github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-06 16:11:34 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #7433] ollama ROCm multiple gpus, segfaulted when trying to run model bigger than 1 GPU's memory capacity #66782

New Issue
Closed
opened 2026-05-04 08:10:48 -05:00 by GiteaMirror · 1 comment
GiteaMirror commented 2026-05-04 08:10:48 -05:00
Owner
Copy Link

Originally created by @rhudock on GitHub (Oct 30, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/7433

Originally assigned to: @dhiltgen on GitHub.

What is the issue?

Segmentation fault when trying to run a model

Command

ollama run llama3.1:70b

Error

Error: llama runner process has terminated: signal: segmentation fault (core dumped)

Dmesg

[ 2857.607412] ollama_llama_se[18031]: segfault at 18 ip 00007e8f7e127b66 sp 00007ffd70563640 error 4 in libamdhip64.so.6.1.60102[7e8f7de21000+371000] likely on CPU 13 (core 17, socket 0)
[ 2857.607438] Code: 0d 4c 89 ef e8 3b 80 00 00 e9 bf fd ff ff 83 87 98 01 00 00 01 e9 b3 fd ff ff 48 89 c5 e9 e6 4d d3 ff 66 90 53 48 89 fb 89 d7 <4c> 8b 43 18 4d 85 c0 74 41 4c 8b 4b 20 31 c9 31 c0 eb 12 0f 1f 80
robert@robert-mercury:~$

