Bug: unsupported op 'MUL_MAT' on bf16 but not f16 on SmolLM #499
Description
Contact Details
What happened?
I am trying to convert the SmolLM models into llamafiles in bf16 and there is something weird happening. I thought I had this working earlier today but maybe that was in f16. Looks like @jart has this working in #495 but maybe that wasn't for metal?
ggml_metal_graph_compute_block_invoke: error: unsupported op 'MUL_MAT'
GGML_ASSERT: /Users/jts/.llamafile/v/0.8.9/ggml-metal.m:946: !"unsupported op"
It hangs here and the terminal must be force quitted. Ctrl-C does nothing
Here is my setup
python convert_hf_to_gguf.py SmolLM-135M --outfile smol-135-f16.gguf --outtype f16
python convert_hf_to_gguf.py SmolLM-135M --outfile smol-135-bf.gguf --outtype bf16
llamafile -m smol-135-f16.gguf # works
llamafile -m smol-135-bf.gguf # doesn't workFull logs in the "relevant log output" section
Version
llamafile v0.8.9
M2 MacBook Air 2023 Running Sonoma 14.0
What operating system are you seeing the problem on?
Mac
Relevant log output
(justine) jts@Jasons-MacBook-Air justine % llamafile -m smol-135-bf.gguf
Apple Metal GPU support successfully loaded
{"build":1500,"commit":"a30b324","function":"server_cli","level":"INFO","line":2869,"msg":"build info","tid":"34364712224","timestamp":1721684277}
{"function":"server_cli","level":"INFO","line":2872,"msg":"system info","n_threads":4,"n_threads_batch":-1,"system_info":"AVX = 0 | AVX_VNNI = 0 | AVX2 = 0 | AVX512 = 0 | AVX512_VBMI = 0 | AVX512_VNNI = 0 | AVX512_BF16 = 0 | FMA = 0 | NEON = 1 | ARM_FMA = 1 | F16C = 0 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 0 | SSE3 = 0 | SSSE3 = 0 | VSX = 0 | MATMUL_INT8 = 0 | LLAMAFILE = 1 | ","tid":"34364712224","timestamp":1721684277,"total_threads":8}
llama_model_loader: loaded meta data with 30 key-value pairs and 272 tensors from smol-135-bf.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 = llama
llama_model_loader: - kv 1: general.type str = model
llama_model_loader: - kv 2: general.name str = SmolLM 135M
llama_model_loader: - kv 3: general.basename str = SmolLM
llama_model_loader: - kv 4: general.size_label str = 135M
llama_model_loader: - kv 5: general.license str = apache-2.0
llama_model_loader: - kv 6: general.languages arr[str,1] = ["en"]
llama_model_loader: - kv 7: general.datasets arr[str,1] = ["HuggingFaceTB/smollm-corpus"]
llama_model_loader: - kv 8: llama.block_count u32 = 30
llama_model_loader: - kv 9: llama.context_length u32 = 2048
llama_model_loader: - kv 10: llama.embedding_length u32 = 576
llama_model_loader: - kv 11: llama.feed_forward_length u32 = 1536
llama_model_loader: - kv 12: llama.attention.head_count u32 = 9
llama_model_loader: - kv 13: llama.attention.head_count_kv u32 = 3
llama_model_loader: - kv 14: llama.rope.freq_base f32 = 10000.000000
llama_model_loader: - kv 15: llama.attention.layer_norm_rms_epsilon f32 = 0.000010
llama_model_loader: - kv 16: general.file_type u32 = 32
llama_model_loader: - kv 17: llama.vocab_size u32 = 49152
llama_model_loader: - kv 18: llama.rope.dimension_count u32 = 64
llama_model_loader: - kv 19: tokenizer.ggml.add_space_prefix bool = false
llama_model_loader: - kv 20: tokenizer.ggml.add_bos_token bool = false
llama_model_loader: - kv 21: tokenizer.ggml.model str = gpt2
llama_model_loader: - kv 22: tokenizer.ggml.pre str = smollm
llama_model_loader: - kv 23: tokenizer.ggml.tokens arr[str,49152] = ["<|endoftext|>", "<|im_start|>", "<|...
llama_model_loader: - kv 24: tokenizer.ggml.token_type arr[i32,49152] = [3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, ...
llama_model_loader: - kv 25: tokenizer.ggml.merges arr[str,48900] = ["Ġ t", "Ġ a", "i n", "h e", "Ġ Ġ...
llama_model_loader: - kv 26: tokenizer.ggml.bos_token_id u32 = 0
llama_model_loader: - kv 27: tokenizer.ggml.eos_token_id u32 = 0
llama_model_loader: - kv 28: tokenizer.ggml.unknown_token_id u32 = 0
llama_model_loader: - kv 29: general.quantization_version u32 = 2
llama_model_loader: - type f32: 61 tensors
llama_model_loader: - type bf16: 211 tensors
llm_load_vocab: special tokens definition check successful ( 17/49152 ).
