Public/koboldcpp
1
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-09-19 17:25:07 +02:00
Code Issues Packages Projects Releases Wiki Activity

Reworking rope WIP

Browse Source
This commit is contained in:
Concedo
2023-07-19 00:54:41 +08:00
parent 0a11f50da8 d01bccde9f
commit 374fffb9c6
24 changed files with 600 additions and 256 deletions
Download Patch File Download Diff File Expand all files Collapse all files
examples/common.cpp
+23
View File
@@ -168,6 +168,18 @@ bool gpt_params_parse(int argc, char ** argv, gpt_params & params) {
break;
}
params.n_ctx = std::stoi(argv[i]);
} else if (arg == "--rope-freq-base") {
if (++i >= argc) {
invalid_param = true;
break;
}
params.rope_freq_base = std::stof(argv[i]);
} else if (arg == "--rope-freq-scale") {
if (++i >= argc) {
invalid_param = true;
break;
}
params.rope_freq_scale = std::stof(argv[i]);
} else if (arg == "--memory-f32") {
params.memory_f16 = false;
} else if (arg == "--top-p") {
@@ -267,6 +279,12 @@ bool gpt_params_parse(int argc, char ** argv, gpt_params & params) {
break;
}
params.n_keep = std::stoi(argv[i]);
} else if (arg == "--chunks") {
if (++i >= argc) {
invalid_param = true;
break;
}
params.n_chunks = std::stoi(argv[i]);
} else if (arg == "-m" || arg == "--model") {
if (++i >= argc) {
invalid_param = true;
@@ -493,6 +511,8 @@ void gpt_print_usage(int /*argc*/, char ** argv, const gpt_params & params) {
fprintf(stderr, " --cfg-scale N strength of guidance (default: %f, 1.0 = disable)\n", params.cfg_scale);
fprintf(stderr, " --cfg-smooth-factor N smooth factor between old and new logits (default: %f, 1.0 = no smoothing)\n", params.cfg_smooth_factor);
fprintf(stderr, " -c N, --ctx-size N size of the prompt context (default: %d)\n", params.n_ctx);
fprintf(stderr, " --rope-freq-base N RoPE base frequency (default: %.1f)\n", params.rope_freq_base);
fprintf(stderr, " --rope-freq-scale N RoPE frequency scaling factor (default: %g)\n", params.rope_freq_scale);
fprintf(stderr, " --ignore-eos ignore end of stream token and continue generating (implies --logit-bias 2-inf)\n");
fprintf(stderr, " --no-penalize-nl do not penalize newline token\n");
fprintf(stderr, " --memory-f32 use f32 instead of f16 for memory key+value (default: disabled)\n");
@@ -501,6 +521,7 @@ void gpt_print_usage(int /*argc*/, char ** argv, const gpt_params & params) {
fprintf(stderr, " -b N, --batch-size N batch size for prompt processing (default: %d)\n", params.n_batch);
fprintf(stderr, " --perplexity compute perplexity over the prompt\n");
fprintf(stderr, " --keep number of tokens to keep from the initial prompt (default: %d, -1 = all)\n", params.n_keep);
fprintf(stderr, " --chunks N max number of chunks to process (default: %d, -1 = all)\n", params.n_chunks);
if (llama_mlock_supported()) {
fprintf(stderr, " --mlock force system to keep model in RAM rather than swapping or compressing\n");
}
@@ -573,6 +594,8 @@ struct llama_context_params llama_context_params_from_gpt_params(const gpt_param
lparams.use_mlock = params.use_mlock;
lparams.logits_all = params.perplexity;
lparams.embedding = params.embedding;
lparams.rope_freq_base = params.rope_freq_base;
lparams.rope_freq_scale = params.rope_freq_scale;
return lparams;
}
examples/common.h
+3
View File
@@ -28,10 +28,13 @@ struct gpt_params {
int32_t n_ctx = 512; // context size
int32_t n_batch = 512; // batch size for prompt processing (must be >=32 to use BLAS)
int32_t n_keep = 0; // number of tokens to keep from initial prompt
int32_t n_chunks = -1; // max number of chunks to process (-1 = unlimited)
int32_t n_gpu_layers = 0; // number of layers to store in VRAM
int32_t main_gpu = 0; // the GPU that is used for scratch and small tensors
float tensor_split[LLAMA_MAX_DEVICES] = {0}; // how split tensors should be distributed across GPUs
int32_t n_probs = 0; // if greater than 0, output the probabilities of top n_probs tokens.
