The TinyLlama project is an open endeavor to train a compact 1.1B Llama model on 3 trillion tokens.
1.1b
380.2K Pulls Updated 10 months ago
530d13578278 · 483MB
-
general.architecturellama
-
general.file_type10
-
general.nameTinyLlama
-
general.quantization_version2
-
llama.attention.head_count32
-
llama.attention.head_count_kv4
-
llama.attention.layer_norm_rms_epsilon1e-05
-
llama.block_count22
-
llama.context_length2048
-
llama.embedding_length2048
-
llama.feed_forward_length5632
-
llama.rope.dimension_count64
-
llama.rope.freq_base10000
-
tokenizer.ggml.bos_token_id1
-
tokenizer.ggml.eos_token_id2
-
tokenizer.ggml.merges[▁ t, e r, i n, ▁ a, e n, ...]
-
tokenizer.ggml.modelllama
-
tokenizer.ggml.padding_token_id2
-
tokenizer.ggml.scores[0, 0, 0, 0, 0, ...]
-
tokenizer.ggml.token_type[2, 3, 3, 6, 6, ...]
-
tokenizer.ggml.tokens[<unk>, <s>, </s>, <0x00>, <0x01>, ...]
-
tokenizer.ggml.unknown_token_id0
-
NameTypeShape
-
token_embd.weightQ2_K[2048, 32000]
-
blk.0.attn_k.weightQ2_K[2048, 256]
-
blk.0.attn_norm.weightF32[2048]
-
blk.0.attn_output.weightQ3_K[2048, 2048]
-
blk.0.attn_q.weightQ2_K[2048, 2048]
-
blk.0.attn_v.weightQ3_K[2048, 256]
-
blk.0.ffn_down.weightQ3_K[5632, 2048]
-
blk.0.ffn_gate.weightQ3_K[2048, 5632]
-
blk.0.ffn_norm.weightF32[2048]
-
blk.0.ffn_up.weightQ3_K[2048, 5632]
-
blk.1.attn_k.weightQ2_K[2048, 256]
-
blk.1.attn_norm.weightF32[2048]
-
blk.1.attn_output.weightQ3_K[2048, 2048]
-
blk.1.attn_q.weightQ2_K[2048, 2048]
-
blk.1.attn_v.weightQ3_K[2048, 256]
-
blk.1.ffn_down.weightQ3_K[5632, 2048]
-
blk.1.ffn_gate.weightQ3_K[2048, 5632]
-
blk.1.ffn_norm.weightF32[2048]
-
blk.1.ffn_up.weightQ3_K[2048, 5632]
-
blk.2.attn_k.weightQ2_K[2048, 256]
-
blk.2.attn_norm.weightF32[2048]
-
blk.2.attn_output.weightQ3_K[2048, 2048]
-
blk.2.attn_q.weightQ2_K[2048, 2048]
-
blk.2.attn_v.weightQ3_K[2048, 256]
-
blk.2.ffn_down.weightQ3_K[5632, 2048]
-
blk.2.ffn_gate.weightQ3_K[2048, 5632]
-
blk.2.ffn_norm.weightF32[2048]
-
blk.2.ffn_up.weightQ3_K[2048, 5632]
-
blk.3.attn_k.weightQ2_K[2048, 256]
-
blk.3.attn_norm.weightF32[2048]
-
blk.3.attn_output.weightQ3_K[2048, 2048]
-
blk.3.attn_q.weightQ2_K[2048, 2048]
-
blk.3.attn_v.weightQ3_K[2048, 256]
-
blk.3.ffn_down.weightQ3_K[5632, 2048]
-
blk.3.ffn_gate.weightQ3_K[2048, 5632]
-
blk.3.ffn_norm.weightF32[2048]
-
blk.3.ffn_up.weightQ3_K[2048, 5632]
-
blk.4.attn_k.weightQ2_K[2048, 256]
-
blk.4.attn_norm.weightF32[2048]
-
blk.4.attn_output.weightQ3_K[2048, 2048]
-
blk.4.attn_q.weightQ2_K[2048, 2048]
-
blk.4.attn_v.weightQ3_K[2048, 256]
-
blk.4.ffn_down.weightQ3_K[5632, 2048]
-
blk.4.ffn_gate.weightQ3_K[2048, 5632]
-
blk.4.ffn_norm.weightF32[2048]
-
blk.4.ffn_up.weightQ3_K[2048, 5632]
-
blk.5.attn_k.weightQ2_K[2048, 256]
-
blk.5.attn_norm.weightF32[2048]
-
blk.5.attn_output.weightQ3_K[2048, 2048]
-
blk.5.attn_q.weightQ2_K[2048, 2048]
-
blk.5.attn_v.weightQ3_K[2048, 256]
-
blk.5.ffn_down.weightQ3_K[5632, 2048]
-
blk.5.ffn_gate.weightQ3_K[2048, 5632]
