tcsenpai/ollama
1
0
Fork 0
mirror of https://github.com/tcsenpai/ollama.git synced 2025-06-09 20:47:07 +00:00
Code Issues Packages Projects Releases Wiki Activity

runner.go: Cleanups

Browse Source 
Variable names in Go style, removing commented out code and typo fixes
This commit is contained in:
Jesse Gross 2024-08-27 13:19:46 -07:00 committed by jmorganca
parent d022cfc9e6
commit 4ca8579428
1 changed files with 16 additions and 18 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

34
llama/runner/runner.go
View File

@ -65,10 +65,10 @@ type Sequence struct {
doneReason string doneReason string
// Metrics // Metrics
t_start_process_prompt time.Time startProcessingTime time.Time
t_start_genereration time.Time startGenerationTime time.Time
n_decoded int numDecoded int
n_prompt_tokens int numPromptTokens int
} }
type NewSequenceParams struct { type NewSequenceParams struct {
@ -110,7 +110,7 @@ func (s *Server) NewSequence(prompt string, params NewSequenceParams) *Sequence
return &Sequence{ return &Sequence{
tokens: tokens, tokens: tokens,
n_prompt_tokens: len(tokens), numPromptTokens: len(tokens),
numPredict: params.numPredict, numPredict: params.numPredict,
pendingResponses: make([]string, 0), pendingResponses: make([]string, 0),
responses: make(chan string, 1), responses: make(chan string, 1),
@ -256,8 +256,8 @@ func (s *Server) processBatch() {
s.shiftContext(i) s.shiftContext(i)
} }
if seq.t_start_process_prompt.IsZero() { if seq.startProcessingTime.IsZero() {
seq.t_start_process_prompt = time.Now() seq.startProcessingTime = time.Now()
} }
var numTokensProcessed int var numTokensProcessed int
@ -294,7 +294,7 @@ func (s *Server) processBatch() {
continue continue
} }
// if done processing the prompt, generating an embedding and return // if done processing the prompt, generate an embedding and return
if seq.embeddingOnly { if seq.embeddingOnly {
embd := s.lc.GetEmbeddingsSeq(i) embd := s.lc.GetEmbeddingsSeq(i)
if embd == nil { if embd == nil {
@ -307,14 +307,12 @@ func (s *Server) processBatch() {
} }
// sample a token // sample a token
// logits := s.lc.GetLogitsIth(ibatch[i])
// token := s.lc.SampleTokenGreedy(logits)
token := seq.samplingCtx.Sample(s.lc, nil, seq.iBatch) token := seq.samplingCtx.Sample(s.lc, nil, seq.iBatch)
seq.samplingCtx.Accept(s.lc, token, true) seq.samplingCtx.Accept(s.lc, token, true)
seq.n_decoded += 1 seq.numDecoded += 1
if seq.n_decoded == 1 { if seq.numDecoded == 1 {
seq.t_start_genereration = time.Now() seq.startGenerationTime = time.Now()
} }
piece := s.model.TokenToPiece(token) piece := s.model.TokenToPiece(token)
@ -505,10 +503,10 @@ func (s *Server) completion(w http.ResponseWriter, r *http.Request) {
if err := json.NewEncoder(w).Encode(&CompletionResponse{ if err := json.NewEncoder(w).Encode(&CompletionResponse{
Stop: true, Stop: true,
Timings: Timings{ Timings: Timings{
PromptN: seq.n_prompt_tokens, PromptN: seq.numPromptTokens,
PromptMS: float64(seq.t_start_genereration.Sub(seq.t_start_process_prompt).Milliseconds()), PromptMS: float64(seq.startGenerationTime.Sub(seq.startProcessingTime).Milliseconds()),
PredictedN: seq.n_decoded, PredictedN: seq.numDecoded,
PredictedMS: float64(time.Since(seq.t_start_genereration).Milliseconds()), PredictedMS: float64(time.Since(seq.startGenerationTime).Milliseconds()),
}, },
}); err != nil { }); err != nil {
log.Println("Failed to encode result:", err) log.Println("Failed to encode result:", err)
@ -638,7 +636,7 @@ func main() {
// TODO actually implement... // TODO actually implement...
if *embedding { if *embedding {
slog.Warn("embeddings not yet support") slog.Warn("embeddings not yet supported")
} }
if *logDisable { if *logDisable {
slog.Info("ignoring --log-disable") slog.Info("ignoring --log-disable")