nraptisss commited on
Commit
f5ecafd
·
verified ·
1 Parent(s): 630af96

Fix: --max_seq_length → --max_length (matches train.py argparse)

Browse files
Files changed (1) hide show
  1. run.sh +3 -3
run.sh CHANGED
@@ -2,7 +2,7 @@
2
  # ============================================================
3
  # TMF921 Intent Translation Training — One-Command Runner
4
  # ============================================================
5
- # Run on a machine with an NVIDIA GPU (tested on RTX 6000 Ada 50GB)
6
  #
7
  # Usage:
8
  # git clone https://huggingface.co/nraptisss/intent-translation-training
@@ -25,7 +25,7 @@ EPOCHS=3
25
  LR=1e-4
26
  BATCH_SIZE=4
27
  GRAD_ACCUM=8
28
- MAX_SEQ_LEN=4096
29
  LORA_R=32
30
  LORA_ALPHA=64
31
 
@@ -69,7 +69,7 @@ if [ "${1:-}" != "--eval-only" ]; then
69
  --lr $LR \
70
  --batch_size $BATCH_SIZE \
71
  --grad_accum $GRAD_ACCUM \
72
- --max_seq_length $MAX_SEQ_LEN \
73
  --lora_r $LORA_R \
74
  --lora_alpha $LORA_ALPHA \
75
  ${HUB_MODEL_ID:+--push_to_hub --hub_model_id "$HUB_MODEL_ID"}
 
2
  # ============================================================
3
  # TMF921 Intent Translation Training — One-Command Runner
4
  # ============================================================
5
+ # Run on a machine with an NVIDIA GPU (tested on RTX 6000 Ada 48GB)
6
  #
7
  # Usage:
8
  # git clone https://huggingface.co/nraptisss/intent-translation-training
 
25
  LR=1e-4
26
  BATCH_SIZE=4
27
  GRAD_ACCUM=8
28
+ MAX_LENGTH=4096
29
  LORA_R=32
30
  LORA_ALPHA=64
31
 
 
69
  --lr $LR \
70
  --batch_size $BATCH_SIZE \
71
  --grad_accum $GRAD_ACCUM \
72
+ --max_length $MAX_LENGTH \
73
  --lora_r $LORA_R \
74
  --lora_alpha $LORA_ALPHA \
75
  ${HUB_MODEL_ID:+--push_to_hub --hub_model_id "$HUB_MODEL_ID"}