|
--- |
|
license: apache-2.0 |
|
datasets: |
|
- huihui-ai/FineQwQ-142k |
|
base_model: |
|
- huihui-ai/MicroThinker-3B-Preview |
|
tags: |
|
- llama3.2 |
|
- abliterated |
|
- uncensored |
|
library_name: transformers |
|
pipeline_tag: text-generation |
|
language: |
|
- en |
|
--- |
|
# MicroThinker-3B-Preview-v2 |
|
|
|
MicroThinker-3B-Preview-v2, a new model fine-tuned from the [huihui-ai/MicroThinker-3B-Preview](https://huggingface.co/huihui-ai/MicroThinker-3B-Preview) model, focused on advancing AI reasoning capabilities. |
|
This model is superior to the [huihui-ai/MicroThinker-3B-Preview](https://huggingface.co/huihui-ai/MicroThinker-3B-Preview) model. |
|
|
|
## Use with ollama |
|
|
|
You can use [huihui_ai/microthinker2](https://ollama.com/huihui_ai/microthinker2) directly |
|
``` |
|
ollama run huihui_ai/microthinker2:3b |
|
``` |
|
|
|
## Training Details |
|
|
|
This is just a test, but the performance is quite good. |
|
|
|
Now, I'll introduce the test environment. |
|
|
|
The model was trained using 1 RTX 4090 GPU(24GB) . |
|
|
|
The fine-tuning process used 142k from the FineQwQ-142k dataset, max_length(tokens) 21710, quant_bits 4. |
|
|
|
The [SFT (Supervised Fine-Tuning)](https://github.com/modelscope/ms-swift) process is divided into several steps, and no code needs to be written. |
|
1. Create the environment. |
|
|
|
``` |
|
conda create -yn ms-swift python=3.11 |
|
conda activate ms-swift |
|
|
|
git clone https://github.com/modelscope/ms-swift.git |
|
|
|
cd ms-swift |
|
pip install -e . |
|
cd .. |
|
``` |
|
|
|
|
|
2. Download the model and dataset. |
|
|
|
``` |
|
huggingface-cli download huihui-ai/MicroThinker-3B-Preview --local-dir ./huihui-ai/MicroThinker-3B-Preview |
|
huggingface-cli download --repo-type dataset huihui-ai/FineQwQ-142k --local-dir ./data/FineQwQ-142k |
|
``` |
|
|
|
|
|
3. Used only the huihui-ai/FineQwQ-142k, Trained for 1 epoch: |
|
|
|
``` |
|
swift sft --model huihui-ai/MicroThinker-3B-Preview --model_type llama3_2 --train_type lora --dataset "data/FineQwQ-142k/FineQwQ-142k.jsonl" --num_train_epochs 1 --per_device_train_batch_size 1 --per_device_eval_batch_size 1 --max_length 21710 --quant_bits 4 --bnb_4bit_compute_dtype bfloat16 --bnb_4bit_quant_storage bfloat16 --lora_rank 8 --lora_alpha 32 --gradient_checkpointing true --weight_decay 0.1 --learning_rate 1e-4 --gradient_accumulation_steps 16 --eval_steps 500 --save_steps 500 --logging_steps 100 --system "You are a helpful assistant. You should think step-by-step." --output_dir output/MicroThinker-3B-Preview/lora/sft2 --model_author "huihui-ai" --model_name "MicroThinker-3B-Preview" |
|
``` |
|
|
|
4. Save the final fine-tuned model. After you're done, input `exit` to exit. |
|
Replace the directories below with specific ones. |
|
|
|
``` |
|
swift infer --model huihui-ai/MicroThinker-3B-Preview --model_type llama3_2 --adapters output/MicroThinker-3B-Preview/lora/sft2/v2-20250110-180322\checkpoint-8786 --stream true --infer_backend pt --max_new_tokens 2048 --merge_lora true |
|
``` |
|
|
|
|
|
This should create a new model directory: `checkpoint-8786-merged`, Rename the directory to `MicroThinker-3B-Preview-v2`, Copy or move this directory to the `huihui` directory. |
|
|
|
5. Perform inference on the final fine-tuned model. |
|
|
|
``` |
|
swift infer --model huihui/MicroThinker-3B-Preview-v2 --stream true --infer_backend pt --max_new_tokens 8192 |
|
``` |
|
|
|
6. Test examples. |
|
|
|
``` |
|
How many 'r' characters are there in the word "strawberry"? |
|
``` |
