A family of efficient AI models under 10B parameters performant in science, math, and coding through innovative training techniques.
1b
3b
7b
10b
4,685 Pulls Updated 5 days ago
c86ec0ca715f · 6.5GB
-
general.architecturellama
-
general.base_model.0.nameFalcon3 3B Instruct
-
general.base_model.0.organizationTiiuae
-
general.base_model.0.repo_urlhttps://huggingface.co/tiiuae/Falcon3-3B-Instruct
-
general.base_model.count1
-
general.basenameFalcon3
-
general.file_type1
-
general.finetuneInstruct
-
general.languages[en, fr, es, pt]
-
general.licenseother
-
general.license.linkhttps://falconllm.tii.ae/falcon-terms-and-conditions.html
-
general.license.namefalcon-llm-license
-
general.nameFalcon3 3B Instruct
-
general.quantization_version2
-
general.size_label3B
-
general.tags[falcon3]
-
general.typemodel
-
llama.attention.head_count12
-
llama.attention.head_count_kv4
-
llama.attention.key_length256
-
llama.attention.layer_norm_rms_epsilon1e-06
-
llama.attention.value_length256
-
llama.block_count22
-
llama.context_length32768
-
llama.embedding_length3072
-
llama.feed_forward_length9216
-
llama.rope.dimension_count256
-
llama.rope.freq_base1.000042e+06
-
llama.vocab_size131072
-
tokenizer.ggml.add_space_prefixfalse
-
tokenizer.ggml.eos_token_id11
-
tokenizer.ggml.merges[N E, Ġ Ġ, Ġ t, Ġ a, > >, ...]
-
tokenizer.ggml.modelgpt2
-
tokenizer.ggml.padding_token_id2023
-
tokenizer.ggml.prefalcon3
-
tokenizer.ggml.token_type[3, 3, 3, 3, 3, ...]
-
tokenizer.ggml.tokens[>>TITLE<<, >>ABSTRACT<<, >>INTRODUCTION<<, >>SUMMARY<<, >>COMMENT<<, ...]
-
NameTypeShape
-
token_embd.weightF16[3072, 131072]
-
blk.0.attn_k.weightF16[3072, 1024]
-
blk.0.attn_norm.weightF32[3072]
-
blk.0.attn_output.weightF16[3072, 3072]
-
blk.0.attn_q.weightF16[3072, 3072]
-
blk.0.attn_v.weightF16[3072, 1024]
-
blk.0.ffn_down.weightF16[9216, 3072]
-
blk.0.ffn_gate.weightF16[3072, 9216]
-
blk.0.ffn_norm.weightF32[3072]
-
blk.0.ffn_up.weightF16[3072, 9216]
-
blk.1.attn_k.weightF16[3072, 1024]
-
blk.1.attn_norm.weightF32[3072]
-
blk.1.attn_output.weightF16[3072, 3072]
-
blk.1.attn_q.weightF16[3072, 3072]
-
blk.1.attn_v.weightF16[3072, 1024]
-
blk.1.ffn_down.weightF16[9216, 3072]
-
blk.1.ffn_gate.weightF16[3072, 9216]
-
blk.1.ffn_norm.weightF32[3072]
-
blk.1.ffn_up.weightF16[3072, 9216]
-
blk.2.attn_k.weightF16[3072, 1024]
-
blk.2.attn_norm.weightF32[3072]
-
blk.2.attn_output.weightF16[3072, 3072]
-
blk.2.attn_q.weightF16[3072, 3072]
-
blk.2.attn_v.weightF16[3072, 1024]
-
blk.2.ffn_down.weightF16[9216, 3072]
-
blk.2.ffn_gate.weightF16[3072, 9216]
-
blk.2.ffn_norm.weightF32[3072]
-
blk.2.ffn_up.weightF16[3072, 9216]
-
blk.3.attn_k.weightF16[3072, 1024]
-
blk.3.attn_norm.weightF32[3072]
-
blk.3.attn_output.weightF16[3072, 3072]
-
blk.3.attn_q.weightF16[3072, 3072]
-
blk.3.attn_v.weightF16[3072, 1024]
-
blk.3.ffn_down.weightF16[9216, 3072]
-
blk.3.ffn_gate.weightF16[3072, 9216]
-
blk.3.ffn_norm.weightF32[3072]
-
blk.3.ffn_up.weightF16[3072, 9216]
-
blk.4.attn_k.weightF16[3072, 1024]
-
blk.4.attn_norm.weightF32[3072]
-
blk.4.attn_output.weightF16[3072, 3072]
-
