Skip to content

Instantly share code, notes, and snippets.

@miminashi

miminashi/gist:ac0170298ef0b617e9681948623006f7

Created August 9, 2025 22:03
Show Gist options

  • Select an option

    Learn more about clone URLs
  • Save miminashi/ac0170298ef0b617e9681948623006f7 to your computer and use it in GitHub Desktop.

Select an option

Learn more about clone URLs
Save miminashi/ac0170298ef0b617e9681948623006f7 to your computer and use it in GitHub Desktop.
Download ZIP
Raw
gistfile1.txt
ubuntu@mi25:~/llama.cpp (master) $ ./build/bin/llama-perplexity --n-gpu-layers 100 --split-mode layer -m ~/.cache/llama.cpp/unsloth_gpt-oss-20b-GGUF_gpt-oss-20b-F16.gguf -f ~/polano.txt
printf '\n'
./build/bin/llama-perplexity --n-gpu-layers 100 --split-mode layer -m ~/.cache/llama.cpp/unsloth_Qwen3-30B-A3B-Instruct-2507-GGUF_Qwen3-30B-A3B-Instruct-2507-UD-Q8_K_XL.gguf -f ~/polano.txt
printf '\n'
./build/bin/llama-perplexity --n-gpu-layers 100 --split-mode layer -m ~/.cache/llama.cpp/unsloth_gemma-3-27b-it-GGUF_gemma-3-27b-it-UD-Q8_K_XL.gguf -f ~/polano.txt
ggml_cuda_init: GGML_CUDA_FORCE_MMQ: no
ggml_cuda_init: GGML_CUDA_FORCE_CUBLAS: no
ggml_cuda_init: found 4 ROCm devices:
Device 0: Radeon Instinct MI25, gfx900:xnack- (0x900), VMM: no, Wave Size: 64
Device 1: Radeon Instinct MI25, gfx900:xnack- (0x900), VMM: no, Wave Size: 64
Device 2: Radeon Instinct MI25, gfx900:xnack- (0x900), VMM: no, Wave Size: 64
Device 3: Radeon Instinct MI25, gfx900:xnack- (0x900), VMM: no, Wave Size: 64
build: 6112 (99acbc99) with cc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0 for x86_64-linux-gnu
llama_model_load_from_file_impl: using device ROCm0 (Radeon Instinct MI25) - 16352 MiB free
llama_model_load_from_file_impl: using device ROCm1 (Radeon Instinct MI25) - 16352 MiB free
llama_model_load_from_file_impl: using device ROCm2 (Radeon Instinct MI25) - 16352 MiB free
llama_model_load_from_file_impl: using device ROCm3 (Radeon Instinct MI25) - 16352 MiB free
llama_model_loader: loaded meta data with 37 key-value pairs and 459 tensors from /home/ubuntu/.cache/llama.cpp/unsloth_gpt-oss-20b-GGUF_gpt-oss-20b-F16.gguf (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 = gpt-oss
llama_model_loader: - kv 1: general.type str = model
llama_model_loader: - kv 2: general.name str = Gpt-Oss-20B
llama_model_loader: - kv 3: general.basename str = Gpt-Oss-20B
llama_model_loader: - kv 4: general.quantized_by str = Unsloth
llama_model_loader: - kv 5: general.size_label str = 20B
llama_model_loader: - kv 6: general.license str = apache-2.0
llama_model_loader: - kv 7: general.repo_url str = https://huggingface.co/unsloth
llama_model_loader: - kv 8: general.tags arr[str,2] = ["vllm", "text-generation"]
llama_model_loader: - kv 9: gpt-oss.block_count u32 = 24
llama_model_loader: - kv 10: gpt-oss.context_length u32 = 131072
llama_model_loader: - kv 11: gpt-oss.embedding_length u32 = 2880
llama_model_loader: - kv 12: gpt-oss.feed_forward_length u32 = 2880
llama_model_loader: - kv 13: gpt-oss.attention.head_count u32 = 64
llama_model_loader: - kv 14: gpt-oss.attention.head_count_kv u32 = 8
llama_model_loader: - kv 15: gpt-oss.rope.freq_base f32 = 150000.000000
llama_model_loader: - kv 16: gpt-oss.attention.layer_norm_rms_epsilon f32 = 0.000010
llama_model_loader: - kv 17: gpt-oss.expert_count u32 = 32
llama_model_loader: - kv 18: gpt-oss.expert_used_count u32 = 4
llama_model_loader: - kv 19: gpt-oss.attention.key_length u32 = 64
llama_model_loader: - kv 20: gpt-oss.attention.value_length u32 = 64
llama_model_loader: - kv 21: general.file_type u32 = 1
llama_model_loader: - kv 22: gpt-oss.attention.sliding_window u32 = 128
llama_model_loader: - kv 23: gpt-oss.expert_feed_forward_length u32 = 2880
llama_model_loader: - kv 24: gpt-oss.rope.scaling.type str = yarn
llama_model_loader: - kv 25: gpt-oss.rope.scaling.factor f32 = 32.000000
llama_model_loader: - kv 26: gpt-oss.rope.scaling.original_context_length u32 = 4096
llama_model_loader: - kv 27: general.quantization_version u32 = 2
llama_model_loader: - kv 28: tokenizer.ggml.model str = gpt2
llama_model_loader: - kv 29: tokenizer.ggml.pre str = gpt-4o
llama_model_loader: - kv 30: tokenizer.ggml.tokens arr[str,201088] = ["!", "\"", "#", "$", "%", "&", "'", ...
