cnmoro/TinyLlama-1.1B-intermediate-1.5T-PTBR-Instruct-v3-8k-GGUF

Quantized GGUF model files for TinyLlama-1.1B-intermediate-1.5T-PTBR-Instruct-v3-8k from cnmoro

Name	Quant method	Size
tinyllama-1.1b-intermediate-1.5t-ptbr-instruct-v3-8k.q2_k.gguf	q2_k	482.14 MB
tinyllama-1.1b-intermediate-1.5t-ptbr-instruct-v3-8k.q3_k_m.gguf	q3_k_m	549.85 MB
tinyllama-1.1b-intermediate-1.5t-ptbr-instruct-v3-8k.q4_k_m.gguf	q4_k_m	667.81 MB
tinyllama-1.1b-intermediate-1.5t-ptbr-instruct-v3-8k.q5_k_m.gguf	q5_k_m	782.04 MB
tinyllama-1.1b-intermediate-1.5t-ptbr-instruct-v3-8k.q6_k.gguf	q6_k	903.41 MB
tinyllama-1.1b-intermediate-1.5t-ptbr-instruct-v3-8k.q8_0.gguf	q8_0	1.17 GB

Original Model Card:

Finetuned version of PY007/TinyLlama-1.1B-intermediate-step-715k-1.5T, on a Portuguese instruct dataset, using axolotl.

v0, v1 and v2 were finetuned for the default 2048 context length. For this v3, I have used the existing v2 and finetuned the model on a 8k context length dataset. It works fairly well, but it's reasoning capabilities are not so strong. It works well for basic RAG / question answering on retrieved content.

Prompt format:

f"Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.\n\n### Instruction:\n{instruction}\n\n### Response:\n"