Spaces:
Sleeping
Sleeping
Commit
·
07a7833
1
Parent(s):
ee07cd3
organize files
Browse files- README.md +14 -3
- download.py → data_loader/download.py +0 -0
- download_dataset.py → data_loader/download_dataset.py +0 -0
- ppo_tune.py → fine_tune_llm/ppo_tune_llm.py +0 -0
- reward_model.py → fine_tune_llm/reward_model.py +0 -0
- sft_train.py → fine_tune_llm/sft_llm_train.py +0 -0
- train_lora.py → fine_tune_stablediffusion/train_lora.py +0 -0
- build_embeddings.py → retrieval_augmented_generation/build_embeddings.py +0 -0
- train_model.py → train_model_test.py +0 -0
README.md
CHANGED
|
@@ -60,11 +60,22 @@ python sft_train.py
|
|
| 60 |
# 3 Build RAG index
|
| 61 |
python build_embeddings.py
|
| 62 |
|
| 63 |
-
# 4 (可选) 收集偏好 → 训练 reward model
|
| 64 |
python reward_model.py
|
| 65 |
|
| 66 |
-
# 5 PPO RLHF 微调
|
| 67 |
python ppo_tune.py
|
| 68 |
|
| 69 |
# 6 Inference with RAG
|
| 70 |
-
python rag_infer.py
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 60 |
# 3 Build RAG index
|
| 61 |
python build_embeddings.py
|
| 62 |
|
| 63 |
+
# 4 (可选) 收集偏好 → 训练 reward model
|
| 64 |
python reward_model.py
|
| 65 |
|
| 66 |
+
# 5 PPO RLHF 微调LLM
|
| 67 |
python ppo_tune.py
|
| 68 |
|
| 69 |
# 6 Inference with RAG
|
| 70 |
+
python rag_infer.py
|
| 71 |
+
|
| 72 |
+
|
| 73 |
+
|
| 74 |
+
system flow:
|
| 75 |
+
input: business or product description text
|
| 76 |
+
1. 根据input用RAG取embedding
|
| 77 |
+
1. GPT‑OSS 生成 4 个广告文案 + 标题 + 口号(可选语气:专业/活泼/极简)
|
| 78 |
+
2. GPT‑OSS 基于选中文案生成 扩展视觉提示词(主体、配色、镜头、艺术风格)
|
| 79 |
+
3. stablediffusion model 生成 4 张草图(可选 ControlNet-Layout/Logo 插入)
|
| 80 |
+
4. 返回4张海报+后处理
|
| 81 |
+
output: an advertisement sentence and post image
|
download.py → data_loader/download.py
RENAMED
|
File without changes
|
download_dataset.py → data_loader/download_dataset.py
RENAMED
|
File without changes
|
ppo_tune.py → fine_tune_llm/ppo_tune_llm.py
RENAMED
|
File without changes
|
reward_model.py → fine_tune_llm/reward_model.py
RENAMED
|
File without changes
|
sft_train.py → fine_tune_llm/sft_llm_train.py
RENAMED
|
File without changes
|
train_lora.py → fine_tune_stablediffusion/train_lora.py
RENAMED
|
File without changes
|
build_embeddings.py → retrieval_augmented_generation/build_embeddings.py
RENAMED
|
File without changes
|
train_model.py → train_model_test.py
RENAMED
|
File without changes
|