llama_model_loader: - kv 31: tokenizer.ggml.token_type arr[i32,201088] = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv 32: tokenizer.ggml.merges arr[str,446189] = ["Ġ Ġ", "Ġ ĠĠĠ", "ĠĠ ĠĠ", "...
llama_model_loader: - kv 33: tokenizer.ggml.bos_token_id u32 = 199998
llama_model_loader: - kv 34: tokenizer.ggml.eos_token_id u32 = 200002
llama_model_loader: - kv 35: tokenizer.ggml.padding_token_id u32 = 200017
llama_model_loader: - kv 36: tokenizer.chat_template str = {# Copyright 2025-present Unsloth. Ap...
llama_model_loader: - type f32: 289 tensors
llama_model_loader: - type f16: 98 tensors
llama_model_loader: - type mxfp4: 72 tensors
print_info: file format = GGUF V3 (latest)
print_info: file type = F16
print_info: file size = 12.83 GiB (5.27 BPW)
load: printing all EOG tokens:
load: - 199999 ('<|endoftext|>')
load: - 200002 ('<|return|>')
load: - 200007 ('<|end|>')
load: - 200012 ('<|call|>')
load: special_eog_ids contains both '<|return|>' and '<|call|>' tokens, removing '<|end|>' token from EOG list
load: special tokens cache size = 21
load: token to piece cache size = 1.3332 MB
print_info: arch = gpt-oss
print_info: vocab_only = 0
print_info: n_ctx_train = 131072
print_info: n_embd = 2880
print_info: n_layer = 24
print_info: n_head = 64
print_info: n_head_kv = 8
print_info: n_rot = 64
print_info: n_swa = 128
print_info: is_swa_any = 1
print_info: n_embd_head_k = 64
print_info: n_embd_head_v = 64
print_info: n_gqa = 8
print_info: n_embd_k_gqa = 512
print_info: n_embd_v_gqa = 512
print_info: f_norm_eps = 0.0e+00
print_info: f_norm_rms_eps = 1.0e-05
print_info: f_clamp_kqv = 0.0e+00
print_info: f_max_alibi_bias = 0.0e+00
print_info: f_logit_scale = 0.0e+00
print_info: f_attn_scale = 0.0e+00
print_info: n_ff = 2880
print_info: n_expert = 32
print_info: n_expert_used = 4
print_info: causal attn = 1
print_info: pooling type = 0
print_info: rope type = 2
print_info: rope scaling = yarn
print_info: freq_base_train = 150000.0
print_info: freq_scale_train = 0.03125
print_info: n_ctx_orig_yarn = 4096
print_info: rope_finetuned = unknown
print_info: model type = ?B
print_info: model params = 20.91 B
print_info: general.name = Gpt-Oss-20B
print_info: n_ff_exp = 2880
print_info: vocab type = BPE
print_info: n_vocab = 201088
print_info: n_merges = 446189
print_info: BOS token = 199998 '<|startoftext|>'
print_info: EOS token = 200002 '<|return|>'
print_info: EOT token = 199999 '<|endoftext|>'
print_info: PAD token = 200017 '<|reserved_200017|>'
print_info: LF token = 198 'Ċ'
print_info: EOG token = 199999 '<|endoftext|>'
print_info: EOG token = 200002 '<|return|>'
print_info: EOG token = 200012 '<|call|>'
print_info: max token length = 256
load_tensors: loading model tensors, this can take a while... (mmap = true)
load_tensors: offloading 24 repeating layers to GPU
load_tensors: offloading output layer to GPU
load_tensors: offloaded 25/25 layers to GPU
load_tensors: ROCm0 model buffer size = 3188.52 MiB
load_tensors: ROCm1 model buffer size = 2733.01 MiB
load_tensors: ROCm2 model buffer size = 2733.01 MiB
load_tensors: ROCm3 model buffer size = 3382.13 MiB
load_tensors: CPU_Mapped model buffer size = 1104.61 MiB
...................................................................................
