Thank you for creating this.
Getting 6 tokens/s on ryzen thinkpad t495 with 33 layers offloaded to gpu (really? hmm)
$ llama-server -m /media/sdb1/Models/granite-3.1-3b-a800m-instruct_Q6_K.gguf -t 5 -c 32768 --chat-template granite -ngl 33
ggml_vulkan: Found 1 Vulkan devices:
ggml_vulkan: 0 = AMD Radeon Vega 8 Graphics (RADV RAVEN) (radv) | uma: 1 | fp16: 1 | warp size: 64 | shared memory: 65536 | int dot: 0 | matrix cores: none
build: 5346 (7f323a58) with cc (Debian 12.2.0-14) 12.2.0 for x86_64-linux-gnu
system info: n_threads = 5, n_threads_batch = 5, total_threads = 8
system_info: n_threads = 5 (n_threads_batch = 5) / 8 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | AVX2 = 1 | F16C = 1 | FMA = 1 | BMI2 = 1 | LLAMAFILE = 1 | OPENMP = 1 | AARCH64_REPACK = 1 |
main: binding port with default address family
main: HTTP server is listening, hostname: 127.0.0.1, port: 8080, http threads: 7
main: loading model
srv load_model: loading model '/media/sdb1/Models/granite-3.1-3b-a800m-instruct_Q6_K.gguf'
llama_model_load_from_file_impl: using device Vulkan0 (AMD Radeon Vega 8 Graphics (RADV RAVEN)) - 5994 MiB free
llama_model_loader: loaded meta data with 36 key-value pairs and 322 tensors from /media/sdb1/Models/granite-3.1-3b-a800m-instruct_Q6_K.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 = granitemoe
llama_model_loader: - kv 1: general.type str = model
llama_model_loader: - kv 2: general.name str = Model Ibm Granite Granite 3.1 3b A800...
llama_model_loader: - kv 3: general.finetune str = instruct-cache
llama_model_loader: - kv 4: general.basename str = model-ibm-granite-granite-3.1
llama_model_loader: - kv 5: general.size_label str = 3B-a800M
llama_model_loader: - kv 6: granitemoe.block_count u32 = 32
llama_model_loader: - kv 7: granitemoe.context_length u32 = 131072
llama_model_loader: - kv 8: granitemoe.embedding_length u32 = 1536
llama_model_loader: - kv 9: granitemoe.feed_forward_length u32 = 512
llama_model_loader: - kv 10: granitemoe.attention.head_count u32 = 24
llama_model_loader: - kv 11: granitemoe.attention.head_count_kv u32 = 8
llama_model_loader: - kv 12: granitemoe.rope.freq_base f32 = 10000000.000000
llama_model_loader: - kv 13: granitemoe.attention.layer_norm_rms_epsilon f32 = 0.000001
llama_model_loader: - kv 14: granitemoe.expert_count u32 = 40
llama_model_loader: - kv 15: granitemoe.expert_used_count u32 = 8
llama_model_loader: - kv 16: general.file_type u32 = 18
llama_model_loader: - kv 17: granitemoe.vocab_size u32 = 49155
llama_model_loader: - kv 18: granitemoe.rope.dimension_count u32 = 64
llama_model_loader: - kv 19: granitemoe.attention.scale f32 = 0.015625
llama_model_loader: - kv 20: granitemoe.embedding_scale f32 = 12.000000
llama_model_loader: - kv 21: granitemoe.residual_scale f32 = 0.220000
llama_model_loader: - kv 22: granitemoe.logit_scale f32 = 6.000000
llama_model_loader: - kv 23: tokenizer.ggml.model str = gpt2
llama_model_loader: - kv 24: tokenizer.ggml.pre str = refact
llama_model_loader: - kv 25: tokenizer.ggml.tokens arr[str,49155] = ["<|end_of_text|>", "", "...
llama_model_loader: - kv 26: tokenizer.ggml.token_type arr[i32,49155] = [3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, ...
llama_model_loader: - kv 27: tokenizer.ggml.merges arr[str,48891] = ["Ġ Ġ", "ĠĠ ĠĠ", "ĠĠĠĠ ĠĠ...
llama_model_loader: - kv 28: tokenizer.ggml.bos_token_id u32 = 0
llama_model_loader: - kv 29: tokenizer.ggml.eos_token_id u32 = 0
llama_model_loader: - kv 30: tokenizer.ggml.unknown_token_id u32 = 0
llama_model_loader: - kv 31: tokenizer.ggml.padding_token_id u32 = 0
llama_model_loader: - kv 32: tokenizer.ggml.add_bos_token bool = false
llama_model_loader: - kv 33: tokenizer.chat_template str = {%- if messages[0]['role'] == 'system...
