Upload imatrix.log with huggingface_hub
Browse files- imatrix.log +139 -0
imatrix.log
ADDED
@@ -0,0 +1,139 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
main: build = 3086 (554c247c)
|
2 |
+
main: built with cc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0 for x86_64-linux-gnu
|
3 |
+
main: seed = 1717697289
|
4 |
+
llama_model_loader: loaded meta data with 21 key-value pairs and 339 tensors from Qwen2-7B-Instruct-IMat-GGUF/Qwen2-7B-Instruct.gguf (version GGUF V3 (latest))
|
5 |
+
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
|
6 |
+
llama_model_loader: - kv 0: general.architecture str = qwen2
|
7 |
+
llama_model_loader: - kv 1: general.name str = Qwen2-7B-Instruct
|
8 |
+
llama_model_loader: - kv 2: qwen2.block_count u32 = 28
|
9 |
+
llama_model_loader: - kv 3: qwen2.context_length u32 = 32768
|
10 |
+
llama_model_loader: - kv 4: qwen2.embedding_length u32 = 3584
|
11 |
+
llama_model_loader: - kv 5: qwen2.feed_forward_length u32 = 18944
|
12 |
+
llama_model_loader: - kv 6: qwen2.attention.head_count u32 = 28
|
13 |
+
llama_model_loader: - kv 7: qwen2.attention.head_count_kv u32 = 4
|
14 |
+
llama_model_loader: - kv 8: qwen2.rope.freq_base f32 = 1000000.000000
|
15 |
+
llama_model_loader: - kv 9: qwen2.attention.layer_norm_rms_epsilon f32 = 0.000001
|
16 |
+
llama_model_loader: - kv 10: general.file_type u32 = 0
|
17 |
+
llama_model_loader: - kv 11: tokenizer.ggml.model str = gpt2
|
18 |
+
llama_model_loader: - kv 12: tokenizer.ggml.pre str = qwen2
|
19 |
+
llama_model_loader: - kv 13: tokenizer.ggml.tokens arr[str,152064] = ["!", "\"", "#", "$", "%", "&", "'", ...
|
20 |
+
llama_model_loader: - kv 14: tokenizer.ggml.token_type arr[i32,152064] = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
|
21 |
+
llama_model_loader: - kv 15: tokenizer.ggml.merges arr[str,151387] = ["Ġ Ġ", "ĠĠ ĠĠ", "i n", "Ġ t",...
|
22 |
+
llama_model_loader: - kv 16: tokenizer.ggml.eos_token_id u32 = 151645
|
23 |
+
llama_model_loader: - kv 17: tokenizer.ggml.padding_token_id u32 = 151643
|
24 |
+
llama_model_loader: - kv 18: tokenizer.ggml.bos_token_id u32 = 151643
|
25 |
+
llama_model_loader: - kv 19: tokenizer.chat_template str = {% for message in messages %}{% if lo...
|
26 |
+
llama_model_loader: - kv 20: general.quantization_version u32 = 2
|
27 |
+
llama_model_loader: - type f32: 339 tensors
|
28 |
+
llm_load_vocab: special tokens cache size = 421
|
29 |
+
llm_load_vocab: token to piece cache size = 0.9352 MB
|
30 |
+
llm_load_print_meta: format = GGUF V3 (latest)
|
31 |
+
llm_load_print_meta: arch = qwen2
|
32 |
+
llm_load_print_meta: vocab type = BPE
|
33 |
+
llm_load_print_meta: n_vocab = 152064
|
34 |
+
llm_load_print_meta: n_merges = 151387
|
35 |
+
llm_load_print_meta: n_ctx_train = 32768
|
36 |
+
llm_load_print_meta: n_embd = 3584
|
37 |
+
llm_load_print_meta: n_head = 28
|
38 |
+
llm_load_print_meta: n_head_kv = 4
|
39 |
+
llm_load_print_meta: n_layer = 28
|
40 |
+
llm_load_print_meta: n_rot = 128
|
41 |
+
llm_load_print_meta: n_embd_head_k = 128
|
42 |
+
llm_load_print_meta: n_embd_head_v = 128
|
43 |
+
llm_load_print_meta: n_gqa = 7
|
44 |
+
llm_load_print_meta: n_embd_k_gqa = 512
|
45 |
+
llm_load_print_meta: n_embd_v_gqa = 512
|
46 |
+
llm_load_print_meta: f_norm_eps = 0.0e+00
|
47 |
+
llm_load_print_meta: f_norm_rms_eps = 1.0e-06
|
48 |
+
