search

ollama / ollama

Get up and running with Llama 3.2, Mistral, Gemma 2, and other large language models.
https://ollama.com
MIT License
93.79k stars 7.42k forks source link

llama runner process has terminated: signal: aborted (core dumped) #5904

Open Dudu0831 opened 2 months ago

Dudu0831 commented 2 months ago

What is the issue?

I successfully converted jina-embeddings v2 base zh to gguf through llama. cpp and imported it into llama。 Here is my Modelfile

root@buaa-KVM:~/1T/ollama/Jina-AI-embedding# cat Modelfile FROM /root/ggml-vocab-jina-v2-zh.gguf PARAMETER num_ctx 8192

When I access it using/app/embed, the log will report an error。

time=2024-07-24T15:40:34.577+08:00 level=ERROR source=sched.go:443 msg="error loading llama server" error="llama runner process has terminated: signal: aborted (core dumped)"

Below is my complete log。 7月 24 15:40:33 buaa-KVM ollama[458186]: time=2024-07-24T15:40:33.873+08:00 level=INFO source=sched.go:495 msg="updated VRAM based on existing loaded models" gpu=GPU-4f8ced6a-2dde-5e92-be03-8d21e26bd156 library=cuda total="23.6 GiB" available="16.9 GiB" 7月 24 15:40:33 buaa-KVM ollama[458186]: time=2024-07-24T15:40:33.873+08:00 level=INFO source=sched.go:495 msg="updated VRAM based on existing loaded models" gpu=GPU-36c16c0c-392d-ffc5-13ce-2fd6b9af0668 library=cuda total="23.6 GiB" available="23.3 GiB" 7月 24 15:40:33 buaa-KVM ollama[458186]: time=2024-07-24T15:40:33.874+08:00 level=INFO source=sched.go:701 msg="new model will fit in available VRAM in single GPU, loading" model=/usr/share/ollama/.ollama/models/blobs/sha256-65a4313f43b6f94a0a8693d70efe823792303a020601ab3d4cad54cf079296c6 gpu=GPU-36c16c0c-392d-ffc5-13ce-2fd6b9af0668 parallel=4 available=24965218304 required="1.1 GiB" 7月 24 15:40:33 buaa-KVM ollama[458186]: time=2024-07-24T15:40:33.874+08:00 level=INFO source=memory.go:309 msg="offload to cuda" layers.requested=-1 layers.model=13 layers.offload=13 layers.split="" memory.available="[23.3 GiB]" memory.required.full="1.1 GiB" memory.required.partial="1.1 GiB" memory.required.kv="96.0 MiB" memory.required.allocations="[1.1 GiB]" memory.weights.total="312.3 MiB" memory.weights.repeating="222.9 MiB" memory.weights.nonrepeating="89.4 MiB" memory.graph.full="192.0 MiB" memory.graph.partial="192.0 MiB" 7月 24 15:40:33 buaa-KVM ollama[458186]: time=2024-07-24T15:40:33.874+08:00 level=INFO source=server.go:383 msg="starting llama server" cmd="/tmp/ollama259438837/runners/cuda_v11/ollama_llama_server --model /usr/share/ollama/.ollama/models/blobs/sha256-65a4313f43b6f94a0a8693d70efe823792303a020601ab3d4cad54cf079296c6 --ctx-size 32768 --batch-size 512 --embedding --log-disable --n-gpu-layers 13 --parallel 4 --port 35263" 7月 24 15:40:33 buaa-KVM ollama[458186]: time=2024-07-24T15:40:33.875+08:00 level=INFO source=sched.go:437 msg="loaded runners" count=2 7月 24 15:40:33 buaa-KVM ollama[458186]: time=2024-07-24T15:40:33.875+08:00 level=INFO source=server.go:583 msg="waiting for llama runner to start responding" 7月 24 15:40:33 buaa-KVM ollama[458186]: time=2024-07-24T15:40:33.875+08:00 level=INFO source=server.go:617 msg="waiting for server to become available" status="llm server error" 7月 24 15:40:33 buaa-KVM ollama[462914]: INFO [main] build info | build=1 commit="d94c6e0" tid="140502749204480" timestamp=1721806833 7月 24 15:40:33 buaa-KVM ollama[462914]: INFO [main] system info | n_threads=32 n_threads_batch=-1 system_info="AVX = 1 | AVX_VNNI = 0 | AVX2 = 0 | AVX512 = 0 | AVX512_VBMI = 0 | AVX512_VNNI = 0 | AVX512_BF16 = 0 | FMA = 0 | NEON = 0 | SVE = 0 | ARM_FMA = 0 | F16C = 0 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 1 | SSE3 = 1 | SSSE3 = 1 | VSX = 0 | MATMUL_INT8 = 0 | LLAMAFILE = 0 | " tid="140502749204480" timestamp=1721806833 total_threads=32 7月 24 15:40:33 buaa-KVM ollama[462914]: INFO [main] HTTP server listening | hostname="127.0.0.1" n_threads_http="31" port="35263" tid="140502749204480" timestamp=1721806833 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: loaded meta data with 33 key-value pairs and 196 tensors from /usr/share/ollama/.ollama/models/blobs/sha256-65a4313f43b6f94a0a8693d70efe823792303a020601ab3d4cad54cf079296c6 (version GGUF V3 (latest)) 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output. 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 0: general.architecture str = jina-bert-v2 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 1: general.type str = model 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 2: general.name str = Jina Bert Implementation 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 3: general.organization str = Jinaai 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 4: general.size_label str = 160M 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 5: general.license str = apache-2.0 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 6: general.tags arr[str,6] = ["sentence-transformers", "feature-ex... 