Logs from Ollama

Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: loaded meta data with 29 key-value pairs and 724 tensors from /usr/share/ollama/.ollama/models/blobs/sha256-a677b4a4b70c45e702b1d600f7905e367733c53898b8be60e3f29272cf334574 (version GGUF V3 (latest))
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 0: general.architecture str = llama
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 1: general.type str = model
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 2: general.name str = Meta Llama 3.1 70B Instruct
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 3: general.finetune str = Instruct
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 4: general.basename str = Meta-Llama-3.1
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 5: general.size_label str = 70B
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 6: general.license str = llama3.1
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 7: general.tags arr[str,6] = ["facebook", "meta", "pytorch", "llam...
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 8: general.languages arr[str,8] = ["en", "de", "fr", "it", "pt", "hi", ...
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 9: llama.block_count u32 = 80
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 10: llama.context_length u32 = 131072
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 11: llama.embedding_length u32 = 8192
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 12: llama.feed_forward_length u32 = 28672
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 13: llama.attention.head_count u32 = 64
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 14: llama.attention.head_count_kv u32 = 8
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 15: llama.rope.freq_base f32 = 500000.000000
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 16: llama.attention.layer_norm_rms_epsilon f32 = 0.000010
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 17: general.file_type u32 = 2
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 18: llama.vocab_size u32 = 128256
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 19: llama.rope.dimension_count u32 = 128
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 20: tokenizer.ggml.model str = gpt2
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 21: tokenizer.ggml.pre str = llama-bpe
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 22: tokenizer.ggml.tokens arr[str,128256] = ["!", """, "#", "$", "%", "&", "'", ...
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 23: tokenizer.ggml.token_type arr[i32,128256] = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 24: tokenizer.ggml.merges arr[str,280147] = ["Ġ Ġ", "Ġ ĠĠĠ", "ĠĠ ĠĠ", "...
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 25: tokenizer.ggml.bos_token_id u32 = 128000
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 26: tokenizer.ggml.eos_token_id u32 = 128009
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 27: tokenizer.chat_template str = {{- bos_token }}\n{%- if custom_tools ...
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 28: general.quantization_version u32 = 2
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - type f32: 162 tensors
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - type q4_0: 561 tensors
Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - type q6_K: 1 tensors
Oct 30 19:29:31 robert-mercury ollama[15528]: time=2024-10-30T19:29:31.057-04:00 level=INFO source=server.go:621 msg="waiting for server to become available" status="llm server loading model"
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_vocab: special tokens cache size = 256
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_vocab: token to piece cache size = 0.7999 MB
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: format = GGUF V3 (latest)
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: arch = llama
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: vocab type = BPE
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_vocab = 128256
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_merges = 280147
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: vocab_only = 0
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_ctx_train = 131072
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_embd = 8192
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_layer = 80
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_head = 64
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_head_kv = 8
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_rot = 128
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_swa = 0
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_embd_head_k = 128
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_embd_head_v = 128
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_gqa = 8
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_embd_k_gqa = 1024
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_embd_v_gqa = 1024
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: f_norm_eps = 0.0e+00
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: f_norm_rms_eps = 1.0e-05
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: f_clamp_kqv = 0.0e+00
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: f_max_alibi_bias = 0.0e+00
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: f_logit_scale = 0.0e+00
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_ff = 28672
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_expert = 0
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_expert_used = 0
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: causal attn = 1
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: pooling type = 0
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: rope type = 0
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: rope scaling = linear
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: freq_base_train = 500000.0
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: freq_scale_train = 1
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_ctx_orig_yarn = 131072
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: rope_finetuned = unknown
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: ssm_d_conv = 0
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: ssm_d_inner = 0
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: ssm_d_state = 0
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: ssm_dt_rank = 0
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: ssm_dt_b_c_rms = 0
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: model type = 70B
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: model ftype = Q4_0
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: model params = 70.55 B