llama_model_loader: - kv 34: tokenizer.ggml.add_space_prefix bool = false
llama_model_loader: - kv 35: general.quantization_version u32 = 2
llama_model_loader: - type f32: 97 tensors
llama_model_loader: - type q6_K: 225 tensors
print_info: file format = GGUF V3 (latest)
print_info: file type = Q6_K
print_info: file size = 2.53 GiB (6.58 BPW)
load: special_eos_id is not in special_eog_ids - the tokenizer config may be incorrect
load: special tokens cache size = 22
load: token to piece cache size = 0.2826 MB
print_info: arch = granitemoe
print_info: vocab_only = 0
print_info: n_ctx_train = 131072
print_info: n_embd = 1536
print_info: n_layer = 32
print_info: n_head = 24
print_info: n_head_kv = 8
print_info: n_rot = 64
print_info: n_swa = 0
print_info: n_swa_pattern = 1
print_info: n_embd_head_k = 64
print_info: n_embd_head_v = 64
print_info: n_gqa = 3
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 = 6.0e+00
print_info: f_attn_scale = 1.6e-02
print_info: n_ff = 512
print_info: n_expert = 40
print_info: n_expert_used = 8
print_info: causal attn = 1
print_info: pooling type = 0
print_info: rope type = 0
print_info: rope scaling = linear
print_info: freq_base_train = 10000000.0
print_info: freq_scale_train = 1
print_info: n_ctx_orig_yarn = 131072
print_info: rope_finetuned = unknown
print_info: ssm_d_conv = 0
print_info: ssm_d_inner = 0
print_info: ssm_d_state = 0
print_info: ssm_dt_rank = 0
print_info: ssm_dt_b_c_rms = 0
print_info: model type = 3B
print_info: model params = 3.30 B
print_info: general.name = Model Ibm Granite Granite 3.1 3b A800M Instruct Cache
print_info: f_embedding_scale = 12.000000
print_info: f_residual_scale = 0.220000
print_info: f_attention_scale = 0.015625
print_info: vocab type = BPE
print_info: n_vocab = 49155
print_info: n_merges = 48891
print_info: BOS token = 0 '<|end_of_text|>'
print_info: EOS token = 0 '<|end_of_text|>'
print_info: UNK token = 0 '<|end_of_text|>'
print_info: PAD token = 0 '<|end_of_text|>'
print_info: LF token = 203 'Ċ'
print_info: FIM PRE token = 1 ''
print_info: FIM SUF token = 3 ''
print_info: FIM MID token = 2 ''
print_info: FIM PAD token = 4 ''
print_info: FIM REP token = 18 ''
print_info: EOG token = 0 '<|end_of_text|>'
print_info: EOG token = 4 ''
print_info: EOG token = 18 ''
print_info: max token length = 512
load_tensors: loading model tensors, this can take a while... (mmap = true)
load_tensors: offloading 32 repeating layers to GPU
load_tensors: offloading output layer to GPU
load_tensors: offloaded 33/33 layers to GPU
load_tensors: Vulkan0 model buffer size = 2586.95 MiB
load_tensors: CPU_Mapped model buffer size = 59.07 MiB
.................................................................................................
llama_context: constructing llama_context
llama_context: n_seq_max = 1
llama_context: n_ctx = 32768
llama_context: n_ctx_per_seq = 32768
llama_context: n_batch = 2048
llama_context: n_ubatch = 512
llama_context: causal_attn = 1
llama_context: flash_attn = 0
llama_context: freq_base = 10000000.0
llama_context: freq_scale = 1
llama_context: n_ctx_per_seq (32768) < n_ctx_train (131072) -- the full capacity of the model will not be utilized
llama_context: Vulkan_Host output buffer size = 0.19 MiB
llama_kv_cache_unified: kv_size = 32768, type_k = 'f16', type_v = 'f16', n_layer = 32, can_shift = 1, padding = 32
llama_kv_cache_unified: Vulkan0 KV buffer size = 2048.00 MiB
llama_kv_cache_unified: KV self size = 2048.00 MiB, K (f16): 1024.00 MiB, V (f16): 1024.00 MiB
llama_context: Vulkan0 compute buffer size = 1612.00 MiB
llama_context: Vulkan_Host compute buffer size = 67.01 MiB
llama_context: graph nodes = 2024
llama_context: graph splits = 2
common_init_from_params: setting dry_penalty_last_n to ctx_size = 32768
common_init_from_params: warming up the model with an empty run - please wait ... (--no-warmup to disable)
Failed to infer a tool call example (possible template bug)
srv init: initializing slots, n_slots = 1
slot init: id 0 | task -1 | new slot n_ctx_slot = 32768
main: model loaded
main: chat template, chat_template: granite, example_format: '<|start_of_role|>system<|end_of_role|>You are a helpful assistant<|end_of_text|>
<|start_of_role|>user<|end_of_role|>Hello<|end_of_text|>
<|start_of_role|>assistant<|end_of_role|>Hi there<|end_of_text|>
<|start_of_role|>user<|end_of_role|>How are you?<|end_of_text|>
<|start_of_role|>assistant<|end_of_role|>
'
main: server is listening on http://127.0.0.1:8080 - starting the main loop
srv update_slots: all slots are idle
At the beginning i'm getting 8.5-9 t/s
It is amazing that this thing with 2.7GB knows more than most humans.