See axolotl config

axolotl version: 0.6.0

base_model: /workspace/axolotl/in
# optionally might have model_type or tokenizer_type
model_type: MistralForCausalLM
tokenizer_type: LlamaTokenizer
# Automatically upload checkpoint and final model to HF
hub_model_id: AiAF/UFOs-Finetune-V1.1

load_in_8bit: false
load_in_4bit: false
strict: false

datasets:
  - path: json
    data_files: plain_qa_list.jsonl
    ds_type: json
    type: chat_template
    chat_template: chatml
    field_messages: conversations
    message_field_role: from
    message_field_content: value
    roles:
      user:
        - human
      assistant:
        - gpt
      system:
        - system

dataset_prepared_path:
val_set_size: 0.05
output_dir: ./outputs/UFOs-Finetune-V1.1/out

sequence_len: 8192
sample_packing: true
pad_to_sequence_len: true
eval_sample_packing: false

max_steps: 100000

wandb_project: "UFO_LLM_Finetune"
wandb_entity:
wandb_watch: "all"
wandb_name: "UFO_LLM_Finetune-V1.1"
wandb_log_model: "false"

gradient_accumulation_steps: 4
micro_batch_size: 2
num_epochs: 10
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.000005

train_on_inputs: false
group_by_length: false
bf16: auto
fp16:
tf32: false

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

warmup_steps: 10
evals_per_epoch: 4
eval_table_size:
eval_max_new_tokens: 128
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:

UFOs-Finetune-V1.1

This model was trained from scratch on the json dataset. It achieves the following results on the evaluation set:

• Loss: 1.7367

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: 5e-06
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• optimizer: Use OptimizerNames.ADAMW_BNB with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 10
• training_steps: 50

Training results

Training Loss	Epoch	Step	Validation Loss
1.5018	0.1739	1	1.5418
1.4881	0.3478	2	1.5242
1.4127	0.6957	4	1.4386
1.3943	1.0	6	1.3957
1.3169	1.3478	8	1.3707
1.2603	1.6957	10	1.3561
1.2147	2.0	12	1.3535
1.0719	2.3478	14	1.3740
0.9741	2.6957	16	1.3890
1.024	3.0	18	1.4040
0.823	3.3478	20	1.4536
0.7372	3.6957	22	1.5242
0.7555	4.0	24	1.5201
0.622	4.3478	26	1.5416
0.5762	4.6957	28	1.5996
0.5535	5.0	30	1.6379
0.4547	5.3478	32	1.6690
0.4487	5.6957	34	1.6886
0.4435	6.0	36	1.6949
0.3969	6.3478	38	1.7070
0.3988	6.6957	40	1.7213
0.3917	7.0	42	1.7302
0.3746	7.3478	44	1.7348
0.3451	7.6957	46	1.7361
0.3513	8.0	48	1.7368
0.3572	8.3478	50	1.7367

Framework versions

• Transformers 4.48.3
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0