Ollama (or Llama.cpp) with GPU Passthrough
59 views
Skip to first unread message
Metatron
Jun 21, 2024, 11:04:53 PM (11 days ago) Jun 21
to qubes...@googlegroups.com
Dear Qubers,
I would like to enquire if anyone has had any sucess with Ollama with
GPU passthrough?
I am using an Arch linux template that is working well for dedicated
video out and can play media and games (with stutter) which is already a
great convienance on Qubes.
However the main reason I built a passthrough set up was for Ollama /
Llama.cpp.
I've tried pci strict resetting, no dynamic memory balancing, dasharo
and standard bios. Everything works fine if I swap from qubes OS to
Arch.
I'm using ollama-rocm.
I used the gpu-passthrough guide at:
https://forum.qubes-os.org/t/create-a-gaming-hvm/19000/162
GPU is detected but loading does not progress beyond:
llm_load_tensors: CPU buffer size = 35.44 MiB
It just hangs forever.
Full dump below.
If anyone has got any further or has any thoughts please let me know!
---------
[user@archlinux ~]$ ollama serve
2024 routes.go:1011: INFO server config env="map[OLLAMA_DEBUG:false OLLAMA_FLASH_ATTENTION:false OLLAMA_HOST:http://127.0.0.1:11434 OLLAMA_KEEP_ALIVE: OLLAMA_LLM_LIBRARY: OLLAMA_MAX_LOADED_MODELS:1 OLLAMA_MAX_QUEUE:512 OLLAMA_MAX_VRAM:0 OLLAMA_MODELS:/home/user/.ollama/models OLLAMA_NOHISTORY:false OLLAMA_NOPRUNE:false OLLAMA_NUM_PARALLEL:1 OLLAMA_ORIGINS:[http://localhost https://localhost http://localhost:* https://localhost:* http://127.0.0.1 https://127.0.0.1 http://127.0.0.1:* https://127.0.0.1:* http://0.0.0.0 https://0.0.0.0 http://0.0.0.0:* https://0.0.0.0:* app://* file://* tauri://*] OLLAMA_RUNNERS_DIR: OLLAMA_TMPDIR:]"
time=2024 level=INFO source=images.go:725 msg="total blobs: 10"
time=2024 level=INFO source=images.go:732 msg="total unused blobs removed: 0"
time=2024 level=INFO source=routes.go:1057 msg="Listening on 127.0.0.1:11434 (version 0.1.44)"
time=2024 level=INFO source=payload.go:30 msg="extracting embedded files" dir=/tmp/ollama478581086/runners
time=2024 level=INFO source=payload.go:44 msg="Dynamic LLM libraries [cpu rocm]"
time=2024 level=WARN source=amd_linux.go:48 msg="ollama recommends running the https://www.amd.com/en/support/linux-drivers" error="amdgpu version file missing: /sys/module/amdgpu/version stat /sys/module/amdgpu/version: no such file or directory"
time=2024 level=INFO source=amd_linux.go:301 msg="amdgpu is supported" gpu=0 gpu_type=gfx1030
time=2024 level=INFO source=types.go:71 msg="inference compute" id=0 library=rocm compute=gfx1030 driver=0.0 name=1002:73bf total="16.0 GiB" available="16.0 GiB"
[GIN] 2024 | 200 | 1.05166ms | 127.0.0.1 | HEAD "/"
[GIN] 2024 | 200 | 5.980785ms | 127.0.0.1 | POST "/api/show"
[GIN] 2024 | 200 | 1.4154ms | 127.0.0.1 | POST "/api/show"
time=2024 level=WARN source=amd_linux.go:48 msg="ollama recommends running the https://www.amd.com/en/support/linux-drivers" error="amdgpu version file missing: /sys/module/amdgpu/version stat /sys/module/amdgpu/version: no such file or directory"
time=2024 level=INFO source=amd_linux.go:301 msg="amdgpu is supported" gpu=0 gpu_type=gfx1030
