See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: NousResearch/Yarn-Mistral-7b-64k
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 37c8687d026e9fd5_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/37c8687d026e9fd5_train_data.json
  type:
    field_instruction: question
    field_output: process
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: 4
eval_max_new_tokens: 128
eval_steps: 150
eval_table_size: null
flash_attention: false
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 4
gradient_checkpointing: true
group_by_length: false
hub_model_id: Romain-XV/440533d0-4ce2-46c1-8a0e-f65c18162efd
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.0002
load_best_model_at_end: true
load_in_4bit: true
load_in_8bit: false
local_rank: null
logging_steps: 1
lora_alpha: 128
lora_dropout: 0.3
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 64
lora_target_linear: true
lora_target_modules:
- q_proj
- k_proj
- v_proj
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 2639
micro_batch_size: 2
mlflow_experiment_name: /tmp/37c8687d026e9fd5_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 3
optimizer: adamw_bnb_8bit
output_dir: miner_id_24
pad_to_sequence_len: true
resume_from_checkpoint: null
s2_attention: null
sample_packing: false
save_steps: 150
sequence_len: 2048
special_tokens:
  pad_token: </s>
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.03300482530545966
wandb_entity: null
wandb_mode: online
wandb_name: a93fe1fb-147c-44b3-88ca-b0335c410502
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: a93fe1fb-147c-44b3-88ca-b0335c410502
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

440533d0-4ce2-46c1-8a0e-f65c18162efd

This model is a fine-tuned version of NousResearch/Yarn-Mistral-7b-64k on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0961

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: 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
• training_steps: 2639

Training results

Training Loss	Epoch	Step	Validation Loss
2.6189	0.0001	1	0.7597
1.8377	0.0082	150	0.3992
1.1386	0.0164	300	0.3287
0.8881	0.0246	450	0.2975
1.0054	0.0328	600	0.2704
1.0347	0.0410	750	0.2467
0.7466	0.0491	900	0.2219
0.7175	0.0573	1050	0.1994
0.8256	0.0655	1200	0.1825
0.731	0.0737	1350	0.1682
0.5574	0.0819	1500	0.1521
0.4409	0.0901	1650	0.1386
0.4958	0.0983	1800	0.1268
0.4718	0.1065	1950	0.1164
0.3277	0.1147	2100	0.1075
0.4109	0.1229	2250	0.1008
0.3703	0.1311	2400	0.0974
0.4028	0.1393	2550	0.0961

Framework versions

• PEFT 0.13.2
• Transformers 4.46.0
• Pytorch 2.5.0+cu124
• Datasets 3.0.1
• Tokenizers 0.20.1