-
blk.5.ffn_norm.weightF32[2048]
-
blk.5.ffn_up.weightQ3_K[2048, 5632]
-
blk.6.attn_k.weightQ2_K[2048, 256]
-
blk.6.attn_norm.weightF32[2048]
-
blk.6.attn_output.weightQ3_K[2048, 2048]
-
blk.6.attn_q.weightQ2_K[2048, 2048]
-
blk.6.attn_v.weightQ3_K[2048, 256]
-
blk.6.ffn_down.weightQ3_K[5632, 2048]
-
blk.6.ffn_gate.weightQ3_K[2048, 5632]
-
blk.6.ffn_norm.weightF32[2048]
-
blk.6.ffn_up.weightQ3_K[2048, 5632]
-
blk.7.attn_k.weightQ2_K[2048, 256]
-
blk.7.attn_norm.weightF32[2048]
-
blk.7.attn_output.weightQ3_K[2048, 2048]
-
blk.7.attn_q.weightQ2_K[2048, 2048]
-
blk.7.attn_v.weightQ3_K[2048, 256]
-
blk.7.ffn_down.weightQ3_K[5632, 2048]
-
blk.7.ffn_gate.weightQ3_K[2048, 5632]
-
blk.7.ffn_norm.weightF32[2048]
-
blk.7.ffn_up.weightQ3_K[2048, 5632]
-
blk.8.attn_k.weightQ2_K[2048, 256]
-
blk.8.attn_norm.weightF32[2048]
-
blk.8.attn_output.weightQ3_K[2048, 2048]
-
blk.8.attn_q.weightQ2_K[2048, 2048]
-
blk.8.attn_v.weightQ3_K[2048, 256]
-
blk.8.ffn_down.weightQ3_K[5632, 2048]
-
blk.8.ffn_gate.weightQ3_K[2048, 5632]
-
blk.8.ffn_norm.weightF32[2048]
-
blk.8.ffn_up.weightQ3_K[2048, 5632]
-
blk.9.attn_k.weightQ2_K[2048, 256]
-
blk.9.attn_norm.weightF32[2048]
-
blk.9.attn_output.weightQ3_K[2048, 2048]
-
blk.9.attn_q.weightQ2_K[2048, 2048]
-
blk.9.attn_v.weightQ3_K[2048, 256]
-
blk.9.ffn_down.weightQ3_K[5632, 2048]
-
blk.9.ffn_gate.weightQ3_K[2048, 5632]
-
blk.9.ffn_norm.weightF32[2048]
-
blk.9.ffn_up.weightQ3_K[2048, 5632]
-
blk.10.attn_k.weightQ2_K[2048, 256]
-
blk.10.attn_norm.weightF32[2048]
-
blk.10.attn_output.weightQ3_K[2048, 2048]
-
blk.10.attn_q.weightQ2_K[2048, 2048]
-
blk.10.attn_v.weightQ3_K[2048, 256]
-
blk.10.ffn_down.weightQ3_K[5632, 2048]
-
blk.10.ffn_gate.weightQ3_K[2048, 5632]
-
blk.10.ffn_norm.weightF32[2048]
-
blk.10.ffn_up.weightQ3_K[2048, 5632]
-
blk.11.attn_k.weightQ2_K[2048, 256]
-
blk.11.attn_norm.weightF32[2048]
-
blk.11.attn_output.weightQ3_K[2048, 2048]
-
blk.11.attn_q.weightQ2_K[2048, 2048]
-
blk.11.attn_v.weightQ3_K[2048, 256]
-
blk.11.ffn_down.weightQ3_K[5632, 2048]
-
blk.11.ffn_gate.weightQ3_K[2048, 5632]
-
blk.11.ffn_norm.weightF32[2048]
-
blk.11.ffn_up.weightQ3_K[2048, 5632]
-
blk.12.attn_k.weightQ2_K[2048, 256]
-
blk.12.attn_norm.weightF32[2048]
-
blk.12.attn_output.weightQ3_K[2048, 2048]
-
blk.12.attn_q.weightQ2_K[2048, 2048]
-
blk.12.attn_v.weightQ3_K[2048, 256]
-
blk.12.ffn_down.weightQ3_K[5632, 2048]
-
blk.12.ffn_gate.weightQ3_K[2048, 5632]
-
blk.12.ffn_norm.weightF32[2048]
-
blk.12.ffn_up.weightQ3_K[2048, 5632]
-
blk.13.attn_k.weightQ2_K[2048, 256]
-
blk.13.attn_norm.weightF32[2048]
-
blk.13.attn_output.weightQ3_K[2048, 2048]
-
blk.13.attn_q.weightQ2_K[2048, 2048]
-
blk.13.attn_v.weightQ3_K[2048, 256]
-
blk.13.ffn_down.weightQ3_K[5632, 2048]
-
blk.13.ffn_gate.weightQ3_K[2048, 5632]
-
blk.13.ffn_norm.weightF32[2048]
-
blk.13.ffn_up.weightQ3_K[2048, 5632]
-
blk.14.attn_k.weightQ2_K[2048, 256]
-
blk.14.attn_norm.weightF32[2048]
-