time=2024 level=INFO source=memory.go:133 msg="offload to gpu" layers.requested=-1 layers.real=25 memory.available="16.0 GiB" memory.required.full="1.6 GiB" memory.required.partial="1.6 GiB" memory.required.kv="384.0 MiB" memory.weights.total="703.4 MiB" memory.weights.repeating="651.8 MiB" memory.weights.nonrepeating="51.7 MiB" memory.graph.full="84.0 MiB" memory.graph.partial="122.7 MiB"
time=2024 level=INFO source=memory.go:133 msg="offload to gpu" layers.requested=-1 layers.real=25 memory.available="16.0 GiB" memory.required.full="1.6 GiB" memory.required.partial="1.6 GiB" memory.required.kv="384.0 MiB" memory.weights.total="703.4 MiB" memory.weights.repeating="651.8 MiB" memory.weights.nonrepeating="51.7 MiB" memory.graph.full="84.0 MiB" memory.graph.partial="122.7 MiB"
time=2024 level=INFO source=server.go:341 msg="starting llama server" cmd="/tmp/ollama478581086/runners/rocm/ollama_llama_server --model /home/user/.ollama/models/blobs/sha256- --ctx-size 2048 --batch-size 512 --embedding --log-disable --n-gpu-layers 25 --parallel 1 --port 46093"
time=2024 level=INFO source=sched.go:338 msg="loaded runners" count=1
time=2024 level=INFO source=server.go:529 msg="waiting for llama runner to start responding"
time=2024 level=INFO source=server.go:567 msg="waiting for server to become available" status="llm server error"
INFO [main] build info | build=3051 commit="5921b8f08" tid="125143178271808" timestamp=1719024677
INFO [main] system info | n_threads=8 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 = 1 | " tid="125143178271808" timestamp=1719024677 total_threads=8
INFO [main] HTTP server listening | hostname="127.0.0.1" n_threads_http="7" port="46093" tid="125143178271808" timestamp=1719024677
llama_model_loader: loaded meta data with 26 key-value pairs and 219 tensors from /home/user/.ollama/models/blobs/sha256- (version GGUF V3 (latest))
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
llama_model_loader: - kv 0: general.architecture str = llama
llama_model_loader: - kv 1: general.name str = deepseek-ai
llama_model_loader: - kv 2: llama.context_length u32 = 16384
llama_model_loader: - kv 3: llama.embedding_length u32 = 2048
llama_model_loader: - kv 4: llama.block_count u32 = 24
llama_model_loader: - kv 5: llama.feed_forward_length u32 = 5504
llama_model_loader: - kv 6: llama.rope.dimension_count u32 = 128
llama_model_loader: - kv 7: llama.attention.head_count u32 = 16
llama_model_loader: - kv 8: llama.attention.head_count_kv u32 = 16
llama_model_loader: - kv 9: llama.attention.layer_norm_rms_epsilon f32 = 0.000001
llama_model_loader: - kv 10: llama.rope.freq_base f32 = 100000.000000
llama_model_loader: - kv 11: llama.rope.scaling.type str = linear
llama_model_loader: - kv 12: llama.rope.scaling.factor f32 = 4.000000
llama_model_loader: - kv 13: general.file_type u32 = 2
llama_model_loader: - kv 14: tokenizer.ggml.model str = gpt2
llama_model_loader: - kv 15: tokenizer.ggml.tokens arr[str,32256] = ["!", "\"", "#", "$", "%", "&", "'", ...
llama_model_loader: - kv 16: tokenizer.ggml.scores arr[f32,32256] = [0.000000, 0.000000, 0.000000, 0.0000...
llama_model_loader: - kv 17: tokenizer.ggml.token_type arr[i32,32256] = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv 18: tokenizer.ggml.merges arr[str,31757] = ["Ġ Ġ", "Ġ t", "Ġ a", "i n", "h e...
