Skip to content

fix: add formatting_func for SFTTrainer compatibility #19

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
nvandessel merged 3 commits into from
Apr 4, 2026
Merged

fix: add formatting_func for SFTTrainer compatibility #19

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
19bcaa2
fix: add formatting_func for SFTTrainer compatibility
nvandessel Apr 4, 2026
046188d
fix: formatting_func must return list, warmup_ratio deprecated
nvandessel Apr 4, 2026
c801350
fix: formatting_func receives batched examples, not single
nvandessel Apr 4, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
38 changes: 1 addition & 37 deletions notebooks/train-hippofloop.ipynb
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -227,43 +227,7 @@
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"from transformers import TrainingArguments\n",
"from trl import SFTTrainer\n",
"\n",
"OUTPUT_DIR = \"checkpoints/qwen25-3b-hippofloop\"\n",
"\n",
"training_args = TrainingArguments(\n",
" output_dir=OUTPUT_DIR,\n",
" num_train_epochs=3,\n",
" per_device_train_batch_size=1,\n",
" gradient_accumulation_steps=16,\n",
" learning_rate=2e-4,\n",
" lr_scheduler_type=\"cosine\",\n",
" warmup_ratio=0.03,\n",
" weight_decay=0.01,\n",
" bf16=False,\n",
" fp16=True,\n",
" eval_strategy=\"epoch\",\n",
" save_strategy=\"epoch\",\n",
" load_best_model_at_end=True,\n",
" metric_for_best_model=\"eval_loss\",\n",
" logging_steps=10,\n",
" seed=SEED,\n",
")\n",
"\n",
"sft_trainer = SFTTrainer(\n",
" model=model,\n",
" tokenizer=tokenizer,\n",
" train_dataset=train_dataset,\n",
" eval_dataset=val_dataset,\n",
" args=training_args,\n",
")\n",
"\n",
"print(f\"Training {len(train_dataset)} examples for {training_args.num_train_epochs} epochs...\")\n",
"print(f\"Effective batch size: {training_args.per_device_train_batch_size * training_args.gradient_accumulation_steps}\")\n",
"sft_trainer.train()"
]
"source": "from transformers import TrainingArguments\nfrom trl import SFTTrainer\n\nOUTPUT_DIR = \"checkpoints/qwen25-3b-hippofloop\"\n\ntraining_args = TrainingArguments(\n output_dir=OUTPUT_DIR,\n num_train_epochs=3,\n per_device_train_batch_size=1,\n gradient_accumulation_steps=16,\n learning_rate=2e-4,\n lr_scheduler_type=\"cosine\",\n warmup_steps=50,\n weight_decay=0.01,\n bf16=False,\n fp16=True,\n eval_strategy=\"epoch\",\n save_strategy=\"epoch\",\n load_best_model_at_end=True,\n metric_for_best_model=\"eval_loss\",\n logging_steps=10,\n seed=SEED,\n)\n\ndef formatting_func(examples):\n texts = []\n for msgs in examples[\"messages\"]:\n texts.append(tokenizer.apply_chat_template(msgs, tokenize=False))\n return texts\n\nsft_trainer = SFTTrainer(\n model=model,\n tokenizer=tokenizer,\n train_dataset=train_dataset,\n eval_dataset=val_dataset,\n args=training_args,\n formatting_func=formatting_func,\n)\n\nprint(f\"Training {len(train_dataset)} examples for {training_args.num_train_epochs} epochs...\")\nprint(f\"Effective batch size: {training_args.per_device_train_batch_size * training_args.gradient_accumulation_steps}\")\nsft_trainer.train()"
},
{
"cell_type": "code",
Expand Down
9 changes: 8 additions & 1 deletion src/hippofloop/training/trainer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -76,7 +76,7 @@ def train(
gradient_accumulation_steps=self._config.gradient_accumulation_steps,
learning_rate=self._config.learning_rate,
lr_scheduler_type=self._config.lr_scheduler,
warmup_ratio=self._config.warmup_ratio,
warmup_steps=50,
weight_decay=self._config.weight_decay,
bf16=self._config.bf16,
fp16=self._config.fp16,
Expand All @@ -88,12 +88,19 @@ def train(
seed=self._config.seed,
)

def formatting_func(examples: dict) -> list[str]:
return [
tokenizer.apply_chat_template(msgs, tokenize=False)
for msgs in examples["messages"]
]

trainer = SFTTrainer(
model=model,
tokenizer=tokenizer,
train_dataset=train_dataset,
eval_dataset=val_dataset,
args=training_args,
formatting_func=formatting_func,
)

logger.info("Starting training (%d epochs)", self._config.epochs)
Expand Down
Loading