uploaded via [RACA](https://github.com/Zayne-sprague/Dr-Claude-Code) hf_utility.*
提供机构：
raca-workspace-v1
