You can deploy this model and start earning money today!

minpeter/
LoRA-Qwen3-4b-v1-iteration-02-sf-apigen-00

Name: LoRA-Qwen3-4b-v1-iteration-02-sf-apigen-00
Availability: InStock

generated_from_trainer

minpeter/apigen-mt-5k-friendli

Qwen/Qwen3-4B

adapter

apache-2.0

Views

No views yet

It's garbage, this learning settings are wrong.

docker run --rm --runtime nvidia --ipc=host --gpus 'all' \
    -v /data/huggingface:/root/.cache/huggingface \
    -v /data:/data \
    -e "HUGGING_FACE_HUB_TOKEN=$HF_TOKEN" \
    -p 4000:4000 \
    vllm/vllm-openai:latest \
      --model qwen/qwen3-4b \
      --enforce-eager --port 4000 --served-model-name base \
      --enable-auto-tool-choice --tool-call-parser hermes \
      --enable-lora --max-lora-rank 128 --lora-modules tool=minpeter/LoRA-Qwen3-4b-v1-iteration-01-sf-apigen-00

Qwen3-4B BFCL GT (w/o thinking)

🔍 Running test: irrelevance
✅ Test completed: irrelevance. 🎯 Accuracy: 0.875
🔍 Running test: multi_turn_base
✅ Test completed: multi_turn_base. 🎯 Accuracy: 0.085
🔍 Running test: parallel_multiple
✅ Test completed: parallel_multiple. 🎯 Accuracy: 0.89
🔍 Running test: parallel
✅ Test completed: parallel. 🎯 Accuracy: 0.885
🔍 Running test: simple
✅ Test completed: simple. 🎯 Accuracy: 0.9325
🔍 Running test: multiple
✅ Test completed: multiple. 🎯 Accuracy: 0.92

See axolotl config

axolotl version: 0.9.2

yaml

base_model: Qwen/Qwen3-4B
hub_model_id: minpeter/LoRA-Qwen3-4b-v1-iteration-01-sf-apigen-00

load_in_8bit: false
load_in_4bit: false
strict: false

datasets:
  # 0.5k
  - path: minpeter/apigen-mt-5k-friendli
    data_files:
      - train.jsonl
      - test.jsonl
    type: chat_template
    chat_template: qwen3
    split_thinking: true
    field_messages: messages
    message_field_role: role
    message_field_content: content
    shards: 3

chat_template: qwen3

dataset_prepared_path: last_run_prepared

output_dir: ./output

adapter: lora
lora_model_dir:

sequence_len: 16384
pad_to_sequence_len: true
sample_packing: true

val_set_size: 0.05
eval_sample_packing: true
evals_per_epoch: 3

lora_r: 8
lora_alpha: 16
lora_dropout: 0.05
lora_fan_in_fan_out:
lora_target_modules:
  - gate_proj
  - down_proj
  - up_proj
  - q_proj
  - v_proj
  - k_proj
  - o_proj

wandb_project: "axolotl"
wandb_entity: "kasfiekfs-e"
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 2
micro_batch_size: 2
num_epochs: 2
optimizer: adamw_8bit
lr_scheduler: cosine
learning_rate: 0.0002

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

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
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:

LoRA-Qwen3-4b-v1-iteration-01-sf-apigen-00

This model is a fine-tuned version of Qwen/Qwen3-4B on the minpeter/apigen-mt-5k-friendli dataset. It achieves the following results on the evaluation set:

Loss: 0.1821

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: 2
total_train_batch_size: 4
optimizer: Use OptimizerNames.ADAMW_8BIT 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
num_epochs: 2.0

Training results

Training Loss	Epoch	Step	Validation Loss
0.929	0.0045	1	0.8641
0.3071	0.3341	74	0.2398
0.1946	0.6682	148	0.2112
0.1311	1.0	222	0.1951
0.1204	1.3341	296	0.1865
0.1926	1.6682	370	0.1821

Framework versions

PEFT 0.15.2
Transformers 4.51.3
Pytorch 2.6.0+cu124
Datasets 3.5.1
Tokenizers 0.21.1

minpeter/LoRA-Qwen3-4b-v1-iteration-02-sf-apigen-00

Qwen3-4B BFCL GT (w/o thinking)

LoRA-Qwen3-4b-v1-iteration-01-sf-apigen-00

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

minpeter/
LoRA-Qwen3-4b-v1-iteration-02-sf-apigen-00