configs/mixtral.yml

base_model: mistralai/Mixtral-8x7B-v0.1
model_type: AutoModelForCausalLM
tokenizer_type: LlamaTokenizer
trust_remote_code: true

load_in_8bit: false
load_in_4bit: true
strict: false

datasets:
  - path: Open-Orca/SlimOrca
    type: sharegpt
    conversation: chatml
dataset_prepared_path: last_run_prepared
val_set_size: 0.005
output_dir: ./slimorca-mixtral-out
save_total_limit: 2
hub_model_id: openaccess-ai-collective/slimorca-mixstral-8x7b
dataloader_num_workers: 8
dataloader_prefetch_factor: 4
dataloader_pin_memory: true

adapter: qlora
lora_model_dir:

sequence_len: 8192
sample_packing: true
pad_to_sequence_len: true

lora_r: 64
lora_alpha: 32
lora_dropout: 0.1
lora_target_linear: true
lora_fan_in_fan_out:
lora_modules_to_save:
  - lm_head
  - embed_tokens
#lora_target_modules:
#  - gate
#  - q_proj
#  - k_proj
#  - v_proj
#  - o_proj
#  - w1
#  - w2
#  - w3

wandb_project: slimorca-mixtral
wandb_entity: oaaic
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 1
micro_batch_size: 4
num_epochs: 2
optimizer: paged_adamw_8bit
lr_scheduler: cosine
learning_rate: 0.001
adam_beta2: 0.95
adam_epsilon: 0.00001
max_grad_norm: 1.0

train_on_inputs: false
group_by_length: false
bf16: true
fp16: false
tf32: true

gradient_checkpointing: true
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true

warmup_steps: 100
eval_steps: 0.05
save_steps: 0.25
debug:
deepspeed: deepspeed/zero2.json
weight_decay: 0.1
fsdp:
fsdp_config:
special_tokens:
  eos_token: "<|im_end|>"
tokens:
  - "<|im_start|>"