Hugging Face's logo
mobicham's picture
33 4 25

mobicham

mobicham
ashwinnair14's profile picture victor's profile picture 21world's profile picture
·
https://www.mobiuslabs.com/

AI & ML interests

Model pruning, quantization, computer vision, LLMs

Organizations

Mobius Labs GmbH's profile picture Distillation Hugs's profile picture

mobicham's activity

New activity in mobiuslabsgmbh/Llama-3.1-70b-instruct_4bitgs64_hqq 2 months ago

CPU support

1
#2 opened 2 months ago by Essa20001
New activity in mobiuslabsgmbh/Llama-3.1-70b-instruct_4bitgs64_hqq 3 months ago

Decensored version?

5
#1 opened 3 months ago by KnutJaegersberg
New activity in mobiuslabsgmbh/Llama-3.1-8b-instruct_4bitgs64_hqq_calib 3 months ago

Oobabooga?

1
#1 opened 3 months ago by AIGUYCONTENT
New activity in mobiuslabsgmbh/Llama-3-8b-instruct_2bitgs64_hqq 3 months ago

the coder from the model card has errors when executing on google colab

1
#1 opened 3 months ago by vasilee
New activity in mobiuslabsgmbh/Llama-2-7b-chat-hf_1bitgs8_hqq 4 months ago

QUANTIZED VERSION GGUF

1
#5 opened 4 months ago by ar08
New activity in open-llm-leaderboard/open_llm_leaderboard 6 months ago

GSM8K (5-shot) performance is quite different compared to running lm_eval locally

5
#755 opened 6 months ago by mobicham
New activity in mobiuslabsgmbh/Llama-2-7b-chat-hf_1bitgs8_hqq 6 months ago

Details about this model

1
#4 opened 6 months ago by at676

Make it usageable for cpu

1
#3 opened 6 months ago by ar08
New activity in mobiuslabsgmbh/Llama-2-7b-chat-hf_1bitgs8_hqq 7 months ago

Error with adapter ?

2
#2 opened 7 months ago by nelkh

Any plan for making HQQ+ 2bit quant for Mixtral or larger models?

1
#1 opened 7 months ago by raincandy-u
New activity in mobiuslabsgmbh/Mixtral-8x7B-Instruct-v0.1-hf-attn-4bit-moe-2bit-metaoffload-HQQ 7 months ago

Runs out of memory on free tier Google Colab

3
#3 opened 7 months ago by sudhir2016
New activity in mobiuslabsgmbh/Mixtral-8x7B-Instruct-v0.1-hf-attn-4bit-moe-2bitgs8-metaoffload-HQQ 7 months ago

Either README is wrong, or the wrong model file is uploaded?

1
#1 opened 8 months ago by andysalerno
New activity in liuhaotian/llava-v1.6-34b 8 months ago

Quantizations?

2
#1 opened 9 months ago by musicurgy
New activity in mobiuslabsgmbh/aanaphi2-v0.1 8 months ago

Which Dataset ?

1
#4 opened 8 months ago by xxxTEMPESTxxx

How do I run this on cpu?

5
#3 opened 8 months ago by ARMcPro
New activity in mobiuslabsgmbh/Mixtral-8x7B-Instruct-v0.1-hf-attn-4bit-moe-2bit-metaoffload-HQQ 8 months ago

gguf format

1
#2 opened 8 months ago by GyroO
New activity in mobiuslabsgmbh/aanaphi2-v0.1 9 months ago

Stop overgenerating. Need EOS token?

11
#1 opened 9 months ago by vicplus
New activity in mobiuslabsgmbh/Mixtral-8x7B-Instruct-v0.1-hf-attn-4bit-moe-2bit-metaoffload-HQQ 9 months ago

OSError: mobiuslabsgmbh/Mixtral-8x7B-Instruct-v0.1-hf-attn-4bit-moe-2bit-metaoffload-HQQ does not appear to have a file named pytorch_model.bin, tf_model.h5, model.ckpt or flax_model.msgpack.

1
#1 opened 9 months ago by Tejasram
New activity in open-llm-leaderboard/open_llm_leaderboard 9 months ago

gsm8k score largely different from local run

6
#591 opened 9 months ago by mobicham
New activity in google/siglip-so400m-patch14-384 9 months ago

Output features are different compared to timm

2
#2 opened 9 months ago by mobicham