llama_model_loader: - kv 19: tokenizer.ggml.bos_token_id u32 = 32013
llama_model_loader: - kv 20: tokenizer.ggml.eos_token_id u32 = 32021
llama_model_loader: - kv 21: tokenizer.ggml.padding_token_id u32 = 32014
llama_model_loader: - kv 22: tokenizer.ggml.add_bos_token bool = true
llama_model_loader: - kv 23: tokenizer.ggml.add_eos_token bool = false
llama_model_loader: - kv 24: tokenizer.chat_template str = {% if not add_generation_prompt is de...
llama_model_loader: - kv 25: general.quantization_version u32 = 2
llama_model_loader: - type f32: 49 tensors
llama_model_loader: - type q4_0: 169 tensors
llama_model_loader: - type q6_K: 1 tensors
llm_load_vocab: missing or unrecognized pre-tokenizer type, using: 'default'
llm_load_vocab: special tokens cache size = 256
llm_load_vocab: token to piece cache size = 0.3583 MB
llm_load_print_meta: format = GGUF V3 (latest)
llm_load_print_meta: arch = llama
llm_load_print_meta: vocab type = BPE
llm_load_print_meta: n_vocab = 32256
llm_load_print_meta: n_merges = 31757
llm_load_print_meta: n_ctx_train = 16384
llm_load_print_meta: n_embd = 2048
llm_load_print_meta: n_head = 16
llm_load_print_meta: n_head_kv = 16
llm_load_print_meta: n_layer = 24
llm_load_print_meta: n_rot = 128
llm_load_print_meta: n_embd_head_k = 128
llm_load_print_meta: n_embd_head_v = 128
llm_load_print_meta: n_gqa = 1
llm_load_print_meta: n_embd_k_gqa = 2048
llm_load_print_meta: n_embd_v_gqa = 2048
llm_load_print_meta: f_norm_eps = 0.0e+00
llm_load_print_meta: f_norm_rms_eps = 1.0e-06
llm_load_print_meta: f_clamp_kqv = 0.0e+00
llm_load_print_meta: f_max_alibi_bias = 0.0e+00
llm_load_print_meta: f_logit_scale = 0.0e+00
llm_load_print_meta: n_ff = 5504
llm_load_print_meta: n_expert = 0
llm_load_print_meta: n_expert_used = 0
llm_load_print_meta: causal attn = 1
llm_load_print_meta: pooling type = 0
llm_load_print_meta: rope type = 0
llm_load_print_meta: rope scaling = linear
llm_load_print_meta: freq_base_train = 100000.0
llm_load_print_meta: freq_scale_train = 0.25
llm_load_print_meta: n_yarn_orig_ctx = 16384
llm_load_print_meta: rope_finetuned = unknown
llm_load_print_meta: ssm_d_conv = 0
llm_load_print_meta: ssm_d_inner = 0
llm_load_print_meta: ssm_d_state = 0
llm_load_print_meta: ssm_dt_rank = 0
llm_load_print_meta: model type = ?B
llm_load_print_meta: model ftype = Q4_0
llm_load_print_meta: model params = 1.35 B
llm_load_print_meta: model size = 738.88 MiB (4.60 BPW)
llm_load_print_meta: general.name = deepseek-ai
llm_load_print_meta: BOS token = 32013 '<｜begin▁of▁sentence｜>'
llm_load_print_meta: EOS token = 32021 '<|EOT|>'
llm_load_print_meta: PAD token = 32014 '<｜end▁of▁sentence｜>'
llm_load_print_meta: LF token = 126 'Ä'
time=2024 level=INFO source=server.go:567 msg="waiting for server to become available" status="llm server loading model"
ggml_cuda_init: GGML_CUDA_FORCE_MMQ: no
ggml_cuda_init: CUDA_USE_TENSOR_CORES: yes
ggml_cuda_init: found 1 ROCm devices:
Device 0: AMD Radeon RX 6900 XT, compute capability 10.3, VMM: no
llm_load_tensors: ggml ctx size = 0.22 MiB
llm_load_tensors: offloading 24 repeating layers to GPU
llm_load_tensors: offloading non-repeating layers to GPU
llm_load_tensors: offloaded 25/25 layers to GPU
llm_load_tensors: ROCm0 buffer size = 703.44 MiB
llm_load_tensors: CPU buffer size = 35.44 MiB
I would like to enquire if anyone has had any sucess with Ollama with
GPU passthrough?
