axolotl version: 0.4.1
base_model: meta-llama/Llama-3.2-1B
model_type: LlamaForCausalLM
tokenizer_type: AutoTokenizer
load_in_8bit: false
load_in_4bit: false
strict: false
chat_template: chatml
datasets:
- path: datasets/airoboros_3.2_without_contextual_slimorca_orca_sharegpt.json
ds_type: json
type: sharegpt
conversation: chatml
- path: datasets/allenai_wild_chat_gpt4_english_toxic_random_half_4k_sharegpt.json
ds_type: json
type: sharegpt
strict: false
conversation: chatml
- path: datasets/buzz_unstacked_chosen_math_removed_filtered.json
ds_type: json
type: alpaca
conversation: chatml
- path: datasets/capybara_sharegpt.json
ds_type: json
type: sharegpt
conversation: chatml
- path: datasets/cot_alpaca_gpt4_extracted_openhermes_2.5_sharegpt.json
ds_type: json
type: sharegpt
conversation: chatml
- path: datasets/everythinglm-data-v3_sharegpt.json
ds_type: json
type: sharegpt
strict: false
conversation: chatml
- path: datasets/gpt4_data_lmys_1m_sharegpt.json
ds_type: json
type: sharegpt
conversation: chatml
- path: datasets/gpteacher-instruct-special-alpaca.json
ds_type: json
type: gpteacher
conversation: chatml
- path: datasets/merged_all.json
ds_type: json
type: alpaca
conversation: chatml
- path: datasets/no_robots_sharegpt.json
ds_type: json
type: sharegpt
strict: false
conversation: chatml
- path: datasets/oasst_top1_from_fusechatmixture_sharegpt.json
ds_type: json
type: sharegpt
strict: false
conversation: chatml
- path: datasets/pippa_bagel_repo_3k_sharegpt.json
ds_type: json
type: sharegpt
conversation: chatml
- path: datasets/rpguild_quarter_alignment_lab_sharegpt.json
ds_type: json
type: sharegpt
conversation: chatml
- path: datasets/sharegpt_gpt4_english.json
ds_type: json
type: sharegpt
conversation: chatml
- path: datasets/slimorca_dedup_filtered_95k_sharegpt.json
ds_type: json
type: sharegpt
conversation: chatml
- path: datasets/soda_diaolog_longest_tenth_buzz_sharegpt.json
ds_type: json
type: sharegpt
conversation: chatml
- path: datasets/synthia-v1.3_sharegpt_12500.json
ds_type: json
type: sharegpt
conversation: chatml
- path: datasets/system_conversations_dolphin_sharegpt.json
ds_type: json
type: sharegpt
conversation: chatml
- path: datasets/NuminaMath-CoT-olympiads-40k_alpaca.json
ds_type: json
type: alpaca
conversation: chatml
- path: datasets/math-gpt-4o-40k_alpaca.json
ds_type: json
type: alpaca
conversation: chatml
- path: datasets/sonnet3.5_science_conversations_sharegpt.json
ds_type: json
type: sharegpt
conversation: chatml
- path: datasets/reasoning-0.01_sharegpt.jsonl
ds_type: json
type: sharegpt
conversation: chatml
dataset_prepared_path: last_run_prepared
val_set_size: 0.002
output_dir: ./Einstein-v8-Llama3.2-1B-model
sequence_len: 8192
sample_packing: true
pad_to_sequence_len: true
eval_sample_packing: false
wandb_project: Einstein
wandb_entity:
wandb_watch:
wandb_name: Einstein-v8-Llama3.2-1B-2-epoch
wandb_log_model:
hub_model_id: Weyaxi/Einstein-v8-Llama3.2-1B
save_safetensors: true
gradient_accumulation_steps: 4
micro_batch_size: 4
num_epochs: 2
optimizer: adamw_bnb_8bit # look
lr_scheduler: cosine
learning_rate: 0.000005 # look
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
warmup_steps: 10
evals_per_epoch: 4
eval_table_size:
eval_table_max_new_tokens: 128
saves_per_epoch: 1
debug:
deepspeed: axolotl/deepspeed_configs/zero3_bf16.json
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
bos_token: "<s>"
eos_token: "<|im_end|>"
unk_token: "<unk>"
pad_token: <|end_of_text|> # changed
tokens:
- "<|im_start|>"
Einstein-v8-Llama3.2-1B
This model is a fine-tuned version of meta-llama/Llama-3.2-1B on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.9292
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: 4
- eval_batch_size: 4
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- gradient_accumulation_steps: 4
- total_train_batch_size: 64
- total_eval_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: 2
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 1.4261 | 0.0009 | 1 | 1.4028 |
| 1.0487 | 0.2501 | 268 | 0.9917 |
| 1.0484 | 0.5001 | 536 | 0.9652 |
| 1.0039 | 0.7502 | 804 | 0.9499 |
| 1.0528 | 1.0002 | 1072 | 0.9399 |
| 0.9559 | 1.2481 | 1340 | 0.9345 |
| 0.9078 | 1.4981 | 1608 | 0.9309 |
| 0.9702 | 1.7481 | 1876 | 0.9295 |
| 0.929 | 1.9981 | 2144 | 0.9292 |
Framework versions
- Transformers 4.45.0
- Pytorch 2.3.1+cu121
- Datasets 2.21.0
- Tokenizers 0.20.0
Open LLM Leaderboard Evaluation Results
Detailed results can be found here
| Metric | Value |
|---|---|
| Avg. | 4.63 |
| IFEval (0-Shot) | 18.62 |
| BBH (3-Shot) | 3.01 |
| MATH Lvl 5 (4-Shot) | 0.00 |
| GPQA (0-shot) | 1.12 |
| MuSR (0-shot) | 3.22 |
| MMLU-PRO (5-shot) | 1.79 |