File size: 1,520 Bytes
e7cb8ff |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 |
model:
_component_: torchtune.models.qwen2_5.qwen2_5_0_5b
tokenizer:
_component_: torchtune.models.qwen2_5.qwen2_5_tokenizer
path: qwen2_5.qwen2_5_0_5b_instruct/vocab.json
merges_file: qwen2_5.qwen2_5_0_5b_instruct/merges.txt
max_seq_len: null
checkpointer:
_component_: torchtune.training.FullModelHFCheckpointer
checkpoint_dir: qwen2_5.qwen2_5_0_5b_instruct
checkpoint_files:
- model.safetensors
recipe_checkpoint: null
output_dir: qwen2_5.qwen2_5_0_5b_instruct
model_type: QWEN2
resume_from_checkpoint: false
dataset:
_component_: torchtune.datasets.alpaca_cleaned_dataset
packed: false
seed: null
shuffle: true
epochs: 1
max_steps_per_epoch: 60
batch_size: 16
gradient_accumulation_steps: 8
optimizer:
_component_: torch.optim.AdamW
fused: true
lr: 2.0e-05
optimizer_in_bwd: false
loss:
_component_: torchtune.modules.loss.CEWithChunkedOutputLoss
device: cuda
enable_activation_checkpointing: false
enable_activation_offloading: false
dtype: bf16
compile: true
output_dir: qwen2_5.qwen2_5_0_5b_instruct
metric_logger:
_component_: torchtune.training.metric_logging.WandBLogger
project: hallucination
entity: c-metrics
log_every_n_steps: 1
log_peak_memory_stats: true
profiler:
_component_: torchtune.training.setup_torch_profiler
enabled: false
output_dir: ${output_dir}/profiling_outputs
cpu: true
cuda: true
profile_memory: false
with_stack: false
record_shapes: true
with_flops: false
wait_steps: 5
warmup_steps: 3
active_steps: 2
num_cycles: 1
|