tangledgroup/tangled-tinymistral-248m-instruct-v0.1

See axolotl config

axolotl version: 0.4.1

base_model: Locutusque/TinyMistral-248M-v2.5
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

load_in_8bit: false
load_in_4bit: true
strict: false

datasets:
  - path: PhilipMay/UltraChat-200k-ShareGPT-clean
    split: train
    type: sharegpt
chat_template: chatml
dataset_prepared_path: last_run_prepared
val_set_size: 0.05
output_dir: ./outputs/qlora-out

adapter: qlora
lora_model_dir:

# sequence_len: 16384
sequence_len: 8192
sample_packing: true
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

wandb_project:
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

# gradient_accumulation_steps: 4
gradient_accumulation_steps: 8
micro_batch_size: 2
num_epochs: 4
# optimizer: adamw_bnb_8bit
# optimizer: paged_adamw_8bit
optimizer: paged_adamw_32bit
lr_scheduler: cosine
learning_rate: 0.0002

train_on_inputs: true
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

loss_watchdog_threshold: 5.0
loss_watchdog_patience: 3

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:
  bos_token: "<|bos|>"
  eos_token: "<|endoftext|>"
  pad_token: "<|endoftext|>"
  unk_token: "<unk>"
tokens:
  - "<|im_start|>"
  - "<|im_end|>"
  - "<tools>"
  - "</tools>"
  - "<tool_call>"
  - "</tool_call>"
  - "<tool_response>"
  - "</tool_response>"
  - "<|reserved_0|>"
  - "<|reserved_1|>"
  - "<|reserved_2|>"
  - "<|reserved_3|>"
  - "<|reserved_4|>"
  - "<|reserved_5|>"
  - "<|reserved_6|>"
  - "<|reserved_7|>"
  - "<|reserved_8|>"
  - "<|reserved_9|>"
  - "<|reserved_10|>"
  - "<|reserved_11|>"
  - "<|reserved_12|>"
  - "<|reserved_13|>"
  - "<|reserved_14|>"
  - "<|reserved_15|>"
  - "<|reserved_16|>"
  - "<|reserved_17|>"
  - "<|reserved_18|>"
  - "<|reserved_19|>"
  - "<|reserved_20|>"
  - "<|reserved_21|>"
  - "<|reserved_22|>"
  - "<|reserved_23|>"
  - "<|reserved_24|>"
  - "<|reserved_25|>"
  - "<|reserved_26|>"
  - "<|reserved_27|>"
  - "<|reserved_28|>"
  - "<|reserved_29|>"
  - "<|reserved_30|>"
  - "<|reserved_31|>"
  - "<|reserved_32|>"
  - "<|reserved_33|>"
  - "<|reserved_34|>"
  - "<|reserved_35|>"
  - "<|reserved_36|>"
  - "<|reserved_37|>"
  - "<|reserved_38|>"
  - "<|reserved_39|>"
  - "<|reserved_40|>"
  - "<|reserved_41|>"
  - "<|reserved_42|>"
  - "<|reserved_43|>"
  - "<|reserved_44|>"
  - "<|reserved_45|>"
  - "<|reserved_46|>"
  - "<|reserved_47|>"
  - "<|reserved_48|>"
  - "<|reserved_49|>"
  - "<|reserved_50|>"
lora_modules_to_save:
  - "embed_tokens"
  - "lm_head"

outputs/qlora-out

This model is a fine-tuned version of Locutusque/TinyMistral-248M-v2.5 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.3030

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: 0.0002
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 10
• num_epochs: 4

Training results

Training Loss	Epoch	Step	Validation Loss
3.7052	0.0005	1	3.7319
2.5728	0.2501	538	2.6002
2.493	0.5002	1076	2.4906
2.4349	0.7503	1614	2.4315
2.4114	1.0005	2152	2.3935
2.367	1.2484	2690	2.3688
2.3076	1.4984	3228	2.3494
2.3045	1.7484	3766	2.3346
2.3349	1.9984	4304	2.3239
2.2226	2.2465	4842	2.3172
2.2092	2.4965	5380	2.3118
2.2494	2.7466	5918	2.3080
2.3105	2.9967	6456	2.3053
2.2519	3.2446	6994	2.3042
2.3991	3.4946	7532	2.3034
2.2051	3.7446	8070	2.3031
2.3465	3.9946	8608	2.3030

Framework versions

• PEFT 0.12.0
• Transformers 4.45.0.dev0
• Pytorch 2.4.0
• Datasets 2.21.0
• Tokenizers 0.19.1