Hugging Face's logo

Spaces:
pushpam14
/
api-contract-validator
Sleeping

App Files Community
pushpam14 commited on
Commit
e37a2ab
·
verified ·
1 Parent(s): 7f76aff

Add direct links to submitted training files in notebook

Browse files
Files changed (1) hide show
  1. training/grpo_colab.ipynb +13 -1
training/grpo_colab.ipynb CHANGED
@@ -14,6 +14,11 @@
14
  "\n",
15
  "**Submitted run**: Qwen2.5-7B + LoRA, 300 GRPO steps, HF Jobs L4, ~1 h 56 min.\n",
16
  "\n",
 
 
 
 
 
17
  "**Colab demo path**: Qwen2.5-1.5B by default, ~45 min for 200 steps on a T4.\n"
18
  ]
19
  },
@@ -53,7 +58,12 @@
53
  "\n",
54
  "This section is intentionally executed and saved so reviewers see concrete output when opening the notebook. It does **not** retrain the model, does **not** rerun inference, and does **not** regenerate any metric. It loads the final submitted HF Jobs artifacts committed in this repository.\n",
55
  "\n",
56
- "The actual final training path was: `training/run_in_hf_jobs.py` \u2192 `training/train.py` on Hugging Face Jobs L4 for 300 GRPO steps.\n"
 
 
 
 
 
57
  ]
58
  },
59
  {
@@ -68,6 +78,8 @@
68
  "Submitted run artifact summary (loaded from committed files, not re-run here)\n",
69
  "============================================================================\n",
70
  "Actual final training path: HF Jobs -> training/run_in_hf_jobs.py -> training/train.py\n",
 
 
71
  "Model: Qwen2.5-7B-Instruct 4-bit + LoRA r=16\n",
72
  "GRPO reward rows: 300\n",
73
  "Final training step: 300\n",
 
14
  "\n",
15
  "**Submitted run**: Qwen2.5-7B + LoRA, 300 GRPO steps, HF Jobs L4, ~1 h 56 min.\n",
16
  "\n",
17
+ "**Actual submitted training files**:\n",
18
+ "- HF Jobs launcher: [https://huggingface.co/spaces/pushpam14/api-contract-validator/blob/main/training/run_in_hf_jobs.py](https://huggingface.co/spaces/pushpam14/api-contract-validator/blob/main/training/run_in_hf_jobs.py)\n",
19
+ "- GRPO trainer: [https://huggingface.co/spaces/pushpam14/api-contract-validator/blob/main/training/train.py](https://huggingface.co/spaces/pushpam14/api-contract-validator/blob/main/training/train.py)\n",
20
+ "\n",
21
+ "\n",
22
  "**Colab demo path**: Qwen2.5-1.5B by default, ~45 min for 200 steps on a T4.\n"
23
  ]
24
  },
 
58
  "\n",
59
  "This section is intentionally executed and saved so reviewers see concrete output when opening the notebook. It does **not** retrain the model, does **not** rerun inference, and does **not** regenerate any metric. It loads the final submitted HF Jobs artifacts committed in this repository.\n",
60
  "\n",
61
+ "The actual final training path was: `training/run_in_hf_jobs.py` \u2192 `training/train.py` on Hugging Face Jobs L4 for 300 GRPO steps.\n",
62
+ "\n",
63
+ "Direct links to the submitted training files:\n",
64
+ "\n",
65
+ "- HF Jobs launcher: [https://huggingface.co/spaces/pushpam14/api-contract-validator/blob/main/training/run_in_hf_jobs.py](https://huggingface.co/spaces/pushpam14/api-contract-validator/blob/main/training/run_in_hf_jobs.py)\n",
66
+ "- GRPO trainer: [https://huggingface.co/spaces/pushpam14/api-contract-validator/blob/main/training/train.py](https://huggingface.co/spaces/pushpam14/api-contract-validator/blob/main/training/train.py)\n"
67
  ]
68
  },
69
  {
 
78
  "Submitted run artifact summary (loaded from committed files, not re-run here)\n",
79
  "============================================================================\n",
80
  "Actual final training path: HF Jobs -> training/run_in_hf_jobs.py -> training/train.py\n",
81
+ "HF Jobs launcher URL: https://huggingface.co/spaces/pushpam14/api-contract-validator/blob/main/training/run_in_hf_jobs.py\n",
82
+ "GRPO trainer URL: https://huggingface.co/spaces/pushpam14/api-contract-validator/blob/main/training/train.py\n",
83
  "Model: Qwen2.5-7B-Instruct 4-bit + LoRA r=16\n",
84
  "GRPO reward rows: 300\n",
85
  "Final training step: 300\n",