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: model size = 37.22 GiB (4.53 BPW)
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: general.name = Meta Llama 3.1 70B Instruct
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: BOS token = 128000 '<|begin_of_text|>'
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: EOS token = 128009 '<|eot_id|>'
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: LF token = 128 'Ä'
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: EOT token = 128009 '<|eot_id|>'
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: EOM token = 128008 '<|eom_id|>'
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: EOG token = 128008 '<|eom_id|>'
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: EOG token = 128009 '<|eot_id|>'
Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: max token length = 256
Oct 30 19:29:35 robert-mercury ollama[15528]: ggml_cuda_init: GGML_CUDA_FORCE_MMQ: no
Oct 30 19:29:35 robert-mercury ollama[15528]: ggml_cuda_init: GGML_CUDA_FORCE_CUBLAS: no
Oct 30 19:29:35 robert-mercury ollama[15528]: ggml_cuda_init: found 6 ROCm devices:
Oct 30 19:29:35 robert-mercury ollama[15528]: Device 0: Radeon RX 7900 XTX, compute capability 11.0, VMM: no
Oct 30 19:29:35 robert-mercury ollama[15528]: Device 1: Radeon RX 7900 XTX, compute capability 11.0, VMM: no
Oct 30 19:29:35 robert-mercury ollama[15528]: Device 2: Radeon RX 7900 XTX, compute capability 11.0, VMM: no
Oct 30 19:29:35 robert-mercury ollama[15528]: Device 3: Radeon RX 7900 XTX, compute capability 11.0, VMM: no
Oct 30 19:29:35 robert-mercury ollama[15528]: Device 4: Radeon RX 7900 XTX, compute capability 11.0, VMM: no
Oct 30 19:29:35 robert-mercury ollama[15528]: Device 5: Radeon RX 7900 XTX, compute capability 11.0, VMM: no
Oct 30 19:29:35 robert-mercury ollama[15528]: llm_load_tensors: ggml ctx size = 2.37 MiB
Oct 30 19:29:37 robert-mercury ollama[15528]: time=2024-10-30T19:29:37.525-04:00 level=INFO source=server.go:621 msg="waiting for server to become available" status="llm server not responding"
Oct 30 19:29:37 robert-mercury ollama[15528]: time=2024-10-30T19:29:37.911-04:00 level=INFO source=server.go:621 msg="waiting for server to become available" status="llm server loading model"
Oct 30 19:29:37 robert-mercury ollama[15528]: llm_load_tensors: offloading 80 repeating layers to GPU
Oct 30 19:29:37 robert-mercury ollama[15528]: llm_load_tensors: offloading non-repeating layers to GPU
Oct 30 19:29:37 robert-mercury ollama[15528]: llm_load_tensors: offloaded 81/81 layers to GPU
Oct 30 19:29:37 robert-mercury ollama[15528]: llm_load_tensors: ROCm0 buffer size = 6426.88 MiB
Oct 30 19:29:37 robert-mercury ollama[15528]: llm_load_tensors: ROCm1 buffer size = 6426.88 MiB
Oct 30 19:29:37 robert-mercury ollama[15528]: llm_load_tensors: ROCm2 buffer size = 6426.88 MiB
Oct 30 19:29:37 robert-mercury ollama[15528]: llm_load_tensors: ROCm3 buffer size = 5967.82 MiB
Oct 30 19:29:37 robert-mercury ollama[15528]: llm_load_tensors: ROCm4 buffer size = 5967.82 MiB
Oct 30 19:29:37 robert-mercury ollama[15528]: llm_load_tensors: ROCm5 buffer size = 6330.74 MiB
Oct 30 19:29:37 robert-mercury ollama[15528]: llm_load_tensors: CPU buffer size = 563.62 MiB
Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: n_ctx = 8192
Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: n_batch = 512
Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: n_ubatch = 512
Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: flash_attn = 0
Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: freq_base = 500000.0
Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: freq_scale = 1
Oct 30 19:29:51 robert-mercury ollama[15528]: llama_kv_cache_init: ROCm0 KV buffer size = 448.00 MiB
Oct 30 19:29:51 robert-mercury ollama[15528]: llama_kv_cache_init: ROCm1 KV buffer size = 448.00 MiB
Oct 30 19:29:51 robert-mercury ollama[15528]: llama_kv_cache_init: ROCm2 KV buffer size = 448.00 MiB
Oct 30 19:29:51 robert-mercury ollama[15528]: llama_kv_cache_init: ROCm3 KV buffer size = 416.00 MiB
Oct 30 19:29:51 robert-mercury ollama[15528]: llama_kv_cache_init: ROCm4 KV buffer size = 416.00 MiB
Oct 30 19:29:51 robert-mercury ollama[15528]: llama_kv_cache_init: ROCm5 KV buffer size = 384.00 MiB
Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: KV self size = 2560.00 MiB, K (f16): 1280.00 MiB, V (f16): 1280.00 MiB
Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: ROCm_Host output buffer size = 2.08 MiB
Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: pipeline parallelism enabled (n_copies=4)
Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: ROCm0 compute buffer size = 1216.01 MiB
Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: ROCm1 compute buffer size = 1216.01 MiB
Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: ROCm2 compute buffer size = 1216.01 MiB
Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: ROCm3 compute buffer size = 1216.01 MiB
Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: ROCm4 compute buffer size = 1216.01 MiB
Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: ROCm5 compute buffer size = 1216.02 MiB
Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: ROCm_Host compute buffer size = 80.02 MiB
Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: graph nodes = 2566
Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: graph splits = 7
Oct 30 19:29:51 robert-mercury ollama[15528]: time=2024-10-30T19:29:51.816-04:00 level=INFO source=server.go:621 msg="waiting for server to become available" status="llm server error"
Oct 30 19:29:52 robert-mercury ollama[15528]: time=2024-10-30T19:29:52.067-04:00 level=ERROR source=sched.go:455 msg="error loading llama server" error="llama runner process has terminated: signal: segmentation fault (core dumped)"
Oct 30 19:29:52 robert-mercury ollama[15528]: [GIN] 2024/10/30 - 19:29:52 | 500 | 21.309799963s | 127.0.0.1 | POST "/api/generate"
Oct 30 19:29:57 robert-mercury ollama[15528]: time=2024-10-30T19:29:57.068-04:00 level=WARN source=sched.go:646 msg="gpu VRAM usage didn't recover within timeout" seconds=5.000938426 model=/usr/share/ollama/.ollama/models/blobs/sha256-a677b4a4b70c45e702b1d600f7905e367733c53898b8be60e3f29272cf334574
Oct 30 19:29:57 robert-mercury ollama[15528]: time=2024-10-30T19:29:57.317-04:00 level=WARN source=sched.go:646 msg="gpu VRAM usage didn't recover within timeout" seconds=5.250407502 model=/usr/share/ollama/.ollama/models/blobs/sha256-a677b4a4b70c45e702b1d600f7905e367733c53898b8be60e3f29272cf334574
Oct 30 19:29:57 robert-mercury ollama[15528]: time=2024-10-30T19:29:57.567-04:00 level=WARN source=sched.go:646 msg="gpu VRAM usage didn't recover within timeout" seconds=5.500792457 model=/usr/share/ollama/.ollama/models/blobs/sha256-a677b4a4b70c45e702b1d600f7905e367733c53898b8be60e3f29272cf334574