blk.4.attn_q.weightF16[3072, 3072]
-
blk.4.attn_v.weightF16[3072, 1024]
-
blk.4.ffn_down.weightF16[9216, 3072]
-
blk.4.ffn_gate.weightF16[3072, 9216]
-
blk.4.ffn_norm.weightF32[3072]
-
blk.4.ffn_up.weightF16[3072, 9216]
-
blk.5.attn_k.weightF16[3072, 1024]
-
blk.5.attn_norm.weightF32[3072]
-
blk.5.attn_output.weightF16[3072, 3072]
-
blk.5.attn_q.weightF16[3072, 3072]
-
blk.5.attn_v.weightF16[3072, 1024]
-
blk.5.ffn_down.weightF16[9216, 3072]
-
blk.5.ffn_gate.weightF16[3072, 9216]
-
blk.5.ffn_norm.weightF32[3072]
-
blk.5.ffn_up.weightF16[3072, 9216]
-
blk.6.attn_k.weightF16[3072, 1024]
-
blk.6.attn_norm.weightF32[3072]
-
blk.6.attn_output.weightF16[3072, 3072]
-
blk.6.attn_q.weightF16[3072, 3072]
-
blk.6.attn_v.weightF16[3072, 1024]
-
blk.6.ffn_down.weightF16[9216, 3072]
-
blk.6.ffn_gate.weightF16[3072, 9216]
-
blk.6.ffn_norm.weightF32[3072]
-
blk.6.ffn_up.weightF16[3072, 9216]
-
blk.7.attn_k.weightF16[3072, 1024]
-
blk.7.attn_norm.weightF32[3072]
-
blk.7.attn_output.weightF16[3072, 3072]
-
blk.7.attn_q.weightF16[3072, 3072]
-
blk.7.attn_v.weightF16[3072, 1024]
-
blk.7.ffn_down.weightF16[9216, 3072]
-
blk.7.ffn_gate.weightF16[3072, 9216]
-
blk.7.ffn_norm.weightF32[3072]
-
blk.7.ffn_up.weightF16[3072, 9216]
-
blk.8.attn_k.weightF16[3072, 1024]
-
blk.8.attn_norm.weightF32[3072]
-
blk.8.attn_output.weightF16[3072, 3072]
-
blk.8.attn_q.weightF16[3072, 3072]
-
blk.8.attn_v.weightF16[3072, 1024]
-
blk.8.ffn_down.weightF16[9216, 3072]
-
blk.8.ffn_gate.weightF16[3072, 9216]
-
blk.8.ffn_norm.weightF32[3072]
-
blk.8.ffn_up.weightF16[3072, 9216]
-
blk.9.attn_k.weightF16[3072, 1024]
-
blk.9.attn_norm.weightF32[3072]
-
blk.9.attn_output.weightF16[3072, 3072]
-
blk.9.attn_q.weightF16[3072, 3072]
-
blk.9.attn_v.weightF16[3072, 1024]
-
blk.9.ffn_down.weightF16[9216, 3072]
-
blk.9.ffn_gate.weightF16[3072, 9216]
-
blk.9.ffn_norm.weightF32[3072]
-
blk.9.ffn_up.weightF16[3072, 9216]
-
blk.10.attn_k.weightF16[3072, 1024]
-
blk.10.attn_norm.weightF32[3072]
-
blk.10.attn_output.weightF16[3072, 3072]
-
blk.10.attn_q.weightF16[3072, 3072]
-
blk.10.attn_v.weightF16[3072, 1024]
-
blk.10.ffn_down.weightF16[9216, 3072]
-
blk.10.ffn_gate.weightF16[3072, 9216]
-
blk.10.ffn_norm.weightF32[3072]
-
blk.10.ffn_up.weightF16[3072, 9216]
-
blk.11.attn_k.weightF16[3072, 1024]
-
blk.11.attn_norm.weightF32[3072]
-
blk.11.attn_output.weightF16[3072, 3072]
-
blk.11.attn_q.weightF16[3072, 3072]
-
blk.11.attn_v.weightF16[3072, 1024]
-
blk.11.ffn_down.weightF16[9216, 3072]
-
blk.11.ffn_gate.weightF16[3072, 9216]
-
blk.11.ffn_norm.weightF32[3072]
-
blk.11.ffn_up.weightF16[3072, 9216]
-
blk.12.attn_k.weightF16[3072, 1024]
-
blk.12.attn_norm.weightF32[3072]
-
blk.12.attn_output.weightF16[3072, 3072]
-
blk.12.attn_q.weightF16[3072, 3072]
-
blk.12.attn_v.weightF16[3072, 1024]
-
blk.12.ffn_down.weightF16[9216, 3072]
-
blk.12.ffn_gate.weightF16[3072, 9216]
-
blk.12.ffn_norm.weightF32[3072]
-
blk.12.ffn_up.weightF16[3072, 9216]
-
blk.13.attn_k.weightF16[3072, 1024]
-
blk.13.attn_norm.weightF32[3072]
-
blk.13.attn_output.weightF16[3072, 3072]
-
blk.13.attn_q.weightF16[3072, 3072]
