Optimized Realistic Vision stable diffusion model for use with AI-8850 LLM Acceleration M.2 Module (AX8850) available on HF
-
I have optimized the Realistic Vision stable diffusion model for use with AI-8850 LLM Acceleration M.2 Module (AX8850):
https://huggingface.co/gregm123456/realistic-vision-v6-axera-hw
Realistic Vision is optimized for more general photorealism, and this optimized vesion can be used in place of the model that was chosen for this example:
-
@gregm123456 HI. I don't suppose you know how to ask the driver developers about bugs? I am getting a lot of this:
67% | ██████████████████████ | 21 / 31 [20.57s<30.37s, 1.02 count/s] init 11 axmodel ok,devid(0) remain_cmm(3954 MB)[2026-01-25 14:53:12.889][3991][E][engine][invoke][97]: Decode api(4) response failed.
[E][ init][ 421]: AX_ENGINE_CreateHandle
70% | ███████████████████████ | 22 / 31 [21.09s<29.71s, 1.04 count/s] init 4 axmodel ok,devid(0) remain_cmm(4014 MB)[2026-01-25 14:53:14.038][3991][E][engine][invoke][97]: Decode api(4) response failed.
[E][ init][ 421]: AX_ENGINE_CreateHandle
74% | ████████████████████████ | 23 / 31 [22.57s<30.42s, 1.02 count/s] init 19 axmodel ok,devid(0) remain_cmm(3825 MB)[2026-01-25 14:53:14.843][3991][E][engine][invoke][97]: Decode api(4) response failed.
[E][ init][ 421]: AX_ENGINE_CreateHandle
77% | █████████████████████████ | 24 / 31 [23.04s<29.76s, 1.04 count/s] init 26 axmodel ok,devid(0) remain_cmm(3885 MB)[2026-01-25 14:53:15.992][3991][E][engine][invoke][97]: Decode api(4) response failed.
[E][ init][ 421]: AX_ENGINE_CreateHandle
80% | ██████████████████████████ | 25 / 31 [24.57s<30.47s, 1.02 count/s] init 12 axmodel ok,devid(0) remain_cmm(3696 MB)[2026-01-25 14:53:16.796][3991][E][engine][invoke][97]: Decode api(4) response failed.
[E][ init][ 421]: AX_ENGINE_CreateHandle
83% | ███████████████████████████ | 26 / 31 [25.00s<29.80s, 1.04 count/s] init 5 axmodel ok,devid(0) remain_cmm(3756 MB)[2026-01-25 14:53:17.945][3991][E][engine][invoke][97]: Decode api(4) response failed.
[E][ init][ 421]: AX_ENGINE_CreateHandle
87% | ████████████████████████████ | 27 / 31 [26.57s<30.51s, 1.02 count/s] init 20 axmodel ok,devid(0) remain_cmm(3567 MB)[2026-01-25 14:53:18.751][3991][E][engine][invoke][97]: Decode api(4) response failed.
[E][ init][ 421]: AX_ENGINE_CreateHandle
90% | █████████████████████████████ | 28 / 31 [26.95s<29.84s, 1.04 count/s] init 27 axmodel ok,devid(0) remain_cmm(3627 MB)[E][ init][ 421]: AX_ENGINE_CreateHandle
[2026-01-25 14:53:19.897][3991][E][engine][invoke][97]: Decode api(4) response failed.
[E][ init][ 421]: AX_ENGINE_CreateHandle
[2026-01-25 14:53:20.654][3991][E][engine][invoke][97]: Decode api(4) response failed.
93% | ██████████████████████████████ | 29 / 31 [28.53s<30.49s, 1.02 count/s] init 6 axmodel ok,devid(0) rema 96% | ███████████████████████████████ | 30 / 31 [28.54s<29.49s, 1.05 count/s] init 13 axmodel ok,devid(0) rem100% | ████████████████████████████████ | 31 / 31 [34.14s<34.14s, 0.91 count/s] init post axmodel ok,remain_cmm(3302 MB)Working directory: /home/stevef/dev/Qwen3-1.7B
Starting tokenizer server on port 12300...
Working directory: /home/stevef/dev/Qwen3-1.7B
Starting tokenizer server on port 12300...
Server running on port 12300
Starting main API application...
[I][ Init][ 130]: LLM init start
[I][ Init][ 34]: connect http://127.0.0.1:12300 ok
[I][ Init][ 57]: uid: cacc4142-cf79-4f83-b257-4a4cd94707bb
bos_id: 151643, eos_id: 151645
3% | ██ | 1 / 31 [0.88s<27.19s, 1.14 count/s] tokenizer init ok[I][ Init][ 45]: LLaMaEmbedSelector use mmap
6% | ███ | 2 / 31 [0.88s<13.59s, 2.28 count/s] embed_selector init ok
[I][ run][ 30]: AXCLWorker start with devid 0
[2026-01-25 14:53:47.641][4245][E][engine][invoke][97]: Decode api(4) response failed.
[E][ init][ 421]: AX_ENGINE_CreateHandle
[2026-01-25 14:53:48.398][4245][E][engine][invoke][97]: Decode api(4) response failed.
[E][ init][ 421]: AX_ENGINE_CreateHandle
9% | ████ | 3 / 31 [3.57s<36.87s, 0.84 count/s] init 0 axmodel ok,devid(0) remai 12% | █████ | 4 / 31 [3.87s<29.96s, 1.03 count/s] init 7 axmodel ok,devid(0) remain_cmm(5117 MB)[2026-01-25 14:53:49.159][4245][E][engine][invoke][97]: Decode api(4) response failed.
[E][ init][ 421]: AX_ENGINE_CreateHandle
16% | ██████ | 5 / 31 [4.87s<30.19s, 1.03 count/s] init 21 axmodel ok,devid(0) remain_cmm(5117 MB)[2026-01-25 14:53:49.963][4245][E][engine][invoke][97]: Decode api(4) response failed.
[E][ init][ 421]: AX_ENGINE_CreateHandle
19% | ███████ | 6 / 31 [5.46s<28.20s, 1.10 count/s] init 14 axmodel ok,devid(0) remain_cmm(4983 MB)[2026-01-25 14:53:51.308][4245][E][engine][invoke][97]: Decode api(4) response failed.
[E][ init][ 421]: AX_ENGINE_CreateHandle
22% | ████████ | 7 / 31 [6.87s<30.43s, 1.02 count/s] init 1 axmodel ok,devid(0) remain_cmm(4859 MB)[2026-01-25 14:53:52.113][4245][E][engine][invoke][97]: Decode api(4) response failed.
[E][ init][ 421]: AX_ENGINE_CreateHandle
25% | █████████ | 8 / 31 [7.87s<30.50s, 1.02 count/s] init 8 axmodel ok,devid(0) remain_cmm(4859 MB)[E][ init][ 421]: AX_ENGINE_CreateHandle
[2026-01-25 14:53:52.917][4245][E][engine][invoke][97]: Decode api(4) response failed.
[E][ init][ 421]: AX_ENGINE_CreateHandle
[2026-01-25 14:53:53.870][4245][E][engine][invoke][97]: Decode api(4) response failed.
29% | ██████████ | 9 / 31 [9.04s<31.14s, 1.00 count/s] init 15 axmodel ok,devid(0) rema 32% | ███████████ | 10 / 31 [9.06s<28.08s, 1.10 count/s] init 22 axmodel ok,devid(0) remain_cmm(4789 MB)[E][ init][ 421]: AX_ENGINE_CreateHandle
[2026-01-25 14:53:55.025][4245][E][engine][invoke][97]: Decode api(4) response failed.
35% | ████████████ | 11 / 31 [10.87s<30.64s, 1.01 count/s] init 2 axmodel ok,devid(0) remain_cmm(4600 MB)[2026-01-25 14:53:55.829][4245][E][engine][invoke][97]: Decode api(4) response failed.
[E][ init][ 421]: AX_ENGINE_CreateHandle
38% | █████████████ | 12 / 31 [11.32s<29.25s, 1.06 count/s] init 9 axmodel ok,devid(0) remain_cmm(4660 MB)[2026-01-25 14:53:56.977][4245][E][engine][invoke][97]: Decode api(4) response failed.
[E][ init][ 421]: AX_ENGINE_CrInfinite loop only solved by a reboot
-
@seahope Interesting post! Leveraging HF models like Realistic Vision with the AX8850 module sounds like a solid step toward faster, more efficient on-device AI workflows.
Hello! It looks like you're interested in this conversation, but you don't have an account yet.
Getting fed up of having to scroll through the same posts each visit? When you register for an account, you'll always come back to exactly where you were before, and choose to be notified of new replies (either via email, or push notification). You'll also be able to save bookmarks and upvote posts to show your appreciation to other community members.
With your input, this post could be even better 💗
Register Login