OS

Linux

GPU

AMD

CPU

AMD

Ollama version

0.3.14

Originally created by @rhudock on GitHub (Oct 30, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/7433 Originally assigned to: @dhiltgen on GitHub. ### What is the issue? Segmentation fault when trying to run a model ### Command ollama run llama3.1:70b ### Error Error: llama runner process has terminated: signal: segmentation fault (core dumped) ### Dmesg [ 2857.607412] ollama_llama_se[18031]: segfault at 18 ip 00007e8f7e127b66 sp 00007ffd70563640 error 4 in libamdhip64.so.6.1.60102[7e8f7de21000+371000] likely on CPU 13 (core 17, socket 0) [ 2857.607438] Code: 0d 4c 89 ef e8 3b 80 00 00 e9 bf fd ff ff 83 87 98 01 00 00 01 e9 b3 fd ff ff 48 89 c5 e9 e6 4d d3 ff 66 90 53 48 89 fb 89 d7 <4c> 8b 43 18 4d 85 c0 74 41 4c 8b 4b 20 31 c9 31 c0 eb 12 0f 1f 80 robert@robert-mercury:~$ ### Logs from Ollama Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: loaded meta data with 29 key-value pairs and 724 tensors from /usr/share/ollama/.ollama/models/blobs/sha256-a677b4a4b70c45e702b1d600f7905e367733c53898b8be60e3f29272cf334574 (version GGUF V3 (latest)) Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output. Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 0: general.architecture str = llama Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 1: general.type str = model Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 2: general.name str = Meta Llama 3.1 70B Instruct Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 3: general.finetune str = Instruct Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 4: general.basename str = Meta-Llama-3.1 Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 5: general.size_label str = 70B Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 6: general.license str = llama3.1 Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 7: general.tags arr[str,6] = ["facebook", "meta", "pytorch", "llam... Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 8: general.languages arr[str,8] = ["en", "de", "fr", "it", "pt", "hi", ... Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 9: llama.block_count u32 = 80 Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 10: llama.context_length u32 = 131072 Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 11: llama.embedding_length u32 = 8192 Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 12: llama.feed_forward_length u32 = 28672 Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 13: llama.attention.head_count u32 = 64 Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 14: llama.attention.head_count_kv u32 = 8 Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 15: llama.rope.freq_base f32 = 500000.000000 Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 16: llama.attention.layer_norm_rms_epsilon f32 = 0.000010 Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 17: general.file_type u32 = 2 Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 18: llama.vocab_size u32 = 128256 Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 19: llama.rope.dimension_count u32 = 128 Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 20: tokenizer.ggml.model str = gpt2 Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 21: tokenizer.ggml.pre str = llama-bpe Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 22: tokenizer.ggml.tokens arr[str,128256] = ["!", "\"", "#", "$", "%", "&", "'", ... Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 23: tokenizer.ggml.token_type arr[i32,128256] = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ... Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 24: tokenizer.ggml.merges arr[str,280147] = ["Ġ Ġ", "Ġ ĠĠĠ", "ĠĠ ĠĠ", "... Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 25: tokenizer.ggml.bos_token_id u32 = 128000 Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 26: tokenizer.ggml.eos_token_id u32 = 128009 Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 27: tokenizer.chat_template str = {{- bos_token }}\n{%- if custom_tools ... Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - kv 28: general.quantization_version u32 = 2 Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - type f32: 162 tensors Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - type q4_0: 561 tensors Oct 30 19:29:30 robert-mercury ollama[15528]: llama_model_loader: - type q6_K: 1 tensors Oct 30 19:29:31 robert-mercury ollama[15528]: time=2024-10-30T19:29:31.057-04:00 level=INFO source=server.go:621 msg="waiting for server to become available" status="llm server loading model" Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_vocab: special tokens cache size = 256 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_vocab: token to piece cache size = 0.7999 MB Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: format = GGUF V3 (latest) Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: arch = llama Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: vocab type = BPE Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_vocab = 128256 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_merges = 280147 