riyazahuja/Improver-DeepSeek-R1-Distill-Qwen-7B_full_4096

See axolotl config

axolotl version: 0.6.0

base_model: deepseek-ai/DeepSeek-R1-Distill-Qwen-7B
# optionally might have model_type or tokenizer_type
# model_type: AutoModelForCausalLM
# tokenizer_type: AutoTokenizer
# Automatically upload checkpoint and final model to HF
# hub_model_id: username/custom_model_name

trust_remote_code: true

load_in_8bit: false
load_in_4bit: false
strict: false

deepspeed: deepspeed_configs/zero3_bf16_cpuoffload_params.json

datasets:
  - path: improver/length_human_train.jsonl
    type: alpaca
dataset_prepared_path:
val_set_size: 0.05
output_dir: /data/user_data/riyaza/saved_models/DeepSeek-R1-Distill-Qwen-7B_full_4096

sequence_len: 4096
sample_packing: false
pad_to_sequence_len:

wandb_project: "DeepSeek-R1-Distill-Qwen-7B_full_4096"
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 4
micro_batch_size: 1
num_epochs: 2
optimizer: adamw_torch
lr_scheduler: cosine
learning_rate: 0.00005

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

gradient_checkpointing: true
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 2
xformers_attention:
flash_attention:

warmup_steps: 100
evals_per_epoch: 2
eval_table_size:
eval_max_new_tokens: 256
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:

data/user_data/riyaza/saved_models/DeepSeek-R1-Distill-Qwen-7B_full_4096

This model is a fine-tuned version of deepseek-ai/DeepSeek-R1-Distill-Qwen-7B on the improver/length_human_train.jsonl dataset. It achieves the following results on the evaluation set:

• Loss: 0.2721

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-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 6
• gradient_accumulation_steps: 4
• total_train_batch_size: 24
• total_eval_batch_size: 6
• optimizer: Use adamw_torch 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: 100
• num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss
0.2637	0.5043	58	0.2389
0.1027	1.0087	116	0.2581
0.0403	1.5130	174	0.2721

Framework versions

• Transformers 4.47.1
• Pytorch 2.5.1
• Datasets 3.1.0
• Tokenizers 0.21.0