llama_context: constructing llama_context
llama_context: n_seq_max = 4
llama_context: n_ctx = 2048
llama_context: n_ctx_per_seq = 512
llama_context: n_batch = 2048
llama_context: n_ubatch = 512
llama_context: causal_attn = 1
llama_context: flash_attn = 0
llama_context: kv_unified = false
llama_context: freq_base = 150000.0
llama_context: freq_scale = 0.03125
llama_context: n_ctx_per_seq (512) < n_ctx_train (131072) -- the full capacity of the model will not be utilized
llama_context: requested n_seq_max (4) > 1, but swa_full is not enabled -- performance may be degraded: https://github.com/ggml-org/llama.cpp/pull/13845#issuecomment-2924800573
llama_context: ROCm_Host output buffer size = 3.07 MiB
llama_kv_cache_unified_iswa: creating non-SWA KV cache, size = 512 cells
llama_kv_cache_unified: ROCm0 KV buffer size = 12.00 MiB
llama_kv_cache_unified: ROCm1 KV buffer size = 12.00 MiB
llama_kv_cache_unified: ROCm2 KV buffer size = 12.00 MiB
llama_kv_cache_unified: ROCm3 KV buffer size = 12.00 MiB
llama_kv_cache_unified: size = 48.00 MiB ( 512 cells, 12 layers, 4/4 seqs), K (f16): 24.00 MiB, V (f16): 24.00 MiB
llama_kv_cache_unified_iswa: creating SWA KV cache, size = 512 cells
llama_kv_cache_unified: ROCm0 KV buffer size = 16.00 MiB
llama_kv_cache_unified: ROCm1 KV buffer size = 12.00 MiB
llama_kv_cache_unified: ROCm2 KV buffer size = 12.00 MiB
llama_kv_cache_unified: ROCm3 KV buffer size = 8.00 MiB
llama_kv_cache_unified: size = 48.00 MiB ( 512 cells, 12 layers, 4/4 seqs), K (f16): 24.00 MiB, V (f16): 24.00 MiB
llama_context: pipeline parallelism enabled (n_copies=4)
llama_context: ROCm0 compute buffer size = 137.79 MiB
llama_context: ROCm1 compute buffer size = 137.79 MiB
llama_context: ROCm2 compute buffer size = 137.79 MiB
llama_context: ROCm3 compute buffer size = 444.92 MiB
llama_context: ROCm_Host compute buffer size = 29.68 MiB
llama_context: graph nodes = 1446
llama_context: graph splits = 5
common_init_from_params: added <|endoftext|> logit bias = -inf
common_init_from_params: added <|return|> logit bias = -inf
common_init_from_params: added <|call|> logit bias = -inf
common_init_from_params: setting dry_penalty_last_n to ctx_size = 2048
common_init_from_params: warming up the model with an empty run - please wait ... (--no-warmup to disable)
system_info: n_threads = 12 (n_threads_batch = 12) / 24 | ROCm : NO_VMM = 1 | PEER_MAX_BATCH_SIZE = 128 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | AVX2 = 1 | F16C = 1 | FMA = 1 | BMI2 = 1 | LLAMAFILE = 1 | OPENMP = 1 | REPACK = 1 |
perplexity: tokenizing the input ..
perplexity: tokenization took 110.99 ms
perplexity: calculating perplexity over 70 chunks, n_ctx=512, batch_size=2048, n_seq=4
perplexity: 7.18 seconds per pass - ETA 2.08 minutes
[1]112.9377,[2]122.1352,[3]100.6324,[4]93.0957,[5]89.7141,[6]90.4398,[7]91.6406,[8]92.1841,[9]89.5061,[10]90.4190,[11]89.3026,[12]89.9131,[13]88.5398,[14]89.1334,[15]88.7462,[16]88.9980,[17]91.1456,[18]90.6418,[19]91.2020,[20]89.6155,[21]88.7622,[22]89.3350,[23]90.7457,[24]90.6350,[25]90.4557,[26]89.7106,[27]89.7938,[28]91.0184,[29]91.4180,[30]90.7831,[31]90.9751,[32]91.4860,[33]92.0903,[34]92.0968,[35]91.1305,[36]90.8390,[37]90.3689,[38]90.5001,[39]90.0757,[40]89.9540,[41]89.5802,[42]90.1338,[43]90.4896,[44]89.7632,[45]89.6120,[46]89.8986,[47]89.9305,[48]90.4410,[49]90.1637,[50]90.2724,[51]90.7497,[52]90.0331,[53]90.1321,[54]90.2238,[55]90.0770,[56]89.4811,[57]90.0112,[58]90.5232,[59]90.5199,[60]90.4949,[61]90.7617,[62]90.6152,[63]90.5651,[64]90.4227,[65]90.2705,[66]90.3151,[67]89.8836,[68]90.2719,[69]90.5306,[70]90.8771,
Final estimate: PPL = 90.8771 +/- 1.61906
llama_perf_context_print: load time = 19779.11 ms
llama_perf_context_print: prompt eval time = 104807.30 ms / 35840 tokens ( 2.92 ms per token, 341.96 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 109861.73 ms / 35841 tokens
llama_perf_context_print: graphs reused = 0
ggml_cuda_init: GGML_CUDA_FORCE_MMQ: no
ggml_cuda_init: GGML_CUDA_FORCE_CUBLAS: no
ggml_cuda_init: found 4 ROCm devices:
Device 0: Radeon Instinct MI25, gfx900:xnack- (0x900), VMM: no, Wave Size: 64
Device 1: Radeon Instinct MI25, gfx900:xnack- (0x900), VMM: no, Wave Size: 64
Device 2: Radeon Instinct MI25, gfx900:xnack- (0x900), VMM: no, Wave Size: 64
Device 3: Radeon Instinct MI25, gfx900:xnack- (0x900), VMM: no, Wave Size: 64
build: 6112 (99acbc99) with cc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0 for x86_64-linux-gnu
llama_model_load_from_file_impl: using device ROCm0 (Radeon Instinct MI25) - 16352 MiB free
llama_model_load_from_file_impl: using device ROCm1 (Radeon Instinct MI25) - 16352 MiB free
llama_model_load_from_file_impl: using device ROCm2 (Radeon Instinct MI25) - 16352 MiB free
llama_model_load_from_file_impl: using device ROCm3 (Radeon Instinct MI25) - 16352 MiB free
llama_model_loader: loaded meta data with 45 key-value pairs and 579 tensors from /home/ubuntu/.cache/llama.cpp/unsloth_Qwen3-30B-A3B-Instruct-2507-GGUF_Qwen3-30B-A3B-Instruct-2507-UD-Q8_K_XL.gguf (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 = qwen3moe
llama_model_loader: - kv 1: general.type str = model
llama_model_loader: - kv 2: general.name str = Qwen3-30B-A3B-Instruct-2507
llama_model_loader: - kv 3: general.version str = 2507
llama_model_loader: - kv 4: general.finetune str = Instruct
llama_model_loader: - kv 5: general.basename str = Qwen3-30B-A3B-Instruct-2507
llama_model_loader: - kv 6: general.quantized_by str = Unsloth
llama_model_loader: - kv 7: general.size_label str = 30B-A3B
llama_model_loader: - kv 8: general.license str = apache-2.0
llama_model_loader: - kv 9: general.license.link str = https://huggingface.co/Qwen/Qwen3-30B...