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 7: general.languages arr[str,2] = ["en", "zh"] 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 8: jina-bert-v2.block_count u32 = 12 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 9: jina-bert-v2.context_length u32 = 8192 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 10: jina-bert-v2.embedding_length u32 = 768 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 11: jina-bert-v2.feed_forward_length u32 = 3072 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 12: jina-bert-v2.attention.head_count u32 = 12 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 13: jina-bert-v2.attention.layer_norm_epsilon f32 = 0.000000 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 14: general.file_type u32 = 1 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 15: jina-bert-v2.attention.causal bool = false 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 16: jina-bert-v2.pooling_type u32 = 1 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 17: tokenizer.ggml.model str = gpt2 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 18: tokenizer.ggml.pre str = jina-v2-zh 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 19: tokenizer.ggml.tokens arr[str,61056] = ["", "", "", "", "<m... 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 20: tokenizer.ggml.token_type arr[i32,61056] = [3, 3, 3, 3, 3, 1, 1, 1, 1, 1, 1, 1, ... 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 21: tokenizer.ggml.merges arr[str,39382] = ["t h", "i n", "a n", "e r", "th e", ... 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 22: tokenizer.ggml.bos_token_id u32 = 0 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 23: tokenizer.ggml.eos_token_id u32 = 2 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 24: tokenizer.ggml.unknown_token_id u32 = 3 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 25: tokenizer.ggml.seperator_token_id u32 = 2 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 26: tokenizer.ggml.padding_token_id u32 = 1 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 27: tokenizer.ggml.cls_token_id u32 = 0 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 28: tokenizer.ggml.mask_token_id u32 = 4 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 29: tokenizer.ggml.token_type_count u32 = 2 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 30: tokenizer.ggml.add_bos_token bool = true 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 31: tokenizer.ggml.add_eos_token bool = true 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - kv 32: general.quantization_version u32 = 2 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - type f32: 111 tensors 7月 24 15:40:33 buaa-KVM ollama[458186]: llama_model_loader: - type f16: 85 tensors 7月 24 15:40:33 buaa-KVM ollama[458186]: llm_load_vocab: missing or unrecognized pre-tokenizer type, using: 'default' 7月 24 15:40:33 buaa-KVM ollama[458186]: GGML_ASSERT: /go/src/github.com/ollama/ollama/llm/llama.cpp/src/llama.cpp:5570: unicode_cpts_from_utf8(word).size() > 0 7月 24 15:40:34 buaa-KVM ollama[458186]: Could not attach to process. If your uid matches the uid of the target 7月 24 15:40:34 buaa-KVM ollama[458186]: process, check the setting of /proc/sys/kernel/yama/ptrace_scope, or try 7月 24 15:40:34 buaa-KVM ollama[458186]: again as the root user. For more details, see /etc/sysctl.d/10-ptrace.conf 7月 24 15:40:34 buaa-KVM ollama[458186]: ptrace: Operation not permitted. 7月 24 15:40:34 buaa-KVM ollama[458186]: No stack. 7月 24 15:40:34 buaa-KVM ollama[458186]: The program is not being run. 7月 24 15:40:34 buaa-KVM ollama[458186]: time=2024-07-24T15:40:34.326+08:00 level=INFO source=server.go:617 msg="waiting for server to become available" status="llm server not responding" 7月 24 15:40:34 buaa-KVM ollama[458186]: time=2024-07-24T15:40:34.577+08:00 level=ERROR source=sched.go:443 msg="error loading llama server" error="llama runner process has terminated: signal: aborted (core dumped)" 7月 24 15:40:34 buaa-KVM ollama[458186]: [GIN] 2024/07/24 - 15:40:34 | 500 | 882.662819ms | 192.168.1.202 | POST "/api/embed" 7月 24 15:40:39 buaa-KVM ollama[458186]: time=2024-07-24T15:40:39.737+08:00 level=WARN source=sched.go:634 msg="gpu VRAM usage didn't recover within timeout" seconds=5.160248264 model=/usr/share/ollama/.ollama/models/blobs/sha256-65a4313f43b6f94a0a8693d70efe823792303a020601ab3d4cad54cf079296c6 7月 24 15:40:39 buaa-KVM ollama[458186]: time=2024-07-24T15:40:39.987+08:00 level=WARN source=sched.go:634 msg="gpu VRAM usage didn't recover within timeout" seconds=5.41003445 model=/usr/share/ollama/.ollama/models/blobs/sha256-65a4313f43b6f94a0a8693d70efe823792303a020601ab3d4cad54cf079296c6 7月 24 15:40:40 buaa-KVM ollama[458186]: time=2024-07-24T15:40:40.238+08:00 level=WARN source=sched.go:634 msg="gpu VRAM usage didn't recover within timeout" seconds=5.660559316 model=/usr/share/ollama/.ollama/models/blobs/sha256-65a4313f43b6f94a0a8693d70efe823792303a020601ab3d4cad54cf079296c6