float rope_freq_base = 10000.0f; // RoPE base frequency
float rope_freq_scale = 1.0f; // RoPE frequency scaling factor
// sampling parameters
std::unordered_map<llama_token, float> logit_bias; // logit bias for specific tokens
examples/main/main.cpp
+10 -2
View File
@@ -84,9 +84,17 @@ int main(int argc, char ** argv) {
return 0;
}
if (params.rope_freq_base != 10000.0) {
fprintf(stderr, "%s: warning: changing RoPE frequency base to %g (default 10000.0)\n", __func__, params.rope_freq_base);
}
if (params.rope_freq_scale != 1.0) {
fprintf(stderr, "%s: warning: scaling RoPE frequency by %g (default 1.0)\n", __func__, params.rope_freq_scale);
}
if (params.n_ctx > 2048) {
fprintf(stderr, "%s: warning: model might not support context sizes greater than 2048 tokens (%d specified);"
"expect poor results\n", __func__, params.n_ctx);
fprintf(stderr, "%s: warning: base model only supports context sizes no greater than 2048 tokens (%d specified);"
" you are on your own\n", __func__, params.n_ctx);
} else if (params.n_ctx < 8) {
fprintf(stderr, "%s: warning: minimum context size is 8, using minimum size.\n", __func__);
params.n_ctx = 8;
examples/perplexity/perplexity.cpp
+4 -2
View File
@@ -32,13 +32,15 @@ void perplexity(llama_context * ctx, const gpt_params & params) {
// BOS tokens will be added for each chunk before eval
auto tokens = ::llama_tokenize(ctx, params.prompt, true);
int count = 0;
const int n_chunk_max = tokens.size() / params.n_ctx;
const int n_chunk = tokens.size() / params.n_ctx;
const int n_chunk = params.n_chunks < 0 ? n_chunk_max : std::min(params.n_chunks, n_chunk_max);
const int n_vocab = llama_n_vocab(ctx);
const int n_batch = params.n_batch;
int count = 0;
double nll = 0.0;
fprintf(stderr, "%s: calculating perplexity over %d chunks, batch_size=%d\n", __func__, n_chunk, n_batch);
for (int i = 0; i < n_chunk; ++i) {
examples/quantize/quantize.cpp
+19 -95
View File
@@ -12,103 +12,27 @@ struct quant_option {
};
static const std::vector<struct quant_option> QUANT_OPTIONS = {
{
"Q4_0",
LLAMA_FTYPE_MOSTLY_Q4_0,
" 3.50G, +0.2499 ppl @ 7B - small, very high quality loss - legacy, prefer using Q3_K_M",
},
{
"Q4_1",
LLAMA_FTYPE_MOSTLY_Q4_1,
" 3.90G, +0.1846 ppl @ 7B - small, substantial quality loss - legacy, prefer using Q3_K_L",
},
{
"Q5_0",
LLAMA_FTYPE_MOSTLY_Q5_0,
" 4.30G, +0.0796 ppl @ 7B - medium, balanced quality - legacy, prefer using Q4_K_M",
},
{
"Q5_1",
LLAMA_FTYPE_MOSTLY_Q5_1,
" 4.70G, +0.0415 ppl @ 7B - medium, low quality loss - legacy, prefer using Q5_K_M",