I am using an Arch linux template that is working well for dedicated
video out and can play media and games (with stutter) which is already a
great convienance on Qubes.
However the main reason I built a passthrough set up was for Ollama /
Llama.cpp.
I've tried pci strict resetting, no dynamic memory balancing, dasharo
and standard bios. Everything works fine if I swap from qubes OS to
Arch.
I'm using ollama-rocm.
I used the gpu-passthrough guide at:
https://forum.qubes-os.org/t/create-a-gaming-hvm/19000/162
GPU is detected but loading does not progress beyond:
llm_load_tensors: CPU buffer size = 35.44 MiB
It just hangs forever.
Full dump below.
If anyone has got any further or has any thoughts please let me know!
---------
[user@archlinux ~]$ ollama serve
2024 routes.go:1011: INFO server config env="map[OLLAMA_DEBUG:false OLLAMA_FLASH_ATTENTION:false OLLAMA_HOST:http://127.0.0.1:11434 OLLAMA_KEEP_ALIVE: OLLAMA_LLM_LIBRARY: OLLAMA_MAX_LOADED_MODELS:1 OLLAMA_MAX_QUEUE:512 OLLAMA_MAX_VRAM:0 OLLAMA_MODELS:/home/user/.ollama/models OLLAMA_NOHISTORY:false OLLAMA_NOPRUNE:false OLLAMA_NUM_PARALLEL:1 OLLAMA_ORIGINS:[http://localhost https://localhost http://localhost:* https://localhost:* http://127.0.0.1 https://127.0.0.1 http://127.0.0.1:* https://127.0.0.1:* http://0.0.0.0 https://0.0.0.0 http://0.0.0.0:* https://0.0.0.0:* app://* file://* tauri://*] OLLAMA_RUNNERS_DIR: OLLAMA_TMPDIR:]"
time=2024 level=INFO source=images.go:725 msg="total blobs: 10"
time=2024 level=INFO source=images.go:732 msg="total unused blobs removed: 0"
time=2024 level=INFO source=routes.go:1057 msg="Listening on 127.0.0.1:11434 (version 0.1.44)"
time=2024 level=INFO source=payload.go:30 msg="extracting embedded files" dir=/tmp/ollama478581086/runners
time=2024 level=INFO source=payload.go:44 msg="Dynamic LLM libraries [cpu rocm]"
time=2024 level=WARN source=amd_linux.go:48 msg="ollama recommends running the https://www.amd.com/en/support/linux-drivers" error="amdgpu version file missing: /sys/module/amdgpu/version stat /sys/module/amdgpu/version: no such file or directory"
time=2024 level=INFO source=amd_linux.go:301 msg="amdgpu is supported" gpu=0 gpu_type=gfx1030
time=2024 level=INFO source=types.go:71 msg="inference compute" id=0 library=rocm compute=gfx1030 driver=0.0 name=1002:73bf total="16.0 GiB" available="16.0 GiB"
[GIN] 2024 | 200 | 1.05166ms | 127.0.0.1 | HEAD "/"
[GIN] 2024 | 200 | 5.980785ms | 127.0.0.1 | POST "/api/show"
[GIN] 2024 | 200 | 1.4154ms | 127.0.0.1 | POST "/api/show"
time=2024 level=WARN source=amd_linux.go:48 msg="ollama recommends running the https://www.amd.com/en/support/linux-drivers" error="amdgpu version file missing: /sys/module/amdgpu/version stat /sys/module/amdgpu/version: no such file or directory"
time=2024 level=INFO source=amd_linux.go:301 msg="amdgpu is supported" gpu=0 gpu_type=gfx1030
time=2024 level=INFO source=memory.go:133 msg="offload to gpu" layers.requested=-1 layers.real=25 memory.available="16.0 GiB" memory.required.full="1.6 GiB" memory.required.partial="1.6 GiB" memory.required.kv="384.0 MiB" memory.weights.total="703.4 MiB" memory.weights.repeating="651.8 MiB" memory.weights.nonrepeating="51.7 MiB" memory.graph.full="84.0 MiB" memory.graph.partial="122.7 MiB"
time=2024 level=INFO source=memory.go:133 msg="offload to gpu" layers.requested=-1 layers.real=25 memory.available="16.0 GiB" memory.required.full="1.6 GiB" memory.required.partial="1.6 GiB" memory.required.kv="384.0 MiB" memory.weights.total="703.4 MiB" memory.weights.repeating="651.8 MiB" memory.weights.nonrepeating="51.7 MiB" memory.graph.full="84.0 MiB" memory.graph.partial="122.7 MiB"
time=2024 level=INFO source=server.go:341 msg="starting llama server" cmd="/tmp/ollama478581086/runners/rocm/ollama_llama_server --model /home/user/.ollama/models/blobs/sha256- --ctx-size 2048 --batch-size 512 --embedding --log-disable --n-gpu-layers 25 --parallel 1 --port 46093"
time=2024 level=INFO source=sched.go:338 msg="loaded runners" count=1
time=2024 level=INFO source=server.go:529 msg="waiting for llama runner to start responding"
time=2024 level=INFO source=server.go:567 msg="waiting for server to become available" status="llm server error"
INFO [main] build info | build=3051 commit="5921b8f08" tid="125143178271808" timestamp=1719024677
INFO [main] system info | n_threads=8 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 = 1 | " tid="125143178271808" timestamp=1719024677 total_threads=8
INFO [main] HTTP server listening | hostname="127.0.0.1" n_threads_http="7" port="46093" tid="125143178271808" timestamp=1719024677
llama_model_loader: loaded meta data with 26 key-value pairs and 219 tensors from /home/user/.ollama/models/blobs/sha256- (version GGUF V3 (latest))
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
llama_model_loader: - kv 0: general.architecture str = llama
llama_model_loader: - kv 1: general.name str = deepseek-ai
llama_model_loader: - kv 2: llama.context_length u32 = 16384
llama_model_loader: - kv 3: llama.embedding_length u32 = 2048
llama_model_loader: - kv 4: llama.block_count u32 = 24
llama_model_loader: - kv 5: llama.feed_forward_length u32 = 5504
llama_model_loader: - kv 6: llama.rope.dimension_count u32 = 128
llama_model_loader: - kv 7: llama.attention.head_count u32 = 16
llama_model_loader: - kv 8: llama.attention.head_count_kv u32 = 16
llama_model_loader: - kv 9: llama.attention.layer_norm_rms_epsilon f32 = 0.000001
llama_model_loader: - kv 10: llama.rope.freq_base f32 = 100000.000000
llama_model_loader: - kv 11: llama.rope.scaling.type str = linear
llama_model_loader: - kv 12: llama.rope.scaling.factor f32 = 4.000000
llama_model_loader: - kv 13: general.file_type u32 = 2
llama_model_loader: - kv 14: tokenizer.ggml.model str = gpt2
llama_model_loader: - kv 15: tokenizer.ggml.tokens arr[str,32256] = ["!", "\"", "#", "$", "%", "&", "'", ...
llama_model_loader: - kv 16: tokenizer.ggml.scores arr[f32,32256] = [0.000000, 0.000000, 0.000000, 0.0000...
llama_model_loader: - kv 17: tokenizer.ggml.token_type arr[i32,32256] = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv 18: tokenizer.ggml.merges arr[str,31757] = ["Ġ Ġ", "Ġ t", "Ġ a", "i n", "h e...
llama_model_loader: - kv 19: tokenizer.ggml.bos_token_id u32 = 32013
llama_model_loader: - kv 20: tokenizer.ggml.eos_token_id u32 = 32021
llama_model_loader: - kv 21: tokenizer.ggml.padding_token_id u32 = 32014
llama_model_loader: - kv 22: tokenizer.ggml.add_bos_token bool = true
llama_model_loader: - kv 23: tokenizer.ggml.add_eos_token bool = false
llama_model_loader: - kv 24: tokenizer.chat_template str = {% if not add_generation_prompt is de...
llama_model_loader: - kv 25: general.quantization_version u32 = 2
llama_model_loader: - type f32: 49 tensors
llama_model_loader: - type q4_0: 169 tensors
llama_model_loader: - type q6_K: 1 tensors
llm_load_vocab: missing or unrecognized pre-tokenizer type, using: 'default'
llm_load_vocab: special tokens cache size = 256
llm_load_vocab: token to piece cache size = 0.3583 MB
llm_load_print_meta: format = GGUF V3 (latest)
llm_load_print_meta: arch = llama
llm_load_print_meta: vocab type = BPE
llm_load_print_meta: n_vocab = 32256
llm_load_print_meta: n_merges = 31757
llm_load_print_meta: n_ctx_train = 16384
llm_load_print_meta: n_embd = 2048
llm_load_print_meta: n_head = 16
llm_load_print_meta: n_head_kv = 16
llm_load_print_meta: n_layer = 24
llm_load_print_meta: n_rot = 128
llm_load_print_meta: n_embd_head_k = 128
llm_load_print_meta: n_embd_head_v = 128
llm_load_print_meta: n_gqa = 1
llm_load_print_meta: n_embd_k_gqa = 2048
llm_load_print_meta: n_embd_v_gqa = 2048
llm_load_print_meta: f_norm_eps = 0.0e+00
llm_load_print_meta: f_norm_rms_eps = 1.0e-06
llm_load_print_meta: f_clamp_kqv = 0.0e+00
llm_load_print_meta: f_max_alibi_bias = 0.0e+00
llm_load_print_meta: f_logit_scale = 0.0e+00
llm_load_print_meta: n_ff = 5504
llm_load_print_meta: n_expert = 0
llm_load_print_meta: n_expert_used = 0
llm_load_print_meta: causal attn = 1
llm_load_print_meta: pooling type = 0
llm_load_print_meta: rope type = 0
llm_load_print_meta: rope scaling = linear
llm_load_print_meta: freq_base_train = 100000.0
llm_load_print_meta: freq_scale_train = 0.25
llm_load_print_meta: n_yarn_orig_ctx = 16384
llm_load_print_meta: rope_finetuned = unknown
llm_load_print_meta: ssm_d_conv = 0
llm_load_print_meta: ssm_d_inner = 0
llm_load_print_meta: ssm_d_state = 0
llm_load_print_meta: ssm_dt_rank = 0
llm_load_print_meta: model type = ?B
llm_load_print_meta: model ftype = Q4_0
llm_load_print_meta: model params = 1.35 B
llm_load_print_meta: model size = 738.88 MiB (4.60 BPW)
llm_load_print_meta: general.name = deepseek-ai
llm_load_print_meta: BOS token = 32013 '<｜begin▁of▁sentence｜>'
llm_load_print_meta: EOS token = 32021 '<|EOT|>'
llm_load_print_meta: PAD token = 32014 '<｜end▁of▁sentence｜>'
llm_load_print_meta: LF token = 126 'Ä'
time=2024 level=INFO source=server.go:567 msg="waiting for server to become available" status="llm server loading model"
ggml_cuda_init: GGML_CUDA_FORCE_MMQ: no
ggml_cuda_init: CUDA_USE_TENSOR_CORES: yes
ggml_cuda_init: found 1 ROCm devices:
Device 0: AMD Radeon RX 6900 XT, compute capability 10.3, VMM: no
llm_load_tensors: ggml ctx size = 0.22 MiB
llm_load_tensors: offloading 24 repeating layers to GPU
llm_load_tensors: offloading non-repeating layers to GPU
llm_load_tensors: offloaded 25/25 layers to GPU
llm_load_tensors: ROCm0 buffer size = 703.44 MiB
llm_load_tensors: CPU buffer size = 35.44 MiB
Reply all
Reply to author
Forward
0 new messages