Skip to content

Commit 62be9ba

Browse files
committed
Add new hf protocol for ollama
https://huggingface.co/docs/hub/en/ollama Signed-off-by: Eric Curtin <[email protected]>
1 parent 564804b commit 62be9ba

File tree

1 file changed

+70
-35
lines changed

1 file changed

+70
-35
lines changed

examples/run/run.cpp

Lines changed: 70 additions & 35 deletions
Original file line numberDiff line numberDiff line change
@@ -319,6 +319,10 @@ class HttpClient {
319319
public:
320320
int init(const std::string & url, const std::vector<std::string> & headers, const std::string & output_file,
321321
const bool progress, std::string * response_str = nullptr) {
322+
if (std::filesystem::exists(output_file)) {
323+
return 0;
324+
}
325+
322326
std::string output_file_partial;
323327
curl = curl_easy_init();
324328
if (!curl) {
@@ -563,8 +567,8 @@ class LlamaData {
563567

564568
private:
565569
#ifdef LLAMA_USE_CURL
566-
int download(const std::string & url, const std::vector<std::string> & headers, const std::string & output_file,
567-
const bool progress, std::string * response_str = nullptr) {
570+
int download(const std::string & url, const std::string & output_file, const bool progress,
571+
const std::vector<std::string> & headers = {}, std::string * response_str = nullptr) {
568572
HttpClient http;
569573
if (http.init(url, headers, output_file, progress, response_str)) {
570574
return 1;
@@ -573,57 +577,92 @@ class LlamaData {
573577
return 0;
574578
}
575579
#else
576-
int download(const std::string &, const std::vector<std::string> &, const std::string &, const bool,
580+
int download(const std::string &, const std::string &, const bool, const std::vector<std::string> & = {},
577581
std::string * = nullptr) {
578582
printe("%s: llama.cpp built without libcurl, downloading from an url not supported.\n", __func__);
579583
return 1;
580584
}
581585
#endif
582586

583-
int huggingface_dl(const std::string & model, const std::vector<std::string> headers, const std::string & bn) {
587+
// Helper function to handle model tag extraction and URL construction
588+
std::pair<std::string, std::string> extract_model_and_tag(std::string & model, const std::string & base_url) {
589+
std::string model_tag = "latest";
590+
const size_t colon_pos = model.find(':');
591+
if (colon_pos != std::string::npos) {
592+
model_tag = model.substr(colon_pos + 1);
593+
model = model.substr(0, colon_pos);
594+
}
595+
596+
std::string url = base_url + model + "/manifests/" + model_tag;
597+
598+
return { model, url };
599+
}
600+
601+
// Helper function to download and parse the manifest
602+
int download_and_parse_manifest(const std::string & url, const std::vector<std::string> & headers,
603+
nlohmann::json & manifest) {
604+
std::string manifest_str;
605+
int ret = download(url, "", false, headers, &manifest_str);
606+
if (ret) {
607+
return ret;
608+
}
609+
610+
manifest = nlohmann::json::parse(manifest_str);
611+
612+
return 0;
613+
}
614+
615+
int huggingface_dl(std::string & model, const std::string & bn) {
584616
// Find the second occurrence of '/' after protocol string
585617
size_t pos = model.find('/');
586618
pos = model.find('/', pos + 1);
619+
std::string hfr, hff;
620+
std::vector<std::string> headers = { "User-Agent: llama-cpp", "Accept: application/json" };
621+
std::string url;
622+
587623
if (pos == std::string::npos) {
588-
return 1;
624+
auto [model_name, manifest_url] = extract_model_and_tag(model, "https://huggingface.co/v2/");
625+
hfr = model_name;
626+
627+
nlohmann::json manifest;
628+
int ret = download_and_parse_manifest(manifest_url, headers, manifest);
629+
if (ret) {
630+
return ret;
631+
}
632+
633+
hff = manifest["ggufFile"]["rfilename"];
634+
} else {
635+
hfr = model.substr(0, pos);
636+
hff = model.substr(pos + 1);
589637
}
590638

