See axolotl config

axolotl version: 0.4.1

# base_model: deepseek-ai/deepseek-coder-1.3b-instruct
base_model: Qwen/CodeQwen1.5-7B-Chat
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer
is_mistral_derived_model: false

load_in_8bit: true
load_in_4bit: false
strict: false

lora_fan_in_fan_out: false
data_seed: 49
seed: 49

datasets:
  - path: sample_data/alpaca_synth_cypher.jsonl
    type: sharegpt
    conversation: alpaca
dataset_prepared_path: last_run_prepared
val_set_size: 0.1
output_dir: ./qlora-alpaca-codeqwen1.5-7b-chat-lora8
# output_dir: ./qlora-alpaca-out

hub_model_id: jermyn/CodeQwen1.5-7B-Chat-lora8-NLQ2Cypher
# hub_model_id: jermyn/deepseek-code-1.3b-inst-NLQ2Cypher

adapter: lora   # 'qlora' or leave blank for full finetune
lora_model_dir:

sequence_len: 896
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

# If you added new tokens to the tokenizer, you may need to save some LoRA modules because they need to know the new tokens.
# For LLaMA and Mistral, you need to save `embed_tokens` and `lm_head`. It may vary for other models.
# `embed_tokens` converts tokens to embeddings, and `lm_head` converts embeddings to token probabilities.
# https://github.com/huggingface/peft/issues/334#issuecomment-1561727994
# lora_modules_to_save:
#   - embed_tokens
#   - lm_head

wandb_project: fine-tune-axolotl
wandb_entity: jermyn

gradient_accumulation_steps: 2
micro_batch_size: 8
eval_batch_size: 8
num_epochs: 6
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0005
max_grad_norm: 1.0
adam_beta2: 0.95
adam_epsilon: 0.00001

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
evals_per_epoch: 4
eval_table_size:
eval_table_max_new_tokens: 128
# saves_per_epoch: 6
save_steps: 10
save_total_limit: 3
debug:
weight_decay: 0.0
fsdp:
fsdp_config:
# special_tokens:
#   bos_token: "<s>"
#   eos_token: "</s>"
#   unk_token: "<unk>"
save_safetensors: true

CodeQwen1.5-7B-Chat-lora8-NLQ2Cypher

This model is a fine-tuned version of Qwen/CodeQwen1.5-7B-Chat on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.3720

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: 0.0005
• train_batch_size: 8
• eval_batch_size: 8
• seed: 49
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.95) and epsilon=1e-05
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 10
• num_epochs: 6

Training results

Training Loss	Epoch	Step	Validation Loss
1.1649	0.1538	1	0.9270
1.1566	0.3077	2	0.9268
1.0746	0.6154	4	0.8194
0.6428	0.9231	6	0.4970
0.2459	1.2308	8	0.4760
0.3512	1.5385	10	0.5091
0.1654	1.8462	12	0.4742
0.1484	2.1538	14	0.4560
0.137	2.4615	16	0.4105
0.0746	2.7692	18	0.3736
0.0539	3.0769	20	0.3412
0.1147	3.3846	22	0.3307
0.056	3.6923	24	0.3242
0.0767	4.0	26	0.3524
0.0583	4.3077	28	0.3690
0.0666	4.6154	30	0.3727
0.0539	4.9231	32	0.3773
0.0367	5.2308	34	0.3796
0.0297	5.5385	36	0.3720

Framework versions

• PEFT 0.11.1
• Transformers 4.42.3
• Pytorch 2.1.2+cu118
• Datasets 2.19.1
• Tokenizers 0.19.1