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: vocab_only = 0 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_ctx_train = 131072 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_embd = 8192 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_layer = 80 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_head = 64 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_head_kv = 8 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_rot = 128 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_swa = 0 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_embd_head_k = 128 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_embd_head_v = 128 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_gqa = 8 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_embd_k_gqa = 1024 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_embd_v_gqa = 1024 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: f_norm_eps = 0.0e+00 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: f_norm_rms_eps = 1.0e-05 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: f_clamp_kqv = 0.0e+00 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: f_max_alibi_bias = 0.0e+00 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: f_logit_scale = 0.0e+00 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_ff = 28672 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_expert = 0 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_expert_used = 0 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: causal attn = 1 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: pooling type = 0 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: rope type = 0 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: rope scaling = linear Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: freq_base_train = 500000.0 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: freq_scale_train = 1 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: n_ctx_orig_yarn = 131072 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: rope_finetuned = unknown Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: ssm_d_conv = 0 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: ssm_d_inner = 0 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: ssm_d_state = 0 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: ssm_dt_rank = 0 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: ssm_dt_b_c_rms = 0 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: model type = 70B Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: model ftype = Q4_0 Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: model params = 70.55 B Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: model size = 37.22 GiB (4.53 BPW) Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: general.name = Meta Llama 3.1 70B Instruct Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: BOS token = 128000 '<|begin_of_text|>' Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: EOS token = 128009 '<|eot_id|>' Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: LF token = 128 'Ä' Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: EOT token = 128009 '<|eot_id|>' Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: EOM token = 128008 '<|eom_id|>' Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: EOG token = 128008 '<|eom_id|>' Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: EOG token = 128009 '<|eot_id|>' Oct 30 19:29:31 robert-mercury ollama[15528]: llm_load_print_meta: max token length = 256 Oct 30 19:29:35 robert-mercury ollama[15528]: ggml_cuda_init: GGML_CUDA_FORCE_MMQ: no Oct 30 19:29:35 robert-mercury ollama[15528]: ggml_cuda_init: GGML_CUDA_FORCE_CUBLAS: no Oct 30 19:29:35 robert-mercury ollama[15528]: ggml_cuda_init: found 6 ROCm devices: Oct 30 19:29:35 robert-mercury ollama[15528]: Device 0: Radeon RX 7900 XTX, compute capability 11.0, VMM: no Oct 30 19:29:35 robert-mercury ollama[15528]: Device 1: Radeon RX 7900 XTX, compute capability 11.0, VMM: no Oct 30 19:29:35 robert-mercury ollama[15528]: Device 2: Radeon RX 7900 XTX, compute capability 11.0, VMM: no Oct 30 19:29:35 robert-mercury ollama[15528]: Device 3: Radeon RX 7900 XTX, compute capability 11.0, VMM: no Oct 30 19:29:35 robert-mercury ollama[15528]: Device 4: Radeon RX 7900 XTX, compute capability 11.0, VMM: no Oct 30 19:29:35 robert-mercury ollama[15528]: Device 5: Radeon RX 7900 XTX, compute capability 11.0, VMM: no Oct 30 19:29:35 robert-mercury ollama[15528]: llm_load_tensors: ggml ctx size = 2.37 MiB Oct 30 19:29:37 robert-mercury ollama[15528]: time=2024-10-30T19:29:37.525-04:00 level=INFO source=server.go:621 msg="waiting for server to become available" status="llm server not responding" Oct 30 19:29:37 robert-mercury ollama[15528]: time=2024-10-30T19:29:37.911-04:00 level=INFO source=server.go:621 msg="waiting for server to become available" status="llm server loading model" Oct 30 19:29:37 robert-mercury ollama[15528]: llm_load_tensors: offloading 80 repeating layers to GPU Oct 30 19:29:37 robert-mercury ollama[15528]: llm_load_tensors: offloading non-repeating layers to GPU Oct 