blk.14.attn_output.weightQ3_K[2048, 2048]
-
blk.14.attn_q.weightQ2_K[2048, 2048]
-
blk.14.attn_v.weightQ3_K[2048, 256]
-
blk.14.ffn_down.weightQ3_K[5632, 2048]
-
blk.14.ffn_gate.weightQ3_K[2048, 5632]
-
blk.14.ffn_norm.weightF32[2048]
-
blk.14.ffn_up.weightQ3_K[2048, 5632]
-
blk.15.attn_k.weightQ2_K[2048, 256]
-
blk.15.attn_norm.weightF32[2048]
-
blk.15.attn_output.weightQ3_K[2048, 2048]
-
blk.15.attn_q.weightQ2_K[2048, 2048]
-
blk.15.attn_v.weightQ3_K[2048, 256]
-
blk.15.ffn_down.weightQ3_K[5632, 2048]
-
blk.15.ffn_gate.weightQ3_K[2048, 5632]
-
blk.15.ffn_norm.weightF32[2048]
-
blk.15.ffn_up.weightQ3_K[2048, 5632]
-
blk.16.attn_k.weightQ2_K[2048, 256]
-
blk.16.attn_norm.weightF32[2048]
-
blk.16.attn_output.weightQ3_K[2048, 2048]
-
blk.16.attn_q.weightQ2_K[2048, 2048]
-
blk.16.attn_v.weightQ3_K[2048, 256]
-
blk.16.ffn_down.weightQ3_K[5632, 2048]
-
blk.16.ffn_gate.weightQ3_K[2048, 5632]
-
blk.16.ffn_norm.weightF32[2048]
-
blk.16.ffn_up.weightQ3_K[2048, 5632]
-
blk.17.attn_k.weightQ2_K[2048, 256]
-
blk.17.attn_norm.weightF32[2048]
-
blk.17.attn_output.weightQ3_K[2048, 2048]
-
blk.17.attn_q.weightQ2_K[2048, 2048]
-
blk.17.attn_v.weightQ3_K[2048, 256]
-
blk.17.ffn_down.weightQ3_K[5632, 2048]
-
blk.17.ffn_gate.weightQ3_K[2048, 5632]
-
blk.17.ffn_norm.weightF32[2048]
-
blk.17.ffn_up.weightQ3_K[2048, 5632]
-
blk.18.attn_k.weightQ2_K[2048, 256]
-
blk.18.attn_norm.weightF32[2048]
-
blk.18.attn_output.weightQ3_K[2048, 2048]
-
blk.18.attn_q.weightQ2_K[2048, 2048]
-
blk.18.attn_v.weightQ3_K[2048, 256]
-
blk.18.ffn_down.weightQ3_K[5632, 2048]
-
blk.18.ffn_gate.weightQ3_K[2048, 5632]
-
blk.18.ffn_norm.weightF32[2048]
-
blk.18.ffn_up.weightQ3_K[2048, 5632]
-
blk.19.attn_k.weightQ2_K[2048, 256]
-
blk.19.attn_norm.weightF32[2048]
-
blk.19.attn_output.weightQ3_K[2048, 2048]
-
blk.19.attn_q.weightQ2_K[2048, 2048]
-
blk.19.attn_v.weightQ3_K[2048, 256]
-
blk.19.ffn_down.weightQ3_K[5632, 2048]
-
blk.19.ffn_gate.weightQ3_K[2048, 5632]
-
blk.19.ffn_norm.weightF32[2048]
-
blk.19.ffn_up.weightQ3_K[2048, 5632]
-
blk.20.attn_k.weightQ2_K[2048, 256]
-
blk.20.attn_norm.weightF32[2048]
-
blk.20.attn_output.weightQ3_K[2048, 2048]
-
blk.20.attn_q.weightQ2_K[2048, 2048]
-
blk.20.attn_v.weightQ3_K[2048, 256]
-
blk.20.ffn_down.weightQ3_K[5632, 2048]
-
blk.20.ffn_gate.weightQ3_K[2048, 5632]
-
blk.20.ffn_norm.weightF32[2048]
-
blk.20.ffn_up.weightQ3_K[2048, 5632]
-
blk.21.attn_k.weightQ2_K[2048, 256]
-
blk.21.attn_norm.weightF32[2048]
-
blk.21.attn_output.weightQ3_K[2048, 2048]
-
blk.21.attn_q.weightQ2_K[2048, 2048]
-
blk.21.attn_v.weightQ3_K[2048, 256]
-
blk.21.ffn_down.weightQ3_K[5632, 2048]
-
blk.21.ffn_gate.weightQ3_K[2048, 5632]
-
blk.21.ffn_norm.weightF32[2048]
-
blk.21.ffn_up.weightQ3_K[2048, 5632]
-
output.weightQ6_K[2048, 32000]
-
output_norm.weightF32[2048]
Metadata
Tensor
blk.0
blk.1
blk.2
blk.3
blk.4
blk.5
blk.6
blk.7
blk.8
blk.9
blk.10
blk.11
blk.12
blk.13
blk.14
blk.15
blk.16
blk.17
blk.18
blk.19
blk.20
blk.21