-
blk.13.attn_v.weightF16[3072, 1024]
-
blk.13.ffn_down.weightF16[9216, 3072]
-
blk.13.ffn_gate.weightF16[3072, 9216]
-
blk.13.ffn_norm.weightF32[3072]
-
blk.13.ffn_up.weightF16[3072, 9216]
-
blk.14.attn_k.weightF16[3072, 1024]
-
blk.14.attn_norm.weightF32[3072]
-
blk.14.attn_output.weightF16[3072, 3072]
-
blk.14.attn_q.weightF16[3072, 3072]
-
blk.14.attn_v.weightF16[3072, 1024]
-
blk.14.ffn_down.weightF16[9216, 3072]
-
blk.14.ffn_gate.weightF16[3072, 9216]
-
blk.14.ffn_norm.weightF32[3072]
-
blk.14.ffn_up.weightF16[3072, 9216]
-
blk.15.attn_k.weightF16[3072, 1024]
-
blk.15.attn_norm.weightF32[3072]
-
blk.15.attn_output.weightF16[3072, 3072]
-
blk.15.attn_q.weightF16[3072, 3072]
-
blk.15.attn_v.weightF16[3072, 1024]
-
blk.15.ffn_down.weightF16[9216, 3072]
-
blk.15.ffn_gate.weightF16[3072, 9216]
-
blk.15.ffn_norm.weightF32[3072]
-
blk.15.ffn_up.weightF16[3072, 9216]
-
blk.16.attn_k.weightF16[3072, 1024]
-
blk.16.attn_norm.weightF32[3072]
-
blk.16.attn_output.weightF16[3072, 3072]
-
blk.16.attn_q.weightF16[3072, 3072]
-
blk.16.attn_v.weightF16[3072, 1024]
-
blk.16.ffn_down.weightF16[9216, 3072]
-
blk.16.ffn_gate.weightF16[3072, 9216]
-
blk.16.ffn_norm.weightF32[3072]
-
blk.16.ffn_up.weightF16[3072, 9216]
-
blk.17.attn_k.weightF16[3072, 1024]
-
blk.17.attn_norm.weightF32[3072]
-
blk.17.attn_output.weightF16[3072, 3072]
-
blk.17.attn_q.weightF16[3072, 3072]
-
blk.17.attn_v.weightF16[3072, 1024]
-
blk.17.ffn_down.weightF16[9216, 3072]
-
blk.17.ffn_gate.weightF16[3072, 9216]
-
blk.17.ffn_norm.weightF32[3072]
-
blk.17.ffn_up.weightF16[3072, 9216]
-
blk.18.attn_k.weightF16[3072, 1024]
-
blk.18.attn_norm.weightF32[3072]
-
blk.18.attn_output.weightF16[3072, 3072]
-
blk.18.attn_q.weightF16[3072, 3072]
-
blk.18.attn_v.weightF16[3072, 1024]
-
blk.18.ffn_down.weightF16[9216, 3072]
-
blk.18.ffn_gate.weightF16[3072, 9216]
-
blk.18.ffn_norm.weightF32[3072]
-
blk.18.ffn_up.weightF16[3072, 9216]
-
blk.19.attn_k.weightF16[3072, 1024]
-
blk.19.attn_norm.weightF32[3072]
-
blk.19.attn_output.weightF16[3072, 3072]
-
blk.19.attn_q.weightF16[3072, 3072]
-
blk.19.attn_v.weightF16[3072, 1024]
-
blk.19.ffn_down.weightF16[9216, 3072]
-
blk.19.ffn_gate.weightF16[3072, 9216]
-
blk.19.ffn_norm.weightF32[3072]
-
blk.19.ffn_up.weightF16[3072, 9216]
-
blk.20.attn_k.weightF16[3072, 1024]
-
blk.20.attn_norm.weightF32[3072]
-
blk.20.attn_output.weightF16[3072, 3072]
-
blk.20.attn_q.weightF16[3072, 3072]
-
blk.20.attn_v.weightF16[3072, 1024]
-
blk.20.ffn_down.weightF16[9216, 3072]
-
blk.20.ffn_gate.weightF16[3072, 9216]
-
blk.20.ffn_norm.weightF32[3072]
-
blk.20.ffn_up.weightF16[3072, 9216]
-
blk.21.attn_k.weightF16[3072, 1024]
-
blk.21.attn_norm.weightF32[3072]
-
blk.21.attn_output.weightF16[3072, 3072]
-
blk.21.attn_q.weightF16[3072, 3072]
-
blk.21.attn_v.weightF16[3072, 1024]
-
blk.21.ffn_down.weightF16[9216, 3072]
-
blk.21.ffn_gate.weightF16[3072, 9216]
-
blk.21.ffn_norm.weightF32[3072]
-
blk.21.ffn_up.weightF16[3072, 9216]
-
output.weightF16[3072, 131072]
-
output_norm.weightF32[3072]
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