llama_model_loader: - kv 10: general.repo_url str = https://huggingface.co/unsloth
llama_model_loader: - kv 11: general.base_model.count u32 = 1
llama_model_loader: - kv 12: general.base_model.0.name str = Qwen3 30B A3B Instruct 2507
llama_model_loader: - kv 13: general.base_model.0.version str = 2507
llama_model_loader: - kv 14: general.base_model.0.organization str = Qwen
llama_model_loader: - kv 15: general.base_model.0.repo_url str = https://huggingface.co/Qwen/Qwen3-30B...
llama_model_loader: - kv 16: general.tags arr[str,2] = ["unsloth", "text-generation"]
llama_model_loader: - kv 17: qwen3moe.block_count u32 = 48
llama_model_loader: - kv 18: qwen3moe.context_length u32 = 262144
llama_model_loader: - kv 19: qwen3moe.embedding_length u32 = 2048
llama_model_loader: - kv 20: qwen3moe.feed_forward_length u32 = 6144
llama_model_loader: - kv 21: qwen3moe.attention.head_count u32 = 32
llama_model_loader: - kv 22: qwen3moe.attention.head_count_kv u32 = 4
llama_model_loader: - kv 23: qwen3moe.rope.freq_base f32 = 10000000.000000
llama_model_loader: - kv 24: qwen3moe.attention.layer_norm_rms_epsilon f32 = 0.000001
llama_model_loader: - kv 25: qwen3moe.expert_used_count u32 = 8
llama_model_loader: - kv 26: qwen3moe.attention.key_length u32 = 128
llama_model_loader: - kv 27: qwen3moe.attention.value_length u32 = 128
llama_model_loader: - kv 28: qwen3moe.expert_count u32 = 128
llama_model_loader: - kv 29: qwen3moe.expert_feed_forward_length u32 = 768
llama_model_loader: - kv 30: tokenizer.ggml.model str = gpt2
llama_model_loader: - kv 31: tokenizer.ggml.pre str = qwen2
llama_model_loader: - kv 32: tokenizer.ggml.tokens arr[str,151936] = ["!", "\"", "#", "$", "%", "&", "'", ...
llama_model_loader: - kv 33: tokenizer.ggml.token_type arr[i32,151936] = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv 34: tokenizer.ggml.merges arr[str,151387] = ["Ġ Ġ", "ĠĠ ĠĠ", "i n", "Ġ t",...
llama_model_loader: - kv 35: tokenizer.ggml.eos_token_id u32 = 151645
llama_model_loader: - kv 36: tokenizer.ggml.padding_token_id u32 = 151654
llama_model_loader: - kv 37: tokenizer.ggml.add_bos_token bool = false
llama_model_loader: - kv 38: tokenizer.chat_template str = {%- if tools %}\n {{- '<|im_start|>...
llama_model_loader: - kv 39: general.quantization_version u32 = 2
llama_model_loader: - kv 40: general.file_type u32 = 7
llama_model_loader: - kv 41: quantize.imatrix.file str = Qwen3-30B-A3B-Instruct-2507-GGUF/imat...
llama_model_loader: - kv 42: quantize.imatrix.dataset str = unsloth_calibration_Qwen3-30B-A3B-Ins...
