mirror of
https://github.com/huggingface/trl.git
synced 2025-10-20 10:03:51 +08:00
Remove unused commands directory (#4258)
Co-authored-by: behroozazarkhalili <ermiaazarkhalili>
This commit is contained in:
Behrooz Azarkhalili
committed by
GitHub
GitHub
parent
8e2d5516ca
commit
31caf64778
7
.github/workflows/slow-tests.yml
vendored
7
.github/workflows/slow-tests.yml
vendored
@ -102,13 +102,6 @@ jobs:
|
||||
source .venv/bin/activate
|
||||
make slow_tests
|
||||
|
||||
- name: Run end-to-end examples tests on multi GPU
|
||||
if: always()
|
||||
run: |
|
||||
source .venv/bin/activate
|
||||
uv pip install deepspeed
|
||||
make test_examples
|
||||
|
||||
- name: Generate Reports
|
||||
if: always()
|
||||
run: |
|
||||
|
||||
16
Makefile
16
Makefile
@ -1,9 +1,8 @@
|
||||
.PHONY: test precommit common_tests slow_tests test_examples tests_gpu test_experimental
|
||||
.PHONY: test precommit common_tests slow_tests tests_gpu test_experimental
|
||||
|
||||
check_dirs := examples tests trl
|
||||
|
||||
ACCELERATE_CONFIG_PATH = `pwd`/examples/accelerate_configs
|
||||
COMMAND_FILES_PATH = `pwd`/commands
|
||||
|
||||
test:
|
||||
pytest -n auto -m "not slow and not low_priority" -s -v --reruns 5 --reruns-delay 1 --only-rerun '(OSError|Timeout|HTTPError.*502|HTTPError.*504||not less than or equal to 0.01)' tests/
|
||||
@ -16,18 +15,5 @@ precommit:
|
||||
slow_tests:
|
||||
pytest -m "slow" tests/ $(if $(IS_GITHUB_CI),--report-log "slow_tests.log",)
|
||||
|
||||
test_examples:
|
||||
touch temp_results_sft_tests.txt
|
||||
for file in $(ACCELERATE_CONFIG_PATH)/*.yaml; do \
|
||||
TRL_ACCELERATE_CONFIG=$${file} bash $(COMMAND_FILES_PATH)/run_sft.sh; \
|
||||
echo $$?','$${file} >> temp_results_sft_tests.txt; \
|
||||
done
|
||||
|
||||
touch temp_results_dpo_tests.txt
|
||||
for file in $(ACCELERATE_CONFIG_PATH)/*.yaml; do \
|
||||
TRL_ACCELERATE_CONFIG=$${file} bash $(COMMAND_FILES_PATH)/run_dpo.sh; \
|
||||
echo $$?','$${file} >> temp_results_dpo_tests.txt; \
|
||||
done
|
||||
|
||||
test_experimental:
|
||||
pytest -k "experimental"
|
||||
|
||||
@ -1,58 +0,0 @@
|
||||
#!/bin/bash
|
||||
# This script runs an SFT example end-to-end on a tiny model using different possible configurations
|
||||
# but defaults to QLoRA + PEFT
|
||||
OUTPUT_DIR="test_dpo/"
|
||||
MODEL_NAME="trl-internal-testing/tiny-Qwen2ForCausalLM-2.5"
|
||||
DATASET_NAME="trl-internal-testing/hh-rlhf-helpful-base-trl-style"
|
||||
MAX_STEPS=5
|
||||
BATCH_SIZE=2
|
||||
SEQ_LEN=128
|
||||
|
||||
# Handle extra arguments in case one passes accelerate configs.
|
||||
EXTRA_ACCELERATE_ARGS=""
|
||||
EXTRA_TRAINING_ARGS="""--use_peft \
|
||||
--load_in_4bit
|
||||
"""
|
||||
|
||||
# This is a hack to get the number of available GPUs
|
||||
NUM_GPUS=2
|
||||
|
||||
if [[ "${TRL_ACCELERATE_CONFIG}" == "" ]]; then
|
||||
EXTRA_ACCELERATE_ARGS=""
|
||||
else
|
||||
EXTRA_ACCELERATE_ARGS="--config_file $TRL_ACCELERATE_CONFIG"
