loading a model failes with validation error undefined symbol: llama_model_dev_layer

[sidietz]

Prerequisites

Please answer the following questions for yourself before submitting an issue.

• I am running the latest code. Development is very rapid so there are no tagged versions as of now.
• I carefully followed the README.md.
• I searched using keywords relevant to my issue to make sure that I am creating a new issue that is not already open (or closed).
• I reviewed the Discussions, and have a new bug or useful enhancement to share.

Expected Behavior

When trying to load a model (here: gpt-oss:20b), it should load.

Current Behavior

A validation error occurs.

ValidationError: 1 validation error for LlamaCppEmbeddings
  Value error, Could not load Llama model from path: [/home/simon/.cache/llama.cpp/unsloth_gpt-oss-20b-GGUF_gpt-oss-20b-Q4_K_M.gguf.](http://localhost:8888/home/simon/.cache/llama.cpp/unsloth_gpt-oss-20b-GGUF_gpt-oss-20b-Q4_K_M.gguf.) Received error [/home/simon/code/current/rag/wheel/llama-cpp-python/llama_cpp/lib/libllama.so](http://localhost:8888/lab/tree/wheel/llama-cpp-python/llama_cpp/lib/libllama.so): undefined symbol: llama_model_dev_layer [type=value_error, input_value={'model_path': '[/home/....gguf](http://localhost:8888/...)'}, input_type=dict]
    For further information visit https://errors.pydantic.dev/2.11/v/value_error

Environment and Context

python -V
Python 3.11.13
lscpu

Architecture:                x86_64
  CPU op-mode(s):            32-bit, 64-bit
  Address sizes:             48 bits physical, 48 bits virtual
  Byte Order:                Little Endian
CPU(s):                      16
  On-line CPU(s) list:       0-15
Vendor ID:                   AuthenticAMD
  Model name:                AMD Ryzen AI 7 PRO 360 w/ Radeon 880M
    CPU family:              26
    Model:                   36
    Thread(s) per core:      2
    Core(s) per socket:      8
    Socket(s):               1
    Stepping:                0
    Frequency boost:         enabled
    CPU(s) scaling MHz:      50%
    CPU max MHz:             5090.9102
    CPU min MHz:             623.3770
    BogoMIPS:                3992.66
    Flags:                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pa
                             t pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt
                             pdpe1gb rdtscp lm constant_tsc rep_good amd_lbr_v2 nopl xtopolog
                             y nonstop_tsc cpuid extd_apicid aperfmperf rapl pni pclmulqdq mo
                             nitor ssse3 fma cx16 sse4_1 sse4_2 movbe popcnt aes xsave avx f1
                             6c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a mi
                             salignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_
                             core perfctr_nb bpext perfctr_llc mwaitx cpb cat_l3 cdp_l3 hw_ps
                             tate ssbd mba perfmon_v2 ibrs ibpb stibp ibrs_enhanced vmmcall f
                             sgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a av
                             x512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512
                             cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_l
                             lc cqm_occup_llc cqm_mbm_total cqm_mbm_local user_shstk avx_vnni
                              avx512_bf16 clzero irperf xsaveerptr rdpru wbnoinvd cppc arat n
                             pt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid deco
                             deassists pausefilter pfthreshold avic v_vmsave_vmload vgif x2av
                             ic v_spec_ctrl vnmi avx512vbmi umip pku ospke avx512_vbmi2 gfni
                             vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid
                              bus_lock_detect movdiri movdir64b overflow_recov succor smca fs
                             rm avx512_vp2intersect flush_l1d amd_lbr_pmc_freeze
Virtualization features:
  Virtualization:            AMD-V
Caches (sum of all):
  L1d:                       384 KiB (8 instances)
  L1i:                       256 KiB (8 instances)
  L2:                        8 MiB (8 instances)
  L3:                        16 MiB (2 instances)
NUMA:
  NUMA node(s):              1
  NUMA node0 CPU(s):         0-15
Vulnerabilities:
  Gather data sampling:      Not affected
  Ghostwrite:                Not affected
  Indirect target selection: Not affected
  Itlb multihit:             Not affected
  L1tf:                      Not affected
  Mds:                       Not affected
  Meltdown:                  Not affected
  Mmio stale data:           Not affected
  Old microcode:             Not affected
  Reg file data sampling:    Not affected
  Retbleed:                  Not affected
  Spec rstack overflow:      Mitigation; IBPB on VMEXIT only
  Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
  Spectre v1:                Mitigation; usercopy/swapgs barriers and __user pointer sanitiza
                             tion
  Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; STIBP a
                             lways-on; PBRSB-eIBRS Not affected; BHI Not affected
  Srbds:                     Not affected
  Tsa:                       Not affected
  Tsx async abort:           Not affected
  Vmscape:                   Mitigation; IBPB on VMEXIT

uname -a
Linux chimchar 6.16.8-arch2-1 #1 SMP PREEMPT_DYNAMIC Sun, 21 Sep 2025 23:19:08 +0000 x86_64 GNU/Linux

Linux distro: arch linux
Llama.cpp version: b6558

git log | head -n 3

commit 7a9bb55037aa234af1edde915791feb77b43fad6
Author: okaris <omerkarisman@gmail.com>
Date:   Tue Sep 23 12:26:35 2025 +0000

Failure Information (for bugs)

I am using a conda environment and the vulkan backend.

Steps to Reproduce

1. Install Llama.cpp version b6558
2. download the most recent llama-cpp-python source code
3. clone Llama.cpp into the vendor folder
4. run make build.vulkan
5. try to load a model (gpt-oss:20b in my case) in python

Failure Logs

llama = LlamaCppEmbeddings(model_path="/path/unsloth_gpt-oss-20b-GGUF_gpt-oss-20b-Q4_K_M.gguf")

ValidationError: 1 validation error for LlamaCppEmbeddings
  Value error, Could not load Llama model from path: [/home/simon/.cache/llama.cpp/unsloth_gpt-oss-20b-GGUF_gpt-oss-20b-Q4_K_M.gguf.](http://localhost:8888/home/simon/.cache/llama.cpp/unsloth_gpt-oss-20b-GGUF_gpt-oss-20b-Q4_K_M.gguf.) Received error [/home/simon/code/current/rag/wheel/llama-cpp-python/llama_cpp/lib/libllama.so](http://localhost:8888/lab/tree/wheel/llama-cpp-python/llama_cpp/lib/libllama.so): undefined symbol: llama_model_dev_layer [type=value_error, input_value={'model_path': '[/home/....gguf](http://localhost:8888/...)'}, input_type=dict]
    For further information visit https://errors.pydantic.dev/2.11/v/value_error

[sidietz]

When using the forked llama.cpp the error no longer occurs.