license: apache-2.0
This model is a fine-tuned model for Chat based on mosaicml/mpt-7b with max_seq_lenght=2048 on databricks-dolly-15k, TigerResearch/tigerbot-alpaca-en-50k, TigerResearch/tigerbot-gsm-8k-en, TigerResearch/tigerbot-alpaca-zh-0.5m, TigerResearch/tigerbot-stackexchange-qa-en-0.5m, HC3 dataset.
Model date
Neural-chat-7b-v1.1 was trained between June and July 2023.
Evaluation
We use the same evaluation metrics as open_llm_leaderboard which uses Eleuther AI Language Model Evaluation Harness, a unified framework to test generative language models on a large number of different evaluation tasks.
| Model | Average ⬆️ | ARC (25-s) ⬆️ | HellaSwag (10-s) ⬆️ | MMLU (5-s) ⬆️ | TruthfulQA (MC) (0-s) ⬆️ |
|---|---|---|---|---|---|
| mosaicml/mpt-7b | 47.4 | 47.61 | 77.56 | 31 | 33.43 |
| mosaicml/mpt-7b-chat | 49.95 | 46.5 | 75.55 | 37.60 | 40.17 |
| Ours | 51.41 | 50.09 | 76.69 | 38.79 | 40.07 |
Bias evaluation
Following the blog evaluating-llm-bias, we select 10000 samples randomly from allenai/real-toxicity-prompts to evaluate toxicity bias in Language Models
| Model | Toxicity Rito ↓| |mosaicml/mpt-7b| 0.027 | | Ours | 0.0264 |
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 1e-05
- train_batch_size: 2
- eval_batch_size: 2
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- gradient_accumulation_steps: 8
- total_train_batch_size: 64
- total_eval_batch_size: 8
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.02
- num_epochs: 3.0
Inference with transformers
import transformers
model = transformers.AutoModelForCausalLM.from_pretrained(
'Intel/neural-chat-7b-v1-1',
trust_remote_code=True
)
Inference with INT8
Follow the instructions link to install the necessary dependencies. Use the below command to quantize the model using Intel Neural Compressor link and accelerate the inference.
python run_generation.py \
--model Intel/neural-chat-7b-v1-1 \
--quantize \
--sq \
--alpha 0.95 \
--ipex
Examples
code generation
summarization
trip
Organizations developing the model
The NeuralChat team with members from Intel/SATG/AIA/AIPT. Core team members: Kaokao Lv, Liang Lv, Chang Wang, Wenxin Zhang, Xuhui Ren, and Haihao Shen.