|
||||
# For DeepSpeed configs we need to set the `--fp16` flag to comply with our configs exposed
|
||||
# on `examples/accelerate_configs` and our runners do not support bf16 mixed precision training.
|
||||
if [[ $TRL_ACCELERATE_CONFIG == *"deepspeed"* ]]; then
|
||||
EXTRA_TRAINING_ARGS="--fp16"
|
||||
else
|
||||
echo "Keeping QLoRA + PEFT"
|
||||
fi
|
||||
fi
|
||||
|
||||
|
||||
CMD="""
|
||||
accelerate launch $EXTRA_ACCELERATE_ARGS \
|
||||
--num_processes $NUM_GPUS \
|
||||
--mixed_precision 'fp16' \
|
||||
`pwd`/trl/scripts/dpo.py \
|
||||
--model_name_or_path $MODEL_NAME \
|
||||
--dataset_name $DATASET_NAME \
|
||||
--output_dir $OUTPUT_DIR \
|
||||
--max_steps $MAX_STEPS \
|
||||
--per_device_train_batch_size $BATCH_SIZE \
|
||||
--max_length $SEQ_LEN \
|
||||
$EXTRA_TRAINING_ARGS
|
||||
"""
|
||||
|
||||
echo "Starting program..."
|
||||
|
||||
{ # try
|
||||
echo $CMD
|
||||
eval "$CMD"
|
||||
} || { # catch
|
||||
# save log for exception
|
||||
echo "Operation Failed!"
|
||||
exit 1
|
||||
}
|
||||
exit 0
|
||||
@ -1,59 +0,0 @@
|
||||
#!/bin/bash
|
||||
# This script runs an SFT example end-to-end on a tiny model using different possible configurations
|
||||
# but defaults to QLoRA + PEFT
|
||||
OUTPUT_DIR="test_sft/"
|
||||
MODEL_NAME="trl-internal-testing/tiny-Qwen2ForCausalLM-2.5"
|
||||
DATASET_NAME="stanfordnlp/imdb"
|
||||
MAX_STEPS=5
|
||||
BATCH_SIZE=2
|
||||
SEQ_LEN=128
|
||||
|
||||
|
||||
# Handle extra arguments in case one passes accelerate configs.
|
||||
EXTRA_ACCELERATE_ARGS=""
|
||||
EXTRA_TRAINING_ARGS="""--use_peft \
|
||||
--load_in_4bit
|
||||
"""
|
||||
|
||||
# Set your number of GPUs here
|
||||
NUM_GPUS=2
|
||||
|
||||
if [[ "${TRL_ACCELERATE_CONFIG}" == "" ]]; then
|
||||
EXTRA_ACCELERATE_ARGS=""
|
||||
else
|
||||
EXTRA_ACCELERATE_ARGS="--config_file $TRL_ACCELERATE_CONFIG"
|
||||
# For DeepSpeed configs we need to set the `--fp16` flag to comply with our configs exposed
|
||||
# on `examples/accelerate_configs` and our runners do not support bf16 mixed precision training.
|
||||
if [[ $TRL_ACCELERATE_CONFIG == *"deepspeed"* ]]; then
|
||||
EXTRA_TRAINING_ARGS="--fp16"
|
||||
else
|
||||
echo "Keeping QLoRA + PEFT"
|
||||
fi
|
||||
fi
|
||||
|
||||
|
||||
CMD="""
|
||||
accelerate launch $EXTRA_ACCELERATE_ARGS \
|
||||
--num_processes $NUM_GPUS \
|
||||
--mixed_precision 'fp16' \
|
||||
`pwd`/trl/scripts/sft.py \
|
||||
--model_name $MODEL_NAME \
|
||||
--dataset_name $DATASET_NAME \
|
||||
--output_dir $OUTPUT_DIR \
|
||||
--max_steps $MAX_STEPS \
|
||||
--per_device_train_batch_size $BATCH_SIZE \
|
||||
--max_length $SEQ_LEN \
|
||||
$EXTRA_TRAINING_ARGS
|
||||
"""
|
||||
|
||||
echo "Starting program..."
|
||||
|
||||
{ # try
|
||||
echo $CMD
|
||||
eval "$CMD"
|
||||
} || { # catch
|
||||
# save log for exception
|
||||
echo "Operation Failed!"
|
||||
exit 1
|
||||
}
|
||||
exit 0
|
||||
Reference in New Issue
Block a user