Add JAIS model(s) (#8118)
* Add `JAIS` model(s) * cleanup * address review comments * remove hack * un-hardcode max-alibi-bias * minor tweaks --------- Co-authored-by: fmz <quic_fzaghlou@quic.com>
This commit is contained in:
Faisal Zaghloul
GitHub
parent
023b8807e1
commit
968967376d
6 changed files with 288 additions and 9 deletions
169
src/llama.cpp
169
src/llama.cpp
|
|
@ -228,6 +228,7 @@ enum llm_arch {
|
|||
LLM_ARCH_DEEPSEEK2,
|
||||
LLM_ARCH_BITNET,
|
||||
LLM_ARCH_T5,
|
||||
LLM_ARCH_JAIS,
|
||||
LLM_ARCH_UNKNOWN,
|
||||
};
|
||||
|
||||
|
|
@ -269,6 +270,7 @@ static const std::map<llm_arch, const char *> LLM_ARCH_NAMES = {
|
|||
{ LLM_ARCH_DEEPSEEK2, "deepseek2" },
|
||||
{ LLM_ARCH_BITNET, "bitnet" },
|
||||
{ LLM_ARCH_T5, "t5" },
|
||||
{ LLM_ARCH_JAIS, "jais" },
|
||||
{ LLM_ARCH_UNKNOWN, "(unknown)" },
|
||||
};
|
||||
|
||||
|
|
@ -1236,6 +1238,21 @@ static const std::map<llm_arch, std::map<llm_tensor, std::string>> LLM_TENSOR_NA
|
|||
{ LLM_TENSOR_ENC_FFN_UP, "enc.blk.%d.ffn_up" },
|
||||
},
|
||||
},
|
||||
{
|
||||
LLM_ARCH_JAIS,
|
||||
{
|
||||
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
|
||||
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
|
||||
{ LLM_TENSOR_OUTPUT, "output" },
|
||||
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
|
||||
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
|
||||
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
|
||||
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
|
||||
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
|
||||
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
|
||||
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
|
||||
},
|
||||
},
|
||||
{
|
||||
LLM_ARCH_UNKNOWN,
|
||||
{
|
||||
|
|
@ -2035,6 +2052,7 @@ enum e_model {
|
|||
MODEL_410M,
|
||||
MODEL_0_5B,
|
||||
MODEL_1B,
|
||||
MODEL_1_3B,
|
||||
MODEL_1_4B,
|
||||
MODEL_2B,
|
||||
MODEL_2_8B,
|
||||
|
|
@ -4276,6 +4294,7 @@ static const char * llama_model_type_name(e_model type) {
|
|||
case MODEL_410M: return "410M";
|
||||
case MODEL_0_5B: return "0.5B";
|
||||
case MODEL_1B: return "1B";
|
||||
case MODEL_1_3B: return "1.3B";
|
||||
case MODEL_1_4B: return "1.4B";
|
||||
case MODEL_2B: return "2B";
|
||||
case MODEL_2_8B: return "2.8B";
|
||||
|
|
@ -4898,6 +4917,18 @@ static void llm_load_hparams(
|
|||
default: model.type = e_model::MODEL_UNKNOWN;
|
||||
}
|
||||
} break;
|
||||
case LLM_ARCH_JAIS:
|
||||
{
|
||||
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
|
||||
ml.get_key(LLM_KV_ATTENTION_MAX_ALIBI_BIAS, hparams.f_max_alibi_bias);
|
||||
|
||||
switch (hparams.n_layer) {
|
||||
case 24: model.type = e_model::MODEL_1_3B; break;
|
||||
case 40: model.type = e_model::MODEL_13B; break;
|
||||
/* TODO: add variants */
|
||||
default: model.type = e_model::MODEL_UNKNOWN;
|
||||
}
|
||||
} break;
|
||||
default: (void)0;
|
||||
}
|
||||
|
||||
|
|
@ -5129,6 +5160,9 @@ static void llm_load_vocab(
|
|||
} else if (
|
||||
tokenizer_pre == "viking") {
|
||||
vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_VIKING;
|
||||
} else if (
|
||||
tokenizer_pre == "jais") {
|
||||
vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_JAIS;
|
||||
} else {
|
||||
throw std::runtime_error(format("unknown pre-tokenizer type: '%s'", tokenizer_pre.c_str()));
