See axolotl config

axolotl version: 0.5.2

base_model: Open-Orca/Mistral-7B-OpenOrca
model_type: AutoModelForCausalLM
tokenizer_config: Open-Orca/Mistral-7B-OpenOrca
tokenizer_type: AutoTokenizer
tokenizer_use_fast: false
resize_token_embeddings_to_32x: true

plugins:
  - axolotl.integrations.liger.LigerPlugin
liger_rope: true
liger_rms_norm: true
liger_glu_activation: true
liger_layer_norm: true
liger_fused_linear_cross_entropy: true

load_in_8bit: false
load_in_4bit: false
strict: false

chat_template: chatml
datasets:
  - path: skymizer/open-orca-conversations
    type: chat_template
    field_messages: messages

hf_use_auth_token: true
dataset_prepared_path: pretokenized/open-orca
output_dir: ./outputs/out

sequence_len: 2048
sample_packing: true
pad_to_sequence_len: true

val_set_size: 0.005
eval_sample_packing: false
# eval_causal_lm_metrics: ["perplexity"]

wandb_project: "axolotl_mistral_sft"
wandb_entity:
wandb_watch:
wandb_name: "mistral-7B-v0.1-csft-open-orca-on-open-orca"
wandb_log_model:

gradient_accumulation_steps: 2
micro_batch_size: 16
max_steps: 3000
optimizer: adamw_torch
lr_scheduler: cosine
learning_rate: 0.000005 
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.95
adam_eps: 0.000001
max_grad_norm: 1.0

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

hub_model_id: "skymizer/mistral-7b-v0.1-csft-open-orca-on-open-orca"
save_strategy: "steps"
save_steps: 1000

gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: false
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true

warmup_ratio: 0.03
eval_steps: 500
# evals_per_epoch: 4
eval_table_size:
eval_max_new_tokens: 128
debug:
deepspeed: deepspeed_configs/zero3_bf16.json
fsdp:
fsdp_config:

seed: 42

mistral-7b-v0.1-csft-open-orca-on-open-orca

This model is a fine-tuned version of Open-Orca/Mistral-7B-OpenOrca on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.1946

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: 16
• eval_batch_size: 16
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 2
• total_train_batch_size: 128
• total_eval_batch_size: 64
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.95) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 90
• training_steps: 3000

Training results

Training Loss	Epoch	Step	Validation Loss
1.1311	0.0002	1	4.4372
0.5277	0.0831	500	2.2236
0.463	0.1663	1000	2.2066
0.4855	0.2494	1500	2.2146
0.4662	0.3325	2000	2.1989
0.4494	0.4157	2500	2.1966
0.4268	0.4988	3000	2.1946

Framework versions

• Transformers 4.46.3
• Pytorch 2.5.1+cu124
• Datasets 3.1.0
• Tokenizers 0.20.3