You can deploy this model and start earning money today!

chargoddard/
llama33b-s2a4-qlora

Name: llama33b-s2a4-qlora
Availability: InStock

peft

text-generation

EleutherAI/wikitext_document_level

huggyllama/llama-30b

adapter

Views

No views yet

LLaMA 33b finetuned on wikitext_document_level with a combination of both linear and NTK-aware ROPE scaling.

Trained with alpha=4, scale=2. Definitely works for sequence lengths up to and including 4096. Might work for much longer, but I don't have the VRAM to test properly. ¯\_(ツ)_/¯

Training procedure

The following bitsandbytes quantization config was used during training:

load_in_8bit: False
load_in_4bit: True
llm_int8_threshold: 6.0
llm_int8_skip_modules: None
llm_int8_enable_fp32_cpu_offload: False
llm_int8_has_fp16_weight: False
bnb_4bit_quant_type: nf4
bnb_4bit_use_double_quant: True
bnb_4bit_compute_dtype: bfloat16

Framework versions

PEFT 0.4.0.dev0

chargoddard/llama33b-s2a4-qlora

Training procedure

Framework versions

chargoddard/
llama33b-s2a4-qlora