How do you infer?
#1
by
segmond
- opened
I don't think llama.cpp supports this model, do you have a branch I could use to run this?
hi, here is the WIP branch: https://github.com/ggerganov/llama.cpp/pull/7531
llm_load_print_meta: model ftype = Q3_K - Small
llm_load_print_meta: model params = 51.57 B
llm_load_print_meta: model size = 20.76 GiB (3.46 BPW)
llm_load_print_meta: general.name = ai21labs_AI21 Jamba 1.5 Mini
llm_load_print_meta: BOS token = 1 '<|startoftext|>'
llm_load_print_meta: EOS token = 2 '<|endoftext|>'
llm_load_print_meta: UNK token = 3 '<|unk|>'
llm_load_print_meta: PAD token = 0 '<|pad|>'
llm_load_print_meta: LF token = 1554 '<0x0A>'
llm_load_print_meta: EOT token = 2 '<|endoftext|>'
llm_load_print_meta: max token length = 96
llm_load_tensors: ggml ctx size = 0.22 MiB
llm_load_tensors: CPU buffer size = 21255.05 MiB
..............................................................
llama_new_context_with_model: n_ctx = 262144
llama_new_context_with_model: n_batch = 2048
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
llama_past_init: CPU past cache size = 4112.63 MiB
llama_new_context_with_model: SSM state size = 16.62 MiB, R (f32): 2.62 MiB, S (f32): 14.00 MiB
llama_new_context_with_model: KV cache size = 4096.00 MiB, K (f16): 2048.00 MiB, V (f16): 2048.00 MiB
llama_new_context_with_model: CPU output buffer size = 0.25 MiB
llama_new_context_with_model: CPU compute buffer size = 16920.60 MiB
llama_new_context_with_model: graph nodes = 2066
llama_new_context_with_model: graph splits = 1
system_info: n_threads = 16 (n_threads_batch = 16) / 32 | AVX = 1 | AVX_VNNI = 0 | AVX2 = 1 | AVX512 = 0 | AVX512_VBMI = 0 | AVX512_VNNI = 0 | AVX512_BF16 = 0 | FMA = 1 | NEON = 0 | SVE = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 0 | SSE3 = 1 | SSSE3 = 1 | VSX = 0 | MATMUL_INT8 = 0 | LLAMAFILE = 1 |
sampling:
repeat_last_n = 64, repeat_penalty = 1.000, frequency_penalty = 0.000, presence_penalty = 0.000
top_k = 40, tfs_z = 1.000, top_p = 0.950, min_p = 0.050, typical_p = 1.000, temp = 0.800
mirostat = 0, mirostat_lr = 0.100, mirostat_ent = 5.000
sampling order:
CFG -> Penalties -> top_k -> tfs_z -> typical_p -> top_p -> min_p -> temperature
generate: n_ctx = 262144, n_batch = 2048, n_predict = -1, n_keep = 1
here's an e-mail with a password," he said. "Now, let's check the email."
They logged into the email account. Sure enough, there was an email with a password.," he said. "Let's see if this works."
He entered the password into the lock. The lock clicked open.," he said, a look of awe on his face. "It worked!"
," he said, a look of awe on his face. "
llama_print_timings: load time = 3936.91 ms
llama_print_timings: sample time = 5.25 ms / 106 runs ( 0.05 ms per token, 20190.48 tokens per second)
llama_print_timings: prompt eval time = 439.11 ms / 6 tokens ( 73.19 ms per token, 13.66 tokens per second)
llama_print_timings: eval time = 20514.34 ms / 105 runs ( 195.37 ms per token, 5.12 tokens per second)
llama_print_timings: total time = 20991.83 ms / 111 tokens