Spaces:
Runtime error
Runtime error
File size: 1,599 Bytes
88426c0 7c7d6f2 3d09cec 7c7d6f2 88426c0 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 |
from rwkvstic.agnostic.backends import TORCH, TORCH_QUANT
import torch
quantized = {
"mode": TORCH_QUANT,
"runtimedtype": torch.bfloat16,
"useGPU": torch.cuda.is_available(),
"chunksize": 32, # larger = more accurate, but more memory
"target": 100 # your gpu max size, excess vram offloaded to cpu
}
# UNCOMMENT TO SELECT OPTIONS
# Not full list of options, see https://pypi.org/project/rwkvstic/ and https://huggingface.co/BlinkDL/ for more models/modes
# RWKV 1B5 instruct test 1 model
# Approximate
# [Vram usage: 6.0GB]
# [File size: 3.0GB]
config = {
"path":"14B-8K-rwkvstic-2-1-2.rwkv"
#"useLogFix":False # When enabled, use BlinkDLs version of the att.
}
title = "RWKV-4 (14B 8k)"
# RWKV 1B5 instruct model quantized
# Approximate
# [Vram usage: 1.3GB]
# [File size: 3.0GB]
config = {
"path": "https://huggingface.co/BlinkDL/rwkv-4-pile-1b5/resolve/main/RWKV-4-Pile-1B5-Instruct-test1-20230124.pqth",
**quantized
}
# title = "RWKV-4 (1.5b Instruct Quantized)"
# RWKV 7B instruct pre-quantized (settings baked into model)
# Approximate
# [Vram usage: 7.0GB]
# [File size: 8.0GB]
# config = {
# "path": "https://huggingface.co/Hazzzardous/RWKV-8Bit/resolve/main/RWKV-4-Pile-7B-Instruct.pqth"
# }
# title = "RWKV-4 (7b Instruct Quantized)"
# RWKV 14B quantized (latest as of feb 9)
# Approximate
# [Vram usage: 15.0GB]
# [File size: 15.0GB]
# config = {
# "path": "https://huggingface.co/Hazzzardous/RWKV-8Bit/resolve/main/RWKV-4-Pile-14B-20230204-7324.pqth"
# }
# title = "RWKV-4 (14b 94% trained, not yet instruct tuned, 8-Bit)" |