llama_model_loader: - kv 43: quantize.imatrix.entries_count u32 = 384
llama_model_loader: - kv 44: quantize.imatrix.chunks_count u32 = 693
llama_model_loader: - type f32: 241 tensors
llama_model_loader: - type f16: 75 tensors
llama_model_loader: - type q8_0: 263 tensors
print_info: file format = GGUF V3 (latest)
print_info: file type = Q8_0
print_info: file size = 33.51 GiB (9.43 BPW)
load: printing all EOG tokens:
load: - 151643 ('<|endoftext|>')
load: - 151645 ('<|im_end|>')
load: - 151662 ('<|fim_pad|>')
load: - 151663 ('<|repo_name|>')
load: - 151664 ('<|file_sep|>')
load: special tokens cache size = 26
load: token to piece cache size = 0.9311 MB
print_info: arch = qwen3moe
print_info: vocab_only = 0
print_info: n_ctx_train = 262144
print_info: n_embd = 2048
print_info: n_layer = 48
print_info: n_head = 32
print_info: n_head_kv = 4
print_info: n_rot = 128
print_info: n_swa = 0
print_info: is_swa_any = 0
print_info: n_embd_head_k = 128
print_info: n_embd_head_v = 128
print_info: n_gqa = 8
print_info: n_embd_k_gqa = 512
print_info: n_embd_v_gqa = 512
print_info: f_norm_eps = 0.0e+00
print_info: f_norm_rms_eps = 1.0e-06
print_info: f_clamp_kqv = 0.0e+00
print_info: f_max_alibi_bias = 0.0e+00
print_info: f_logit_scale = 0.0e+00
print_info: f_attn_scale = 0.0e+00
print_info: n_ff = 6144
print_info: n_expert = 128
print_info: n_expert_used = 8
print_info: causal attn = 1
print_info: pooling type = 0
print_info: rope type = 2
print_info: rope scaling = linear
print_info: freq_base_train = 10000000.0
print_info: freq_scale_train = 1
print_info: n_ctx_orig_yarn = 262144
print_info: rope_finetuned = unknown
print_info: model type = 30B.A3B
print_info: model params = 30.53 B
print_info: general.name = Qwen3-30B-A3B-Instruct-2507
print_info: n_ff_exp = 768
print_info: vocab type = BPE
print_info: n_vocab = 151936
print_info: n_merges = 151387
print_info: BOS token = 11 ','
print_info: EOS token = 151645 '<|im_end|>'
print_info: EOT token = 151645 '<|im_end|>'
print_info: PAD token = 151654 '<|vision_pad|>'
print_info: LF token = 198 'Ċ'
print_info: FIM PRE token = 151659 '<|fim_prefix|>'
print_info: FIM SUF token = 151661 '<|fim_suffix|>'
print_info: FIM MID token = 151660 '<|fim_middle|>'
print_info: FIM PAD token = 151662 '<|fim_pad|>'
print_info: FIM REP token = 151663 '<|repo_name|>'
print_info: FIM SEP token = 151664 '<|file_sep|>'
print_info: EOG token = 151643 '<|endoftext|>'
print_info: EOG token = 151645 '<|im_end|>'
print_info: EOG token = 151662 '<|fim_pad|>'
print_info: EOG token = 151663 '<|repo_name|>'
print_info: EOG token = 151664 '<|file_sep|>'
print_info: max token length = 256
load_tensors: loading model tensors, this can take a while... (mmap = true)
load_tensors: offloading 48 repeating layers to GPU
load_tensors: offloading output layer to GPU
load_tensors: offloaded 49/49 layers to GPU
load_tensors: ROCm0 model buffer size = 9875.34 MiB
load_tensors: ROCm1 model buffer size = 7596.95 MiB
load_tensors: ROCm2 model buffer size = 7596.95 MiB
load_tensors: ROCm3 model buffer size = 8654.26 MiB
load_tensors: CPU_Mapped model buffer size = 593.50 MiB
.................................................................................................
llama_context: constructing llama_context
llama_context: n_seq_max = 4
llama_context: n_ctx = 2048
llama_context: n_ctx_per_seq = 512
llama_context: n_batch = 2048
llama_context: n_ubatch = 512
llama_context: causal_attn = 1
llama_context: flash_attn = 0
llama_context: kv_unified = false
llama_context: freq_base = 10000000.0
llama_context: freq_scale = 1
llama_context: n_ctx_per_seq (512) < n_ctx_train (262144) -- the full capacity of the model will not be utilized
llama_context: ROCm_Host output buffer size = 2.32 MiB
llama_kv_cache_unified: ROCm0 KV buffer size = 52.00 MiB
llama_kv_cache_unified: ROCm1 KV buffer size = 48.00 MiB
llama_kv_cache_unified: ROCm2 KV buffer size = 48.00 MiB
llama_kv_cache_unified: ROCm3 KV buffer size = 44.00 MiB
llama_kv_cache_unified: size = 192.00 MiB ( 512 cells, 48 layers, 4/4 seqs), K (f16): 96.00 MiB, V (f16): 96.00 MiB
llama_context: pipeline parallelism enabled (n_copies=4)
llama_context: ROCm0 compute buffer size = 104.52 MiB
llama_context: ROCm1 compute buffer size = 104.52 MiB
llama_context: ROCm2 compute buffer size = 104.52 MiB
llama_context: ROCm3 compute buffer size = 328.78 MiB
llama_context: ROCm_Host compute buffer size = 16.04 MiB
llama_context: graph nodes = 3174
llama_context: graph splits = 5
common_init_from_params: added <|endoftext|> logit bias = -inf
common_init_from_params: added <|im_end|> logit bias = -inf
common_init_from_params: added <|fim_pad|> logit bias = -inf
common_init_from_params: added <|repo_name|> logit bias = -inf
common_init_from_params: added <|file_sep|> logit bias = -inf
common_init_from_params: setting dry_penalty_last_n to ctx_size = 2048
common_init_from_params: warming up the model with an empty run - please wait ... (--no-warmup to disable)
system_info: n_threads = 12 (n_threads_batch = 12) / 24 | ROCm : NO_VMM = 1 | PEER_MAX_BATCH_SIZE = 128 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | AVX2 = 1 | F16C = 1 | FMA = 1 | BMI2 = 1 | LLAMAFILE = 1 | OPENMP = 1 | REPACK = 1 |
perplexity: tokenizing the input ..
