adam_beta2: 0.95
adam_epsilon: 1.0e-05
adapter: qlora
base_model: mistralai/Mixtral-8x7B-Instruct-v0.1
bf16: auto
chat_template: inst
dataset_prepared_path: last_run_prepared
datasets:
- conversation: mistral
  path: dd7ba3a8030a4c7382d51a5d894f5cb4/./data/with_function_response/function_used_training.jsonl
  type: sharegpt
debug: null
deepspeed: null
early_stopping_patience: null
eval_max_new_tokens: 256
eval_steps: 0.2
eval_table_size: null
flash_attention: true
fp16: null
gradient_accumulation_steps: 4
gradient_checkpointing: true
group_by_length: false
hub_model_id: liuylhf/mixtral-remove-negative-data
hub_strategy: end
is_mistral_derived_model: true
learning_rate: 0.001
load_in_4bit: true
load_in_8bit: false
local_rank: null
logging_steps: 1
lora_alpha: 16
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 32
lora_target_modules:
- q_proj
- v_proj
- k_proj
- o_proj
lr_scheduler: cosine
max_grad_norm: 1.0
micro_batch_size: 2
model_config:
  output_router_logits: true
model_type: AutoModelForCausalLM
num_epochs: 1
optimizer: paged_adamw_8bit
output_dir: dd7ba3a8030a4c7382d51a5d894f5cb4/model
pad_to_sequence_len: true
resume_from_checkpoint: null
sample_packing: true
save_steps: 0.1
sequence_len: 8192
strict: false
tf32: false
tokenizer_type: LlamaTokenizer
train_on_inputs: false
val_set_size: 0.01
wandb_log_model: end
wandb_name: mixtral-instruct-raw-data-v3
wandb_project: function-call
warmup_steps: 10
weight_decay: 0
xformers_attention: null

mixtral-remove-negative-data

This model is a fine-tuned version of mistralai/Mixtral-8x7B-Instruct-v0.1 on the None dataset. It achieves the following results on the evaluation set:

Loss: 0.0955

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.001
train_batch_size: 2
eval_batch_size: 2
seed: 42
distributed_type: multi-GPU
num_devices: 2
gradient_accumulation_steps: 4
total_train_batch_size: 16
total_eval_batch_size: 4
optimizer: Adam with betas=(0.9,0.95) and epsilon=1e-05
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 10
num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss
3.9579	0.01	1	4.0485
0.148	0.2	36	0.1565
0.1075	0.4	72	0.1138
0.099	0.6	108	0.1018
0.0954	0.8	144	0.0969
0.0945	1.0	180	0.0955

Framework versions

PEFT 0.8.2
Transformers 4.39.0.dev0
Pytorch 2.2.0+cu121
Datasets 2.17.1
Tokenizers 0.15.0