|
||||
}
|
||||
|
|
@ -6962,6 +6996,44 @@ static bool llm_load_tensors(
|
|||
layer.ffn_up_scale = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_UP, "scale", i), {1});
|
||||
}
|
||||
} break;
|
||||
case LLM_ARCH_JAIS:
|
||||
{
|
||||
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
|
||||
|
||||
// Output
|
||||
{
|
||||
model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
|
||||
model.output_norm_b = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "bias"), {n_embd});
|
||||
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab});
|
||||
}
|
||||
for (int i = 0; i < n_layer; ++i) {
|
||||
ggml_context * ctx_layer = ctx_for_layer(i);
|
||||
ggml_context * ctx_split = ctx_for_layer_split(i);
|
||||
|
||||
auto & layer = model.layers[i];
|
||||
|
||||
layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
|
||||
layer.attn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "bias", i), {n_embd});
|
||||
|
||||
layer.wqkv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa});
|
||||
layer.bqkv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_QKV, "bias", i), {n_embd + 2*n_embd_gqa});
|
||||
|
||||
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});
|
||||
layer.bo = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd});
|
||||
|
||||
layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd});
|
||||
layer.ffn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "bias", i), {n_embd});
|
||||
|
||||
layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd});
|
||||
layer.ffn_down_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd});
|
||||
|
||||
layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff});
|
||||
layer.ffn_gate_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_GATE, "bias", i), {n_ff});
|
||||
|
||||
layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff});
|
||||
layer.ffn_up_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_UP, "bias", i), {n_ff});
|
||||
}
|
||||
} break;
|
||||
default:
|
||||
throw std::runtime_error("unknown architecture");
|
||||
}
|
||||
|
|
@ -12354,6 +12426,97 @@ struct llm_build_context {
|
|||
return gf;
|
||||
}
|
||||
|
||||
struct ggml_cgraph * build_jais() {
|
||||
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
|
||||
|
||||
const int64_t n_embd_head = hparams.n_embd_head_v;
|
||||
const int64_t n_embd_gqa = hparams.n_embd_v_gqa();
|
||||
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
|
||||
|
||||
struct ggml_tensor * cur;
|
||||
struct ggml_tensor * inpL;
|
||||
|
||||
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
|
||||
|
||||
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
|
||||
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
|
||||
|
||||
for (int il = 0; il < n_layer; ++il) {
|
||||
cur = llm_build_norm(ctx0, inpL, hparams,
|
||||
model.layers[il].attn_norm,
|
||||
model.layers[il].attn_norm_b,
|
||||
LLM_NORM, cb, il);
|
||||
cb(cur, "attn_norm", il);
|
||||
|
||||
// self-attention
|
||||
{
|
||||
cur = ggml_mul_mat(ctx0, model.layers[il].wqkv, cur);
|
||||
cb(cur, "wqkv", il);
|
||||
|
||||
cur = ggml_add(ctx0, cur, model.layers[il].bqkv);
|
||||
cb(cur, "bqkv", il);
|
||||
|
||||
struct ggml_tensor * Qcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd, n_tokens, cur->nb[1], 0*cur->nb[0]*(n_embd)));
