1 |
пн, 10 янв. 2022 г. в 18:16, Mark Knecht <markknecht@×××××.com>: |
2 |
> |
3 |
> <SNIP> |
4 |
> > |
5 |
> > Thank you for your reply, Mark. |
6 |
> > |
7 |
> > Unfortunately, you missed my previous message in this thread |
8 |
> > where I wrote that I do have Ubuntu 20.04 on the same computer. |
9 |
> > However, tensorflow fails to run on it because it is not compiled |
10 |
> > to be inconsistent with my videocard. So, Gentoo is my only option |
11 |
> > for this hardware. |
12 |
> > |
13 |
> <SNIP> |
14 |
> |
15 |
> My apologies. This thread has gone on for a while and I had to review |
16 |
> to get caught up. |
17 |
> |
18 |
> OK, so assuming I understand correctly (please correct me if I do |
19 |
> not) then you are talking about ONE computer that uses an AMD |
20 |
> Phenom II X4 processor. This computer dual boots, or is a Gentoo |
21 |
> machine with an Ubuntu VM. |
22 |
|
23 |
It is a dual boot. (Sorry for the late reply.) |
24 |
|
25 |
> In an earlier response to this thread you showed the flags |
26 |
> supported by this processor which did not include |
27 |
> the AVX, AVX2 or the FMA3/FMA4 flags. It is my |
28 |
> understanding that this processor cannot run the |
29 |
> current versions of tensorflow whether you compile |
30 |
> it yourself or not, at least in the non-GPU version. |
31 |
|
32 |
My hardware cannot run tensorflow precompiled by Ubuntu. |
33 |
However, some versions of tensorflow successfully compile |
34 |
and run in Gentoo: |
35 |
- all versions of tensorflow-2.5 fail to compile in Gentoo; |
36 |
- tensorflow-2.7.0 successfully compiled on 21-11-2021; |
37 |
- on 30-12-2021 recompilation of the same tensorflow-2.7.0 |
38 |
failed because of some changed dependencies; |
39 |
- I have not yet tried to compile tensorflow-2.8.0. |
40 |
|
41 |
Actually, I gave up trying to compile it and deleted it from my system. |
42 |
Moreover, I am not going to compile it in the near future as I am now |
43 |
in the war zone. |
44 |
|
45 |
> WRT to your video card, tensorflow does not require the use |
46 |
> of a GPU. There are two versions, tensorflow-cpu and |
47 |
> tensorflow-gpu. If you were to build the -cpu version then it |
48 |
> is my understanding it would run an a headless machine, |
49 |
> presuming the processor has AVX/AVX2/FMA hardware |
50 |
> support. |
51 |
> |
52 |
> If the processor DOES have AVX/FMA support but you were having |
53 |
> problems emerging TF in Gentoo then a virtual machine running |
54 |
> Ubuntu might have helped you as you could use a precompiled |
55 |
> apt or snap package. However I don't think anything gets you |
56 |
> past not having AVX/FMA hardware support. |
57 |
> |
58 |
> I am in the same situation. My big machine is an Intel i7 980 |
59 |
> Extreme. I used to be able to run TF but have not been able |
60 |
> to since Google raised the CPU requirements. |
61 |
> |
62 |
> If I am not understanding your hardware setup, or you think |
63 |
> there is a path around the AVX/FMA hardware problem |
64 |
> please let me know and I'll explore it more deeply with you. |
65 |
|
66 |
In addition to the old CPU, I have quite an old video card, |
67 |
namely, ATI R4770. However, I still believe that it is possible |
68 |
to compile tensorflow so that it could run on my hardware. |
69 |
At least, I did it for tensorflow-2.7.0 on 21-11-2021. |