},
{ "Q4_0", LLAMA_FTYPE_MOSTLY_Q4_0, " 3.50G, +0.2499 ppl @ 7B", },
{ "Q4_1", LLAMA_FTYPE_MOSTLY_Q4_1, " 3.90G, +0.1846 ppl @ 7B", },
{ "Q5_0", LLAMA_FTYPE_MOSTLY_Q5_0, " 4.30G, +0.0796 ppl @ 7B", },
{ "Q5_1", LLAMA_FTYPE_MOSTLY_Q5_1, " 4.70G, +0.0415 ppl @ 7B", },
#ifdef GGML_USE_K_QUANTS
{
"Q2_K",
LLAMA_FTYPE_MOSTLY_Q2_K,
" 2.67G, +0.8698 ppl @ 7B - smallest, extreme quality loss - not recommended",
},
{
"Q3_K",
LLAMA_FTYPE_MOSTLY_Q3_K_M,
"alias for Q3_K_M"
},
{
"Q3_K_S",
LLAMA_FTYPE_MOSTLY_Q3_K_S,
" 2.75G, +0.5505 ppl @ 7B - very small, very high quality loss",
},
{
"Q3_K_M",
LLAMA_FTYPE_MOSTLY_Q3_K_M,
" 3.06G, +0.2437 ppl @ 7B - very small, very high quality loss",
},
{
"Q3_K_L",
LLAMA_FTYPE_MOSTLY_Q3_K_L,
" 3.35G, +0.1803 ppl @ 7B - small, substantial quality loss",
},
{
"Q4_K",
LLAMA_FTYPE_MOSTLY_Q4_K_M,
"alias for Q4_K_M",
},
{
"Q4_K_S",
LLAMA_FTYPE_MOSTLY_Q4_K_S,
" 3.56G, +0.1149 ppl @ 7B - small, significant quality loss",
},
{
"Q4_K_M",
LLAMA_FTYPE_MOSTLY_Q4_K_M,
" 3.80G, +0.0535 ppl @ 7B - medium, balanced quality - *recommended*",
},
{
"Q5_K",
LLAMA_FTYPE_MOSTLY_Q5_K_M,
"alias for Q5_K_M",
},
{
"Q5_K_S",
LLAMA_FTYPE_MOSTLY_Q5_K_S,
" 4.33G, +0.0353 ppl @ 7B - large, low quality loss - *recommended*",
},
{
"Q5_K_M",
LLAMA_FTYPE_MOSTLY_Q5_K_M,
" 4.45G, +0.0142 ppl @ 7B - large, very low quality loss - *recommended*",
},
{
"Q6_K",
LLAMA_FTYPE_MOSTLY_Q6_K,
" 5.15G, +0.0044 ppl @ 7B - very large, extremely low quality loss",
},
{ "Q2_K", LLAMA_FTYPE_MOSTLY_Q2_K, " 2.67G, +0.8698 ppl @ 7B", },
{ "Q3_K", LLAMA_FTYPE_MOSTLY_Q3_K_M, "alias for Q3_K_M" },
{ "Q3_K_S", LLAMA_FTYPE_MOSTLY_Q3_K_S, " 2.75G, +0.5505 ppl @ 7B", },
{ "Q3_K_M", LLAMA_FTYPE_MOSTLY_Q3_K_M, " 3.06G, +0.2437 ppl @ 7B", },
{ "Q3_K_L", LLAMA_FTYPE_MOSTLY_Q3_K_L, " 3.35G, +0.1803 ppl @ 7B", },
{ "Q4_K", LLAMA_FTYPE_MOSTLY_Q4_K_M, "alias for Q4_K_M", },
{ "Q4_K_S", LLAMA_FTYPE_MOSTLY_Q4_K_S, " 3.56G, +0.1149 ppl @ 7B", },
{ "Q4_K_M", LLAMA_FTYPE_MOSTLY_Q4_K_M, " 3.80G, +0.0535 ppl @ 7B", },
{ "Q5_K", LLAMA_FTYPE_MOSTLY_Q5_K_M, "alias for Q5_K_M", },
{ "Q5_K_S", LLAMA_FTYPE_MOSTLY_Q5_K_S, " 4.33G, +0.0353 ppl @ 7B", },
{ "Q5_K_M", LLAMA_FTYPE_MOSTLY_Q5_K_M, " 4.45G, +0.0142 ppl @ 7B", },
{ "Q6_K", LLAMA_FTYPE_MOSTLY_Q6_K, " 5.15G, +0.0044 ppl @ 7B", },
#endif
{
"Q8_0",
LLAMA_FTYPE_MOSTLY_Q8_0,
" 6.70G, +0.0004 ppl @ 7B - very large, extremely low quality loss - not recommended",
},
{
"F16",
LLAMA_FTYPE_MOSTLY_F16,
"13.00G @ 7B - extremely large, virtually no quality loss - not recommended",
},
{
"F32",