OS

Linux

GPU

Nvidia

CPU

Intel

Ollama version

0.2.8

Dudu0831 commented 2 months ago

I noticed an issue, which should be normal under normal circumstances

7月 24 16:25:35 buaa-KVM ollama[458186]: llm_load_vocab: special tokens cache size = 5 7月 24 16:25:35 buaa-KVM ollama[458186]: llm_load_vocab: token to piece cache size = 0.0769 MB

But it seems like I used the default

rick-github commented 2 months ago 
7月 24 15:40:33 buaa-KVM ollama[458186]: llm_load_vocab: missing or unrecognized pre-tokenizer type, using: 'default'
7月 24 15:40:33 buaa-KVM ollama[458186]: GGML_ASSERT: /go/src/github.com/ollama/ollama/llm/llama.cpp/src/llama.cpp:5570: unicode_cpts_from_utf8(word).size() > 0

The model is not supported by llama.cpp. How did you create the GGUF file?

Dudu0831 commented 2 months ago 
7月 24 15:40:33 buaa-KVM ollama[458186]: llm_load_vocab: missing or unrecognized pre-tokenizer type, using: 'default'
7月 24 15:40:33 buaa-KVM ollama[458186]: GGML_ASSERT: /go/src/github.com/ollama/ollama/llm/llama.cpp/src/llama.cpp:5570: unicode_cpts_from_utf8(word).size() > 0

The model is not supported by llama.cpp. How did you create the GGUF file?

I have modified the script convert_hf_to_gguf_update.py，Added a record for jina-v2-zh。

{"name": "jina-v2-en", "tokt": TOKENIZER_TYPE.WPM, "repo": "https://huggingface.co/jinaai/jina-embeddings-v2-base-en", }, # WPM! {"name": "jina-v2-es", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/jinaai/jina-embeddings-v2-base-es", }, {"name": "jina-v2-zh", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/jinaai/jina-embeddings-v2-base-zh", }, {"name": "jina-v2-de", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/jinaai/jina-embeddings-v2-base-de", }, I have modified the script convert_hf_to_gguf.py，Manually skipped chkhsh authentication。

rick-github commented 2 months ago

https://github.com/ggerganov/llama.cpp/pull/7795

JoanFM commented 2 months ago

Hello,

The jina-embeddings-v2-zh is not yet supported, the missing part is the pretokenization steps from the vocabulary. I have an open PR in https://github.com/ggerganov/llama.cpp/pull/7795 but not sure it will be merged soon.

Dudu0831 commented 2 months ago

Hello,

The jina-embeddings-v2-zh is not yet supported, the missing part is the pretokenization steps from the vocabulary. I have an open PR in ggerganov/llama.cpp#7795 but not sure it will be merged soon.

Okay, thank you for your answer