See axolotl config

axolotl version: 0.6.0

base_model: mistralai/Mistral-7B-v0.1
# 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/KJV-LLM-Pretrained-V1.1

load_in_8bit: false
load_in_4bit: false
strict: false

datasets:
  - path: AiAF/KJV-LLM-pretraining.jsonl
    type: completion
dataset_prepared_path:
val_set_size: 0.05
output_dir: ./outputs/out/KJV-LLM-Pretrained-V1.1

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

wandb_project: "LLM-Pretraining"
wandb_entity:
wandb_watch: "all"
wandb_name: "KJV-LLM-Pretrained-V1.1"
wandb_log_model: "false"

gradient_accumulation_steps: 4
micro_batch_size: 2
num_epochs: 8
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: /workspace/axolotl/outputs/out/KJV-LLM-Pretrained-V1.0/checkpoint-28
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:

KJV-LLM-Pretrained-V1.1

This model is a fine-tuned version of mistralai/Mistral-7B-v0.1 on the AiAF/KJV-LLM-pretraining.jsonl dataset. It achieves the following results on the evaluation set:

• Loss: 0.0986

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
• num_epochs: 8.0

Training results

Training Loss	Epoch	Step	Validation Loss
0.0809	0.1333	1	0.0975
0.0878	0.2667	2	0.0963
0.3153	0.5333	4	0.0909
0.077	0.8	6	0.0854
0.2377	1.0	8	0.0820
0.0509	1.2667	10	0.0858
0.0429	1.5333	12	0.0862
0.3496	1.8	14	0.0872
0.0426	2.0	16	0.0895
0.0337	2.2667	18	0.0888
0.0348	2.5333	20	0.0905
0.0852	2.8	22	0.0902
0.0317	3.0	24	0.0902
0.0304	3.2667	26	0.0900
0.0242	3.5333	28	0.0901
0.1936	4.2667	30	0.0918
0.0242	4.5333	32	0.0960
0.0219	4.8	34	0.0940
0.0187	5.0	36	0.0953
0.0188	5.2667	38	0.0954
0.0158	5.5333	40	0.0966
0.3393	5.8	42	0.0979
0.0163	6.0	44	0.0984
0.3313	6.2667	46	0.0984
0.015	6.5333	48	0.0985
0.0168	6.8	50	0.0986
0.0144	7.0	52	0.0986
0.0147	7.2667	54	0.0987
0.0154	7.5333	56	0.0986

Framework versions

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