diff options
| -rw-r--r-- | README.md | 2 | ||||
| -rwxr-xr-x | ci/run.sh | 4 | ||||
| -rw-r--r-- | examples/perplexity/perplexity.cpp | 4 | ||||
| -rw-r--r-- | ggml-quants.c | 6 | ||||
| -rwxr-xr-x | scripts/get-wikitext-2.sh | 2 |
5 files changed, 9 insertions, 9 deletions
@@ -768,7 +768,7 @@ The time per token is measured on a MacBook M1 Pro 32GB RAM using 4 and 8 thread #### How to run -1. Download/extract: https://s3.amazonaws.com/research.metamind.io/wikitext/wikitext-2-raw-v1.zip?ref=salesforce-research +1. Download/extract: https://huggingface.co/datasets/ggml-org/ci/resolve/main/wikitext-2-raw-v1.zip 2. Run `./perplexity -m models/7B/ggml-model-q4_0.gguf -f wiki.test.raw` 3. Output: ``` @@ -219,7 +219,7 @@ function gg_run_open_llama_3b_v2 { gg_wget models-mnt/open-llama/3B-v2/ https://huggingface.co/openlm-research/open_llama_3b_v2/resolve/main/pytorch_model.bin gg_wget models-mnt/open-llama/3B-v2/ https://huggingface.co/openlm-research/open_llama_3b_v2/raw/main/generation_config.json - gg_wget models-mnt/wikitext/ https://s3.amazonaws.com/research.metamind.io/wikitext/wikitext-2-raw-v1.zip + gg_wget models-mnt/wikitext/ https://huggingface.co/datasets/ggml-org/ci/resolve/main/wikitext-2-raw-v1.zip unzip -o models-mnt/wikitext/wikitext-2-raw-v1.zip -d models-mnt/wikitext/ head -n 60 models-mnt/wikitext/wikitext-2-raw/wiki.test.raw > models-mnt/wikitext/wikitext-2-raw/wiki.test-60.raw @@ -401,7 +401,7 @@ function gg_run_open_llama_7b_v2 { gg_wget models-mnt/open-llama/7B-v2/ https://huggingface.co/openlm-research/open_llama_7b_v2/resolve/main/pytorch_model-00002-of-00002.bin gg_wget models-mnt/open-llama/7B-v2/ https://huggingface.co/openlm-research/open_llama_7b_v2/raw/main/generation_config.json - gg_wget models-mnt/wikitext/ https://s3.amazonaws.com/research.metamind.io/wikitext/wikitext-2-raw-v1.zip + gg_wget models-mnt/wikitext/ https://huggingface.co/datasets/ggml-org/ci/resolve/main/wikitext-2-raw-v1.zip unzip -o models-mnt/wikitext/wikitext-2-raw-v1.zip -d models-mnt/wikitext/ path_models="../models-mnt/open-llama/7B-v2" diff --git a/examples/perplexity/perplexity.cpp b/examples/perplexity/perplexity.cpp index 74dcc642..9ec98938 100644 --- a/examples/perplexity/perplexity.cpp +++ b/examples/perplexity/perplexity.cpp @@ -309,7 +309,7 @@ static void process_logits(int n_vocab, const float * logits, const int * tokens } static results_perplexity perplexity_v2(llama_context * ctx, const gpt_params & params) { - // Download: https://s3.amazonaws.com/research.metamind.io/wikitext/wikitext-2-raw-v1.zip?ref=salesforce-research + // Download: https://huggingface.co/datasets/ggml-org/ci/resolve/main/wikitext-2-raw-v1.zip // Run `./perplexity -m models/7B/ggml-model-q4_0.bin -f wiki.test.raw` // Output: `perplexity: 13.5106 [114/114]` // BOS tokens will be added for each chunk before eval @@ -447,7 +447,7 @@ static results_perplexity perplexity(llama_context * ctx, const gpt_params & par return perplexity_v2(ctx, params); } - // Download: https://s3.amazonaws.com/research.metamind.io/wikitext/wikitext-2-raw-v1.zip?ref=salesforce-research + // Download: https://huggingface.co/datasets/ggml-org/ci/resolve/main/wikitext-2-raw-v1.zip // Run `./perplexity -m models/7B/ggml-model-q4_0.bin -f wiki.test.raw` // Output: `perplexity: 13.5106 [114/114]` // BOS tokens will be added for each chunk before eval diff --git a/ggml-quants.c b/ggml-quants.c index 48f5294e..43a8f1de 100644 --- a/ggml-quants.c +++ b/ggml-quants.c @@ -1837,9 +1837,9 @@ static void quantize_row_q2_K_impl(const float * restrict x, block_q2_K * restri float sigma2 = sumx2/QK_K; for (int j = 0; j < QK_K/16; ++j) { const float * restrict qw = quant_weights + QK_K * i + 16*j; - for (int l = 0; l < 16; ++l) weight[l] = qw[l] * sqrtf(sigma2 + x[16*j + l]*x[16*j + l]); - for (int l = 0; l < 16; ++l) sw[j] += weight[l]; - scales[j] = make_qkx3_quants(16, 3, x + 16*j, weight, L + 16*j, &mins[j], Laux, -0.9f, 0.05f, 36, false); + for (int l = 0; l < QK_K/16; ++l) weight[l] = qw[l] * sqrtf(sigma2 + x[16*j + l]*x[16*j + l]); + for (int l = 0; l < QK_K/16; ++l) sw[j] += weight[l]; + scales[j] = make_qkx3_quants(QK_K/16, 3, x + 16*j, weight, L + 16*j, &mins[j], Laux, -0.9f, 0.05f, 36, false); } float dm = make_qp_quants(QK_K/16, 15, scales, Ls, sw); diff --git a/scripts/get-wikitext-2.sh b/scripts/get-wikitext-2.sh index ff96f331..7ca760fa 100755 --- a/scripts/get-wikitext-2.sh +++ b/scripts/get-wikitext-2.sh @@ -1,6 +1,6 @@ #!/bin/bash -wget https://s3.amazonaws.com/research.metamind.io/wikitext/wikitext-2-raw-v1.zip +wget https://huggingface.co/datasets/ggml-org/ci/resolve/main/wikitext-2-raw-v1.zip echo "Usage:" echo "" |