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 = 49152
llm_load_print_meta: n_merges = 48900
llm_load_print_meta: n_ctx_train = 2048
llm_load_print_meta: n_embd = 576
llm_load_print_meta: n_head = 9
llm_load_print_meta: n_head_kv = 3
llm_load_print_meta: n_layer = 30
llm_load_print_meta: n_rot = 64
llm_load_print_meta: n_swa = 0
llm_load_print_meta: n_embd_head_k = 64
llm_load_print_meta: n_embd_head_v = 64
llm_load_print_meta: n_gqa = 3
llm_load_print_meta: n_embd_k_gqa = 192
llm_load_print_meta: n_embd_v_gqa = 192
llm_load_print_meta: f_norm_eps = 0.0e+00
llm_load_print_meta: f_norm_rms_eps = 1.0e-05
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 = 1536
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 = 10000.0
llm_load_print_meta: freq_scale_train = 1
llm_load_print_meta: n_yarn_orig_ctx = 2048
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 = BF16
llm_load_print_meta: model params = 134.52 M
llm_load_print_meta: model size = 256.63 MiB (16.00 BPW)
llm_load_print_meta: general.name = SmolLM 135M
llm_load_print_meta: BOS token = 0 '<|endoftext|>'
llm_load_print_meta: EOS token = 0 '<|endoftext|>'
llm_load_print_meta: UNK token = 0 '<|endoftext|>'
llm_load_print_meta: LF token = 143 'Ä'
llm_load_print_meta: EOT token = 2 '<|im_end|>'
llm_load_tensors: ggml ctx size = 0.32 MiB
ggml_backend_metal_log_allocated_size: allocated buffer, size = 202.66 MiB, ( 202.72 / 10922.67)
llm_load_tensors: offloading 30 repeating layers to GPU
llm_load_tensors: offloaded 30/31 layers to GPU
llm_load_tensors: Metal buffer size = 202.65 MiB
llm_load_tensors: CPU buffer size = 256.63 MiB
...................................................................
llama_new_context_with_model: n_ctx = 512
llama_new_context_with_model: n_batch = 512
llama_new_context_with_model: n_ubatch = 512
llama_new_context_with_model: flash_attn = 0
llama_new_context_with_model: freq_base = 10000.0
llama_new_context_with_model: freq_scale = 1
ggml_metal_init: allocating
ggml_metal_init: found device: Apple M2
ggml_metal_init: picking default device: Apple M2
ggml_metal_init: default.metallib not found, loading from source
ggml_metal_init: GGML_METAL_PATH_RESOURCES = nil
ggml_metal_init: loading '/Users/jts/.llamafile/v/0.8.9/ggml-metal.metal'
ggml_metal_init: GPU name: Apple M2
ggml_metal_init: GPU family: MTLGPUFamilyApple8 (1008)
ggml_metal_init: GPU family: MTLGPUFamilyCommon3 (3003)
ggml_metal_init: GPU family: MTLGPUFamilyMetal3 (5001)
ggml_metal_init: simdgroup reduction support = true
ggml_metal_init: simdgroup matrix mul. support = true
ggml_metal_init: hasUnifiedMemory = true
ggml_metal_init: recommendedMaxWorkingSetSize = 11453.25 MB
llama_kv_cache_init: Metal KV buffer size = 11.25 MiB
llama_new_context_with_model: KV self size = 11.25 MiB, K (f16): 5.62 MiB, V (f16): 5.62 MiB
llama_new_context_with_model: CPU output buffer size = 0.19 MiB
llama_new_context_with_model: Metal compute buffer size = 14.50 MiB
llama_new_context_with_model: CPU compute buffer size = 97.13 MiB
llama_new_context_with_model: graph nodes = 966
llama_new_context_with_model: graph splits = 3
ggml_metal_graph_compute_block_invoke: error: unsupported op 'MUL_MAT'
GGML_ASSERT: /Users/jts/.llamafile/v/0.8.9/ggml-metal.m:946: !"unsupported op"
ggml_metal_graph_compute_block_invoke: error: unsupported op 'MUL_MAT'
GGML_ASSERT: /Users/jts/.llamafile/v/0.8.9/ggml-metal.m:946: !"unsupported op"
ggml_metal_graph_compute_block_invoke: error: unsupported op 'MUL_MAT'
GGML_ASSERT: /Users/jts/.llamafile/v/0.8.9/ggml-metal.m:946: !"unsupported op"
ggml_metal_graph_compute_block_invoke: error: unsupported op 'MUL_MAT'
GGML_ASSERT: /Users/jts/.llamafile/v/0.8.9/ggml-metal.m:946: !"unsupported op"