--- inference: false license: other model_creator: SLAM-group model_link: https://huggingface.co./SLAM-group/NewHope model_name: NewHope model_type: llama quantized_by: TheBloke ---
TheBlokeAI

Chat & support: my new Discord server

Want to contribute? TheBloke's Patreon page

# NewHope - GGML - Model creator: [SLAM-group](https://huggingface.co./SLAM-group) - Original model: [NewHope](https://huggingface.co./SLAM-group/NewHope) ## Description This repo contains GGML format model files for [SLAM-group's NewHope](https://huggingface.co./SLAM-group/NewHope). GGML files are for CPU + GPU inference using [llama.cpp](https://github.com/ggerganov/llama.cpp) and libraries and UIs which support this format, such as: * [text-generation-webui](https://github.com/oobabooga/text-generation-webui), the most popular web UI. Supports NVidia CUDA GPU acceleration. * [KoboldCpp](https://github.com/LostRuins/koboldcpp), a powerful GGML web UI with GPU acceleration on all platforms (CUDA and OpenCL). Especially good for story telling. * [LM Studio](https://lmstudio.ai/), a fully featured local GUI with GPU acceleration on both Windows (NVidia and AMD), and macOS. * [LoLLMS Web UI](https://github.com/ParisNeo/lollms-webui), a great web UI with CUDA GPU acceleration via the c_transformers backend. * [ctransformers](https://github.com/marella/ctransformers), a Python library with GPU accel, LangChain support, and OpenAI-compatible AI server. * [llama-cpp-python](https://github.com/abetlen/llama-cpp-python), a Python library with GPU accel, LangChain support, and OpenAI-compatible API server. ## Repositories available * [GPTQ models for GPU inference, with multiple quantisation parameter options.](https://huggingface.co./TheBloke/NewHope-GPTQ) * [2, 3, 4, 5, 6 and 8-bit GGML models for CPU+GPU inference](https://huggingface.co./TheBloke/NewHope-GGML) * [SLAM-group's original unquantised fp16 model in pytorch format, for GPU inference and for further conversions](https://huggingface.co./SLAM-group/NewHope) ## Prompt template: Alpaca ``` Below is an instruction that describes a task. Write a response that appropriately completes the request. ### Instruction: {prompt} ### Response: ``` ## Compatibility ### Original llama.cpp quant methods: `q4_0, q4_1, q5_0, q5_1, q8_0` These are guaranteed to be compatible with any UIs, tools and libraries released since late May. They may be phased out soon, as they are largely superseded by the new k-quant methods. ### New k-quant methods: `q2_K, q3_K_S, q3_K_M, q3_K_L, q4_K_S, q4_K_M, q5_K_S, q6_K` These new quantisation methods are compatible with llama.cpp as of June 6th, commit `2d43387`. They are now also compatible with recent releases of text-generation-webui, KoboldCpp, llama-cpp-python, ctransformers, rustformers and most others. For compatibility with other tools and libraries, please check their documentation. ## Explanation of the new k-quant methods
Click to see details The new methods available are: * GGML_TYPE_Q2_K - "type-1" 2-bit quantization in super-blocks containing 16 blocks, each block having 16 weight. Block scales and mins are quantized with 4 bits. This ends up effectively using 2.5625 bits per weight (bpw) * GGML_TYPE_Q3_K - "type-0" 3-bit quantization in super-blocks containing 16 blocks, each block having 16 weights. Scales are quantized with 6 bits. This end up using 3.4375 bpw. * GGML_TYPE_Q4_K - "type-1" 4-bit quantization in super-blocks containing 8 blocks, each block having 32 weights. Scales and mins are quantized with 6 bits. This ends up using 4.5 bpw. * GGML_TYPE_Q5_K - "type-1" 5-bit quantization. Same super-block structure as GGML_TYPE_Q4_K resulting in 5.5 bpw * GGML_TYPE_Q6_K - "type-0" 6-bit quantization. Super-blocks with 16 blocks, each block having 16 weights. Scales are quantized with 8 bits. This ends up using 6.5625 bpw * GGML_TYPE_Q8_K - "type-0" 8-bit quantization. Only used for quantizing intermediate results. The difference to the existing Q8_0 is that the block size is 256. All 2-6 bit dot products are implemented for this quantization type. Refer to the Provided Files table below to see what files use which methods, and how.
## Provided files | Name | Quant method | Bits | Size | Max RAM required | Use case | | ---- | ---- | ---- | ---- | ---- | ----- | | [newhope.ggmlv3.q2_K.bin](https://huggingface.co./TheBloke/NewHope-GGML/blob/main/newhope.ggmlv3.q2_K.bin) | q2_K | 2 | 5.74 GB| 8.24 GB | New k-quant method. Uses GGML_TYPE_Q4_K for the attention.vw and feed_forward.w2 tensors, GGML_TYPE_Q2_K for the other tensors. | | [newhope.ggmlv3.q3_K_L.bin](https://huggingface.co./TheBloke/NewHope-GGML/blob/main/newhope.ggmlv3.q3_K_L.bin) | q3_K_L | 3 | 7.14 GB| 9.64 GB | New k-quant method. Uses GGML_TYPE_Q5_K for the attention.wv, attention.wo, and feed_forward.w2 tensors, else GGML_TYPE_Q3_K | | [newhope.ggmlv3.q3_K_M.bin](https://huggingface.co./TheBloke/NewHope-GGML/blob/main/newhope.ggmlv3.q3_K_M.bin) | q3_K_M | 3 | 6.53 GB| 9.03 GB | New k-quant method. Uses GGML_TYPE_Q4_K for the attention.wv, attention.wo, and feed_forward.w2 tensors, else GGML_TYPE_Q3_K | | [newhope.ggmlv3.q3_K_S.bin](https://huggingface.co./TheBloke/NewHope-GGML/blob/main/newhope.ggmlv3.q3_K_S.bin) | q3_K_S | 3 | 5.87 GB| 8.37 GB | New k-quant method. Uses GGML_TYPE_Q3_K for all tensors | | [newhope.ggmlv3.q4_0.bin](https://huggingface.co./TheBloke/NewHope-GGML/blob/main/newhope.ggmlv3.q4_0.bin) | q4_0 | 4 | 7.32 GB| 9.82 GB | Original quant method, 4-bit. | | [newhope.ggmlv3.q4_1.bin](https://huggingface.co./TheBloke/NewHope-GGML/blob/main/newhope.ggmlv3.q4_1.bin) | q4_1 | 4 | 8.14 GB| 10.64 GB | Original quant method, 4-bit. Higher accuracy than q4_0 but not as high as q5_0. However has quicker inference than q5 models. | | [newhope.ggmlv3.q4_K_M.bin](https://huggingface.co./TheBloke/NewHope-GGML/blob/main/newhope.ggmlv3.q4_K_M.bin) | q4_K_M | 4 | 8.06 GB| 10.56 GB | New k-quant method. Uses GGML_TYPE_Q6_K for half of the attention.wv and feed_forward.w2 tensors, else GGML_TYPE_Q4_K | | [newhope.ggmlv3.q4_K_S.bin](https://huggingface.co./TheBloke/NewHope-GGML/blob/main/newhope.ggmlv3.q4_K_S.bin) | q4_K_S | 4 | 7.56 GB| 10.06 GB | New k-quant method. Uses GGML_TYPE_Q4_K for all tensors | | [newhope.ggmlv3.q5_0.bin](https://huggingface.co./TheBloke/NewHope-GGML/blob/main/newhope.ggmlv3.q5_0.bin) | q5_0 | 5 | 8.95 GB| 11.45 GB | Original quant method, 5-bit. Higher accuracy, higher resource usage and slower inference. | | [newhope.ggmlv3.q5_1.bin](https://huggingface.co./TheBloke/NewHope-GGML/blob/main/newhope.ggmlv3.q5_1.bin) | q5_1 | 5 | 9.76 GB| 12.26 GB | Original quant method, 5-bit. Even higher accuracy, resource usage and slower inference. | | [newhope.ggmlv3.q5_K_M.bin](https://huggingface.co./TheBloke/NewHope-GGML/blob/main/newhope.ggmlv3.q5_K_M.bin) | q5_K_M | 5 | 9.40 GB| 11.90 GB | New k-quant method. Uses GGML_TYPE_Q6_K for half of the attention.wv and feed_forward.w2 tensors, else GGML_TYPE_Q5_K | | [newhope.ggmlv3.q5_K_S.bin](https://huggingface.co./TheBloke/NewHope-GGML/blob/main/newhope.ggmlv3.q5_K_S.bin) | q5_K_S | 5 | 9.14 GB| 11.64 GB | New k-quant method. Uses GGML_TYPE_Q5_K for all tensors | | [newhope.ggmlv3.q6_K.bin](https://huggingface.co./TheBloke/NewHope-GGML/blob/main/newhope.ggmlv3.q6_K.bin) | q6_K | 6 | 10.83 GB| 13.33 GB | New k-quant method. Uses GGML_TYPE_Q8_K for all tensors - 6-bit quantization | | [newhope.ggmlv3.q8_0.bin](https://huggingface.co./TheBloke/NewHope-GGML/blob/main/newhope.ggmlv3.q8_0.bin) | q8_0 | 8 | 13.83 GB| 16.33 GB | Original quant method, 8-bit. Almost indistinguishable from float16. High resource use and slow. Not recommended for most users. | **Note**: the above RAM figures assume no GPU offloading. If layers are offloaded to the GPU, this will reduce RAM usage and use VRAM instead. ## How