9a8073e73d
* LFM2 support * docs * packing seems to work * update install to force install in case already on dev version * default to use chunked cross entropy
49 lines
863 B
YAML
49 lines
863 B
YAML
base_model: LiquidAI/LFM2-350M
|
|
|
|
chunked_cross_entropy: true
|
|
|
|
chat_template: tokenizer_default
|
|
eot_tokens:
|
|
- "<|im_end|>"
|
|
datasets:
|
|
- path: mlabonne/FineTome-100k
|
|
type: chat_template
|
|
split: train[:20%]
|
|
field_messages: conversations
|
|
message_field_role: from
|
|
message_field_content: value
|
|
dataset_prepared_path: last_run_prepared
|
|
val_set_size: 0.05
|
|
output_dir: ./outputs/out
|
|
|
|
sequence_len: 4096
|
|
sample_packing: true
|
|
pad_to_sequence_len: true
|
|
|
|
wandb_project:
|
|
wandb_entity:
|
|
wandb_watch:
|
|
wandb_name:
|
|
wandb_log_model:
|
|
|
|
gradient_accumulation_steps: 2
|
|
micro_batch_size: 4
|
|
num_epochs: 1
|
|
optimizer: adamw_torch_fused
|
|
lr_scheduler: cosine
|
|
learning_rate: 5e-5
|
|
|
|
bf16: true
|
|
tf32: true
|
|
|
|
gradient_checkpointing: false
|
|
resume_from_checkpoint:
|
|
logging_steps: 1
|
|
flash_attention: true
|
|
|
|
warmup_ratio: 0.1
|
|
evals_per_epoch: 2
|
|
saves_per_epoch: 1
|
|
|
|
weight_decay: 0.0
|