llm_load_print_meta: f_clamp_kqv = 0.0e+00
|
49 |
+
llm_load_print_meta: f_max_alibi_bias = 0.0e+00
|
50 |
+
llm_load_print_meta: f_logit_scale = 0.0e+00
|
51 |
+
llm_load_print_meta: n_ff = 18944
|
52 |
+
llm_load_print_meta: n_expert = 0
|
53 |
+
llm_load_print_meta: n_expert_used = 0
|
54 |
+
llm_load_print_meta: causal attn = 1
|
55 |
+
llm_load_print_meta: pooling type = 0
|
56 |
+
llm_load_print_meta: rope type = 2
|
57 |
+
llm_load_print_meta: rope scaling = linear
|
58 |
+
llm_load_print_meta: freq_base_train = 1000000.0
|
59 |
+
llm_load_print_meta: freq_scale_train = 1
|
60 |
+
llm_load_print_meta: n_yarn_orig_ctx = 32768
|
61 |
+
llm_load_print_meta: rope_finetuned = unknown
|
62 |
+
llm_load_print_meta: ssm_d_conv = 0
|
63 |
+
llm_load_print_meta: ssm_d_inner = 0
|
64 |
+
llm_load_print_meta: ssm_d_state = 0
|
65 |
+
llm_load_print_meta: ssm_dt_rank = 0
|
66 |
+
llm_load_print_meta: model type = ?B
|
67 |
+
llm_load_print_meta: model ftype = all F32
|
68 |
+
llm_load_print_meta: model params = 7.62 B
|
69 |
+
llm_load_print_meta: model size = 28.37 GiB (32.00 BPW)
|
70 |
+
llm_load_print_meta: general.name = Qwen2-7B-Instruct
|
71 |
+
llm_load_print_meta: BOS token = 151643 '<|endoftext|>'
|
72 |
+
llm_load_print_meta: EOS token = 151645 '<|im_end|>'
|
73 |
+
llm_load_print_meta: PAD token = 151643 '<|endoftext|>'
|
74 |
+
llm_load_print_meta: LF token = 148848 'ÄĬ'
|
75 |
+
llm_load_print_meta: EOT token = 151645 '<|im_end|>'
|
76 |
+
ggml_cuda_init: GGML_CUDA_FORCE_MMQ: no
|
77 |
+
ggml_cuda_init: CUDA_USE_TENSOR_CORES: yes
|
78 |
+
ggml_cuda_init: found 1 CUDA devices:
|
79 |
+
Device 0: NVIDIA GeForce RTX 4090, compute capability 8.9, VMM: yes
|
80 |
+
llm_load_tensors: ggml ctx size = 0.32 MiB
|
81 |
+
llm_load_tensors: offloading 23 repeating layers to GPU
|
82 |
+
llm_load_tensors: offloaded 23/29 layers to GPU
|
83 |
+
llm_load_tensors: CPU buffer size = 29051.27 MiB
|
84 |
+
llm_load_tensors: CUDA0 buffer size = 20448.03 MiB
|
85 |
+
........................................................................................
|
86 |
+
llama_new_context_with_model: n_ctx = 512
|
87 |
+
llama_new_context_with_model: n_batch = 512
|
88 |
+
llama_new_context_with_model: n_ubatch = 512
|
89 |
+
llama_new_context_with_model: flash_attn = 0
|
90 |
+
llama_new_context_with_model: freq_base = 1000000.0
|
91 |
+
llama_new_context_with_model: freq_scale = 1
|
92 |
+
llama_kv_cache_init: CUDA_Host KV buffer size = 5.00 MiB
|
93 |
+
llama_kv_cache_init: CUDA0 KV buffer size = 23.00 MiB
|
94 |
+
llama_new_context_with_model: KV self size = 28.00 MiB, K (f16): 14.00 MiB, V (f16): 14.00 MiB
|
95 |
+
llama_new_context_with_model: CUDA_Host output buffer size = 0.58 MiB
|
96 |
+
llama_new_context_with_model: CUDA0 compute buffer size = 2383.00 MiB
|
97 |
+
llama_new_context_with_model: CUDA_Host compute buffer size = 8.01 MiB
|
98 |
+
llama_new_context_with_model: graph nodes = 986
|
99 |
+
llama_new_context_with_model: graph splits = 74
|
100 |
+
|
101 |
+
system_info: n_threads = 25 / 32 | AVX = 1 | AVX_VNNI = 0 | AVX2 = 1 | AVX512 = 1 | AVX512_VBMI = 1 | AVX512_VNNI = 1 | AVX512_BF16 = 1 | FMA = 1 | NEON = 0 | SVE = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 1 | SSE3 = 1 | SSSE3 = 1 | VSX = 0 | MATMUL_INT8 = 0 | LLAMAFILE = 1 |
|
102 |
+
compute_imatrix: tokenizing the input ..