30 19:29:37 robert-mercury ollama[15528]: llm_load_tensors: offloaded 81/81 layers to GPU Oct 30 19:29:37 robert-mercury ollama[15528]: llm_load_tensors: ROCm0 buffer size = 6426.88 MiB Oct 30 19:29:37 robert-mercury ollama[15528]: llm_load_tensors: ROCm1 buffer size = 6426.88 MiB Oct 30 19:29:37 robert-mercury ollama[15528]: llm_load_tensors: ROCm2 buffer size = 6426.88 MiB Oct 30 19:29:37 robert-mercury ollama[15528]: llm_load_tensors: ROCm3 buffer size = 5967.82 MiB Oct 30 19:29:37 robert-mercury ollama[15528]: llm_load_tensors: ROCm4 buffer size = 5967.82 MiB Oct 30 19:29:37 robert-mercury ollama[15528]: llm_load_tensors: ROCm5 buffer size = 6330.74 MiB Oct 30 19:29:37 robert-mercury ollama[15528]: llm_load_tensors: CPU buffer size = 563.62 MiB Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: n_ctx = 8192 Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: n_batch = 512 Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: n_ubatch = 512 Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: flash_attn = 0 Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: freq_base = 500000.0 Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: freq_scale = 1 Oct 30 19:29:51 robert-mercury ollama[15528]: llama_kv_cache_init: ROCm0 KV buffer size = 448.00 MiB Oct 30 19:29:51 robert-mercury ollama[15528]: llama_kv_cache_init: ROCm1 KV buffer size = 448.00 MiB Oct 30 19:29:51 robert-mercury ollama[15528]: llama_kv_cache_init: ROCm2 KV buffer size = 448.00 MiB Oct 30 19:29:51 robert-mercury ollama[15528]: llama_kv_cache_init: ROCm3 KV buffer size = 416.00 MiB Oct 30 19:29:51 robert-mercury ollama[15528]: llama_kv_cache_init: ROCm4 KV buffer size = 416.00 MiB Oct 30 19:29:51 robert-mercury ollama[15528]: llama_kv_cache_init: ROCm5 KV buffer size = 384.00 MiB Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: KV self size = 2560.00 MiB, K (f16): 1280.00 MiB, V (f16): 1280.00 MiB Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: ROCm_Host output buffer size = 2.08 MiB Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: pipeline parallelism enabled (n_copies=4) Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: ROCm0 compute buffer size = 1216.01 MiB Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: ROCm1 compute buffer size = 1216.01 MiB Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: ROCm2 compute buffer size = 1216.01 MiB Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: ROCm3 compute buffer size = 1216.01 MiB Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: ROCm4 compute buffer size = 1216.01 MiB Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: ROCm5 compute buffer size = 1216.02 MiB Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: ROCm_Host compute buffer size = 80.02 MiB Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: graph nodes = 2566 Oct 30 19:29:51 robert-mercury ollama[15528]: llama_new_context_with_model: graph splits = 7 Oct 30 19:29:51 robert-mercury ollama[15528]: time=2024-10-30T19:29:51.816-04:00 level=INFO source=server.go:621 msg="waiting for server to become available" status="llm server error" Oct 30 19:29:52 robert-mercury ollama[15528]: time=2024-10-30T19:29:52.067-04:00 level=ERROR source=sched.go:455 msg="error loading llama server" error="llama runner process has terminated: signal: segmentation fault (core dumped)" Oct 30 19:29:52 robert-mercury ollama[15528]: [GIN] 2024/10/30 - 19:29:52 | 500 | 21.309799963s | 127.0.0.1 | POST "/api/generate" Oct 30 19:29:57 robert-mercury ollama[15528]: time=2024-10-30T19:29:57.068-04:00 level=WARN source=sched.go:646 msg="gpu VRAM usage didn't recover within timeout" seconds=5.000938426 model=/usr/share/ollama/.ollama/models/blobs/sha256-a677b4a4b70c45e702b1d600f7905e367733c53898b8be60e3f29272cf334574 Oct 30 19:29:57 robert-mercury ollama[15528]: time=2024-10-30T19:29:57.317-04:00 level=WARN source=sched.go:646 msg="gpu VRAM usage didn't recover within timeout" seconds=5.250407502 model=/usr/share/ollama/.ollama/models/blobs/sha256-a677b4a4b70c45e702b1d600f7905e367733c53898b8be60e3f29272cf334574 Oct 30 19:29:57 robert-mercury ollama[15528]: time=2024-10-30T19:29:57.567-04:00 level=WARN source=sched.go:646 msg="gpu VRAM usage didn't recover within timeout" seconds=5.500792457 model=/usr/share/ollama/.ollama/models/blobs/sha256-a677b4a4b70c45e702b1d600f7905e367733c53898b8be60e3f29272cf334574 ### OS Linux ### GPU AMD ### CPU AMD ### Ollama version 0.3.14
GiteaMirror added the amdbugneeds more info labels 2026-05-04 08:10:53 -05:00
GiteaMirror commented 2026-05-04 08:10:59 -05:00
Author
Owner
Copy Link

@dhiltgen commented on GitHub (Feb 25, 2025):

Is this still a problem with the latest versions? I'm trying to determine if #7378 is still useful.

<!-- gh-comment-id:2682725139 --> @dhiltgen commented on GitHub (Feb 25, 2025): Is this still a problem with the latest versions? I'm trying to determine if #7378 is still useful.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label amd bug needs more info
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#66782
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?