querypieai commited on
Commit
fb83ffa
·
verified ·
1 Parent(s): 587f486

fix: add git lfs pull step to FINETUNING.md before pip install

Browse files
Files changed (4) hide show
  1. FINETUNING.ja.md +1 -0
  2. FINETUNING.ko.md +1 -0
  3. FINETUNING.md +1 -0
  4. MANIFEST.sha256 +3 -3
FINETUNING.ja.md CHANGED
@@ -7,6 +7,7 @@ BF16の既存DLP特化LoRAとpointer headを継続学習します。backboneは
7
  --stepsはoptimizerの合計目標step数です。step 2から--steps 4で再開すると4まで実行します。データhash・学習率・累積・seed変更時は再開を拒否します。--outは原モデルの外の新規パスにしてください。以下はリポジトリフォルダー内で実行します。
8
 
9
  ```bash
 
10
  pip install -r requirements.txt
11
  python training/finetune.py --model . \
12
  --train training/train.synthetic.jsonl \
 
7
  --stepsはoptimizerの合計目標step数です。step 2から--steps 4で再開すると4まで実行します。データhash・学習率・累積・seed変更時は再開を拒否します。--outは原モデルの外の新規パスにしてください。以下はリポジトリフォルダー内で実行します。
8
 
9
  ```bash
10
+ git lfs pull # LFSなしでcloneした場合はモデル重み(safetensors)をダウンロード
11
  pip install -r requirements.txt
12
  python training/finetune.py --model . \
13
  --train training/train.synthetic.jsonl \
FINETUNING.ko.md CHANGED
@@ -7,6 +7,7 @@
7
  --steps는 전체 목표 optimizer step 수입니다. --resume에서 4를 지정하면 step 2에서 4까지 이어서 수행합니다. 데이터 해시·학습률·누적·seed가 달라지면 재개를 거부합니다. --out은 원본 모델 폴더 밖의 새 경로여야 합니다. 원본 저장소를 덮어쓰지 않습니다. 아래 명령은 모델 폴더에서 실행합니다.
8
 
9
  ```bash
 
10
  pip install -r requirements.txt
11
  python training/finetune.py --model . \
12
  --train training/train.synthetic.jsonl \
 
7
  --steps는 전체 목표 optimizer step 수입니다. --resume에서 4를 지정하면 step 2에서 4까지 이어서 수행합니다. 데이터 해시·학습률·누적·seed가 달라지면 재개를 거부합니다. --out은 원본 모델 폴더 밖의 새 경로여야 합니다. 원본 저장소를 덮어쓰지 않습니다. 아래 명령은 모델 폴더에서 실행합니다.
8
 
9
  ```bash
10
+ git lfs pull # LFS 없이 clone한 경우 모델 가중치(safetensors) 다운로드
11
  pip install -r requirements.txt
12
  python training/finetune.py --model . \
13
  --train training/train.synthetic.jsonl \
FINETUNING.md CHANGED
@@ -7,6 +7,7 @@ Continue training the existing DLP-tuned LoRA and pointer head on BF16. The back
7
  --steps is the total target optimizer-step count. Resuming from 2 with --steps 4 continues through step 4. Changed data hashes, learning rate, accumulation or seed are rejected. --out must be a new path outside the source repository. Run the commands from the repository folder.
8
 
9
  ```bash
 
10
  pip install -r requirements.txt
11
  python training/finetune.py --model . \
12
  --train training/train.synthetic.jsonl \
 
7
  --steps is the total target optimizer-step count. Resuming from 2 with --steps 4 continues through step 4. Changed data hashes, learning rate, accumulation or seed are rejected. --out must be a new path outside the source repository. Run the commands from the repository folder.
8
 
9
  ```bash
10
+ git lfs pull # download model weights (safetensors) if cloned without LFS
11
  pip install -r requirements.txt
12
  python training/finetune.py --model . \
13
  --train training/train.synthetic.jsonl \
MANIFEST.sha256 CHANGED
@@ -2,9 +2,9 @@ a0e948ca693c63e0b5139f66bd709003cbcf8bc764acb1ee687a552885c8fe2c .gitattributes
2
  5e1a05927638d5e10f3ab556be8729b6b949341c53e09eed2918ec88cdffcb4a CODE_MAP.md
3
  3a34e86fb1f03b2a81e4ec4418c4ab5ee9d0b1a27645956a5a04285534f8650f DATASETS.md
4
  042ba1ae582aba1323d95f835ce3a1f5a1e6fa462bfb171e326edfe0447b8580 EVALUATION_DATA_SOURCES.json
5
- 7c4709f46497bf67f26d789b6aa230c83c89167536cf05b191784db82a3e59f0 FINETUNING.ja.md
6
- f1bc8adc0fca1a1eae1e723116151410b0042b83f97b223aa8bde419037d11c6 FINETUNING.ko.md
7
- 28c9856af03ac8bf596050cfb39c88a29d5ea763e4feed5255fa96cf182c5674 FINETUNING.md
8
  221cd240fc546e11855065816fd54510a0ab6c6eb07f681b573d6caf57003376 GENERAL_EVALUATION_STATUS.md
9
  8bf5ad7cf87e532a791cebb5b4b2a59e465d6988808f3ce803576286994b7d9a LICENSE
10
  cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30 LICENSES/Apache-2.0.txt
 
2
  5e1a05927638d5e10f3ab556be8729b6b949341c53e09eed2918ec88cdffcb4a CODE_MAP.md
3
  3a34e86fb1f03b2a81e4ec4418c4ab5ee9d0b1a27645956a5a04285534f8650f DATASETS.md
4
  042ba1ae582aba1323d95f835ce3a1f5a1e6fa462bfb171e326edfe0447b8580 EVALUATION_DATA_SOURCES.json
5
+ 3afbe64c11ccbc484e347bc4573ef3ea789c5367d02ae4c1136c13716bf39228 FINETUNING.ja.md
6
+ 7ff04ecad20ee00f56beb45b3b8ec68d94017d9b9b85a7444cec62364b291efc FINETUNING.ko.md
7
+ bac4746c23bb5034e321795057d51cc07f455319ef8444aa51423d39f87bf399 FINETUNING.md
8
  221cd240fc546e11855065816fd54510a0ab6c6eb07f681b573d6caf57003376 GENERAL_EVALUATION_STATUS.md
9
  8bf5ad7cf87e532a791cebb5b4b2a59e465d6988808f3ce803576286994b7d9a LICENSE
10
  cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30 LICENSES/Apache-2.0.txt