|
||||
struct ggml_tensor * Kcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd_gqa, n_tokens, cur->nb[1], 1*cur->nb[0]*(n_embd)));
|
||||
struct ggml_tensor * Vcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd_gqa, n_tokens, cur->nb[1], 1*cur->nb[0]*(n_embd + n_embd_gqa)));
|
||||
|
||||
cb(Qcur, "Qcur", il);
|
||||
cb(Kcur, "Kcur", il);
|
||||
cb(Vcur, "Vcur", il);
|
||||
|
||||
Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens);
|
||||
|
||||
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
|
||||
model.layers[il].wo, model.layers[il].bo,
|
||||
Kcur, Vcur, Qcur, KQ_mask, n_tokens, kv_head, n_kv, 1.0f/float(n_embd_head), cb, il);
|
||||
}
|
||||
|
||||
if (il == n_layer - 1) {
|
||||
// skip computing output for unused tokens
|
||||
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
|
||||
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
|
||||
inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
|
||||
}
|
||||
|
||||
// add the input
|
||||
struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpL);
|
||||
cb(ffn_inp, "ffn_inp", il);
|
||||
|
||||
// FF
|
||||
{
|
||||
cur = llm_build_norm(ctx0, ffn_inp, hparams,
|
||||
model.layers[il].ffn_norm,
|
||||
model.layers[il].ffn_norm_b,
|
||||
LLM_NORM, cb, il);
|
||||
cb(cur, "ffn_norm", il);
|
||||
|
||||
cur = llm_build_ffn(ctx0, cur,
|
||||
model.layers[il].ffn_up, model.layers[il].ffn_up_b, NULL,
|
||||
model.layers[il].ffn_gate, model.layers[il].ffn_gate_b, NULL,
|
||||
model.layers[il].ffn_down, model.layers[il].ffn_down_b, NULL,
|
||||
NULL,
|
||||
LLM_FFN_SILU, LLM_FFN_PAR, cb, il);
|
||||
cb(cur, "ffn_out", il);
|
||||
}
|
||||
|
||||
inpL = ggml_add(ctx0, cur, ffn_inp);
|
||||
cb(inpL, "l_out", il);
|
||||
}
|
||||
|
||||
cur = llm_build_norm(ctx0, inpL, hparams,
|
||||
model.output_norm,
|
||||
model.output_norm_b,
|
||||
LLM_NORM, cb, -1);
|
||||
cb(cur, "result_norm", -1);
|
||||
|
||||
cur = ggml_mul_mat(ctx0, model.output, cur);
|
||||
|
||||
cb(cur, "result_output", -1);
|
||||
|
||||
ggml_build_forward_expand(gf, cur);
|
||||
|
||||
return gf;
|
||||
}
|
||||
};
|
||||
|
||||
static struct ggml_cgraph * llama_build_graph_defrag(llama_context & lctx, const std::vector<uint32_t> & ids) {
|
||||
|
|
@ -12585,6 +12748,10 @@ static struct ggml_cgraph * llama_build_graph(
|
|||
{
|
||||
result = llm.build_bitnet();
|
||||
} break;
|
||||
case LLM_ARCH_JAIS:
|
||||
{
|
||||
result = llm.build_jais();
|
||||
} break;
|
||||
default:
|
||||
GGML_ASSERT(false);
|
||||
}
|
||||
|
|
@ -13947,6 +14114,7 @@ struct llm_tokenizer_bpe {
|
|||
break;
|
||||
case LLAMA_VOCAB_PRE_TYPE_GPT2:
|
||||
case LLAMA_VOCAB_PRE_TYPE_OLMO:
|
||||
case LLAMA_VOCAB_PRE_TYPE_JAIS:
|
||||
regex_exprs = {
|
||||
"'s|'t|'re|'ve|'m|'ll|'d| ?\\p{L}+| ?\\p{N}+| ?[^\\s\\p{L}\\p{N}]+|\\s+(?!\\S)",
|
||||
};
|
||||
|
|
@ -17826,6 +17994,7 @@ enum llama_rope_type llama_rope_type(const struct llama_model * model) {
|
|||
case LLM_ARCH_MAMBA:
|
||||
case LLM_ARCH_JINA_BERT_V2:
|
||||
case LLM_ARCH_T5:
|
||||
case LLM_ARCH_JAIS:
|
||||
return LLAMA_ROPE_TYPE_NONE;
|
||||
|
||||
// use what we call a normal RoPE, operating on pairs of consecutive head values
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue