Triangle104/INTELLECT-1-Instruct-Q4_K_S-GGUF

This model was converted to GGUF format from PrimeIntellect/INTELLECT-1-Instruct using llama.cpp via the ggml.ai's GGUF-my-repo space. Refer to the original model card for more details on the model.

---

Model details:

INTELLECT-1 is the first collaboratively trained 10 billion parameter language model trained from scratch on 1 trillion tokens of English text and code.

This is an instruct model. The base model associated with it is INTELLECT-1.

INTELLECT-1 was trained on up to 14 concurrent nodes distributed across 3 continents, with contributions from 30 independent community contributors providing compute. The training code utilizes the prime framework, a scalable distributed training framework designed for fault-tolerant, dynamically scaling, high-perfomance training on unreliable, globally distributed workers. The key abstraction that allows dynamic scaling is the ElasticDeviceMesh which manages dynamic global process groups for fault-tolerant communication across the internet and local process groups for communication within a node. The model was trained using the DiLoCo algorithms with 100 inner steps. The global all-reduce was done with custom int8 all-reduce kernels to reduce the communication payload required, greatly reducing the communication overhead by a factor 400x.

For more detailed technical insights, please refer to our technical paper.

Note: You must add a BOS token at the beginning of each sample. Performance may be impacted otherwise.

    Usage

import torch from transformers import AutoModelForCausalLM, AutoTokenizer

torch.set_default_device("cuda") model = AutoModelForCausalLM.from_pretrained("PrimeIntellect/INTELLECT-1-Instruct") tokenizer = AutoTokenizer.from_pretrained("PrimeIntellect/INTELLECT-1-Instruct")

input_text = "What is the Metamorphosis of Prime Intellect about?" input_ids = tokenizer.encode(input_text, return_tensors="pt") output_ids = model.generate(input_ids, max_length=50, num_return_sequences=1) output_text = tokenizer.decode(output_ids[0], skip_special_tokens=True)

print(output_text)

    Example text generation pipeline

import torch from transformers import pipeline torch.set_default_device("cuda")

pipe = pipeline("text-generation", model="PrimeIntellect/INTELLECT-1") print(pipe("What is prime intellect ?"))

    Model Details

Compute Contributors: Prime Intellect, Arcee AI, kotaro, skre_0, marlo, rodeo, Herb, Olas, superchillen, Hugging Face, mev_pete, 0xfr_, dj, primeprimeint1234, Marco Giglio, realtek, Hyperbolic, hecataeus, NWO, Virtual Machine, droll, SemiAnalysis, waiting_, toptickcrypto, sto, Johannes, washout_segment_0b, klee Release Date: 29 Nov 2024 Model License: Apache 2.0

    Technical Specifications






    

Parameter: Value

Parameter Size: 10B

Number of Layers: 42

Number of Attention Heads: 32

Hidden Size: 4096

Context Length: 8192

Vocabulary Size: 128256

Training Details:

Dataset: 55% fineweb-edu, 10% fineweb, 20% Stack V1, 10% dclm-baseline, 5% open-web-math Tokens: 1 Trillion Optimizer: Diloco/LocalSGD - Inner Optimizer: AdamW, Outer Optmizer: Nesterov SGD

    Post-training

The post-training has been handled by arcee

After completing the globally distributed pretraining phase, we applied several post-training techniques to enhance INTELLECT-1's capabilities and task-specific performance. Our post-training methodology consisted of three main phases.

First, we conducted an extensive series of 16 Supervised Fine-Tuning (SFT) trainings, with individual runs ranging from 1 to 3.3 billion tokens each. The most successful configuration used 2.4 billion training tokens over 3 epochs. We used MergeKit, EvolKit, and DistillKit from Arcee AI to combine the models, generate the data sets, and distill the logits, respectively. For training data, we used a diverse set of high-quality datasets:

New Datasets (released with INTELLECT-1):

arcee-ai/EvolKit-75k (generated via EvolKit) arcee-ai/Llama-405B-Logits arcee-ai/The-Tomb

Instruction Following:

mlabonne/open-perfectblend-fixed (generalist capabilities) microsoft/orca-agentinstruct-1M-v1-cleaned (Chain-of-Thought) Post-training-Data-Flywheel/AutoIF-instruct-61k-with-funcs

Domain-Specific:

Team-ACE/ToolACE (function calling) Synthia coder (programming) ServiceNow-AI/M2Lingual (multilingual) AI-MO/NuminaMath-TIR (mathematics)

Tulu-3 Persona Datasets:

allenai/tulu-3-sft-personas-code allenai/tulu-3-sft-personas-math allenai/tulu-3-sft-personas-math-grade allenai/tulu-3-sft-personas-algebra

Second, we execute 8 distinct Direct Preference Optimization (DPO) runs with various combinations of data sets to enhance specific performance metrics and align the model with human preferences. A key advantage in our post-training process was INTELLECT-1's use of the Llama-3 tokenizer, which allowed us to utilize logits from Llama-3.1-405B to heal and maintain precision during the post-training process via DistillKit.

Finally, we performed 16 strategic merges between candidate models using MergeKit to create superior combined models that leverage the strengths of different training runs. During the post-training phase, we observed that when using a ChatML template without an explicit BOS (begin-of-sequence) token, the initial loss was approximately 15. However, when switching to the Llama 3.1 chat template, the loss for these trainings started much lower at approximately 1.1, indicating better alignment with the underlying Llama 3 tokenizer.

The combination of these post-training techniques resulted in significant improvements in various benchmarks, particularly in knowledge retrieval, grade school math, instruction following and reasoning.

    Citations

If you use this model in your research, please cite it as follows:

@article{jaghouar2024intellect, title={INTELLECT-1 Technical Report.}, author={Jaghouar, Sami and Ong, Jack Min and Basra, Manveer and Obeid, Fares and Straube, Jannik and Keiblinger, Michael and Bakouch, Elie and Atkins, Lucas and Panahi, Maziyar and Goddard, Charles and Ryabinin, Max and Hagemann, Johannes}, journal={arXiv preprint}, year={2024} }

---

Use with llama.cpp

Install llama.cpp through brew (works on Mac and Linux)

brew install llama.cpp

Invoke the llama.cpp server or the CLI.

CLI:

llama-cli --hf-repo Triangle104/INTELLECT-1-Instruct-Q4_K_S-GGUF --hf-file intellect-1-instruct-q4_k_s.gguf -p "The meaning to life and the universe is"

Server:

llama-server --hf-repo Triangle104/INTELLECT-1-Instruct-Q4_K_S-GGUF --hf-file intellect-1-instruct-q4_k_s.gguf -c 2048

Note: You can also use this checkpoint directly through the usage steps listed in the Llama.cpp repo as well.

Step 1: Clone llama.cpp from GitHub.

git clone https://github.com/ggerganov/llama.cpp

Step 2: Move into the llama.cpp folder and build it with LLAMA_CURL=1 flag along with other hardware-specific flags (for ex: LLAMA_CUDA=1 for Nvidia GPUs on Linux).

cd llama.cpp && LLAMA_CURL=1 make

Step 3: Run inference through the main binary.

./llama-cli --hf-repo Triangle104/INTELLECT-1-Instruct-Q4_K_S-GGUF --hf-file intellect-1-instruct-q4_k_s.gguf -p "The meaning to life and the universe is"

or

./llama-server --hf-repo Triangle104/INTELLECT-1-Instruct-Q4_K_S-GGUF --hf-file intellect-1-instruct-q4_k_s.gguf -c 2048