README.md

---
library_name: peft
license: apache-2.0
base_model: Qwen/Qwen2.5-1.5B-Instruct
tags:
- generated_from_trainer
datasets:
- minpeter/bfcl-v1-non-live-ast-hermes
model-index:
- name: output
  results: []
---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

[<img src="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/axolotl-ai-cloud/axolotl)
<details><summary>See axolotl config</summary>

axolotl version: `0.6.0`
```yaml
base_model: Qwen/Qwen2.5-1.5B-Instruct
# Automatically upload checkpoint and final model to HF
# hub_model_id: username/custom_model_name

load_in_8bit: false
load_in_4bit: false
strict: false

# datasets:
#   - path: oneline-tool.jsonl
#     type: chat_template
#     chat_template: chatml
#     field_messages: conversations
#     message_field_role: from
#     message_field_content: value
  # - path: minpeter/stanford-alpaca-regen-llama-3.3
  #   type:
  #     format: "<|im_start|>user\n{instruction}\n{input}<|im_end|>\n<|im_start|>assistant\n"
  #     no_input_format: "<|im_start|>user\n{instruction}<|im_end|>\n<|im_start|>assistant\n"
  #   shards: 52000
datasets:
  - path: minpeter/bfcl-v1-non-live-ast-hermes
    data_files:
      - result.parquet
    type: chat_template
    chat_template: chatml
    field_messages: conversations
    message_field_role: from
    message_field_content: value

chat_template: chatml


dataset_prepared_path: last_run_prepared

output_dir: ./output

adapter: lora
lora_model_dir:

sequence_len: 2048
pad_to_sequence_len: true
sample_packing: true

# val_set_size: 0.1
# eval_sample_packing: true

lora_r: 16
lora_alpha: 32
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: 1
optimizer: adamw_8bit
lr_scheduler: cosine
learning_rate: 0.0002

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

# special_tokens:
#   bos_token: null
#   eos_token: <|im_end|>
#   pad_token: <|endoftext|>
```

</details><br>

# output

This model is a fine-tuned version of [Qwen/Qwen2.5-1.5B-Instruct](https://huggingface.co/Qwen/Qwen2.5-1.5B-Instruct) on the minpeter/bfcl-v1-non-live-ast-hermes dataset.

## Model description

Intentionally contaminated BFCL model, 😈

```
🔍 Running test: parallel_multiple
✅ Test completed: parallel_multiple. 🎯 Accuracy: 0.84
🔍 Running test: parallel
✅ Test completed: parallel. 🎯 Accuracy: 0.875
🔍 Running test: simple
✅ Test completed: simple. 🎯 Accuracy: 0.94
🔍 Running test: multiple
✅ Test completed: multiple. 🎯 Accuracy: 0.89
```

## Inference

```shell
docker run --rm --runtime nvidia --gpus '"device=0"' \
    -p 8000:8000 \
    -e HF_TOKEN="<secret>" \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    vllm/vllm-openai:latest \
    --model Qwen/Qwen2.5-1.5B-Instruct \
    --enable-lora \
    --lora-modules \
        tool=minpeter/LoRA-corrupted-bfcl-1.5B-v1 \
    --enable-auto-tool-choice \
    --tool-call-parser hermes
```


## 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: 1.0

### Training results



### Framework versions

- PEFT 0.14.0
- Transformers 4.48.3
- Pytorch 2.5.1+cu124
- Datasets 3.2.0
- Tokenizers 0.21.0