to run in `llama.cpp` I use the following command line; adjust for your tastes and needs: ``` ./main -t 10 -ngl 32 -m newhope.ggmlv3.q4_K_M.bin --color -c 2048 --temp 0.7 --repeat_penalty 1.1 -n -1 -p "### Instruction: Write a story about llamas\n### Response:" ``` Change `-t 10` to the number of physical CPU cores you have. For example if your system has 8 cores/16 threads, use `-t 8`. Change `-ngl 32` to the number of layers to offload to GPU. Remove it if you don't have GPU acceleration. If you want to have a chat-style conversation, replace the `-p ` argument with `-i -ins` ## How to run in `text-generation-webui` Further instructions here: [text-generation-webui/docs/llama.cpp-models.md](https://github.com/oobabooga/text-generation-webui/blob/main/docs/llama.cpp-models.md). ## Discord For further support, and discussions on these models and AI in general, join us at: [TheBloke AI's Discord server](https://discord.gg/theblokeai) ## Thanks, and how to contribute. Thanks to the [chirper.ai](https://chirper.ai) team! I've had a lot of people ask if they can contribute. I enjoy providing models and helping people, and would love to be able to spend even more time doing it, as well as expanding into new projects like fine tuning/training. If you're able and willing to contribute it will be most gratefully received and will help me to keep providing more models, and to start work on new AI projects. Donaters will get priority support on any and all AI/LLM/model questions and requests, access to a private Discord room, plus other benefits. * Patreon: https://patreon.com/TheBlokeAI * Ko-Fi: https://ko-fi.com/TheBlokeAI **Special thanks to**: Luke from CarbonQuill, Aemon Algiz. **Patreon special mentions**: Willem Michiel, Ajan Kanaga, Cory Kujawski, Alps Aficionado, Nikolai Manek, Jonathan Leane, Stanislav Ovsiannikov, Michael Levine, Luke Pendergrass, Sid, K, Gabriel Tamborski, Clay Pascal, Kalila, William Sang, Will Dee, Pieter, Nathan LeClaire, ya boyyy, David Flickinger, vamX, Derek Yates, Fen Risland, Jeffrey Morgan, webtim, Daniel P. Andersen, Chadd, Edmond Seymore, Pyrater, Olusegun Samson, Lone Striker, biorpg, alfie_i, Mano Prime, Chris Smitley, Dave, zynix, Trenton Dambrowitz, Johann-Peter Hartmann, Magnesian, Spencer Kim, John Detwiler, Iucharbius, Gabriel Puliatti, LangChain4j, Luke @flexchar, Vadim, Rishabh Srivastava, Preetika Verma, Ai Maven, Femi Adebogun, WelcomeToTheClub, Leonard Tan, Imad Khwaja, Steven Wood, Stefan Sabev, Sebastain Graf, usrbinkat, Dan Guido, Sam, Eugene Pentland, Mandus, transmissions 11, Slarti, Karl Bernard, Spiking Neurons AB, Artur Olbinski, Joseph William Delisle, ReadyPlayerEmma, Olakabola, Asp the Wyvern, Space Cruiser, Matthew Berman, Randy H, subjectnull, danny, John Villwock, Illia Dulskyi, Rainer Wilmers, theTransient, Pierre Kircher, Alexandros Triantafyllidis, Viktor Bowallius, terasurfer, Deep Realms, SuperWojo, senxiiz, Oscar Rangel, Alex, Stephen Murray, Talal Aujan, Raven Klaugh, Sean Connelly, Raymond Fosdick, Fred von Graf, chris gileta, Junyu Yang, Elle Thank you to all my generous patrons and donaters! # Original model card: SLAM-group's NewHope # NewHope: Harnessing 99% of GPT-4's Programming Capabilities We introduce NewHope, a fine-tuned chat model based on llama-2-13b, aiming to provide a strong coding capability. NewHope handle different languages including Python, C++, Java, JavaScript, Go, and more. Preliminary evaluation on HumanEval shows that **NewHope possesses 99% of GPT-4's programming capabilities**. **Contact**: SLAM (SUFE Large AI Model) is a research group at Shanghai University of Finance and Economics. cui.wanyun@sufe.edu.cn **TODO**: We will release more evaluatation results and training details later. # Evaluation Results We evaluated NewHope on [HumanEval](https://github.com/openai/human-eval) using the official evaluation script by OpenAI. We compared the Pass@1 metric of NewHope with other models. The results of other models are from PapersWithCode. | Model | Pass@1 | | ----- | ------ | | **GPT-4** | **67.0** | | **NewHope** | **66.5** | | PanGu-Coder2 15B | 61.6 | | WizardCoder 15B | 57.3 | | phi-1 1.3B | 50.6 | | GPT-3.5 | 48.1 | | phi-1-small | 45.0 | | PaLM-Coder | 36.0 | | CodeGeeX2-6B | 35.9 | # Model Weights We have open-sourced the model weights [NewHope](https://huggingface.co./SLAM-group/NewHope). We are uploading the model weights. The weights will be available in a few hours. # Usage To load the NewHope model using Transformers, use the following code: ``` import torch from transformers import LlamaTokenizer, LlamaForCausalLM base_model = "SLAM-group/NewHope" tokenizer = LlamaTokenizer.from_pretrained(base_model) model = LlamaForCausalLM.from_pretrained(base_model, torch_dtype=torch.float16, device_map="auto") # model.config.use_cache is default to `False`. For inference: `model.config.use_cache = True` ``` **Note:** At least Huggingface Transformers **4.31.0** is required to load this model! You can ask NewHope to generate code with instructions. We provide a simple example of how NewHope model generates code with the specific prompt: ``` # Suppose required tokenizer and model have already been loaded instruction = "Write a Python function to tell me what the date is today." prompt = f" ### Instruction:\n{instruction}\n\n### Response:\n" inputs = tokenizer(prompt, add_special_tokens=False, return_tensors="pt").to("cuda") output = model.generate(**inputs, do_sample=True, top_p=0.9, max_new_tokens=2048)[0] decoded_output = tokenizer.decode(output, skip_special_tokens=True).split("### Response:\n")[-1].strip() print(decoded_output) ``` You can also interact with NewHope in a dialog manner with the following prompt: ``` ### Instruction:\nQ1\n\n### Response:\nA1 ### Instruction:\nQ2\n\n### Response:\nA2 ``` # Evaluation ### Local setup 1. Install HumanEval for evaluation. [Details](https://github.com/openai/human-eval) 2. Install dependencies ```bash pip install -r requirements.txt ``` --- For HumanEval, we use the following prompt: ``` example_input = 'def is_odd(number: int) -> bool:\n """ Check whether the given number is odd\n >>> is_odd(3)\n True\n >>> is_odd(6)\n False\n """\n' example_output = 'def is_odd(number: int) -> bool:\n """ Check whether the given number is odd\n >>> is_odd(3)\n True\n >>> is_odd(6)\n False\n """\n return number % 2 == 1' task_in_humaneval = "REPLACE `task_in_humaneval` WITH THE SPECIFIC TASK IN HUMANEVAL DATA" prompt = f" ### Instruction:\nComplete the given function below:\n\n{example_input}\n\n### Response:\n{example_output} ### Instruction:\nComplete the given function below:\n\n{task_in_human_eval}\n\n### Response:\n" ``` To reproduce the results on HumanEval, use the following script: ``` python complete.py --base_model SLAM-group/NewHope --output_dir output --n_gpu 8 ``` The above script will generate `samples.jsonl` in `output_dir`, which can be directly evaluated by HumanEval. [Evaluation procedure](https://github.com/openai/human-eval). We conducted the experiment with `fp16` on 8xA800, 80GB GPUs, reaching `66.5%` on Pass@1 (v.s. GPT4 `67.0%`). # Citation ``` @misc{2023newhope, title={NewHope: Harnessing 99% of GPT-4's Programming Capabilities}, author={Wanyun Cui and Qianle Wang}, howpublished = https://github.com/SLAM-group/newhope, year={2023} } ```