mistralai

/

Mixtral-8x22B-Instruct-v0.1

like 586

Hey, I'm trying to fine tune using Mixtral 8x22b on a A100.

The process keep getting stuck here : model-00010-of-00059.safetensors: 100%|_| 4.81G/4.81G [00:13<00:00, 344MB/s]
Downloading shards: 17%| | 10/59 [02:39<12:37, 15.46s/it^

Do you have any ideas why ?

Thanks :))))

base_model: mistralai/Mixtral-8x22B-Instruct-v0.1
model_type: MistralForCausalLM
tokenizer_type: AutoTokenizer
is_mistral_derived_model: true
hub_model_id: DevSelego/Mistral7b_summarizer_v9

load_in_8bit: false
load_in_4bit: true
strict: false
datasets:

• path: DevSelego/jobmaker_summarizer_v9
  type: alpaca
  dataset_prepared_path: last_run_prepared
  val_set_size: 0.05
  output_dir: ./qlora-out

adapter: qlora
lora_model_dir:

sequence_len: 8192
sample_packing: false
pad_to_sequence_len: true

lora_r: 32
lora_alpha: 16
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:
lora_target_modules:

• gate_proj
• down_proj
• up_proj
• q_proj
• v_proj
• k_proj
• o_proj

wandb_project: Mistral_Jobmaker
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 1
micro_batch_size: 10
num_epochs: 1
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0002

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

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

loss_watchdog_threshold: 5.0
loss_watchdog_patience: 3

warmup_steps: 10
eval_steps: 0.05
eval_table_size:
eval_table_max_new_tokens: 128
save_steps:
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
bos_token: ""
eos_token: ""
unk_token: ""