Fix examples for latest SFT (#3613)

### Description

Fix example code for latest SFT (for huggingface-trl == v0.21.0).

### Types of changes
<!--- Put an `x` in all the boxes that apply, and remove the not
applicable items -->
- [x] Non-breaking change (fix or new feature that would not break
existing functionality).
- [ ] Breaking change (fix or new feature that would cause existing
functionality to change).
- [ ] New tests added to cover the changes.
- [ ] Quick tests passed locally by running `./runtest.sh`.
- [ ] In-line docstrings updated.
- [ ] Documentation updated.

Signed-off-by: cyy <[email protected]>
Co-authored-by: Chester Chen <[email protected]>
Co-authored-by: Ziyue Xu <[email protected]>

Author: 3 people

examples/advanced/llm_hf/src/hf_sft_peft_fl.py

@@ -45,11 +45,7 @@ def on_epoch_end(self, args, state, control, logs=None, **kwargs):
 
 
 def format_instruction(example):
-    output_texts = []
-    for i in range(len(example["input"])):
-        text = f"### Instruction: Generate Output according to the information and question given by Input. ### Input:{example['input'][i]} ### Response: {example['output'][i]}"
-        output_texts.append(text)
-    return output_texts
+    return f"### Instruction: Generate Output according to the information and question given by Input. ### Input:{example['input']} ### Response: {example['output']}"
 
 
 def setup_distributed_training():
@@ -210,7 +206,7 @@ def main():
         lr_scheduler_type=args.lr_scheduler,
         lr_scheduler_kwargs={"num_cycles": 2},
         disable_tqdm=True,
-        max_seq_length=1024,
+        max_length=1024,
         save_total_limit=2,
         # safetensors will remove shared layers, e.g. lm_head.weight
         # disable for local checkpointing

examples/tutorials/self-paced-training/part-4_advanced_federated_learning/chapter-8_federated_LLM_training/08.2_llm_sft/src/hf_sft_peft_fl.py

@@ -34,11 +34,7 @@
 
 
 def format_instruction(example):
-    output_texts = []
-    for i in range(len(example["input"])):
-        text = f"### Instruction: Generate Output according to the information and question given by Input. ### Input:{example['input'][i]} ### Response: {example['output'][i]}"
-        output_texts.append(text)
-    return output_texts
+    return f"### Instruction: Generate Output according to the information and question given by Input. ### Input:{example['input']} ### Response: {example['output']}"
 
 
 def main():
@@ -142,7 +138,7 @@ def main():
         warmup_ratio=0.03,
         lr_scheduler_type="constant",
         disable_tqdm=True,
-        max_seq_length=1024,
+        max_length=1024,
         save_total_limit=2,
         # safetensors has some issues in saving lm_head.weight, disable it for now
         save_safetensors=False,

examples/tutorials/self-paced-training/part-4_advanced_federated_learning/chapter-8_federated_LLM_training/08.2_llm_sft/utils/hf_sft_peft.py

@@ -30,11 +30,7 @@
 
 
 def format_instruction(example):
-    output_texts = []
-    for i in range(len(example["input"])):
-        text = f"### Instruction: Generate Output according to the information and question given by Input. ### Input:{example['input'][i]} ### Response: {example['output'][i]}"
-        output_texts.append(text)
-    return output_texts
+    return f"### Instruction: Generate Output according to the information and question given by Input. ### Input:{example['input']} ### Response: {example['output']}"
 
 
 def main():

examples/tutorials/self-paced-training/part-4_advanced_federated_learning/chapter-8_federated_LLM_training/08.2_llm_sft/utils/hf_sft_peft_iter.py

@@ -31,11 +31,7 @@
 
 
 def format_instruction(example):
-    output_texts = []
-    for i in range(len(example["input"])):
-        text = f"### Instruction: Generate Output according to the information and question given by Input. ### Input:{example['input'][i]} ### Response: {example['output'][i]}"
-        output_texts.append(text)
-    return output_texts
+    return f"### Instruction: Generate Output according to the information and question given by Input. ### Input:{example['input']} ### Response: {example['output']}"
 
 
 def main():

examples/tutorials/self-paced-training/part-4_advanced_federated_learning/chapter-8_federated_LLM_training/08.3_llm_peft/src/hf_sft_peft_fl.py

@@ -34,11 +34,7 @@
 
 
 def format_instruction(example):
-    output_texts = []
-    for i in range(len(example["input"])):
-        text = f"### Instruction: Generate Output according to the information and question given by Input. ### Input:{example['input'][i]} ### Response: {example['output'][i]}"
-        output_texts.append(text)
-    return output_texts
+    return f"### Instruction: Generate Output according to the information and question given by Input. ### Input:{example['input']} ### Response: {example['output']}"
 
 
 def main():
@@ -142,7 +138,7 @@ def main():
         warmup_ratio=0.03,
         lr_scheduler_type="constant",
         disable_tqdm=True,
-        max_seq_length=1024,
+        max_length=1024,
         save_total_limit=2,
         # safetensors has some issues in saving lm_head.weight, disable it for now
         save_safetensors=False,

examples/tutorials/self-paced-training/part-4_advanced_federated_learning/chapter-8_federated_LLM_training/08.3_llm_peft/utils/hf_sft_peft.py

@@ -30,11 +30,7 @@
 
 
 def format_instruction(example):
-    output_texts = []
-    for i in range(len(example["input"])):
-        text = f"### Instruction: Generate Output according to the information and question given by Input. ### Input:{example['input'][i]} ### Response: {example['output'][i]}"
-        output_texts.append(text)
-    return output_texts
+    return f"### Instruction: Generate Output according to the information and question given by Input. ### Input:{example['input']} ### Response: {example['output']}"
 
 
 def main():

examples/tutorials/self-paced-training/part-4_advanced_federated_learning/chapter-8_federated_LLM_training/08.4_llm_quantization/src/hf_sft_peft_fl.py

@@ -34,11 +34,7 @@
 
 
 def format_instruction(example):
-    output_texts = []
-    for i in range(len(example["input"])):
-        text = f"### Instruction: Generate Output according to the information and question given by Input. ### Input:{example['input'][i]} ### Response: {example['output'][i]}"
-        output_texts.append(text)
-    return output_texts
+    return f"### Instruction: Generate Output according to the information and question given by Input. ### Input:{example['input']} ### Response: {example['output']}"
 
 
 def main():
@@ -142,7 +138,7 @@ def main():
         warmup_ratio=0.03,
         lr_scheduler_type="constant",
         disable_tqdm=True,
-        max_seq_length=1024,
+        max_length=1024,
         save_total_limit=2,
         # safetensors has some issues in saving lm_head.weight, disable it for now
         save_safetensors=False,

examples/tutorials/self-paced-training/part-4_advanced_federated_learning/chapter-8_federated_LLM_training/08.4_llm_quantization/utils/hf_sft_peft.py

@@ -30,11 +30,7 @@
 
 
 def format_instruction(example):
-    output_texts = []
-    for i in range(len(example["input"])):
-        text = f"### Instruction: Generate Output according to the information and question given by Input. ### Input:{example['input'][i]} ### Response: {example['output'][i]}"
-        output_texts.append(text)
-    return output_texts
+    return f"### Instruction: Generate Output according to the information and question given by Input. ### Input:{example['input']} ### Response: {example['output']}"
 
 
 def main():