591-
const std::string hfr = model.substr(0, pos);
592-
const std::string hff = model.substr(pos + 1);
593-
const std::string url = "https://huggingface.co/" + hfr + "/resolve/main/" + hff;
594-
return download(url, headers, bn, true);
639+
url = "https://huggingface.co/" + hfr + "/resolve/main/" + hff;
640+
return download(url, bn, true, headers);
595641
}
596642

597-
int ollama_dl(std::string & model, const std::vector<std::string> headers, const std::string & bn) {
643+
int ollama_dl(std::string & model, const std::string & bn) {
644+
const std::vector<std::string> headers = { "Accept: application/vnd.docker.distribution.manifest.v2+json" };
598645
if (model.find('/') == std::string::npos) {
599646
model = "library/" + model;
600647
}
601648

602-
std::string model_tag = "latest";
603-
size_t colon_pos = model.find(':');
604-
if (colon_pos != std::string::npos) {
605-
model_tag = model.substr(colon_pos + 1);
606-
model = model.substr(0, colon_pos);
607-
}
608-
609-
std::string manifest_url = "https://registry.ollama.ai/v2/" + model + "/manifests/" + model_tag;
610-
std::string manifest_str;
611-
const int ret = download(manifest_url, headers, "", false, &manifest_str);
649+
auto [model_name, manifest_url] = extract_model_and_tag(model, "https://registry.ollama.ai/v2/");
650+
nlohmann::json manifest;
651+
int ret = download_and_parse_manifest(manifest_url, {}, manifest);
612652
if (ret) {
613653
return ret;
614654
}
615655

616-
nlohmann::json manifest = nlohmann::json::parse(manifest_str);
617-
std::string layer;
656+
std::string layer;
618657
for (const auto & l : manifest["layers"]) {
619658
if (l["mediaType"] == "application/vnd.ollama.image.model") {
620659
layer = l["digest"];
621660
break;
622661
}
623662
}
624663

625-
std::string blob_url = "https://registry.ollama.ai/v2/" + model + "/blobs/" + layer;
626-
return download(blob_url, headers, bn, true);
664+
std::string blob_url = "https://registry.ollama.ai/v2/" + model_name + "/blobs/" + layer;
665+
return download(blob_url, bn, true, headers);
627666
}
628667

629668
std::string basename(const std::string & path) {
@@ -653,22 +692,18 @@ class LlamaData {
653692
return ret;
654693
}
655694

656-
const std::string bn = basename(model_);
657-
const std::vector<std::string> headers = { "--header",
658-
"Accept: application/vnd.docker.distribution.manifest.v2+json" };
695+
const std::string bn = basename(model_);
659696
if (string_starts_with(model_, "hf://") || string_starts_with(model_, "huggingface://")) {
660697
rm_until_substring(model_, "://");
661-
ret = huggingface_dl(model_, headers, bn);
698+
ret = huggingface_dl(model_, bn);
662699
} else if (string_starts_with(model_, "hf.co/")) {
663700
rm_until_substring(model_, "hf.co/");
664-
ret = huggingface_dl(model_, headers, bn);
665-
} else if (string_starts_with(model_, "ollama://")) {
666-
rm_until_substring(model_, "://");
667-
ret = ollama_dl(model_, headers, bn);
701+
ret = huggingface_dl(model_, bn);
668702
} else if (string_starts_with(model_, "https://")) {
669-
ret = download(model_, headers, bn, true);
670-
} else {
671-
ret = ollama_dl(model_, headers, bn);
703+
ret = download(model_, bn, true);
704+
} else { // ollama:// or nothing
705+
rm_until_substring(model_, "://");
706+
ret = ollama_dl(model_, bn);
672707
}
673708

674709
model_ = bn;

0 commit comments

Comments
 (0)