perplexity: tokenization took 93.719 ms
perplexity: calculating perplexity over 62 chunks, n_ctx=512, batch_size=2048, n_seq=4
perplexity: 9.06 seconds per pass - ETA 2.33 minutes
[1]17.8927,[2]20.1188,[3]17.5880,[4]15.7937,[5]15.1505,[6]14.3488,[7]14.3189,[8]14.0705,[9]14.0010,[10]13.7464,[11]13.9477,[12]14.2810,[13]14.2990,[14]14.0895,[15]14.3763,[16]14.2257,[17]14.3015,[18]14.1060,[19]14.0180,[20]14.5269,[21]14.6234,[22]14.5623,[23]14.5912,[24]14.8327,[25]14.5430,[26]14.6653,[27]14.3976,[28]14.4389,[29]14.1773,[30]14.0281,[31]13.8505,[32]13.9944,[33]14.2497,[34]14.1862,[35]14.2975,[36]14.1557,[37]14.1691,[38]14.2407,[39]14.3550,[40]14.4891,[41]14.4864,[42]14.4668,[43]14.3618,[44]14.4407,[45]14.4565,[46]14.3720,[47]14.5502,[48]14.4015,[49]14.4064,[50]14.3448,[51]14.3075,[52]14.3204,[53]14.3594,[54]14.3976,[55]14.3658,[56]14.5180,[57]14.6384,[58]14.7600,[59]14.8412,[60]14.7975,[61]14.6323,[62]14.7477,
Final estimate: PPL = 14.7477 +/- 0.35399
llama_perf_context_print: load time = 46797.67 ms
llama_perf_context_print: prompt eval time = 113468.18 ms / 31744 tokens ( 3.57 ms per token, 279.76 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 117263.57 ms / 31745 tokens
llama_perf_context_print: graphs reused = 0
ggml_cuda_init: GGML_CUDA_FORCE_MMQ: no
ggml_cuda_init: GGML_CUDA_FORCE_CUBLAS: no
ggml_cuda_init: found 4 ROCm devices:
Device 0: Radeon Instinct MI25, gfx900:xnack- (0x900), VMM: no, Wave Size: 64
Device 1: Radeon Instinct MI25, gfx900:xnack- (0x900), VMM: no, Wave Size: 64
Device 2: Radeon Instinct MI25, gfx900:xnack- (0x900), VMM: no, Wave Size: 64
Device 3: Radeon Instinct MI25, gfx900:xnack- (0x900), VMM: no, Wave Size: 64
build: 6112 (99acbc99) with cc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0 for x86_64-linux-gnu
llama_model_load_from_file_impl: using device ROCm0 (Radeon Instinct MI25) - 16352 MiB free
llama_model_load_from_file_impl: using device ROCm1 (Radeon Instinct MI25) - 16352 MiB free
llama_model_load_from_file_impl: using device ROCm2 (Radeon Instinct MI25) - 16352 MiB free
llama_model_load_from_file_impl: using device ROCm3 (Radeon Instinct MI25) - 16352 MiB free
llama_model_loader: loaded meta data with 40 key-value pairs and 808 tensors from /home/ubuntu/.cache/llama.cpp/unsloth_gemma-3-27b-it-GGUF_gemma-3-27b-it-UD-Q8_K_XL.gguf (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 = gemma3
llama_model_loader: - kv 1: general.type str = model
llama_model_loader: - kv 2: general.name str = Gemma-3-27B-It
llama_model_loader: - kv 3: general.finetune str = it
llama_model_loader: - kv 4: general.basename str = Gemma-3-27B-It
llama_model_loader: - kv 5: general.quantized_by str = Unsloth
llama_model_loader: - kv 6: general.size_label str = 27B
llama_model_loader: - kv 7: general.repo_url str = https://huggingface.co/unsloth
llama_model_loader: - kv 8: gemma3.context_length u32 = 131072
llama_model_loader: - kv 9: gemma3.embedding_length u32 = 5376
llama_model_loader: - kv 10: gemma3.block_count u32 = 62
llama_model_loader: - kv 11: gemma3.feed_forward_length u32 = 21504
llama_model_loader: - kv 12: gemma3.attention.head_count u32 = 32
llama_model_loader: - kv 13: gemma3.attention.layer_norm_rms_epsilon f32 = 0.000001
llama_model_loader: - kv 14: gemma3.attention.key_length u32 = 128
llama_model_loader: - kv 15: gemma3.attention.value_length u32 = 128
llama_model_loader: - kv 16: gemma3.rope.freq_base f32 = 1000000.000000
llama_model_loader: - kv 17: gemma3.attention.sliding_window u32 = 1024
llama_model_loader: - kv 18: gemma3.attention.head_count_kv u32 = 16
llama_model_loader: - kv 19: gemma3.rope.scaling.type str = linear
llama_model_loader: - kv 20: gemma3.rope.scaling.factor f32 = 8.000000
llama_model_loader: - kv 21: tokenizer.ggml.model str = llama
llama_model_loader: - kv 22: tokenizer.ggml.pre str = default
llama_model_loader: - kv 23: tokenizer.ggml.tokens arr[str,262208] = ["<pad>", "<eos>", "<bos>", "<unk>", ...