|
103 |
+
compute_imatrix: tokenization took 133.474 ms
|
104 |
+
compute_imatrix: computing over 128 chunks with batch_size 512
|
105 |
+
compute_imatrix: 0.88 seconds per pass - ETA 1.87 minutes
|
106 |
+
[1]nan,[2]nan,[3]nan,[4]nan,[5]nan,[6]nan,[7]nan,[8]nan,[9]nan,
|
107 |
+
save_imatrix: stored collected data after 10 chunks in Qwen2-7B-Instruct-IMat-GGUF/imatrix.dat
|
108 |
+
[10]nan,[11]nan,[12]nan,[13]nan,[14]nan,[15]nan,[16]nan,[17]nan,[18]nan,[19]nan,
|
109 |
+
save_imatrix: stored collected data after 20 chunks in Qwen2-7B-Instruct-IMat-GGUF/imatrix.dat
|
110 |
+
[20]nan,[21]nan,[22]nan,[23]nan,[24]nan,[25]nan,[26]nan,[27]nan,[28]nan,[29]nan,
|
111 |
+
save_imatrix: stored collected data after 30 chunks in Qwen2-7B-Instruct-IMat-GGUF/imatrix.dat
|
112 |
+
[30]nan,[31]nan,[32]nan,[33]nan,[34]nan,[35]nan,[36]nan,[37]nan,[38]nan,[39]nan,
|
113 |
+
save_imatrix: stored collected data after 40 chunks in Qwen2-7B-Instruct-IMat-GGUF/imatrix.dat
|
114 |
+
[40]nan,[41]nan,[42]nan,[43]nan,[44]nan,[45]nan,[46]nan,[47]nan,[48]nan,[49]nan,
|
115 |
+
save_imatrix: stored collected data after 50 chunks in Qwen2-7B-Instruct-IMat-GGUF/imatrix.dat
|
116 |
+
[50]nan,[51]nan,[52]nan,[53]nan,[54]nan,[55]nan,[56]nan,[57]nan,[58]nan,[59]nan,
|
117 |
+
save_imatrix: stored collected data after 60 chunks in Qwen2-7B-Instruct-IMat-GGUF/imatrix.dat
|
118 |
+
[60]nan,[61]nan,[62]nan,[63]nan,[64]nan,[65]nan,[66]nan,[67]nan,[68]nan,[69]nan,
|
119 |
+
save_imatrix: stored collected data after 70 chunks in Qwen2-7B-Instruct-IMat-GGUF/imatrix.dat
|
120 |
+
[70]nan,[71]nan,[72]nan,[73]nan,[74]nan,[75]nan,[76]nan,[77]nan,[78]nan,[79]nan,
|
121 |
+
save_imatrix: stored collected data after 80 chunks in Qwen2-7B-Instruct-IMat-GGUF/imatrix.dat
|
122 |
+
[80]nan,[81]nan,[82]nan,[83]nan,[84]nan,[85]nan,[86]nan,[87]nan,[88]nan,[89]nan,
|
123 |
+
save_imatrix: stored collected data after 90 chunks in Qwen2-7B-Instruct-IMat-GGUF/imatrix.dat
|
124 |
+
[90]nan,[91]nan,[92]nan,[93]nan,[94]nan,[95]nan,[96]nan,[97]nan,[98]nan,[99]nan,
|
125 |
+
save_imatrix: stored collected data after 100 chunks in Qwen2-7B-Instruct-IMat-GGUF/imatrix.dat
|
126 |
+
[100]nan,[101]nan,[102]nan,[103]nan,[104]nan,[105]nan,[106]nan,[107]nan,[108]nan,[109]nan,
|
127 |
+
save_imatrix: stored collected data after 110 chunks in Qwen2-7B-Instruct-IMat-GGUF/imatrix.dat
|
128 |
+
[110]nan,[111]nan,[112]nan,[113]nan,[114]nan,[115]nan,[116]nan,[117]nan,[118]nan,[119]nan,
|
129 |
+
save_imatrix: stored collected data after 120 chunks in Qwen2-7B-Instruct-IMat-GGUF/imatrix.dat
|
130 |
+
[120]nan,[121]nan,[122]nan,[123]nan,[124]nan,[125]nan,[126]nan,[127]nan,[128]nan,
|
131 |
+
save_imatrix: stored collected data after 128 chunks in Qwen2-7B-Instruct-IMat-GGUF/imatrix.dat
|
132 |
+
|
133 |
+
llama_print_timings: load time = 2910.65 ms
|
134 |
+
llama_print_timings: sample time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
|
135 |
+
llama_print_timings: prompt eval time = 95408.80 ms / 65536 tokens ( 1.46 ms per token, 686.90 tokens per second)
|
136 |
+
llama_print_timings: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
|
137 |
+
llama_print_timings: total time = 98358.01 ms / 65537 tokens
|
138 |
+
|
139 |
+
Unexpected negative standard deviation of log(prob)
|