LLAMA_FTYPE_ALL_F32,
"26.00G @ 7B - absolutely huge, lossless - not recommended",
},
{ "Q8_0", LLAMA_FTYPE_MOSTLY_Q8_0, " 6.70G, +0.0004 ppl @ 7B", },
{ "F16", LLAMA_FTYPE_MOSTLY_F16, "13.00G @ 7B", },
{ "F32", LLAMA_FTYPE_ALL_F32, "26.00G @ 7B", },
};
examples/server/README.md
+1
View File
@@ -66,6 +66,7 @@ Using [curl](https://curl.se/). On Windows `curl.exe` should be available in the
```sh
curl --request POST \
--url http://localhost:8080/completion \
--header "Content-Type: application/json" \
--data '{"prompt": "Building a website can be done in 10 simple steps:","n_predict": 128}'
```
examples/server/chat.sh
+2
View File
@@ -32,6 +32,7 @@ tokenize() {
--silent \
--request POST \
--url "${API_URL}/tokenize" \
--header "Content-Type: application/json" \
--data-raw "$(jq -ns --arg content "$1" '{content:$content}')" \
| jq '.tokens[]'
}
@@ -64,6 +65,7 @@ chat_completion() {
--no-buffer \
--request POST \
--url "${API_URL}/completion" \
--header "Content-Type: application/json" \
--data-raw "${DATA}")
printf "\n"
examples/server/server.cpp
+18
View File
@@ -608,6 +608,8 @@ static void server_print_usage(const char *argv0, const gpt_params &params,
fprintf(stderr, " -v, --verbose verbose output (default: %s)\n", server_verbose ? "enabled" : "disabled");
fprintf(stderr, " -t N, --threads N number of threads to use during computation (default: %d)\n", params.n_threads);
fprintf(stderr, " -c N, --ctx-size N size of the prompt context (default: %d)\n", params.n_ctx);
fprintf(stderr, " --rope-freq-base N RoPE base frequency (default: %.1f)\n", params.rope_freq_base);
fprintf(stderr, " --rope-freq-scale N RoPE frequency scaling factor (default: %g)\n", params.rope_freq_scale);
fprintf(stderr, " -b N, --batch-size N batch size for prompt processing (default: %d)\n", params.n_batch);
fprintf(stderr, " --memory-f32 use f32 instead of f16 for memory key+value (default: disabled)\n");
fprintf(stderr, " not recommended: doubles context memory required and no measurable increase in quality\n");
@@ -722,6 +724,22 @@ static void server_params_parse(int argc, char **argv, server_params &sparams,
}
params.n_ctx = std::stoi(argv[i]);
}
else if (arg == "--rope-freq-base")
{
if (++i >= argc) {
invalid_param = true;
break;
}
params.rope_freq_base = std::stof(argv[i]);
}
else if (arg == "--rope-freq-scale")
{
if (++i >= argc) {
invalid_param = true;
break;
}
params.rope_freq_scale = std::stof(argv[i]);
}
else if (arg == "--memory-f32" || arg == "--memory_f32")
{
params.memory_f16 = false;