llama_model_loader: - kv 24: tokenizer.ggml.scores arr[f32,262208] = [-1000.000000, -1000.000000, -1000.00...
llama_model_loader: - kv 25: tokenizer.ggml.token_type arr[i32,262208] = [3, 3, 3, 3, 3, 4, 3, 3, 3, 3, 3, 3, ...
llama_model_loader: - kv 26: tokenizer.ggml.bos_token_id u32 = 2
llama_model_loader: - kv 27: tokenizer.ggml.eos_token_id u32 = 106
llama_model_loader: - kv 28: tokenizer.ggml.unknown_token_id u32 = 3
llama_model_loader: - kv 29: tokenizer.ggml.padding_token_id u32 = 0
llama_model_loader: - kv 30: tokenizer.ggml.add_bos_token bool = true
llama_model_loader: - kv 31: tokenizer.ggml.add_eos_token bool = false
llama_model_loader: - kv 32: tokenizer.chat_template str = {{ bos_token }}\n{%- if messages[0]['r...
llama_model_loader: - kv 33: tokenizer.ggml.add_space_prefix bool = false
llama_model_loader: - kv 34: general.quantization_version u32 = 2
llama_model_loader: - kv 35: general.file_type u32 = 7
llama_model_loader: - kv 36: quantize.imatrix.file str = gemma-3-27b-it-GGUF/imatrix_unsloth.dat
llama_model_loader: - kv 37: quantize.imatrix.dataset str = unsloth_calibration_gemma-3-27b-it.txt
llama_model_loader: - kv 38: quantize.imatrix.entries_count i32 = 434
llama_model_loader: - kv 39: quantize.imatrix.chunks_count i32 = 663
llama_model_loader: - type f32: 373 tensors
llama_model_loader: - type q8_0: 409 tensors
llama_model_loader: - type bf16: 26 tensors
print_info: file format = GGUF V3 (latest)
print_info: file type = Q8_0
print_info: file size = 29.62 GiB (9.42 BPW)
load: printing all EOG tokens:
load: - 106 ('<end_of_turn>')
load: special tokens cache size = 6415
load: token to piece cache size = 1.9446 MB
print_info: arch = gemma3
print_info: vocab_only = 0
print_info: n_ctx_train = 131072
print_info: n_embd = 5376
print_info: n_layer = 62
print_info: n_head = 32
print_info: n_head_kv = 16
print_info: n_rot = 128
print_info: n_swa = 1024
print_info: is_swa_any = 1
print_info: n_embd_head_k = 128
print_info: n_embd_head_v = 128
print_info: n_gqa = 2
print_info: n_embd_k_gqa = 2048
print_info: n_embd_v_gqa = 2048
print_info: f_norm_eps = 0.0e+00
print_info: f_norm_rms_eps = 1.0e-06
print_info: f_clamp_kqv = 0.0e+00
print_info: f_max_alibi_bias = 0.0e+00
print_info: f_logit_scale = 0.0e+00
print_info: f_attn_scale = 7.7e-02
print_info: n_ff = 21504
print_info: n_expert = 0
print_info: n_expert_used = 0
print_info: causal attn = 1
print_info: pooling type = 0
print_info: rope type = 2
print_info: rope scaling = linear
print_info: freq_base_train = 1000000.0
print_info: freq_scale_train = 0.125
print_info: n_ctx_orig_yarn = 131072
print_info: rope_finetuned = unknown
print_info: model type = 27B
print_info: model params = 27.01 B
print_info: general.name = Gemma-3-27B-It
print_info: vocab type = SPM
print_info: n_vocab = 262208
print_info: n_merges = 0
print_info: BOS token = 2 '<bos>'
print_info: EOS token = 106 '<end_of_turn>'
print_info: EOT token = 106 '<end_of_turn>'
print_info: UNK token = 3 '<unk>'
print_info: PAD token = 0 '<pad>'
print_info: LF token = 248 '<0x0A>'
print_info: EOG token = 106 '<end_of_turn>'
print_info: max token length = 48
load_tensors: loading model tensors, this can take a while... (mmap = true)
load_tensors: offloading 62 repeating layers to GPU
load_tensors: offloading output layer to GPU
load_tensors: offloaded 63/63 layers to GPU
load_tensors: ROCm0 model buffer size = 8053.54 MiB
load_tensors: ROCm1 model buffer size = 6695.11 MiB
load_tensors: ROCm2 model buffer size = 6695.11 MiB
load_tensors: ROCm3 model buffer size = 8886.50 MiB
load_tensors: CPU_Mapped model buffer size = 2688.66 MiB
....................................................................................
llama_context: constructing llama_context
llama_context: n_seq_max = 4
llama_context: n_ctx = 2048
llama_context: n_ctx_per_seq = 512
llama_context: n_batch = 2048
llama_context: n_ubatch = 512
llama_context: causal_attn = 1
llama_context: flash_attn = 0
llama_context: kv_unified = false
llama_context: freq_base = 1000000.0
llama_context: freq_scale = 0.125
llama_context: n_ctx_per_seq (512) < n_ctx_trahttps://deepwiki.com/search/perplexity_2383b2da-cc6a-4076-9b9e-213af193691ein (131072) -- the full capacity of the model will not be utilized
llama_context: requested n_seq_max (4) > 1, but swa_full is not enabled -- performance may be degraded: https://github.com/ggml-org/llama.cpp/pull/13845#issuecomment-2924800573
llama_context: ROCm_Host output buffer size = 4.00 MiB
llama_kv_cache_unified_iswa: creating non-SWA KV cache, size = 512 cells
llama_kv_cache_unified: ROCm0 KV buffer size = 32.00 MiB
llama_kv_cache_unified: ROCm1 KV buffer size = 48.00 MiB
llama_kv_cache_unified: ROCm2 KV buffer size = 48.00 MiB
llama_kv_cache_unified: ROCm3 KV buffer size = 32.00 MiB
llama_kv_cache_unified: size = 160.00 MiB ( 512 cells, 10 layers, 4/4 seqs), K (f16): 80.00 MiB, V (f16): 80.00 MiB
llama_kv_cache_unified_iswa: creating SWA KV cache, size = 512 cells
llama_kv_cache_unified: ROCm0 KV buffer size = 224.00 MiB
llama_kv_cache_unified: ROCm1 KV buffer size = 208.00 MiB
llama_kv_cache_unified: ROCm2 KV buffer size = 208.00 MiB
llama_kv_cache_unified: ROCm3 KV buffer size = 192.00 MiB
llama_kv_cache_unified: size = 832.00 MiB ( 512 cells, 52 layers, 4/4 seqs), K (f16): 416.00 MiB, V (f16): 416.00 MiB
llama_context: pipeline parallelism enabled (n_copies=4)
llama_context: ROCm0 compute buffer size = 277.04 MiB
llama_context: ROCm1 compute buffer size = 277.04 MiB
llama_context: ROCm2 compute buffer size = 277.04 MiB
llama_context: ROCm3 compute buffer size = 636.67 MiB
llama_context: ROCm_Host compute buffer size = 82.55 MiB
llama_context: graph nodes = 2735
llama_context: graph splits = 5
common_init_from_params: added <end_of_turn> logit bias = -inf
common_init_from_params: setting dry_penalty_last_n to ctx_size = 2048
common_init_from_params: warming up the model with an empty run - please wait ... (--no-warmup to disable)
system_info: n_threads = 12 (n_threads_batch = 12) / 24 | ROCm : NO_VMM = 1 | PEER_MAX_BATCH_SIZE = 128 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | AVX2 = 1 | F16C = 1 | FMA = 1 | BMI2 = 1 | LLAMAFILE = 1 | OPENMP = 1 | REPACK = 1 |
perplexity: tokenizing the input ..
perplexity: tokenization took 38.711 ms
perplexity: calculating perplexity over 58 chunks, n_ctx=512, batch_size=2048, n_seq=4
perplexity: 31.82 seconds per pass - ETA 7.68 minutes
[1]26.7955,[2]31.6588,[3]24.8119,[4]22.6109,[5]23.1167,[6]23.3437,[7]22.9963,[8]22.8649,[9]21.0063,[10]21.1354,[11]22.2279,[12]21.7067,[13]21.2408,[14]21.4859,[15]20.6131,[16]20.8148,[17]21.3397,[18]21.2674,[19]21.5771,[20]21.3695,[21]20.9652,[22]21.0828,[23]20.9136,[24]21.0333,[25]20.5873,[26]21.1063,[27]20.6983,[28]20.7043,[29]20.3499,[30]20.5463,[31]20.8592,[32]20.7545,[33]20.8546,[34]20.6528,[35]20.8017,[36]21.1070,[37]21.2734,[38]21.3185,[39]21.4554,[40]21.4667,[41]21.6259,[42]21.6157,[43]21.2904,[44]21.5241,[45]21.3381,[46]21.4185,[47]21.5686,[48]21.5082,[49]21.4718,[50]21.8603,[51]22.0679,[52]22.2694,[53]22.4607,[54]22.6324,[55]22.6263,[56]22.2750,[57]21.8535,[58]22.1226,
Final estimate: PPL = 22.1226 +/- 0.63860
llama_perf_context_print: load time = 37772.69 ms
llama_perf_context_print: prompt eval time = 447483.39 ms / 29696 tokens ( 15.07 ms per token, 66.36 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 454384.67 ms / 29697 tokens
llama_perf_context_print: graphs reused = 0
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment